[Federal Register Volume 61, Number 79 (Tuesday, April 23, 1996)]
[Notices]
[Pages 17960-18011]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 96-9711]




[[Page 17959]]


_______________________________________________________________________

Part II





Environmental Protection Agency





_______________________________________________________________________



Proposed Guidelines for Carcinogen Risk Assessment; Notice

  Federal Register / Vol. 61, No. 79 / Tuesday, April 23, 1996 / 
Notices  

[[Page 17960]]



ENVIRONMENTAL PROTECTION AGENCY

[FRL-5460-3]


Proposed Guidelines for Carcinogen Risk Assessment

AGENCY: Environmental Protection Agency (EPA).

ACTION: Notice of Availability and Opportunity to Comment on Proposed 
Guidelines for Carcinogen Risk Assessment.

-----------------------------------------------------------------------

SUMMARY: The U.S. Environmental Protection Agency (EPA) is today 
publishing a document entitled Proposed Guidelines for Carcinogen Risk 
Assessment (hereafter ``Proposed Guidelines''). These Proposed 
Guidelines were developed as part of an interoffice guidelines 
development program by a Technical Panel of the Risk Assessment Forum 
within EPA's Office of Research and Development. These Proposed 
Guidelines are a revision of EPA's 1986 Guidelines for Carcinogen Risk 
Assessment (hereafter ``1986 cancer guidelines'') published on 
September 24, 1986 (51 FR 33992). When final, these guidelines will 
replace the 1986 guidelines.
    In a future Federal Register notice, the Agency intends to publish 
for comment how it will implement the Proposed Guidelines once they are 
finalized. The plans will propose and seek comment on how the 
Guidelines will be used for Agency carcinogen risk assessment and, in 
particular, will address the impact of the Guidelines on the Agency's 
existing assessments, and any mechanisms for handling reassessments 
under finalized Guidelines.

DATES: The Proposed Guidelines are being made available for a 120-day 
public review and comment period. Comments must be in writing and must 
be postmarked by August 21, 1996. See Addresses section for guidance on 
submitting comments.

ADDRESSES: The Proposed Guidelines will be made available in the 
following ways:
    (1) The electronic version will be accessible on EPA's Office of 
Research and Development home page on the Internet at http://
www.epa.gov/ORD
    (2) 3\1/2\'' high-density computer diskettes in Wordperfect 5.1 
format will be available from ORD Publications, Technology Transfer and 
Support Division, National Risk Management Research Laboratory, 
Cincinnati, OH; telephone: 513-569-7562; fax: 513-569-7566. Please 
provide the EPA No. (EPA/600/P-92/003Ca) when ordering.
    (3) This notice contains the full draft document. In addition, 
copies of the draft will be available for inspection at EPA 
headquarters and regional libraries, through the U.S. Government 
Depository Library program, and for purchase from the National 
Technical Information Service (NTIS), Springfield, VA; telephone: 703-
487-4650, fax: 703-321-8547. Please provide the NTIS PB No. (PB96-
157599) ($35.00) when ordering.
SUBMITTING COMMENTS: Comments on the Proposed Guidelines may be mailed 
or delivered to the Technical Information Staff (8623), NCEA-WA/OSG, 
U.S. Environmental Protection Agency, 401 M Street, S.W., Washington, 
DC 20460. Comments should be in writing and must be postmarked by the 
date indicated. Please submit one unbound original with pages numbered 
consecutively, and three copies. For attachments, provide an index, 
number pages consecutively with the comment, and submit an unbound 
original and three copies.
    Please note that all technical comments received in response to 
this notice will be placed in a public record. For that reason, 
commenters should not submit personal information (such as medical data 
or home address), Confidential Business Information, or information 
protected by copyright. Due to limited resources, acknowledgments will 
not be sent.

FOR FURTHER INFORMATION CONTACT: Technical Information Staff, 
Operations and Support Group, National Center for Environmental 
Assessment--Washington Office, telephone: 202-260-7345. Email inquiries 
may be sent to [email protected].

SUPPLEMENTARY INFORMATION: In 1983, the National Academy of Sciences 
(NAS)/National Research Council (NRC) published its report entitled 
Risk Assessment in the Federal Government: Managing the Process (NRC, 
1983). In that report, the NRC recommended that Federal regulatory 
agencies establish ``inference guidelines'' to ensure consistency and 
technical quality in risk assessments and to ensure that the risk 
assessment process was maintained as a scientific effort separate from 
risk management. The 1986 cancer guidelines were issued on September 
24, 1986 (51 FR 33992). The Proposed Guidelines published today 
continue the guidelines development process. These guidelines set forth 
principles and procedures to guide EPA scientists in the conduct of 
Agency cancer risk assessments and to inform Agency decisionmakers and 
the public about these procedures.
    Both the 1986 guidelines and the current proposal contain inference 
guidance in the form of default inferences to bridge gaps in knowledge 
and data. Research conducted in the past decade has elucidated much 
about the nature of carcinogenic processes and continues to provide new 
information. The intent of this proposal is to take account of 
knowledge available now and to provide flexibility for the future in 
assessing data and employing default inferences, recognizing that the 
guidelines cannot always anticipate future research findings. Because 
methods and knowledge are expected to change more rapidly than 
guidelines can practicably be revised, the Agency will update specific 
assessment procedures with peer-reviewed supplementary, technical 
documents as needed. Further revision of the guidelines themselves will 
take place when extensive changes are necessary.
    Since 1986, the EPA has sponsored several workshops about revising 
the cancer guidelines (U.S. EPA, 1989b, 1989c, 1994a). The Society for 
Risk Analysis conducted a workshop on the subject in connection with 
its 1992 annual meeting (Anderson et al., 1993). Participants in the 
most recent workshop in 1994 reviewed an earlier version of the 
guidelines proposed here and made numerous recommendations about 
individual issues as well as broad recommendations about explanations 
and perspectives that should be added. Most recently, the Committee on 
the Environment and Natural Resources of the Office of Science and 
Technology Policy reviewed the guidelines at a meeting held on August 
15, 1995. The EPA appreciates the efforts of all participants in the 
process and has tried to address their recommendations in this 
proposal.
    In addition, the recommendations of the NRC (1994) in Science and 
Judgment in Risk Assessment have been addressed. Responses to these 
recommendations are given generally in Appendix B as well as being 
embodied in the Proposed Guidelines. Responses that explain the major 
default assumptions adopted under these guidelines and the policy for 
using and departing from these default assumptions appear in Section 
1.3.
    The Science Advisory Board also will review these Proposed 
Guidelines at a meeting to be announced in a future Federal Register 
notice. Following these reviews Agency staff will prepare summaries of 
the public and SAB comments. Appropriate comments will be incorporated, 
and the revised Guidelines will be submitted to the Risk Assessment 
Forum for review. The

[[Page 17961]]

Agency will consider comments from the public, the SAB, and the Risk 
Assessment Forum in its recommendations to the EPA Administrator.

Major Changes From the 1986 Guidelines

Characterizations

    Increased emphasis on providing characterization discussions for 
the hazard, dose response, and exposure sections is part of the 
proposal. These discussions will summarize the assessments to explain 
the extent and weight of evidence, major points of interpretation and 
rationale, and strengths and weaknesses of the evidence and the 
analysis, and to discuss alternative conclusions and uncertainties that 
deserve serious consideration (U.S. EPA, 1995). They serve as starting 
materials for the risk characterization process which completes the 
risk assessment.

Weighing Evidence of Hazard

    A major change is in the way hazard evidence is weighed in reaching 
conclusions about the human carcinogenic potential of agents. In the 
1986 cancer guidelines, tumor findings in animals or humans were the 
dominant components of decisions. Other information about an agent's 
properties, its structure-activity relationships to other carcinogenic 
agents, and its activities in studies of carcinogenic processes was 
often limited and played only a modulating role as compared with tumor 
findings. In this proposal, decisions come from considering all of the 
evidence. This change recognizes the growing sophistication of research 
methods, particularly in their ability to reveal the modes of action of 
carcinogenic agents at cellular and subcellular levels as well as 
toxicokinetic and metabolic processes. The effect of the change on the 
assessment of individual agents will depend greatly on the availability 
of new kinds of data on them in keeping with the state of the art. If 
these new kinds of data are not forthcoming from public and private 
research on agents, assessments under these guidelines will not differ 
significantly from assessments under former guidelines.
    Weighing of the evidence includes addressing the likelihood of 
human carcinogenic effects of the agent and the conditions under which 
such effects may be expressed, as these are revealed in the 
toxicological and other biologically important features of the agent. 
(Consideration of actual human exposure and risk implications are done 
separately; they are not parts of the hazard characterization). In this 
respect, the guidelines incorporate recommendations of the NRC (1994). 
In that report, the NRC recommends expansion of the former concept of 
hazard identification, which rests on simply a finding of carcinogenic 
potential, to a concept of characterization that includes dimensions of 
the expression of this potential. For example, an agent might be 
observed to be carcinogenic via inhalation exposure and not via oral 
exposure, or its carcinogenic activity might be secondary to another 
toxic effect. In addition, the consideration of evidence includes the 
mode(s) of action of the agent apparent from the available data as a 
basis for approaching dose response assessment.

Classification Descriptors

    To express the weight of evidence for carcinogenic hazard 
potential, the 1986 cancer guidelines provided summary rankings for 
human and animal cancer studies. These summary rankings were integrated 
to place the overall evidence in classification groups A through E, 
Group A being associated with the greatest probability of human 
carcinogenicity and Group E with evidence of noncarcinogenicity in 
humans. Data other than tumor findings played a modifying role after 
initial placement of an agent into a group.
    These Proposed Guidelines take a different approach, consistent 
with the change in the basic approach to weighing evidence. No interim 
classification of tumor findings followed by modifications with other 
data takes place. Instead, the conclusion reflects the weighing of 
evidence in one step. Moreover, standard descriptors of conclusions are 
employed rather than letter designations, and these are incorporated 
into a brief narrative description of their informational basis. The 
narrative with descriptors replaces the previous letter designation. 
The descriptors are in three categories: ``known/likely,'' ``cannot be 
determined,'' or ``not likely.'' For instance, using a descriptor in 
context, a narrative could say that an agent is likely to be 
carcinogenic by inhalation exposure and not likely to be carcinogenic 
by oral exposure. The narrative explains the kinds of evidence 
available and how they fit together in drawing conclusions, and points 
out significant issues/strengths/limitations of the data and 
conclusions. Subdescriptors are used to further refine the conclusion. 
The narrative also summarizes the mode of action information underlying 
a recommended approach to dose response assessment.
    In considering revision of the former classification method, the 
Agency has examined other possibilities that would retain the use of 
letter and number designation of weights of evidence. The use of 
standard descriptors within a narrative presentation is proposed for 
three primary reasons. First, the proposed method permits inclusion of 
explanations of data and of their strengths and limitations. This is 
more consistent with current policy emphasis on risk characterization. 
Second, it would take a large set of individual number or letter codes 
to cover differences in the nature of contributing information (animal, 
human, other), route of exposure, mode of action, and relative overall 
weight. When such a set becomes large--10 to 30 codes--it is too large 
to be a good communication device, because people cannot remember the 
definitions of the codes so they have to be explained in narrative. 
Third, it is impossible to predefine the course of cancer research and 
the kinds of data that may become available. A flexible system is 
needed to accommodate change in the underlying data and inferences, and 
a system of codes might become out of date, as has the one in the 1986 
cancer guidelines.

Dose Response Assessment

    The approach to dose response assessment calls for analysis that 
follows the conclusions reached in the hazard assessment as to 
potential mode(s) of action. The assessment begins by analyzing the 
empirical data in the range of observation. When animal studies are the 
basis of the analysis, the estimation of a human equivalent dose 
utilizes toxicokinetic data, if appropriate and adequate data are 
available. Otherwise, default procedures are applied. For oral dose, 
the default is to scale daily applied doses experienced for a lifetime 
in proportion to body weight raised to the 0.75 power. For inhalation 
dose, the default methodology estimates respiratory deposition of 
particles and gases and estimates internal doses of gases with 
different absorption characteristics. These two defaults are a change 
from the 1986 cancer guidelines which provided a single scaling factor 
of body weight raised to the 0.66 power. Another change from the 1986 
guidelines is that response data on effects of the agent on 
carcinogenic processes are analyzed (nontumor data) in addition to data 
on tumor incidence. If appropriate, the analyses of data on tumor 
incidence and on precursor effects may be combined, using

[[Page 17962]]

precursor data to extend the dose response curve below the tumor data. 
Even if combining data is not appropriate, study of the dose response 
for effects believed to be part of the carcinogenic influence of the 
agent may assist in thinking about the relationship of exposure and 
response in the range of observation and at exposure levels below the 
range of observation.
    Whenever data are sufficient, a biologically based or case-specific 
dose response model is developed to relate dose and response data in 
the range of empirical observation. Otherwise, as a standard, default 
procedure, a model is used to curve-fit the data. The lower 95% 
confidence limit on a dose associated with an estimated 10% increased 
tumor or relevant nontumor response (LED10) is identified. This 
generally serves as the point of departure for extrapolating the 
relationship to environmental exposure levels of interest when the 
latter are outside the range of observed data. The environmental 
exposures of interest may be measured ones or levels of risk management 
interest in considering potential exposure control options. Other 
points of departure may be more appropriate for certain data sets; as 
described in the guidance, these may be used instead of the LED10. 
Additionally, the LED10 is available for comparison with parallel 
analyses of other carcinogenic agents or of noncancer effects of agents 
and for gauging and explaining the magnitude of subsequent 
extrapolation to low-dose levels. The LED10, rather than the 
ED10 (the estimate of a 10% increased response), is the proposed 
standard point of departure for two reasons. One is to permit easier 
comparison with the benchmark dose procedure for noncancer health 
assessment--also based on the lower limit on dose. Another is that the 
lower limit, as opposed to the central estimate, accounts for 
uncertainty in the experimental data. The issue of using a lower limit 
or central estimate was discussed at a workshop held on the benchmark 
procedure for noncancer assessment (Barnes et al., 1995) and at a 
workshop on a previous version of this proposal (U.S. EPA, 1994b). The 
latter workshop recommended a central estimate; the benchmark workshop 
recommended a lower limit.
    The second step of dose response assessment is extrapolation to 
lower dose levels, if needed. This is based on a biologically based or 
case-specific model if supportable by substantial data. Otherwise, 
default approaches are applied that accord with the view of mode(s) of 
action of the agent. These include approaches that assume linearity or 
nonlinearity of the dose response relationship or both. The default 
approach for linearity is to extend a straight line to zero dose, zero 
response. The default approach for nonlinearity is to use a margin of 
exposure analysis rather than estimating the probability of effects at 
low doses. A margin of exposure analysis explains the biological 
considerations for comparing the observed data with the environmental 
exposure levels of interest and helps in deciding on an acceptable 
level of exposure in accordance with applicable management factors.
    The use of straight line extrapolation for a linear default is a 
change from the 1986 guidelines which used the ``linearized 
multistage'' (LMS) procedure. This change is made because the former 
modeling procedure gave an appearance of specific knowledge and 
sophistication unwarranted for a default. The proposed approach is also 
more like that employed by the Food and Drug Administration (U.S. FDA, 
1987). The numerical results of the straight line and LMS procedures 
are not significantly different (Krewski et al., 1984). The use of a 
margin of exposure approach is included as a new default procedure to 
accommodate cases in which there is sufficient evidence of a nonlinear 
dose response, but not enough evidence to construct a mathematical 
model for the relationship. (The Agency will continue to seek a 
modeling method to apply in these cases. If a modeling approach is 
developed, it will be subject to peer review and public notice in the 
context of a supplementary document for these guidelines.)
    The public is invited to provide comments to be considered in EPA 
decisions about the content of the final guidelines. After the public 
comment period, the EPA Science Advisory Board will be asked to review 
and provide advice on the guidelines and issues raised in comments. EPA 
asks those who respond to this notice to include their views on the 
following:
    (1) The proposed guidance for characterization of hazard, including 
the weight of evidence descriptors and weight of evidence narrative 
which are major features of the proposal. There are three categories of 
descriptors: ``known/likely,'' ``cannot be determined,'' and ``not 
likely'' which are further refined by subdescriptors. It is felt that 
these three descriptors will satisfactorily delineate the types of 
evidence bearing on carcinogenicity as they are used with 
subdescriptors in the context of a narrative of data and rationale. 
However, an issue that has been discussed by external peer reviewers 
and by EPA staff is whether the descriptor-subdescriptor called 
``cannot be determined--suggestive evidence'' should become a separate, 
fourth category called ``suggestive.'' The EPA may choose this course 
in the final guidelines and requests comment. In considering this 
issue, commenters may wish to refer not only to Sections 2.6.2. and 
2.7.2. which cover the descriptors and narrative, but also to case 
study example #6 in Section 2.6.3. and example narrative #2 in Appendix 
A of the proposal. EPA asks commenters on this question to address the 
rationale (science as well as policy) for leaving the categories of 
descriptors as proposed or making the fourth category. How might the 
coverage of a ``suggestive'' category be defined in order to be most 
useful?
    (2) The use of mode of action information in hazard 
characterization and to guide dose response assessment is a central 
part of the proposed approach to bringing new research on carcinogenic 
processes to bear in assessments of environmental agents (Sections 
1.3.2., 2.3.2., 2.5., 3.1.). The appropriate use of this information 
now and in the future is important. EPA requests comment on the 
treatment of such information in the proposal, including reliance on 
peer review as a part of the judgmental process on its application.
    (3) Uses of nontumor data in the dose response assessment and the 
methodological and science policy issues posed are new to these 
guidelines (Sections 1.3.2., 3.1.2.). EPA requests comment on both 
issues.
    (4) Dose response assessment is proposed to be considered in two 
parts--range of observed data and range of extrapolation (Section 
3.1.). The lower 95% confidence limit on a dose associated with a 10% 
response (tumor or nontumor response) is proposed as a default point of 
departure, marking the beginning of extrapolation. This is a parallel 
to the benchmark procedure for evaluating dose-response of noncancer 
health endpoints (Barnes et al., 1995). An alternative is to use the 
central estimate of a 10% response. Another alternative is to use a 1%, 
instead of a 10%, response when the observed data are tumor incidence 
data. Does the generally larger sample size of tumor effect studies 
support using a 1% response as compared with using 10% for smaller 
studies? Are there other approaches for the point of departure that 
might be considered?
    (5) Discussions of default assumptions and other responses to the 
1994 NRC report Science and Judgment in Risk

[[Page 17963]]

Assessment appear in Section 1.3.1. and Appendix B of the proposal, 
respectively. Comments are requested on responses to the NRC 
recommendations and how the guidelines as a whole address them.
    Dated: April 10, 1996.
Carol M. Browner,
Administrator.

Contents

List of Figures
1. Introduction
    1.1. Purpose and Scope of the Guidelines
    1.2. Organization and Application of the Guidelines
    1.2.1. Organization
    1.2.2. Application
    1.3. Use of Default Assumptions
    1.3.1. Default Assumptions
    1.3.2. Major Defaults
    1.3.2.1. Is the Presence or Absence of Effects Observed in a 
Human Population Predictive of Effects in Another Exposed Human 
Population?
    1.3.2.2. Is the Presence or Absence of Effects Observed in an 
Animal Population Predictive of Effects in Exposed Humans?
    1.3.2.3. How Do Metabolic Pathways Relate Across Species?
    1.3.2.4. How Do Toxicokinetic Processes Relate Across Species?
    1.3.2.5. What Is the Correlation of the Observed Dose Response 
Relationship to the Relationship at Lower Doses?
    1.4. Characterizations
2. Hazard Assessment
    2.1. Overview of Hazard Assessment and Characterization
    2.1.1. Analyses of Data
    2.1.2. Cross-Cutting Topics for Data Integration
    2.1.2.1. Conditions of Expression
    2.1.2.2. Mode of Action
    2.1.3. Presentation of Results
    2.2. Analysis of Tumor Data
    2.2.1. Human Data
    2.2.1.1. Types of Studies
    2.2.1.2. Criteria for Assessing Adequacy of Epidemiologic 
Studies
    2.2.1.3. Criteria for Causality
    2.2.1.4. Assessment of Evidence of Carcinogenicity from Human 
Data
    2.2.2. Animal Data
    2.2.2.1. Long-Term Carcinogenicity Studies
    2.2.2.2. Other Studies
    2.2.3. Structural Analogue Data
    2.3. Analysis of Other Key Data
    2.3.1. Physicochemical Properties
    2.3.2. Structure-Activity Relationships
    2.3.3. Comparative Metabolism and Toxicokinetics
    2.3.4. Toxicological and Clinical Findings
    2.3.5. Mode of Action-Related Endpoints and Short-Term Tests
    2.3.5.1. Direct DNA Effects
    2.3.5.2. Secondary DNA Effects
    2.3.5.3. Nonmutagenic and Other Effects
    2.3.5.4. Criteria for Judging Mode of Action
    2.4. Biomarker Information
    2.5. Mode of Action--Implications for Hazard Characterization 
and Dose Response
    2.6. Weight of Evidence Evaluation for Potential Human 
Carcinogenicity
    2.6.1. Weight of Evidence Analysis
    2.6.2. Descriptors for Classifying Weight of Evidence
    2.6.3. Case Study Examples
    2.7. Presentation of Results
    2.7.1. Technical Hazard Characterization
    2.7.2. Weight of Evidence Narrative
3. Dose Response Assessment
    3.1. Dose Response Relationship
    3.1.1. Analysis in the Range of Observation
    3.1.2. Analysis in the Range of Extrapolation
    3.1.3. Use of Toxicity Equivalence Factors and Relative Potency 
Estimates
    3.2. Response Data
    3.3. Dose Data
    3.3.1. Interspecies Adjustment of Dose
    3.3.2. Toxicokinetic Analyses
    3.3.3. Route-to-Route Extrapolation
    3.3.4. Dose Averaging
    3.4. Discussion of Uncertainties
    3.5. Technical Dose Response Characterization
4. Technical Exposure Characterization
5. Risk Characterization
    5.1. Purpose
    5.2. Application
    5.3. Presentation of Risk Characterization Summary
    5.4. Content of Risk Characterization Summary
Appendix A
Appendix B
Appendix C
References

List of Figures

Figure 1-1. Decisions on Dose Response Assessment Approaches for the 
Range of Extrapolation
Figure 1-2. Risk Characterization
Figure 2-1. Factors for Weighing Human Evidence
Figure 2-2. Factors for Weighing Animal Evidence
Figure 2-3. Factors for Weighing Other Key Evidence
Figure 2-4. Factors for Weighing Totality of Evidence
Figure 3-1. Graphical Presentation of Data and Extrapolations

1. Introduction

1.1. Purpose and Scope of the Guidelines

    These guidelines revise and replace United States Environmental 
Protection Agency (EPA) Guidelines for Carcinogen Risk Assessment 
published in 51 FR 33992, September 24, 1986. The guidelines provide 
EPA staff and decisionmakers with guidance and perspectives to develop 
and use risk assessments. They also provide basic information to the 
public about the Agency's risk assessment methods.
    The guidelines encourage both regularity in procedures to support 
consistency in scientific components of Agency decisionmaking and 
innovation to remain up-to-date in scientific thinking. In balancing 
these goals, the Agency relies on input from the general scientific 
community through established scientific peer review processes. The 
guidelines incorporate basic principles and science policies based on 
evaluation of the currently available information. As more is 
discovered about carcinogenesis, the need will arise to make 
appropriate changes in risk assessment guidance. The Agency will revise 
these guidelines when extensive changes are due. In the interim, the 
Agency will issue special reports, after appropriate peer review, to 
supplement and update guidance on single topics, (e.g., U.S. EPA, 
1991b)

1.2. Organization and Application of the Guidelines

1.2.1. Organization
    Publications of the Office of Science and Technology Policy (OSTP, 
1985) and the National Research Council (NRC, 1983, 1994) provide 
information and general principles about risk assessment. Risk 
assessment uses available scientific information on the properties of 
an agent \1\ and its effects in biological systems to provide an 
evaluation of the potential for harm as a consequence of environmental 
exposure to the agent. Risk assessment is one of the scientific 
analyses available for consideration, with other analyses, in 
decisionmaking on environmental protection. The 1983 and 1994 NRC 
documents organize risk assessment information into four areas: hazard 
identification, dose response assessment, exposure assessment, and risk 
characterization. This structure appears in these guidelines, which 
additionally emphasize characterization of evidence and conclusions in 
each part of the assessment. In particular, the guidelines adopt the 
approach of the NRC's 1994 report in adding a dimension of 
characterization to the hazard identification step. Added to the 
identification of hazard is an evaluation of the conditions under which 
its expression is anticipated. The risk assessment questions addressed 
in these guidelines are:
---------------------------------------------------------------------------

    \1\ The term ``agent'' refers generally to any chemical 
substance, mixture, or physical or biological entity being assessed, 
unless otherwise noted.
---------------------------------------------------------------------------

     For hazard--Can the agent present a carcinogenic hazard to 
humans, and if so, under what circumstances?
     For dose response--At what levels of exposure might 
effects occur?
     For exposure--What are the conditions of human exposure?
     For risk--What is the character of the risk? How well do 
data support conclusions about the nature and extent of the risk?

[[Page 17964]]

1.2.2. Application
    The guidelines apply within the framework of policies provided by 
applicable EPA statutes and do not alter such policies. The guidelines 
cover assessment of available data. They do not imply that one kind of 
data or another is prerequisite for regulatory action concerning any 
agent. Risk management applies directives of regulatory legislation, 
which may require consideration of potential risk, or solely hazard or 
exposure potential, along with social, economic, technical, and other 
factors in decisionmaking. Risk assessments support decisions, but to 
maintain their integrity as decisionmaking tools, they are not 
influenced by consideration of the social or economic consequences of 
regulatory action.
    Not every EPA assessment has the same scope or depth. Agency staff 
often conduct screening-level assessments for priority-setting or 
separate assessments of hazard or exposure for ranking purposes or to 
decide whether to invest resources in collecting data for a full 
assessment. Moreover, a given assessment of hazard and dose response 
may be used with more than one exposure assessment that may be 
conducted separately and at different times as the need arises in 
studying environmental problems in various media. The guidelines apply 
to these various situations in appropriate detail given the scope and 
depth of the particular assessment. For example, a screening assessment 
may be based almost entirely on structure-activity relationships and 
default assumptions. As more data become available, assessments can 
replace or modify default assumptions accordingly. These guidelines do 
not require that all of the kinds of data covered here be available for 
either assessment or decisionmaking. The level of detail of an 
assessment is a matter of Agency management policy regarding the 
applicable decisionmaking framework.

1.3. Use of Default Assumptions

    The National Research Council, in its 1983 report on the science of 
risk assessment (NRC, 1983), recognized that default assumptions are 
necessarily made in risk assessments where gaps exist in general 
knowledge or in available data for a particular agent. These default 
assumptions are inferences based on general scientific knowledge of the 
phenomena in question and are also matters of policy concerning the 
appropriate way to bridge uncertainties that concern potential risk to 
human health (or, more generally, to environmental systems) from the 
agent under assessment.
    EPA's 1986 guidelines for cancer risk assessment (EPA, 1986) were 
developed in response to the 1983 NRC report. The guidelines contained 
a number of default assumptions. They also encouraged research and 
analysis that would lead to new risk assessment methods and data and 
anticipated that these would replace defaults. The 1986 guidelines did 
not explicitly discuss how to depart from defaults. In practice, the 
agency's assessments routinely have employed defaults and, until 
recently, only occasionally departed from them.
    In its 1994 report on risk assessment, the NRC supported continued 
use of default assumptions (NRC, 1994). The NRC report thus validated a 
central premise of the approach to risk assessment that EPA had evolved 
in preceding years--the making of science policy inferences to bridge 
gaps in knowledge--while at the same time recommending that EPA develop 
more systematic and transparent guidelines to inform the public of the 
default inferences EPA uses in practice. It recommended that the EPA 
review and update the 1986 guidelines in light of evolving scientific 
information and experience in practice in applying those guidelines, 
and that the EPA explain the science and policy considerations 
underlying current views as to the appropriate defaults and provide 
general criteria to guide preparers and reviewers of risks assessments 
in deciding when to depart from a default. Pursuant to this 
recommendation, the following discussion presents descriptions of the 
major defaults and their rationales. In addition, it presents general 
policy guidance on using and departing from defaults in specific risk 
assessments.
1.3.1. Default Assumptions
    The 1994 NRC report contains several recommendations regarding 
flexibility and the use of default options:
     EPA should continue to regard the use of default options 
as a reasonable way to deal with uncertainty about underlying 
mechanisms in selecting methods and models for use in risk assessment.
     EPA should explicitly identify each use of a default 
option in risk assessments.
     EPA should clearly state the scientific and policy basis 
for each default option.
     The Agency should consider attempting to give greater 
formality to its criteria for a departure from default options in order 
to give greater guidance to the public and to lessen the possibility of 
ad hoc, undocumented departures from default options that would 
undercut the scientific credibility of the Agency's risk assessments. 
At the same time, the Agency should be aware of the undesirability of 
having its guidelines evolve into inflexible rules.
     EPA should continue to use the Science Advisory Board and 
other expert bodies. In particular, the Agency should continue to make 
the greatest possible use of peer review, workshops, and other devices 
to ensure broad peer and scientific participation to guarantee that its 
risk assessment decisions will be based on the best science available 
through a process that allows full public discussion and peer 
participation by the scientific community.
    In the 1983 report (p. 28), NAS defined the use of ``inference 
options'' (default options) as a means to bridge inherent uncertainties 
in risk assessment. These options exist when the assessment encounters 
either ``missing or ambiguous information on a particular substance'' 
or ``gaps in current scientific theory.'' Since there is no instance in 
which a set of data on an agent or exposure is complete, all risk 
assessments must use general knowledge and policy guidance to bridge 
data gaps. Animal toxicity data are used, for example, to substitute 
for human data because we do not test human beings. The report 
described the components of risk assessment in terms of questions 
encountered during analysis for which inferences must be made. The 
report noted (p. 36) that many components ``* * * lack definitive 
scientific answers, that the degree of scientific consensus concerning 
the best answer varies (some are more controversial than others), and 
that the inference options available for each component differ in their 
degree of conservatism. The choices encountered in risk assessment 
rest, to various degrees, on a mixture of scientific fact and 
consensus, on informed scientific judgment, and on policy 
determinations (the appropriate degree of conservatism)* * *.'' The 
report did not note that the mix varies significantly from case to 
case. For instance, a question that arises in hazard identification is 
how to use experimental animal data when the routes of exposure differ 
between animals and humans. A spectrum of inferences could be made, 
ranging from the most conservative, or risk adverse one that effects in 
animals from one route may be seen in humans by another route, to an 
intermediate, conditional inference that such translation of effects 
will be assumed if the agent is absorbed by humans through the second 
route, to

[[Page 17965]]

a nonconservative view that no inference is possible and the agent's 
effects in animals must be tested by the second route. The choice of an 
inference, as the report observed, comes from more than scientific 
thinking alone. While the report focused mainly on the idea of 
conservatism of public health as a science policy rationale for making 
the choice, it did not evaluate other considerations. These include 
such things as the matters of time and resources and whether the 
analysis is for an important decision required to be made soon or is 
simply a screening or ranking effort. For a screening analysis, one 
might make several ``worst case'' inferences to determine if, even 
under those conditions, risk is low enough that a problem can be 
eliminated from further consideration. In the above discussion 
concerning inferences about route-to-route extrapolation, one might use 
the most conservative one for screening.
    These revised guidelines retain the use of default assumptions as 
recommended in the 1994 report. Generally, these defaults remain public 
health conservative, but in some instances, they have been modified to 
reflect the evolution of scientific knowledge since 1986.
    In addition, the guidelines reflect evaluation of experience in 
practice in applying defaults and departing from them in individual 
risk assessments conducted under the 1986 guidelines. The application 
and departure from defaults and the principles to be used in these 
judgments have been matters of debate among practitioners and reviewers 
of risk assessments. Some observers believe that in practice EPA risk 
assessors have been too resistant to considering departures; others 
question whether proposed departures have been adequately supported. 
Some cases in which departures have been considered have been generally 
accepted, while others have been controversial. The guidelines here are 
intended to be both explicit and more flexible than in the past 
concerning the basis for making departures from defaults, recognizing 
that expert judgment and peer review are essential elements of the 
process.
    In response to the recommendations of the 1994 report, these 
guidelines call for identification of the default assumptions used 
within assessments and for highlighting significant issues about 
defaults within characterization summaries of component analyses in 
assessment documents. As to the use of peer review to aid in making 
judgments about applying or departing from defaults, we agree with the 
NRC recommendation. The Agency has long made use of workshops, peer 
review of documents and guidelines, and consultations as well as formal 
peer review by the Science Advisory Board (SAB). In 1994, the 
Administrator of EPA published formal guidance for peer review of EPA 
scientific work products that increases the amount of peer review for 
risk assessments as well as other work, as a response to the NRC report 
and to SAB recommendations (U.S. EPA, 1994b).
    The 1994 NRC report recommended that EPA should consider adopting 
principles or criteria that would give greater formality and 
transparency to decisions to depart from defaults. The report named 
several possible criteria for such principles (p. 7): ``* * * 
[P]rotecting the public health, ensuring scientific validity, 
minimizing serious errors in estimating risks, maximizing incentives 
for research, creating an orderly and predictable process, and 
fostering openness and trustworthiness. There might be additional 
relevant criteria* * *.'' The report indicated, however, that the 
committee members had not reached consensus on a single criterion to 
address the key issue of how much certainty or proof a risk assessor 
must have in order to justify departing from a default. Appendix N of 
the report contains two presentations of alternative views held by some 
committee members on this issue. One view, known as ``plausible 
conservatism,'' suggested that departures from defaults should not be 
made unless new information improves the understanding of a biological 
process to the point that relevant experts reach consensus that the 
conservative default assumption concerning that process is no longer 
plausible. The same criterion was recommended where the underlying 
scientific mechanism is well understood, but where a default is used to 
address missing data. In this case, the default should not be replaced 
with case-specific data unless it is the consensus of relevant experts 
that the proffered data make the default assumption no longer 
plausible. Another view, known as the ``maximum use of scientific 
information'' approach, acknowledged that the initial choice of 
defaults should be conservative but argued that conservatism should not 
be a factor in determining whether to depart from the default in favor 
of an alternate biological theory or alternate data. According to this 
view, it should not be necessary to reach expert consensus that the 
default assumption had been rendered implausible; it should be 
sufficient that risk assessors find the alternate approach more 
plausible than the default.
    The EPA is not adopting a list of formal decision criteria in the 
sense of a checklist based on either view. It would not be helpful to 
generate a checklist of uniform criteria for several reasons. First, 
risk assessments are highly variable in content and purpose. Screening 
assessments may be purposely ``worst case'' in their default 
assumptions to eliminate problems from further investigation. 
Subsequent risk assessments based on a fuller data set can discard 
worst-case default assumptions in favor of plausibly conservative 
assumptions and progressively replace or modify the latter with data. 
No uniform checklist will fit all cases. Second, a checklist would 
likely become more a source of rote discussion than of enlightenment 
about the process.
    Instead, these guidelines use a combination of principles and 
process in the application of and departure from default assumptions. 
The guidelines provide a framework of default assumptions to allow risk 
assessment to proceed when current scientific theory or available case-
specific data do not provide firm answers in a particular case, as the 
1983 report outlined. Some of the default assumptions bridge large gaps 
in fundamental knowledge which will be filled by basic research on the 
causes of cancer and on other biological processes, rather than by 
agent-specific testing. Other default assumptions bridge smaller data 
gaps that can feasibly be filled for a single agent, such as whether a 
metabolic pathway in test animals is like (default) or unlike that in 
humans.
    The decision to use a default, or not, is a choice considering 
available information on an underlying scientific process and agent-
specific data, depending on which kind of default it is. Generally, if 
a gap in basic understanding exists, or if agent-specific data are 
missing, the default is used without pause. If data are present, their 
evaluation may reveal inadequacies that also lead to use of the 
default. If data support a plausible alternative to the default, but no 
more strongly than they support the default, both the default and its 
alternative are carried through the assessment and characterized for 
the risk manager. If data support an alternative to the default as the 
more reasonable judgment, the data are used. (This framework of choices 
is not wholly applicable to screening assessments. As mentioned above, 
screening assessments may appropriately use ``worst case'' inferences 
to determine if, even under

[[Page 17966]]

those conditions, risk is low enough that a problem can be eliminated 
from further consideration.)
    Scientific peer review, peer consultative workshops and similar 
processes are the principal ways determining the strength of thinking 
and generally accepted views within the scientific community about the 
application of and departure from defaults and about judgments 
concerning the plausibility and persuasiveness of data in a particular 
case. The choices made are explicitly discussed in the assessment, and 
if a particular choice raises a significant issue, it is highlighted in 
the risk characterization.
    The discussion of major defaults in these guidelines together with 
the explicit discussion of the choice of inferences within the 
assessment and the processes of peer review and peer consultation will 
serve the several goals stated in the 1994 report. One is to encourage 
research, since results of research efforts will be considered. Another 
is to allow timely decisionmaking, when time is a constraint, by 
supporting completion of the risk assessment using defaults as needed. 
Another is to be flexible, using new science as it develops. Finally, 
the use of public processes of peer consultation and peer review will 
ensure that discipline of thought is maintained to support trust in 
assessment results.
    Experience has shown that the most difficult part of the framework 
of choices is the judgment of whether a data analysis is both 
biologically plausible and persuasive as applied to the case at hand. 
There is no set of rules for making this judgment in all cases. Two 
criteria that apply in these guidelines are that the underlying 
scientific principle has been generally accepted within the scientific 
community and that supportive experiments are available that test the 
application of the principle to the agent under review. For example, 
mutagenicity through reactivity with DNA has been generally accepted as 
a carcinogenic influence for many years. This acceptance, together with 
evidence of such mutagenicity in experiments on an agent, provides 
plausible and persuasive support for the inference that mutagenicity is 
a mode of action for the agent.
    Judgments about plausibility and persuasiveness of analyses vary 
according to the scientific nature of the default. An analysis of data 
may replace a default or modify it. An illustration of the former is 
development of EPA science policy on the issue of the relevance for 
humans of male rat kidney neoplasia involving alpha 2u globulin (U.S. 
EPA, 1991b). The 1991 EPA policy gives guidance on the kind of 
experimental findings that demonstrate whether the alpha 2u globulin 
mechanism is present and responsible for carcinogenicity in a 
particular case. Before this policy guidance was issued, the default 
assumption was that neoplasia in question was relevant to humans and 
indicated the potential for hazard to humans. A substantial body of 
data was developed by public and private research groups as a 
foundation for the view that the alpha 2u globulin-induced response was 
not relevant to humans. These studies first addressed the alpha 2u 
globulin mechanism in the rat and whether this mechanism has a 
counterpart in the human being, both were large research efforts. The 
resulting data presented difficulties; some reviewers were concerned 
that the mechanism in the rat appeared to be understood only in 
outline, not in detail, and felt that the data were insufficient to 
show the lack of a counterpart mechanism in humans. It was particularly 
difficult to support a negative such as the nonexistence of a mechanism 
in humans because so little is known about what the mechanisms are in 
humans. Despite these concerns, in its 1991 policy guidance, EPA 
concluded that the alpha 2u globulin-induced response in rats should be 
regarded as not relevant to humans (i.e., as not indicating human 
hazard).
    One lesson in the development and peer review of this policy is 
that if the default concerns an inherently complex biological question, 
large amounts of work will be required to replace the default. A second 
is that addressing a negative is difficult. A third is that ``proof'' 
in the strict sense of having laid all reasonable doubt to rest is not 
required. Instead, an alternative may displace a default when it is 
generally accepted in peer review as the most reasonable judgment. The 
issue of relevance may not always be so difficult. It would be an 
experimentally easier task, for example, to determine whether 
carcinogenesis in an animal species is due to a metabolite of the agent 
in question that is not produced in humans.
    When scientific processes are understood but case-specific data are 
missing, defaults can be constructed to be modified by experimental 
data, even if data do not suffice to replace them entirely. For 
example, the approaches adopted in these guidelines for scaling dose 
from experimental animals to humans are constructed to be either 
modified or replaced as data become available on toxicokinetic 
parameters for the particular agent being assessed. Similarly, the 
selection of an approach or approaches for dose response assessment is 
based on a series of decisions that consider the nature and adequacy of 
available data in choosing among alternative modeling and default 
approaches.
    The 1994 NRC report notes (p. 6) that ``[a]s scientific knowledge 
increases, the science policy choices made by the Agency and Congress 
should have less impact on regulatory decisionmaking. Better data and 
increased understanding of biological mechanisms should enable risk 
assessments that are less dependent on conservative default assumptions 
and more accurate as predictions of human risk.'' Undoubtedly, this is 
the trend as scientific understanding increases. However, some gaps in 
knowledge and data will doubtless continue to be encountered in 
assessment of even data-rich cases, and it will remain necessary for 
risk assessments to continue using defaults within the framework set 
forth here.
1.3.2. Major Defaults
    This discussion covers the major default assumptions commonly 
employed in a cancer risk assessment and adopted in these guidelines. 
They are predominantly inferences necessary to use data observed under 
empirical conditions to estimate events and outcomes under 
environmental conditions. Several inferential issues arise when effects 
seen in a subpopulation of humans or animals are used to qualitatively 
infer potential effects in the population of environmentally exposed 
humans. Several more inferential issues arise in extrapolating the 
exposure-effect relationship observed empirically to lower-exposure 
environmental conditions. The following issues cover the major default 
areas. Typically, an issue has some subissues; they are introduced 
here, but are discussed in greater detail in subsequent sections.
     Is the presence or absence of effects observed in a human 
population predictive of effects in another exposed human population?
     Is the presence or absence of effects observed in an 
animal population predictive of effects in exposed humans?
     How do metabolic pathways relate across species?
     How do toxicokinetic processes relate across species?
     What is the correlation of the observed dose response 
relationship to the relationship at lower doses?

[[Page 17967]]

    1.3.2.1. Is the Presence or Absence of Effects Observed in a Human 
Population Predictive of Effects in Another Exposed Human Population? 
When cancer effects in exposed humans are attributed to exposure to an 
exogenous agent, the default assumption is that such data are 
predictive of cancer in any other exposed human population. Studies 
either attributing cancer effects in humans to exogenous agents or 
reporting no effects are often studies of occupationally exposed 
humans. By sex, age, and general health, workers are not representative 
of the general population exposed environmentally to the same agents. 
In such studies there is no opportunity to observe whether infants and 
children, males, or females who are under represented in the study, or 
people whose health is not good, would respond differently. Therefore, 
it is understood that this assumption could still underestimate the 
response of certain sensitive human subpopulations, i.e. biologically 
vulnerable parts of the population may be left out of risk assessments 
(NRC, 1993a, 1994). Consequently, this is a default that does not err 
on the side of public health conservatism, as the 1994 NRC report also 
recognizes.
    On the one hand, if effects are seen in a worker population, this 
may be in fact indicative of heightened effects in sensitive 
subpopulations. There is not enough knowledge yet to form a basis for 
any generally applicable, qualitative inference to compensate for this 
knowledge gap. In these guidelines, this problem is left to case-by-
case analysis, to be attended to as future research and information on 
particular agents allow. When information on a sensitive subpopulation 
exists, it will be used. The topic of variability is addressed further 
in the discussion of quantitative default assumptions about dose 
response relationships below. On the other hand, when cancer effects 
are not found in an exposed human population, this information by 
itself is not generally sufficient to conclude that the agent poses no 
carcinogenic hazard to this or other populations of potentially exposed 
humans. This is because epidemiologic studies usually have low power to 
detect and attribute responses (section 2.2.1.). This may be 
particularly true when extrapolating null results from a healthy, 
worker population to other potentially sensitive exposed humans. Again, 
the problem is left to case-by-case analysis.
    1.3.2.2. Is the Presence or Absence of Effects Observed in an 
Animal Population Predictive of Effects in Exposed Humans? The default 
assumption is that positive effects in animal cancer studies indicate 
that the agent under study can have carcinogenic potential in humans. 
Thus, if no adequate human data are present, positive effects in animal 
cancer studies are a basis for assessing the carcinogenic hazard to 
humans. This assumption is a public health conservative policy, and it 
is both appropriate and necessary given that we do not test for 
carcinogenicity in humans. The assumption is supported by the fact that 
nearly all of the agents known to cause cancer in humans are 
carcinogenic in animals in tests with adequate protocols (IARC, 1994; 
Tomatis et al., 1989; Huff, 1994). Moreover, almost one-third of human 
carcinogens were identified subsequent to animal testing (Huff, 1993). 
Further support is provided by research on the molecular biology of 
cancer processes, which has shown that the mechanisms of control of 
cell growth and differentiation are remarkably homologous among species 
and highly conserved in evolution. Nevertheless, the same research 
tools that have enabled recognition of the nature and commonality of 
cancer processes at the molecular level also have the power to reveal 
differences and instances in which animal responses are not relevant to 
humans (Linjinsky, 1993; U.S. EPA, 1991b). Under these guidelines, 
available mode of action information is studied for its implications in 
both hazard and dose response assessment and its effect on default 
assumptions.
    There may be instances in which the use of an animal model would 
identify a hazard in animals that is not truly a hazard in humans 
(e.g., the alpha-2u-globulin association with renal neoplasia in male 
rats (U.S. EPA, 1991b)). The extent to which animal studies may yield 
false positive indications for humans is a matter of scientific debate. 
To demonstrate that a response in animals is not relevant to any human 
situation, adequate data to assess the relevancy issue must be 
available.
    Animal studies are conducted at high doses in order to provide 
statistical power, the highest dose being one that is minimally toxic 
(maximum tolerated dose). Consequently, the question often arises 
whether a carcinogenic effect at the highest dose may be a consequence 
of cell killing with compensatory cell replication or of general 
physiological disruption, rather than inherent carcinogenicity of the 
tested agent. There is little doubt that this may happen in some cases, 
but skepticism exists among some scientists that it is a pervasive 
problem (Ames and Gold, 1990; Melnick et al., 1993a; Melnick et al., 
1993b; Barrett, 1993). In light of this question, the default 
assumption is that effects seen at the highest dose tested are 
appropriate for assessment, but it is necessary that the experimental 
conditions be scrutinized. If adequate data demonstrate that the 
effects are solely the result of excessive toxicity rather than 
carcinogenicity of the tested agent per se, then the effects may be 
regarded as not appropriate to include in assessment of the potential 
for human carcinogenicity of the agent. This is a matter of expert 
judgment, considering all of the data available about the agent 
including effects in other toxicity studies, structure-activity 
relationships, and effects on growth control and differentiation.
    When cancer effects are not found in well-conducted animal cancer 
studies in two or more appropriate species and other information does 
not support the carcinogenic potential of the agent, these data provide 
a basis for concluding that the agent is not likely to possess human 
carcinogenic potential, in the absence of human data to the contrary. 
This default assumption about lack of cancer effects is not public 
health conservative. For instance, the tested animal species may not be 
predictive of effects in humans; arsenic shows only minimal or no 
effect in animals, while it is clearly positive in humans. (Other 
information, such as absence of mutagenic activity or absence of 
carcinogenic activity among structural analogues, can increase the 
confidence that negative results in animal studies indicate a lack of 
human hazard.) Also, it is recognized that animal studies (and 
epidemiologic studies as well) have very low power to detect cancer 
effects. Detection of a 10% tumor incidence is generally the limit of 
power with currently conducted animal studies (with the exception of 
rare tumors that are virtually markers for a particular agent, e.g., 
angiosarcoma caused by vinyl chloride).
    Target organs of carcinogenesis for agents that cause cancer in 
both animals and humans are most often concordant at one or more sites 
(Tomatis et al., 1989; Huff, 1994). However, concordance by site is not 
uniform. The default assumption is that target organ concordance is not 
a prerequisite for evaluating the implications of animal study results 
for humans. This is a public health conservative science policy. The 
mechanisms of control of cell growth and differentiation are concordant 
among species, but there are marked differences among species in the 
way control is managed in various tissues. For example, in humans, 
mutation of the tumor suppressor gene

[[Page 17968]]

p53 is one of the most frequently observed genetic changes in tumors. 
This tumor suppressor is also observed to be operating in some rodent 
tissues, but other growth control mechanisms predominate in rodents. 
Thus, an animal response may be due to changes in a control that are 
relevant to humans, but appear in animals in a different way. However, 
it is appropriate under these guidelines to consider the influences of 
route of exposure, metabolism, and, particularly, hormonal modes of 
action that may either support or not support target organ concordance 
between animals and humans. When data allow, these influences are 
considered in deciding whether the default remains appropriate in 
individual instances (NRC, 1994, p. 121). An exception to the basic 
default of not assuming site concordance exists in the context of 
toxicokinetic modeling. Site concordance is inherently assumed when 
these models are used to estimate delivered dose in humans based on 
animal data.
    As in the approach of the National Toxicology Program and the 
International Agency for Research on Cancer, the default is to include 
benign tumors observed in animal studies in the assessment of animal 
tumor incidence if they have the capacity to progress to the 
malignancies with which they are associated. This treats the benign and 
malignant tumors as representative of related responses to the test 
agent, which is scientifically appropriate. This is a science policy 
decision that is somewhat more conservative of public health than not 
including benign tumors in the assessment. Nonetheless, in assessing 
findings from animal studies, a greater proportion of malignancy is 
weighed more heavily than a response with a greater proportion of 
benign tumors. Greater frequency of malignancy of a particular tumor 
type in comparison with other tumor responses observed in an animal 
study is also a factor to be considered in selecting the response to be 
used in dose response assessment.
    Benign tumors that are not observed to progress to malignancy are 
assessed on a case-by-case basis. There is a range of possibilities for 
their overall significance. They may deserve attention because they are 
serious health problems even though they are not malignant; for 
instance, benign tumors may be a health risk because of their effect on 
the function of a target tissue such as the brain. They may be 
significant indicators of the need for further testing of an agent if 
they are observed in a short term test protocol, or such an observation 
may add to the overall weight of evidence if the same agent causes 
malignancies in a long term study. Knowledge of the mode of action 
associated with a benign tumor response may aid in the interpretation 
of other tumor responses associated with the same agent. In other 
cases, observation of a benign tumor response alone may have no 
significant health hazard implications when other sources of evidence 
show no suggestion of carcinogenicity.
    1.3.2.3. How Do Metabolic Pathways Relate Across Species? The 
default assumption is that there is a similarity of the basic pathways 
of metabolism and the occurrence of metabolites in tissues in regard to 
the species-to-species extrapolation of cancer hazard and risk. If 
comparative metabolism studies were to show no similarity between the 
tested species and humans and a metabolite(s) were the active form, 
there would be less support for an inference that the animal 
response(s) relates to humans. In other cases, parameters of metabolism 
may vary quantitatively between species; this becomes part of deciding 
on an appropriate human equivalent dose based on animal studies, 
optimally in the context of a toxicokinetic model.
    1.3.2.4. How Do Toxicokinetic Processes Relate Across Species? A 
major issue is how to estimate human equivalent doses in extrapolating 
from animal studies. As a default for oral exposure, a human equivalent 
dose is estimated from data on another species by an adjustment of 
animal oral dose by a scaling factor of body weight to the 0.75 power. 
This adjustment factor is used because it represents scaling of 
metabolic rate across animals of different size. Because the factor 
adjusts for a parameter that can be improved on and brought into more 
sophisticated toxicokinetic modeling, when such data become available, 
the default assumption of 0.75 power can be refined or replaced.
    For inhalation exposure, a human equivalent dose is estimated by 
default methodologies that provide estimates of lung deposition and of 
internal dose. The methodologies can be refined to more sophisticated 
forms with data on toxicokinetic and metabolic parameters of the 
specific agent. This default assumption, like the one with oral 
exposure, is selected in part because it lays a foundation for 
incorporating better data. The use of information to improve dose 
estimation from applied, to internal, to delivered dose is encouraged, 
including use of toxicokinetic modeling instead of any default, where 
data are available. Health conservatism is not an element in choosing 
the default.
    For a route-to-route of exposure extrapolation, the default 
assumption is that an agent that causes internal tumors by one route of 
exposure will be carcinogenic by another route if it is absorbed by the 
second route to give an internal dose. This is a qualitative assumption 
and is considered to be public health conservative. The rationale is 
that for internal tumors an internal dose is significant no matter what 
the route of exposure. Additionally, the metabolism of the agent will 
be qualitatively the same for an internal dose. The issue of 
quantitative extrapolation of the dose-response relationship from one 
route to another is addressed case by case. Quantitative extrapolation 
is complicated by considerations such as first-pass metabolism, but is 
approachable with empirical data. Adequate data are necessary to 
demonstrate that an agent will act differently by one route versus 
another route of exposure.
    1.3.2.5. What Is the Correlation of the Observed Dose Response 
Relationship to the Relationship at Lower Doses? The overriding 
preferred approach is to use a biologically based or case-specific 
model for both the observed range and extrapolation below that range 
when there are sufficient data. While biologically based models are 
still under development, it is likely that they will be used more 
frequently in the future. The default procedure for the observed range 
of data, when the preferred approach cannot be used, is to use a curve-
fitting model.
    In the absence of data supporting a biologically based or case-
specific model for extrapolation outside of the observed range, the 
choice of approach is based on the view of mode of action of the agent 
arrived at in the hazard assessment. A linear default approach is used 
when the mode of action information is supportive of linearity or, 
alternatively, is insufficient to support a nonlinear mode of action. 
The linear approach is used when a view of the mode of action indicates 
a linear response, for example, when a conclusion is made that an agent 
directly causes alterations in DNA, a kind of interaction that not only 
theoretically requires one reaction, but also is likely to be additive 
to ongoing, spontaneous gene mutation. Other kinds of activity may have 
linear implications, e.g., linear rate-limiting steps, that support a 
linear procedure also. The linear approach is to draw a straight line 
between a point of departure from observed data, generally, as a 
default, the LED10, and the origin (zero dose, zero response). 
Other points of

[[Page 17969]]

departure may be more appropriate for certain data sets; these may be 
used instead of the LED10. This approach is generally considered 
to be public health conservative. The LED10 is the lower 95% limit 
on a dose that is estimated to cause a 10% response. This level is 
chosen to account (conservatively) for experimental variability. 
Additionally, it is chosen because it rewards experiments with better 
designs in regard to number of doses and dose spacing, since these 
generally will have narrower confidence limits. It is also an 
appropriate representative of the lower end of the observed range 
because the limit of detection of studies of tumor effect is about 10%.
    The linear default is thought to generally produce an upper bound 
on potential risk at low doses, e.g., a 1/100,000 to 1/1,000,000 risk; 
the straight line approach gives numerical results about the same as a 
linearized multistage procedure (Krewski et al., 1984). This upper 
bound is thought to cover the range of human variability although, in 
some cases, it may not completely do so (Bois et al., 1995). The EPA 
considers the linear default to be inherently conservative of public 
health, without addition of another factor for human variability. In 
any case, the size of such a factor would be hard to determine since a 
good empirical basis on which to construct an estimate does not 
currently exist. The question of what may be the actual variability in 
human sensitivity is one that the 1994 NRC report discussed as did the 
1993 NRC report on pesticides in children and infants. The NRC has 
recommended research on the question, and the EPA and other agencies 
have begun such research.
    When adequate data on mode of action show that linearity is not the 
most reasonable working judgment and provide sufficient evidence to 
support a nonlinear mode of action, the default changes to a different 
approach--a margin of exposure analysis--which assumes that 
nonlinearity is more reasonable. The departure point is again generally 
the LED10. A margin of exposure analysis compares the LED10 
with the dose associated with the environmental exposure(s) of interest 
by computing the ratio between the two.
    The purpose of a margin of exposure analysis is to provide the risk 
manager with all available information on how much reduction in risk 
may be associated with reduction in exposure from the point of 
departure. This is to support the risk manager's decision as to what 
constitutes an acceptable margin of exposure, given requirements of the 
statute under which the decision is being made. There are several 
factors to be considered. (For perspective, keep in mind that a 
sufficient basis to support this nonlinear procedure often will include 
data on responses that are precursors to tumor effects. This means that 
the point of departure may well be from these biological response data 
rather than tumor incidence data, e.g., hormone levels, mitogenic 
effects.) One factor to consider is the slope of the dose response 
curve at the point of departure. A steeper slope implies an apparent 
greater reduction in risk as exposure decreases. This may support a 
smaller margin of exposure. Conversely, a shallow slope may support use 
of a greater margin of exposure. A second factor is the nature of the 
response used in the assessment--A precursor effect or frank toxicity 
or tumor response. The latter two may support a greater margin of 
exposure. A third factor is the nature and extent of human variability 
in sensitivity to the phenomenon. A fourth factor is the agent's 
persistence in the body. Greater variability or persistence argue for 
greater margins of exposure. A fifth factor is human sensitivity to the 
phenomenon as compared with experimental animals. The size of the 
margin of exposure that is acceptable would increase or decrease as 
this factor increases or decreases. If human variability cannot be 
estimated based on data, it should be considered to be at least 10-
fold. Similarly, if comparison of species sensitivities cannot be 
estimated from available data, humans can be considered to be 10-fold 
more sensitive. If it is found that humans are less sensitive than 
animals a factor that is a fraction no smaller than \1/10\ may be 
assumed. The 10-fold factors are moderately conservative, traditional 
ones used for decades in the assessment of toxicological effects. It 
should not be assumed that the numerical factors are the sole 
components for determination of an acceptable margin of exposure. Each 
case calls for individual judgment. It should be noted that for cancer 
assessment the margin of exposure analysis begins from a point of 
departure that is adjusted for toxicokinetic differences between 
species to give a human equivalent dose. Since the traditional factor 
for interspecies difference is thought to contain a measure for 
toxicokinetics as well as sensitivity to effect, the result of 
beginning with a human equivalent dose is to add some conservatism. The 
ultimate judgment whether a particular margin of exposure is acceptable 
is a risk management decision under applicable law, rather than being 
inherent in the risk assessment. Nonetheless, the risk assessor is 
responsible for providing scientific rationale to support the the 
decision.
    When the mode of action information indicates that the dose 
response may be adequately described by both a linear and a nonlinear 
approach, then the default is to present both the linear and margin of 
exposure analyses. An assessment may use both linear and nonlinear 
approaches either for responses that are thought to result from 
different modes of action or for presenting considerations for a 
response that appears to be very different at high and low doses due to 
influence of separate modes of action. Also, separate approaches may be 
used for different induced responses (i.e. tumor types) from the same 
agent. These would also be carried forward and presented in the 
assessment. Figure 1-1 presents the decision points in deciding on a 
dose response approach or approaches.

BILLING CODE 6560-50-P

[[Page 17970]]

[GRAPHIC] [TIFF OMITTED] TN23AP96.000



BILLING CODE 6560-50-C

    A default assumption is made that cumulative dose received over a 
lifetime, expressed as a lifetime average daily dose, is an appropriate 
measure of dose. This assumes that a high dose of such an agent 
received over a shorter period of time is equivalent to a low dose 
spread over a lifetime. This is thought to be a relatively public 
health conservative assumption and has empirical support (Monro, 1992). 
An example of effects of short-term, high exposure that results in 
subsequent cancer development is treatment of cancer patients with 
certain chemotherapeutic agents. An example of cancer from long-term 
exposure to an agent of relatively low potency is smoking. Whether the 
cumulative dose measure is exactly the correct measure in both such 
instances is not certain and should be assessed case by case and 
altered when data are available to support another approach. Other 
measures of dose that consider dose rate and duration are appropriate, 
e.g., when an agent acts by causing cell toxicity or hormone 
disruption. In these cases both agent concentration and duration are 
likely to be important, because such effects are generally observed to 
be reversible at cessation of short-term exposure.

1.4. Characterizations

    The risk characterization process first summarizes findings on 
hazard, dose response, and exposure characterizations, then develops an 
integrative analysis of the whole risk case. It ends in a nontechnical 
Risk Characterization Summary. The Risk Characterization Summary is a 
presentation for risk managers who may or may not be familiar with the 
scientific details of cancer assessment. It also provides information 
for other interested readers. The initial steps in the risk 
characterization process are to make building blocks in the form of 
characterizations of the assessments of hazard, dose response, and 
exposure. The individual assessments and characterizations are then 
integrated to arrive at risk estimates for exposure scenarios of 
interest. There are two reasons for individually characterizing the 
hazard, dose response, and exposure assessments. One is that they are 
often done by different people than those who do the integrative 
analyses. The second is that there is very often a lapse of time 
between the conduct of hazard and dose response analyses and the 
conduct of exposure assessment and integrative analysis. Thus, it is 
necessary to capture characterizations of assessments as the 
assessments are done to avoid the need to go back and reconstruct them. 
Figure 1-2 shows the relationships of analyses. The figure does not 
necessarily correspond to the number of documents involved; there may 
be one or several. ``Integrative analysis'' is a generic term. At EPA, 
the documents of various programs that contain integrative analyses 
have other names such as the ``Staff Paper'' that discusses air quality 
criteria issues. In the following sections, the elements of this figure 
are discussed.

BILLING CODE 6560-50-P

[[Page 17971]]

[GRAPHIC] [TIFF OMITTED] TN23AP96.001



BILLING CODE 6560-50-C

[[Page 17972]]

2. Hazard Assessment

2.1. Overview of Hazard Assessment and Characterization

2.1.1. Analyses of Data
    The purpose of hazard assessment is to review and evaluate data 
pertinent to two questions: (1) whether an agent may pose a 
carcinogenic hazard to human beings and (2) under what circumstances an 
identified hazard may be expressed (NRC, 1994, p. 142). Hazard 
assessment is composed of analyses of a variety of data that may range 
from observations of tumor responses to analysis of structure-activity 
relationships. The purpose of the assessment is not simply to assemble 
these separate evaluations; its purpose is to construct a total case 
analysis examining the biological story the data reveal as a whole 
about carcinogenic effects, mode of action, and implications of these 
for human hazard and dose response evaluation. Weight of evidence 
conclusions come from the combined strength and coherence of inferences 
appropriately drawn from all of the available evidence. To the extent 
that data permit, hazard assessment addresses the mode of action 
question as both an initial step in considering appropriate approaches 
to dose response assessment and as a part of identifying human hazard 
potential.
    The topics in this section include analysis of tumor data, both 
animal and human, and analysis of other key information about 
properties and effects that relate to carcinogenic potential. The 
section addresses how information can be used to evaluate potential 
modes of action. It also provides guidance on performing a weight of 
evidence evaluation.
2.1.2. Cross-Cutting Topics for Data Integration
    Two topics are included in the analysis of each kind of available 
data: first, gathering information from available data about the 
conditions of expression of hazard and second, gathering perspectives 
on the agent's potential mode of action.
    2.1.2.1. Conditions of Expression. Information on the significance 
of the route of exposure may be available from human or animal studies 
on the agent itself or on structural analogues. This information may be 
found in studies of the agent or analogue for toxicological endpoints 
other than cancer under acute or subchronic or chronic exposure 
regimens. Studies of metabolism or toxicokinetics of the agent 
similarly may provide pertinent data.
    Each kind of data is also examined for information on conditions 
that affect expression of carcinogenic effect such as presence or 
absence of metabolic pathways. If carcinogenicity is secondary to 
another toxic effect, the physiological or tissue changes that mark the 
other toxicity are examined. Comparison of metabolic processes and 
toxicity processes in humans and animals also bears on the relevance of 
animal responses to human hazard. Included in the examination are the 
questions of the potential range of human variability and whether any 
special sensitivity may occur because of age, sex, preexisting disease, 
or other condition.
    2.1.2.2. Mode of Action. Information on an agent's potential 
mode(s) of action is important in considering the relevance of animal 
effects to assessment of human hazard. It also plays an important role 
in selecting dose response approach(es), which are generally either 
biologically based models or case-specific models incorporating mode of 
action data or default procedures based on more limited data that 
support inferences about the likely shape of the dose response curve.
    Each kind of data may provide some insight about mode of action and 
insights are gathered from each to be considered together as discussed 
in section 2.4. In Appendix C, is a background discussion of some of 
the development of views about carcinogenic processes.
    2.1.3. Presentation of Results. Presentation of the results of 
hazard assessment follows Agency guidance as discussed in section 2.7. 
The results are presented in a technical hazard characterization that 
serves as a support to later risk characterization. It includes:
     a summary of the evaluations of hazard data,
     the rationales for its conclusions, and
     an explanation of the significant strengths or limitations 
of the conclusions.
    Another presentation feature is the use of a weight of evidence 
narrative that includes both a conclusion about the weight of evidence 
of carcinogenic potential and a summary of the data on which the 
conclusion rests. This narrative is a brief summary that replaces the 
alphanumerical classification system used in EPA's previous guidelines.

2.2. Analysis of Tumor Data

    Evidence of carcinogenicity comes from finding tumor increases in 
humans or laboratory animals exposed to a given agent, or from finding 
tumors following exposure to structural analogues to the compound under 
review. The significance of observed or anticipated tumor effects is 
evaluated in reference to all of the other key data on the agent. This 
section contains guidance for analyzing human and animal studies to 
decide whether there is an association between exposure to an agent or 
a structural analogue and occurrence of tumors. Note that the use of 
the term ``tumor'' here is generic, meaning malignant neoplasms or a 
combination of malignant and corresponding benign neoplasms.
    Observation of only benign neoplasias may or may not have 
significance. Benign tumors that are not observed to progress to 
malignancy are assessed on a case-by-case basis. There is a range of 
possibilities for their overall significance. They may deserve 
attention because they are serious health problems even though they are 
not malignant; for instance, benign tumors may be a health risk because 
of their effect on the function of a target tissue such as the brain. 
They may be significant indicators of the need for further testing of 
an agent if they are observed in a short term test protocol, or such an 
observation may add to the overall weight of evidence if the same agent 
causes malignancies in a long term study. Knowledge of the mode of 
action associated with a benign tumor response may aid in the 
interpretation of other tumor responses associated with the same agent. 
In other cases, observation of a benign tumor response alone may have 
no significant health hazard implications when other sources of 
evidence show no suggestion of carcinogenicity.
2.2.1. Human Data
    Human data may come from epidemiologic studies or case reports. 
Epidemiology is the study of the distributions and causes of disease 
within human populations. The goals of cancer epidemiology are to 
identify differences in cancer risk between different groups in a 
population or between different populations, and then to determine the 
extent to which these differences in risk can be attributed causally to 
specific exposures to exogenous or endogenous factors. Epidemiologic 
data are extremely useful in risk assessment because they provide 
direct evidence that a substance produces cancer in humans, thereby 
avoiding the problem of species to species inference. Thus, when 
available human data are extensive and of good quality, they are 
generally preferable over animal data and should be given

[[Page 17973]]

greater weight in hazard characterization and dose response assessment, 
although both are utilized.
    Null results from a single epidemiologic study cannot prove the 
absence of carcinogenic effects because they can arise either from 
being truly negative or from inadequate statistical power, inadequate 
design, imprecise estimates, or confounding factors. However, null 
results from a well-designed and well-conducted epidemiologic study 
that contains usable exposure data can help to define upper limits for 
the estimated dose of concern for human exposure if the overall weight 
of the evidence indicates that the agent is potentially carcinogenic in 
humans.
    Epidemiology can also complement experimental evidence in 
corroborating or clarifying the carcinogenic potential of the agent in 
question. For example, observations from epidemiologic studies that 
elevated cancer incidence occurs at sites corresponding to those at 
which laboratory animals experience increased tumor incidence can 
strengthen the weight of evidence of human carcinogenicity. On the 
other hand, strong nonpositive epidemiologic data alone or in 
conjunction with compelling mechanistic information can lend support to 
a conclusion that animal responses may not be predictive of a human 
response. Furthermore, the advent of biochemical or molecular 
epidemiology may help improve understanding of the mechanisms of human 
carcinogenesis.
    2.2.1.1. Types of Studies. The major types of cancer epidemiologic 
studies are analytical epidemiologic studies and descriptive or 
correlation epidemiologic studies. Each study type has well-known 
strengths and weaknesses that affect interpretation of study results as 
summarized below (Kelsey et al., 1986; Lilienfeld and Lilienfeld, 1979; 
Mausner and Kramer, 1985; Rothman, 1986).
    Analytical epidemiologic studies are most useful for identifying an 
association between human exposure and adverse health effects. 
Analytical study designs include case-control studies and cohort 
studies. In case-control studies, groups of individuals with (cases) 
and without (controls) a particular disease are identified and compared 
to determine differences in exposure. In cohort studies, a group of 
``exposed'' and ``nonexposed'' individuals are identified and studied 
over time to determine differences in disease occurrence. Cohort 
studies can either be performed prospectively or retrospectively from 
historical records.
    Descriptive or correlation epidemiologic studies (sometimes called 
ecological studies) examine differences in disease rates among 
populations in relation to age, gender, race, and differences in 
temporal or environmental conditions. In general, these studies can 
only identify patterns or trends in disease occurrence over time or in 
different geographical locations but cannot ascertain the causal agent 
or degree of exposure. These studies, however, are often very useful 
for generating hypotheses for further research.
    Biochemical or molecular epidemiologic studies are studies in which 
laboratory methods are incorporated in analytical investigations. The 
application of techniques for measuring cellular and molecular 
alterations due to exposure to specific environmental agents may allow 
conclusions to be drawn about the mechanisms of carcinogenesis. The use 
of biological biomarkers in epidemiology may improve assessment of 
exposure and internal dose.
    Case reports describe a particular effect in an individual or group 
of individuals who were exposed to a substance. These reports are often 
anecdotal or highly selected in nature and are of limited use for 
hazard assessment. However, reports of cancer cases can identify 
associations particularly when there are unique features such as an 
association with an uncommon tumor (e.g., vinyl chloride and 
angiosarcoma or diethylstilbestrol and clear-cell carcinoma of the 
vagina).
    2.2.1.2. Criteria for Assessing Adequacy of Epidemiologic Studies. 
Criteria for assessing the adequacy of epidemiologic studies are well 
recognized. Characteristics that are desirable in these studies include 
(1) clear articulation of study objectives or hypothesis, (2) proper 
selection and characterization of the exposed and control groups, (3) 
adequate characterization of exposure, (4) sufficient length of follow-
up for disease occurrence, (5) valid ascertainment of the causes of 
cancer morbidity and mortality, (6) proper consideration of bias and 
confounding factors, (7) adequate sample size to detect an effect, (8) 
clear, well-documented, and appropriate methodology for data collection 
and analysis, (9) adequate response rate and methodology for handling 
missing data, and (10) complete and clear documentation of results. 
Ideally, these conditions should be satisfied, where appropriate, but 
rarely can a study meet all of them. No single criterion determines the 
overall adequacy of a study. The following discussions highlight the 
major factors included in an analysis of epidemiologic studies.
    Population Issues. The ideal comparison would be between two 
populations that differ only in exposure to the agent in question. 
Because this is seldom the case, it is important to identify sources of 
bias inherent in a study's design or data collection methods. Bias can 
arise from several sources, including noncomparability between 
populations of factors such as general health (McMichael, 1976), diet, 
lifestyle, or geographic location; differences in the way case and 
control individuals recall past events; differences in data collection 
that result in unequal ascertainment of health effects in the 
populations; and unequal follow-up of individuals. Both acceptance of 
studies for assessment and judgment of their strengths or weaknesses 
depend on identifying their sources of bias and the effects on study 
results.
    Exposure Issues. For epidemiologic data to be useful in determining 
whether there is an association between health effects and exposure to 
an agent, there must be adequate characterization of exposure 
information. In general, greater weight should be given to studies with 
more precise and specific exposure estimates.
    Questions to address about exposure are: What can one reliably 
conclude about the level, duration, route, and frequency of exposure of 
individuals in one population as compared with another? How sensitive 
are study results to uncertainties in these parameters?
    Actual exposure measurements are not available for many 
retrospective studies. Therefore, surrogates are often used to 
reconstruct exposure parameters when historical measurements are not 
available. These may involve attributing exposures to job 
classifications in a workplace or to broader occupational or geographic 
groupings. Use of surrogates carries a potential for misclassification 
in that individuals may be placed in the incorrect exposure group. 
Misclassification generally leads to reduced ability of a study to 
detect differences between study and referent populations.
    When either current or historical monitoring data are available, 
the exposure evaluation includes consideration of the error bounds of 
the monitoring and analytic methods and whether the data are from 
routine or accidental exposures. The potentials for misclassification 
and measurement errors are amenable to both qualitative and 
quantitative analysis. These are essential analyses for judging a 
study's results because exposure estimation is

[[Page 17974]]

the most critical part of a retrospective study.
    Biological markers potentially offer excellent measures of exposure 
(Hulka and Margolin, 1992; Peto and Darby, 1994). Validated markers of 
exposure such as alkylated hemoglobin from exposure to ethylene oxide 
(van Sittert et al., 1985) or urinary arsenic (Enterline et al., 1987) 
can greatly improve estimates of dose. Markers closely identified with 
effects promise to greatly increase the ability of studies to 
distinguish real effects from bias at low levels of relative risk 
between populations (Taylor et al., 1994; Biggs et al., 1993) and to 
resolve problems of confounding risk factors.
    Confounding Factors. Because epidemiologic studies are mostly 
observational, it is not possible to guarantee the control of 
confounding variables, which may affect the study outcome. A 
confounding variable is a risk factor, independent of the putative 
agent, that is distributed unequally among the exposed and unexposed 
populations (e.g., smoking habits, lifestyle). Adjustment for possible 
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. The influence of a potential confounding factor is limited by 
the effect of the exposure of interest. For example, a twofold effect 
of an exposure requires that the confounder effect be at least as big. 
The latter may not be possible due to the presentation of the data or 
because needed information was not collected during the study. In this 
case, indirect comparisons may be possible. For example, in the absence 
of data on smoking status among individuals in the study population, an 
examination of the possible contribution of cigarette smoking to 
increased lung cancer risk may be based on information from other 
sources such as the American Cancer Society's longitudinal studies 
(Hammand, 1966; Garfinkel and Silverberg, 1991). The effectiveness of 
adjustments contributes to the ability to draw inferences from a study.
    Different studies involving exposure to an agent may have different 
confounding factors. If consistent increases in cancer risk are 
observed across a collection of studies with different confounding 
factors, the inference that the agent under investigation was the 
etiologic factor is strengthened, even though complete adjustment for 
confounding factors cannot be made and no single study supports a 
strong inference.
    It also may be the case that the agent of interest is a risk factor 
in conjunction with another agent. This relationship may be revealed in 
a collection of studies such as in the case of asbestos exposure and 
smoking.
    Sensitivity. Sensitivity, or the ability of a study to detect real 
effects, is a function of several factors. Greater size of the study 
population(s) (sample size) increases sensitivity, as does greater 
exposure (levels and duration) of the population members. Because of 
the often long latency period in cancer development, sensitivity also 
depends on whether adequate time has elapsed since exposure began for 
effects to occur. A unique feature that can be ascribed to the effects 
of a particular agent (such as a tumor type that is seen only rarely in 
the absence of the agent) can increase sensitivity by permitting 
separation of bias and confounding factors from real effects. 
Similarly, a biomarker particular to the agent can permit these 
distinctions. Statistical reanalyses of data, particularly an 
examination of different exposure indices, can give insight on 
potential exposure-response relationships. These are all factors to 
explore in statistical analysis of the data.
    Statistical Considerations. The analysis applies appropriate 
statistical methods to ascertain whether or not there is any 
significant association between exposure and effects. A description of 
the method or methods should include the reasons for their selection. 
Statistical analyses of the potential effects of bias or confounding 
factors are part of addressing the significance of an association, or 
lack of one, and whether a study is able to detect any effect.
    The analysis augments examination of the results for the whole 
population with exploration of the results for groups with 
comparatively greater exposure or time since first exposure. This may 
support identifying an association or establishing a dose response 
trend. When studies show no association, such exploration may apply to 
determining an upper limit on potential human risk for consideration 
alongside results of animal tumor effects studies.
    Combining Statistical Evidence Across Studies. Meta-analysis is a 
means of comparing and synthesizing studies dealing with similar health 
effects and risk factors. It is intended to introduce consistency and 
comprehensiveness into what otherwise might be a more subjective review 
of the literature. When utilized appropriately, meta-analysis can 
enhance understanding of associations between sources and their effects 
that may not be apparent from examination of epidemiologic studies 
individually. Whether to conduct a meta-analysis depends on several 
issues. These include the importance of formally examining sources of 
heterogeneity, the refinement of the estimate of the magnitude of an 
effect, and the need for information beyond that provided by individual 
studies or a narrative review. Meta-analysis may not be useful in some 
circumstances. These include when the relationship between exposure and 
disease is obvious without a more formal analysis, when there are only 
a few studies of the key health outcomes, when there is insufficient 
information from available studies related to disease, risk estimate, 
or exposure classification, or when there are substantial confounding 
or other biases that cannot be adjusted for in the analysis (Blair et 
al., 1995; Greenland, 1987; Peto, 1992).
    2.2.1.3. Criteria for Causality. A causal interpretation is 
enhanced for studies to the extent that they meet the criteria 
described below. None of the criteria is conclusive by itself, and the 
only criterion that is essential is the temporal relationship. These 
criteria are modeled after those developed by Bradford Hill in the 
examination of cigarette smoking and lung cancer (Rothman, 1986) and 
they need to be interpreted in the light of all other information on 
the agent being assessed.
     Temporal relationship: The development of cancers require 
certain latency periods, and while latency periods vary, existence of 
such periods is generally acknowledged. Thus, the disease has to occur 
within a biologically reasonable time after initial exposure. This 
feature must be present if causality is to be considered.
     Consistency: Associations occur in several independent 
studies of a similar exposure in different populations, or associations 
occur consistently for different subgroups in the same study. This 
feature usually constitutes strong evidence for a causal interpretation 
when the same bias or confounding is not also duplicated across 
studies.
     Magnitude of the association: A causal relationship is 
more credible when the risk estimate is large and precise (narrow 
confidence intervals).
     Biological gradient: The risk ratio (i.e., the ratio of 
the risk of disease or death among the exposed to the risk of the 
unexposed) increases with increasing exposure or dose. A strong dose 
response relationship across several categories of exposure, latency, 
and duration is supportive for causality given that confounding is 
unlikely to be correlated with exposure. The absence of a dose response 
relationship,

[[Page 17975]]

however, is not by itself evidence against a causal relationship.
     Specificity of the association: The likelihood of a causal 
interpretation is increased if an exposure produces a specific effect 
(one or more tumor types also found in other studies) or if a given 
effect has a unique exposure.
     Biological plausibility: The association makes sense in 
terms of biological knowledge. Information is considered from animal 
toxicology, toxicokinetics, structure-activity relationship analysis, 
and short-term studies of the agent's influence on events in the 
carcinogenic process considered.
     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.
    2.2.1.4. Assessment of Evidence of Carcinogenicity from Human Data. 
In the evaluation of carcinogenicity based on epidemiologic studies, it 
is necessary to critically evaluate each study for the confidence in 
findings and conclusions as discussed under section 2.2.1.2. All 
studies that are properly conducted, whether yielding positive or null 
results, or even suggesting protective carcinogenic effects, should be 
considered in assessing the totality of the human evidence. Although a 
single study may be indicative of a cause-effect relationship, 
confidence in inferring a causal relationship is increased when several 
independent studies are concordant in showing the association, when the 
association is strong, and when other criteria for causality are also 
met. Conclusions about the overall evidence for carcinogenicity from 
available studies in humans should be summarized along with a 
discussion of strengths or limitations of the conclusions.
2.2.2. Animal Data
    Various kinds of whole animal test systems are currently used or 
are under development for evaluating potential carcinogenicity. Cancer 
studies involving chronic exposure for most of the life span of an 
animal are generally accepted for evaluation of tumor effects (Tomatis 
et al., 1989; Rall, 1991; Allen et al., 1988; but see Ames and Gold, 
1990). Other studies of special design are useful for observing 
formation of preneoplastic lesions or tumors or investigating specific 
modes of action.
    2.2.2.1. Long-Term Carcinogenicity Studies. The objective of long-
term carcinogenesis bioassays is to determine the carcinogenic 
potential and dose response relationships of the test agent. Long-term 
rodent studies are designed to examine the production of tumors as well 
as preneoplastic lesions and other indications of chronic toxicity that 
may provide evidence of treatment-related effects and insights into the 
way the test agent produces tumors. Current standardized long-term 
studies in rodents test at least 50 animals per sex per dose group in 
each of three treatment groups and in a concurrent control group, 
usually for 18 to 24 months, depending on the rodent species tested 
(OECD, 1981; U.S. EPA, 1983a; U.S. EPA, 1983b; U.S. EPA, 1983c). The 
high dose in long-term studies is generally selected to provide the 
maximum ability to detect treatment-related carcinogenic effects while 
not compromising the outcome of the study due to excessive toxicity or 
inducing inappropriate toxicokinetics (e.g., overwhelming 
detoxification or absorption mechanisms). The purpose of two or more 
lower doses is to provide some information on the shape of the dose 
response curve. Similar protocols have been and continue to be used by 
many laboratories worldwide.
    All available studies of tumor effects in whole animals are 
considered, at least preliminarily. The analysis discards studies 
judged to be wholly inadequate in protocol, conduct, or results. 
Criteria for the technical adequacy of animal carcinogenicity studies 
have been published and should be used as guidance to judge the 
acceptability of individual studies (NTP, 1984; OSTP, 1985). Care is 
taken to include studies that provide some evidence bearing on 
carcinogenicity or help interpret effects noted in other studies even 
if they have some limitations of protocol or conduct. Such limited, but 
not wholly inadequate, studies can contribute as their deficiencies 
permit. The findings of long-term rodent bioassays are always 
interpreted in conjunction with results of prechronic studies along 
with toxicokinetic and metabolism studies and other pertinent 
information, if available. Evaluation of tumor effects requires 
consideration of both biological and statistical significance of the 
findings (Haseman, 1984, 1985, 1990, 1995). The following sections 
highlight the major issues in the evaluation of long-term 
carcinogenicity studies.
    Dosing issues. In order to obtain the most relevant information 
from a long-term carcinogenicity study, it is important to require 
maximization of exposure to the test material. At the same time, there 
is a need for caution in using excessive high dose levels that would 
confound the interpretation of study results to humans. The high dose 
is conventionally defined as a dose that produces some toxic effects 
without either unduly affecting mortality from effects other than 
cancer or producing significant adverse effects on the nutrition and 
health of the test animals (OECD, 1981; NRC, 1993b). It should be noted 
that practical upper limits have been established to avoid the use of 
excessive high doses in long-term carcinogenicity studies (e.g., 5% of 
the test substance in the feed for dietary studies [OECD, 1981]).
    Evaluating the appropriateness of the high dose in carcinogenicity 
studies is based on scientific judgment using all available relevant 
information. In general, if the test agent does not appear to cause any 
specific target organ toxicity or perturbation of physiological 
function, an adequate high dose would be a dose that causes no more 
than 10% reduction of body weight gain over the life span of the 
animals. On the other hand, significant increases in mortality from 
effects other than cancer is accepted as clear evidence of frank 
toxicity, which indicates that an adequate high dose may have been 
exceeded. Other signs of treatment-related toxicity that may indicate 
that an adequate high dose has been exceeded include the following: (a) 
Reduction of body weight gain of 10% or greater, (b) significant 
increases in abnormal behavioral and clinical signs, (c) significant 
changes in hematology or clinical chemistry, (d) saturation of 
absorption and detoxification mechanisms, or (e) marked changes in 
organ weight, morphology, and histopathology.
    For dietary studies, weight gain reductions should be evaluated as 
to whether there is a palatability problem or an issue with food 
efficiency; certainly, the latter is a toxic manifestation. In the case 
of inhalation studies with respirable particles, evidence of impairment 
of normal clearance of particles from the lung should be considered 
along with other signs of toxicity to the respiratory airways to 
determine whether the high exposure concentration has been 
appropriately selected. For dermal studies, evidence of skin irritation 
may indicate that an adequate high dose has been reached.
    Interpretation of carcinogenicity study results is profoundly 
affected by exposure conditions, especially by inappropriate dose 
selection. This is particularly important in studies that are 
nonpositive for carcinogenicity, since failure to reach a sufficient 
dose reduces the sensitivity of a study. A lack of tumorigenic 
responses at exposure levels that cause significant impairment

[[Page 17976]]

of animal survival may also not be acceptable as negative findings 
because of the reduced sensitivity of the study. On the other hand, 
overt toxicity or inappropriate toxicokinetics due to excessive high 
doses may result in tumor effects that are secondary to the toxicity 
rather than directly attributable to the agent.
    There are several possible outcomes regarding the study 
interpretation of the significance and relevance of tumorigenic effects 
associated with exposure or dose levels below, at, or above an adequate 
high dose. General guidance is given here that should not be taken as 
prescriptive; for each case, the information at hand is evaluated and a 
rationale should be given for the position taken.
     Adequate high dose: If an adequate high dose has been 
utilized, tumor effects are judged positive or negative depending on 
the presence or absence of significant tumor incidence increases, 
respectively.
     Excessive high dose: If toxicity or mortality is excessive 
at the high dose, interpretation depends on the finding of tumors or 
not.
    (a) Studies that show tumor effects only at excessive doses may be 
compromised and may or may not carry weight, depending on the 
interpretation in the context of other study results and other lines of 
evidence. Results of such studies, however, are generally not 
considered suitable for risk extrapolation.
    (b) Studies that show tumors at lower doses, even though the high 
dose is excessive and may be discounted, should be evaluated on their 
own merits.
    (c) If a study does not show an increase in tumor incidence at a 
toxic high dose and appropriately spaced lower doses are used without 
such toxicity or tumors, the study is generally judged as negative for 
carcinogenicity.
     Inadequate high dose: Studies of inadequate sensitivity 
where an adequate high dose has not been reached may be used to bound 
the dose range where carcinogenic effects might be expected.
    Statistical Considerations. The main aim of statistical evaluation 
is to determine whether exposure to the test agent is associated with 
an increase of tumor development. Statistical analysis of a long-term 
study should be performed for each tumor type separately. The incidence 
of benign and malignant lesions of the same cell type, usually within a 
single tissue or organ, are considered separately and are combined when 
scientifically defensible (McConnell et al., 1986).
    Trend tests and pairwise comparison tests are the recommended tests 
for determining whether chance, rather than a treatment-related effect, 
is a plausible explanation for an apparent increase in tumor incidence. 
A trend test such as the Cochran-Armitage test (Snedecor and Cochran, 
1967) asks whether the results in all dose groups together increase as 
dose increases. A pairwise comparison test such as the Fisher exact 
test (Fisher, 1932) asks whether an incidence in one dose group is 
increased over the control group. By convention, for both tests a 
statistically significant comparison is one for which p <0.05 that the 
increased incidence is due to chance. Significance in either kind of 
test is sufficient to reject the hypothesis that chance accounts for 
the result. A statistically significant response may or may not be 
biologically significant and vice versa. The selection of a 
significance level is a policy choice based on a trade-off between the 
risks of false positives and false negatives. A significance level of 
greater or less than 5% is examined to see if it confirms other 
scientific information. When the assessment departs from a simple 5% 
level, this should be highlighted in the risk characterization. A two-
tailed test or a one-tailed test can be used. In either case a 
rationale is provided.
    Considerations of multiple comparisons should also be taken into 
account. Haseman (1983) analyzes typical animal bioassays testing both 
sexes of two species and concludes that, because of multiple 
comparisons, a single tumor increase for a species-sex-site combination 
that is statistically significant at the 1% level for common tumors or 
5% for rare tumors corresponds to a 7-8% significance level for the 
study as a whole. Therefore, animal bioassays presenting only one 
significant result that falls short of the 1% level for a common tumor 
may be treated with caution.
    Concurrent and Historical Controls. The standard for determining 
statistical significance of tumor incidence comes from a comparison of 
tumors in dosed animals as compared with concurrent control animals. 
Additional insights about both statistical and biological significance 
can come from an examination of historical control data (Tarone, 1982; 
Haseman, 1995). Historical control data can add to the analysis 
particularly by enabling identification of uncommon tumor types or high 
spontaneous incidence of a tumor in a given animal strain. 
Identification of common or uncommon situations prompts further thought 
about the meaning of the response in the current study in context with 
other observations in animal studies and with other evidence about the 
carcinogenic potential of the agent. These other sources of information 
may reinforce or weaken the significance given to the response in the 
hazard assessment. Caution should be exercised in simply looking at the 
ranges of historical responses because the range ignores differences in 
survival of animals among studies and is related to the number of 
studies in the database.
    In analyzing results for uncommon tumors in a treated group that 
are not statistically significant in comparison to concurrent controls, 
the analyst can use the experience of historical controls to conclude 
that the result is in fact unlikely to be due to chance. In analyzing 
results for common tumors, a different set of considerations comes into 
play. Generally speaking, statistically significant increases in tumors 
should not be discounted simply because incidence rates in the treated 
groups are within the range of historical controls or because incidence 
rates in the concurrent controls are somewhat lower than average. 
Random assignment of animals to groups and proper statistical 
procedures provide assurance that statistically significant results are 
unlikely to be due to chance alone. However, caution should be used in 
interpreting results that are barely statistically significant or in 
which incidence rates in concurrent controls are unusually low in 
comparison with historical controls.
    In cases where there may be reason to discount the biological 
relevance to humans of increases in common animal tumors, such 
considerations should be weighed on their own merits and clearly 
distinguished from statistical concerns.
    When historical control data are used, the discussion needs to 
address several issues that affect comparability of historical and 
concurrent control data. Among these issues are the following: genetic 
drift in the laboratory strains; differences in pathology examination 
at different times and in different laboratories (e.g., in criteria for 
evaluating lesions; variations in the techniques for preparation or 
reading of tissue samples among laboratories); comparability of animals 
from different suppliers. The most relevant historical data come from 
the same laboratory and same supplier, gathered within 2 or 3 years one 
way or the other of the study under review; other data should be used 
only with extreme caution.
    Assessment of Evidence of Carcinogenicity from Long-Term Animal 
Studies. In general, observation of tumor effects under different 
circumstances lends support to the

[[Page 17977]]

significance of the findings for animal carcinogenicity. Significance 
is a function of the number of factors present, and for a factor such 
as malignancy, the severity of the observed pathology. The following 
observations add significance to the tumor findings:
     uncommon tumor types
     tumors at multiple sites
     tumors by more than one route of administration
     tumors in multiple species, strains, or both sexes
     progression of lesions from preneoplastic to benign to 
malignant
     reduced latency of neoplastic lesions
     metastases
     unusual magnitude of tumor response
     proportion of malignant tumors
     dose-related increases
    These guidelines adopt the science policy position that tumor 
findings in animals indicate that an agent may produce such effects in 
humans. Moreover, the absence of tumor findings in well-conducted, 
long-term animal studies in at least two species provides reasonable 
assurance that an agent may not be a carcinogenic concern for humans. 
Each of these is a default assumption that may be adopted, when 
appropriate, after evaluation of tumor data and other key evidence.
    Site concordance of tumor effects between animals and humans is an 
issue to be considered in each case. Thus far, there is evidence that 
growth control mechanisms at the level of the cell are homologous among 
mammals, but there is no evidence that these mechanisms are site 
concordant. Moreover, agents observed to produce tumors in both humans 
and animals have produced tumors either at the same (e.g., vinyl 
chloride) or different sites (e.g., benzene) (NRC, 1994). Hence, site 
concordance is not assumed a priori. On the other hand, certain 
processes with consequences for particular tissue sites (e.g., 
disruption of thyroid function) may lead to an anticipation of site 
concordance.
    2.2.2.2. Other Studies. Various intermediate-term studies often use 
protocols that screen for carcinogenic or preneoplastic effects, 
sometimes in a single tissue. Some involve the development of various 
proliferative lesions, like foci of alteration in the liver 
(Goldsworthy et al., 1986). Others use tumor endpoints, like the 
induction of lung adenomas in the sensitive strain A mouse (Maronpot et 
al., 1986) or tumor induction in initiation-promotion studies using 
various organs such as the bladder, intestine, liver, lung, mammary 
gland, and thyroid (Ito et al., 1992). In these tests, the selected 
tissue is, in a sense, the test system rather than the whole animal. 
Important information concerning the steps in the carcinogenic process 
and mode of action can be obtained from ``start/stop'' experiments. In 
these protocols, an agent is given for a period of time to induce 
particular lesions or effects, then stopped to evaluate the progression 
or reversibility of processes (Todd, 1986; Marsman and Popp, 1994).
    Assays in genetically engineered rodents may provide insight into 
the chemical and gene interactions involved in carcinogenesis (Tennant 
et al., 1995a). These mechanistically based approaches involve 
activated oncogenes that are introduced (transgenic) or tumor 
suppressor genes that are deleted (knocked-out). If appropriate genes 
are selected, not only may these systems provide information on 
mechanisms, but the rodents typically show tumor development earlier 
than the standard bioassay. Transgenic mutagenesis assays also 
represent a mechanistic approach for assessing the mutagenic properties 
of agents as well as developing quantitative linkages between exposure, 
internal dose, and mutation related to tumor induction (Morrison and 
Ashby, 1994; Sisk et al., 1994; Hayward et al., 1995). These systems 
use a stable genomic integration of a lambda shuttle vector that 
carries a lacI target gene and a lacZ reporter gene.
    The support that these studies give to a determination of 
carcinogenicity rests on their contribution to the consistency of other 
evidence about an agent. For instance, benzoyl peroxide has promoter 
activity on the skin, but the overall evidence may be less supportive 
(Kraus et al., 1995). These studies also may contribute information 
about mode of action. One needs to recognize the limitations of these 
experimental protocols such as short duration, limited histology, lack 
of complete development of tumors, or experimental manipulation of the 
carcinogenic process that may limit their contribution to the overall 
assessment. Generally, their results are appropriate as aids in the 
assessment for interpreting other toxicological evidence (e.g., rodent 
chronic bioassays), especially regarding potential modes of action. 
With sufficient validation, these studies may partially or wholly 
replace chronic bioassays in the future (Tennant et al., 1995).
2.2.3. Structural Analogue Data
    For some chemical classes, there is significant information 
available on the carcinogenicity of analogues, largely in rodent 
bioassays. Analogue effects are instructive in investigating 
carcinogenic potential of an agent as well as identifying potential 
target organs, exposures associated with effects, and potential 
functional class effects or modes of action. All appropriate studies 
are included and analyzed, whether indicative of a positive effect or 
not. Evaluation includes tests in various animal species, strains, and 
sexes; with different routes of administration; and at various doses, 
as data are available. Confidence in conclusions is a function of how 
similar the analogues are to the agent under review in structure, 
metabolism, and biological activity. This confidence needs to be 
considered to ensure a balanced position.

2.3. Analysis of Other Key Data

    The physical, chemical, and structural properties of an agent, as 
well as data on endpoints that are thought to be critical elements of 
the carcinogenic process, provide valuable insights into the likelihood 
of human cancer risk. The following sections provide guidance for 
analyses of these data.
2.3.1. Physicochemical Properties
    Physicochemical properties affect an agent's absorption, tissue 
distribution (bioavailability), biotransformation, and degradation in 
the body and are important determinants of hazard potential (and dose 
response analysis). Properties to analyze include, but are not limited 
to, the following: molecular weight, size, and shape; valence state; 
physical state (gas, liquid, solid); water or lipid solubility, which 
can influence retention and tissue distribution; and potential for 
chemical degradation or stabilization in the body.
    An agent's potential for chemical reaction with cellular 
components, particularly with DNA and proteins, is also important. The 
agent's molecular size and shape, electrophilicity, and charge 
distribution are considered in order to decide whether they would 
facilitate such reactions.
2.3.2. Structure-Activity Relationships
    Structure-activity relationship (SAR) analyses and models can be 
used to predict molecular properties, surrogate biological endpoints, 
and carcinogenicity. Overall, these analyses provide valuable initial 
information on agents, which may strengthen or weaken the concern for 
an agent's carcinogenic potential.
    Currently, SAR analysis is useful for chemicals and metabolites 
that are believed to initiate carcinogenesis through covalent 
interaction with DNA (i.e., DNA-reactive, mutagenic, electrophilic, or 
proelectrophilic

[[Page 17978]]

chemicals) (Ashby and Tennant, 1991). For organic chemicals, the 
predictive capability of SAR analysis combined with other toxicity 
information has been demonstrated (Ashby and Tennant, 1994). The 
following parameters are useful in comparing an agent to its structural 
analogues and congeners that produce tumors and affect related 
biological processes such as receptor binding and activation, 
mutagenicity, and general toxicity (Woo and Arcos, 1989):
     nature and reactivity of the electrophilic moiety or 
moieties present,
     potential to form electrophilic reactive intermediate(s) 
through chemical, photochemical, or metabolic activation,
     contribution of the carrier molecule to which the 
electrophilic moiety(ies) is attached,
     physicochemical properties (e.g., physical state, 
solubility, octanol-water partition coefficient, half-life in aqueous 
solution),
     structural and substructural features (e.g., electronic, 
stearic, molecular geometric),
     metabolic pattern (e.g., metabolic pathways and activation 
and detoxification ratio), and
     possible exposure route(s) of the agent.
    Suitable SAR analysis of non-DNA-reactive chemicals and of DNA-
reactive chemicals that do not appear to bind covalently to DNA 
requires knowledge or postulation of the probable mode(s) of action of 
closely related carcinogenic structural analogues (e.g., receptor-
mediated, cytotoxicity-related). Examination of the physicochemical and 
biochemical properties of the agent may then provide the rest of the 
information needed in order to make an assessment of the likelihood of 
the agent's activity by that mode of action.
2.3.3. Comparative Metabolism and Toxicokinetics
    Studies of the absorption, distribution, biotransformation, and 
excretion of agents permit comparisons among species to assist in 
determining the implications of animal responses for human hazard 
assessment, supporting identification of active metabolites, 
identifying changes in distribution and metabolic pathway or pathways 
over a dose range, and making comparisons among different routes of 
exposure.
    If extensive 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 a determination of tissue dosimetry, species-
to-species extrapolation of dose, and route-to-route extrapolation 
(Connolly and Andersen, 1991; see section 3.2.2). If it is not contrary 
to available data, it is assumed as a default that toxicokinetic and 
metabolic processes are qualitatively comparable between species. 
Discussion of the defaults regarding quantitative comparison and their 
modifications appears in section 3.
    The qualitative question of whether an agent is absorbed by a 
particular route of exposure is important for weight of evidence 
classification discussed in section 2.7.1. Decisions whether route of 
exposure is a limiting factor on expression of any hazard, in that 
absorption does not occur by a route, are based on studies in which 
effects of the agent, or its structural analogues, have been observed 
by different routes, on physical-chemical properties, or on 
toxicokinetics studies.
    Adequate metabolism and pharmacokinetic data can be applied toward 
the following as data permit. Confidence in conclusions is enhanced 
when in vivo data are available.
     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 appropriately focus SAR analysis, 
analysis of potential modes of action, and estimation of internal dose 
in dose response assessment (D'Souza et al., 1987; Krewski et al., 
1987).
     Identifying and comparing the relative activities of 
metabolic pathways in animals with those in humans. This analysis can 
provide insights for extrapolating results of animal studies to humans.
     Describing anticipated distribution within the body and 
possibly identifying target organs. Use of water solubility, molecular 
weight, and structure analysis can support qualitative inferences about 
anticipated 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 approaches.
     Identifying changes in toxicokinetics and metabolic 
pathways with increases in dose. These changes may result in important 
differences in disposition of the agent or its generation of active 
forms of the agent between high and low dose levels. These studies play 
an important role in providing a rationale for dose selection in 
carcinogenicity studies.
     Determining bioavailability via different routes of 
exposure by analyzing uptake processes under various exposure 
conditions. This analysis supports identification of hazards for 
untested routes. 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 these areas, 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 exposure. The analysis takes into 
account the presence of subpopulations of individuals who are 
particularly vulnerable to the effects of an agent because of 
toxicokinetic or metabolic differences (genetically or environmentally 
determined) (Bois et al., 1995).
2.3.4. Toxicological and Clinical Findings
    Toxicological findings in experimental animals and clinical 
observations in humans are an important resource to the cancer hazard 
assessment. Such findings provide information on physiological effects, 
effects on enzymes, hormones, and other important macromolecules as 
well as on target organs for toxicity. Given that the cancer process 
represents defects in terminal differentiation, growth control, and 
cell death, developmental studies of agents may provide an 
understanding of the activity of an agent that carries over to cancer 
assessment. Toxicity studies in animals by different routes of 
administration support comparison of absorption and metabolism by those 
routes. Data on human variability in standard clinical tests may 
provide insight into the range of human sensitivity and common 
mechanisms to agents that affect the tested parameters.
2.3.5. Mode of Action-Related Endpoints and Short-Term Tests
    A myriad of biochemical and biological endpoints relevant to the 
carcinogenic process provide important information in determining 
whether a cancer hazard exists and include, but are not limited to, 
mutagenesis, inhibition of gap junctional intercellular communication, 
increased cell proliferation, inhibition of programmed cell death, 
receptor activation, and immunosuppression. These precursor effects are 
discussed below.
    2.3.5.1. Direct DNA Effects. Because cancer is the result of 
multiple genetic

[[Page 17979]]

defects in genes controlling proliferation and tissue homeostasis 
(Vogelstein et al., 1988), the ability of an agent to affect DNA is of 
obvious importance. It is well known that many carcinogens are 
electrophiles that interact directly with DNA, resulting in DNA damage 
and adducts, and subsequent mutations (referred to in these guidelines 
as direct DNA effects) that are thought to contribute to the 
carcinogenic process (Shelby and Zeiger, 1990; Tinwell and Ashby, 
1991). Thus, studies of these phenomena continue to be important in the 
assessment of cancer hazard. The EPA has published testing guidelines 
for detecting the ability of agents to affect DNA or chromosomes (EPA, 
1991a). Information on agents that induce mutations in animal germ 
cells also deserves attention; several human carcinogens have been 
shown to be positive in rodent tests for the induction of genetic 
damage in both somatic and germ cells (Shelby, 1995).
    2.3.5.2. Secondary DNA Effects. Similarly of interest are secondary 
mechanisms that either increase mutation rates or the number of 
dividing cells. An increase in mutations might be due to cytotoxic 
exposures causing regenerative proliferation or mitogenic influences, 
either of which could result in clonal expansion of initiated cells 
(Cohen and Ellwein, 1990). An agent might interfere with the enzymes 
involved in DNA repair and recombination (Barrett and Lee, 1992). Also, 
programmed cell death (apoptosis) can potentially be blocked by an 
agent, thereby permitting replication of damaged cells. For example, 
peroxisome proliferators may act by suppressing apoptosis pathways 
(Shulte-Hermann et al., 1993; Bayly et al., 1994). An agent may also 
generate reactive oxygen species that produce oxidative damage to DNA 
and other important macromolecules that become important elements of 
the carcinogenic process (Kehrer, 1993; Clayson et al., 1994; Chang et 
al., 1988). Damage to certain critical DNA repair genes or other genes 
(e.g., the p53 gene) may result in genomic instability, which 
predisposes cells to further genetic alterations and increases the 
probability of neoplastic progression independent of any exogenous 
agent (Harris and Hollstein, 1993; Levine, 1994).
    The loss or gain of chromosomes (i.e., aneuploidy) is an effect 
that can result in genomic instability (Fearon and Vogelstein, 1990; 
Cavenee et al., 1986). Although the relationship between induced 
aneuploidy and carcinogenesis is not completely established, several 
carcinogens have been shown to induce aneuploidy (Gibson et al., 1995; 
Barrett, 1992). Agents that cause aneuploidy interfere with the normal 
process of chromosome segregation and lead to chromosomal losses, 
gains, or aberrations by interacting with the proteins (e.g., 
microtubules) needed for chromosome movement.
    2.3.5.3. Nonmutagenic and Other Effects. A failure to detect DNA 
damage and mutation induction in several test systems suggests that a 
carcinogenic agent may act by another mode of action.
    It is possible for an agent to alter gene expression 
(transcriptional, translational, or post-translational modifications) 
by means not involving mutations (Barrett, 1995). For example, 
perturbation of DNA methylation patterns may cause effects that 
contribute to carcinogenesis (Jones, 1986; Goodman and Counts, 1993; 
Holliday, 1987). Overexpression of genes by amplification has been 
observed in certain tumors (Vainio et al., 1992). Other mechanisms may 
involve cellular reprogramming through hormonal mechanisms or receptor-
mediated mechanisms (Ashby et al., 1994; Barrett, 1992).
    Gap-junctional intercellular communication is widely believed to 
play a role in tissue and organ development and in the maintenance of a 
normal cellular phenotype within tissues. A growing body of evidence 
suggests that chemical interference with gap-junctional intercellular 
communication is a contributing factor in tumor development; many 
carcinogens have been shown to inhibit this communication. Thus, such 
information may provide useful mechanistic data in evaluating cancer 
hazard (Swierenga and Yamasaki, 1992; Yamasaki, 1995).
    Both cell death and cell proliferation are mandatory for the 
maintenance of homeostasis in normal tissue. The balance between the 
two directly affects the survival and growth of initiated cells, as 
well as preneoplastic and tumor cell populations (i.e., increase in 
cell proliferation or decrease in cell death) (Bellamy et al., 1995; 
Cohen and Ellwein, 1990, 1991; Cohen et al., 1991). In studies of 
proliferative effects, distinctions should be made between mitogenesis 
and regenerative proliferation (Cohen and Ellwein, 1990, 1991; Cohen et 
al., 1991). In applying information from studies on cell proliferation 
and apoptosis to risk assessment, it is important to identify the 
tissues and target cells involved, to measure effects in both normal 
and neoplastic tissue, to distinguish between apoptosis and necrosis, 
and to determine the dose that affects these processes.
    2.3.5.4. Criteria for Judging Mode of Action. Criteria that are 
applicable for judging the adequacy of mechanistically based data 
include the following:
     mechanistic relevance of the data to carcinogenicity,
     number of studies of each endpoint,
     consistency of results in different test systems and 
different species,
     similar dose response relationships for tumor and mode of 
action-related effects,
     tests conducted in accordance with generally accepted 
protocols, and
     degree of consensus and general acceptance among 
scientists regarding interpretation of the significance and specificity 
of the tests.
    Although important information can be gained from in vitro test 
systems, a higher level of confidence is generally given to data that 
are derived from in vivo systems, particularly those results that show 
a site concordance with the tumor data.

2.4. Biomarker Information

    Various endpoints can serve as biological markers of events in 
biological systems or samples. In some cases, these molecular or 
cellular effects (e.g., DNA or protein adducts, mutation, chromosomal 
aberrations, levels of thyroid stimulating hormone) can be measured in 
blood, body fluids, cells and tissues to serve as biomarkers of 
exposure in both animals and humans (Callemen et al., 1978; Birner et 
al., 1990). As such, they can do the following:
     act as an internal surrogate measure of chemical dose, 
representing as appropriate, either recent (e.g., serum concentration) 
or accumulated (e.g., hemoglobin adducts) exposure,
     help identify doses at which elements of the carcinogenic 
process are operating,
     aid in interspecies extrapolations when data are available 
from both experimental animal and human cells, and
     under certain circumstances, provide insights into the 
possible shape of the dose response curve below levels where tumor 
incidences are observed (e.g., Choy, 1993).
    Genetic and other findings (like changes in proto-oncogenes and 
tumor suppressor genes in preneoplastic and neoplastic tissue or 
possibly measures of endocrine disruption) can indicate the potential 
for disease and as such serve as biomarkers of effect. They, too, can 
be used in different ways:
     The spectrum of genetic changes in proliferative lesions 
and tumors

[[Page 17980]]

following chemical administration to experimental animals can be 
determined and compared with those in spontaneous tumors in control 
animals, in animals exposed to other agents of varying structural and 
functional activities, and in persons exposed to the agent under study.
     They may provide a linkage to tumor response.
     They may help to identify subpopulations of individuals 
who may be at an elevated risk for cancer, e.g., cytochrome P450 2D6/
debrisoquine sensitivity for lung cancer (Caporaso et al., 1989) or 
inherited colon cancer syndromes (Kinzler et al., 1991; Peltomaki et 
al., 1993).
     As with biomarkers of exposure, it may be justified in 
some cases to use these endpoints for dose response assessment or to 
provide insight into the potential shape of the dose response curve at 
doses below those at which tumors are induced experimentally.
    In applying biomarker data to cancer assessment (particularly 
assessments based on epidemiologic data), one should consider the 
following:
     routes of exposure
     exposure to mixtures
     time after exposure
     sensitivity and specificity of biomarkers
     dose response relationships.

2.5. Mode of Action--Implications for Hazard Characterization and Dose 
Response

    The interaction of the biology of the organism and the chemical 
properties of the agent determine whether there is an adverse effect. 
Thus, mode of action analysis is based on physical, chemical, and 
biological information that helps to explain critical events in an 
agent's influence on development of tumors. The entire range of 
information developed in the assessment is reviewed to arrive at a 
reasoned judgment. An agent may work by more than one mode of action 
both at different sites and at the same tumor site. It is felt that at 
least some information bearing on mode of action (e.g., SAR, screening 
tests for mutagenicity) is present for most agents undergoing 
assessment of carcinogenicity, even though certainty about exact 
molecular mechanisms may be rare.
    Inputs to mode of action analysis include tumor data in humans, 
animals, and among structural analogues as well as the other key data. 
The more complete the data package and generic knowledge about a given 
mode of action, the more confidence one has and the more one can 
replace or refine default science policy positions with relevant 
information. Making reasoned judgments is generally based on a data-
rich source of chemical, chemical class, and tumor type-specific 
information. Many times there will be conflicting data and gaps in the 
information base; one must carefully evaluate these uncertainties 
before reaching any conclusion.
    Some of the questions that need to be addressed include the 
following:
     Has a body of data been developed on the agent that fits 
with a generally accepted mode of action?
     Has the mode of action been published and gained general 
scientific acceptance through peer-reviewed research or is it still 
speculative?
     Is the mode of action consistent with generally agreed-
upon principles and understanding of carcinogenesis?
     Is the mode of action reasonably anticipated or assumed, 
in the absence of specific data, to operate in humans? How is this 
question influenced by information on comparative uptake, metabolism, 
and excretion patterns across animals and humans?
     Do humans appear to be more or less sensitive to the mode 
of action than are animals?
     Does the agent affect DNA, directly or indirectly?
     Are there important determinants in carcinogenicity other 
than effects on DNA, such as changes in cell proliferation, apoptosis, 
gene expression, immune surveillance, or other influences?
    In making decisions about potential modes of action and the 
relevance of animal tumor findings to humans (Ashby et al., 1990), very 
often the results of chronic animal studies may give important clues. 
Some of the important factors to review include the following:
     tumor types, e.g., those responsive to endocrine 
influence, those produced by reactive carcinogens (Ashby and Tennant, 
1991),
     number of tumor sites, sexes, studies, and species 
affected or unaffected (Tennant, 1993),
     influence of route of exposure; spectrum of tumors; local 
or systemic sites,
     target organ or system toxicity, e.g., urinary chemical 
changes associated with stone formation, effects on immune 
surveillance,
     presence of proliferative lesions, e.g., hepatic foci, 
hyperplasias,
     progression of lesions from preneoplastic to benign to 
malignant with dose and time,
     ratio of malignant to benign tumors as a function of dose 
and time,
     time of appearance of tumors after commencing exposure,
     tumors invading locally, metastasizing, producing death,
     tumors at sites in laboratory animals with high or low 
spontaneous historical incidence,
     biomarkers in tumor cells, both induced and spontaneous, 
e.g., DNA or protein adducts, mutation spectra, chromosome changes, 
oncogene activation, and
     shape of the dose response in the range of tumor 
observation, e.g., linear vs. profound change in slope.
    Some of the myriad of ways that information from chronic animal 
studies influences mode of action judgments include the following. 
Multisite and multispecies tumor effects are often associated with 
mutagenic agents. Tumors restricted to one sex/species may suggest an 
influence restricted to gender, strain, or species. Late onset of 
tumors that are primarily benign or are at sites with a high historical 
background incidence or show reversal of lesions on cessation of 
exposure may point to a growth-promoting mode of action. The 
possibility that an agent may act differently in different tissues or 
have more than one mode of action in a single tissue must also be kept 
in mind.
    Simple knowledge of sites of tumor increase in rodent studies can 
give preliminary clues as to mode of action. Experience at the National 
Toxicology Program (NTP) indicates that substances that are DNA 
reactive and produce gene mutations may be unique in producing tumors 
in certain anatomical sites, while tumors at other sites may arise from 
both mutagenic or nonmutagenic influences (Ashby and Tennant, 1991; 
Huff et al., 1991).
    Effects on tumor sites in rodents and other mode of action 
information has been explored for certain agents (Alison et al., 1994; 
Clayson, 1989; ECETOC, 1991; MacDonald et al., 1994; McClain, 1994; 
Tischer et al., 1991; ILSI, 1995; Cohen and Ellwein, 1991; FASEB, 1994; 
Havu et al., 1990; U.S. EPA, 1991; Li et al., 1987; Grasso and Hinton, 
1991; Larson et al., 1994; IARC, 1990; Jack et al., 1983; Stitzel et 
al., 1989; Ingram and Grasso, 1991; Bus and Popp, 1987; Prahalada et 
al., 1994; Yamada et al., 1994; Hill et al., 1989; Burek et al., 1988).
    The selection of a dose response extrapolation procedure for cancer 
risk estimation considers mode of action information. When information 
is extensive and there is considerable certainty in a given mode of 
action, a biologically based or case-specific model that incorporates 
data on processes involved is preferred. Obviously, use of such a model 
requires

[[Page 17981]]

the existence of substantial data on component parameters of the mode 
of action, and judgments on its applicability must be made on a case-
by-case basis.
    In the absence of information to develop a biologically based or 
case-specific model, understanding of mode of action should be employed 
to the extent possible in deciding upon one of three science policy 
defaults: Low-dose linear extrapolation, nonlinear, and both 
procedures. The overall choice of the default(s) depends upon weighing 
the various inputs and deciding which best reflect the mode of action 
understanding. A rationale accompanies whichever default or defaults 
are chosen.
    A default assumption of linearity is appropriate when the evidence 
supports a mode of action of gene mutation due to DNA reactivity or 
supports another mode of action that is anticipated to be linear. Other 
elements of empirical data may also support an inference of linearity, 
e.g., the background of human exposure to an agent might be such that 
added human exposure is on the linear part of a dose response curve 
that is sublinear overall. The default assumption of linearity is also 
appropriate as the ultimate default when evidence shows no DNA 
reactivity or other support for linearity, but neither is it sufficient 
evidence of a nonlinear mode of action to support a nonlinear 
procedure.
    A default assumption of nonlinearity is appropriate when there is 
no evidence for linearity and sufficient evidence to support an 
assumption of nonlinearity and a nonlinear procedure. The mode of 
action may lead to a dose response relationship that is nonlinear, with 
response falling much more quickly than linearly with dose, or being 
most influenced by individual differences in sensitivity. 
Alternatively, the mode of action may theoretically have a threshold, 
e.g., the carcinogenicity may be a secondary effect of toxicity that is 
itself a threshold phenomenon.
    Both linear and nonlinear procedures may be used in particular 
cases. If a mode of action analysis finds substantial support for 
differing modes of action for different tumor sites, an appropriate 
procedure is used for each. Both procedures may also be appropriate to 
discuss implications of complex dose response relationships. For 
example, if it is apparent that an agent is both DNA reactive and is 
highly active as a promotor at high doses, and there are insufficient 
data for modeling, both linear and nonlinear default procedures may be 
needed to decouple and consider the contribution of both phenomena.

2.6. Weight of Evidence Evaluation for Potential Human Carcinogenicity

    A weight of evidence evaluation is a collective evaluation of all 
pertinent information so that the full impact of biological 
plausibility and coherence are adequately considered. Identification 
and characterization of human carcinogenicity is based on human and 
experimental data, the nature, advantages and limitations of which have 
been discussed in the preceding sections.
    The subsequent sections outline: (1) the basics of weighing 
individual lines of evidence and combining the entire body of evidence 
to make an informed judgment, (2) classification descriptors of cancer 
hazard, and (3) some case study examples to illustrate how the 
principles of guidance can be applied to arrive at a classification.
2.6.1. Weight of Evidence Analysis
    Judgment about the weight of evidence involves considerations of 
the quality and adequacy of data and consistency of responses induced 
by the agent in question. The weight of evidence judgment requires 
combined input of relevant disciplines. Initial views of one kind of 
evidence may change significantly when other information is brought to 
the interpretation. For example, a positive animal carcinogenicity 
finding may be diminished by other key data; a weak association in 
epidemiologic studies may be bolstered by consideration of other key 
data and animal findings. Factors typically considered are illustrated 
in figures below. Generally, no single weighing factor on either side 
determines the overall weight. The factors are not scored mechanically 
by adding pluses and minuses; they are judged in combination.
    Human Evidence. Analyzing the contribution of evidence from a body 
of human data requires examining available studies and weighing them in 
the context of well-accepted criteria for causation (see section 
2.2.1). A judgment is made about how closely they satisfy these 
criteria, individually and jointly, and how far they deviate from them. 
Existence of temporal relationships, consistent results in independent 
studies, strong association, reliable exposure data, presence of dose-
related responses, freedom from biases and confounding factors, and 
high level of statistical significance are among the factors leading to 
increased confidence in a conclusion of causality.
    Generally, the weight of human evidence increases with the number 
of adequate studies that show comparable results on populations exposed 
to the same agent under different conditions. The analysis takes into 
account all studies of high quality, whether showing positive 
associations or null results, or even protective effects. In weighing 
positive studies against null studies, possible reasons for 
inconsistent results should be sought, and results of studies that are 
judged to be of high quality are given more weight than those from 
studies judged to be methodologically less sound. See figure 2-1.

BILLING CODE 6560-50-P

[[Page 17982]]

[GRAPHIC] [TIFF OMITTED] TN23AP96.002



    Generally, no single factor is determinative. For example, the 
strength of association is one of the causal criteria. A strong 
association (i.e., a large relatively risk) is more likely to indicate 
causality than a weak association. However, finding of a large excess 
risk in a single study must be balanced against the lack of consistency 
as reflected by null results from other equally well designed and well 
conducted studies. In this situation, the positive association of a 
single study may either suggest the presence of chance, bias or 
confounding, or reflect different exposure conditions. On the other 
hand, evidence of weak but consistent associations across several 
studies suggests either causality or the same confounder may be 
operating in all of these studies.
    Animal Evidence. Evidence from long-term or other carcinogenicity 
studies in laboratory animals constitutes the second major class of 
information bearing on carcinogenicity. See figure 2-2. As discussed in 
section 2.2.2., each relevant study must be reviewed and evaluated as 
to its adequacy of design and conduct as well as the statistical 
significance and biological relevance of its findings. Factors that 
usually increase confidence in the predictivity of animal findings are 
those of (1) multiplicity of observations in independent studies; (2) 
severity of lesions, latency, and lesion progression; (3) consistency 
in observations.


[[Page 17983]]

[GRAPHIC] [TIFF OMITTED] TN23AP96.003



    Other Key Evidence. Additional information bearing on the 
qualitative assessment of carcinogenic potential may be gained from 
comparative pharmacokinetic and metabolism studies, genetic toxicity 
studies, SAR analysis, and other studies of an agent's properties. See 
figure 2-3. Information from these studies helps to elucidate potential 
modes of action and biological fate and disposition. The knowledge 
gained supports interpretation of cancer studies in humans and animals 
and provides a separate source of information about carcinogenic 
potential.


[[Page 17984]]

[GRAPHIC] [TIFF OMITTED] TN23AP96.004



    Totality of Evidence. In reaching a view of the entire weight of 
evidence, all data and inferences are merged. Figure 2-4 indicates the 
generalities. In fact, possible weights of evidence span a broad 
continuum that cannot be capsulized. Most of the time the data in 
various lines of evidence fall in the middle of the weights represented 
in the four figures in this section.


[[Page 17985]]

[GRAPHIC] [TIFF OMITTED] TN23AP96.005



BILLING CODE 6560-50-C
    The following section and the weight of evidence narrative 
discussed in 2.7.2. provide a way to state a conclusion and capture 
this complexity in a consistent way.
2.6.2. Descriptors for Classifying Weight of Evidence
    Hazard classification uses three categories of descriptors for 
human carcinogenic potential: ``known/likely,'' ``cannot be 
determined,'' and ``not likely.'' Each category has associated 
subdescriptors to further define the conclusion. The descriptors are 
not meant to replace an explanation of the nuances of the biological 
evidence, but rather to summarize it. Each category spans a wide 
variety of potential data sets and weights of evidence. There will 
always be gray areas, gradations, and borderline cases. That is why the 
descriptors are presented only in the context of a weight of evidence 
narrative whose format is given in section 2.7.2. Using them within a 
narrative preserves and presents the complexity that is an essential 
part of the hazard classification. Applying a descriptor is a matter of 
judgment and cannot be reduced to a formula. Risk managers should 
consider the entire range of information included in the narrative 
rather than focusing simply on the descriptor.
    A single agent may be categorized in more than one way if, for 
instance, the agent is likely to be carcinogenic by one route of 
exposure but not by another (section 2.3.3).
    The descriptors and subdescriptors are standardized and are to be 
used consistently from case to case. The discussions below explain 
descriptors and subdescriptors which appear in italics, and along with 
Appendix A and section 2.6.3, illustrate their use.
``Known/Likely''
    This category of descriptors is appropriate when the available 
tumor effects and other key data are adequate to convincingly 
demonstrate carcinogenic potential for humans; it includes:
     Agents known to be carcinogenic in humans based on either 
epidemiologic evidence or a combination of epidemiologic and 
experimental evidence, demonstrating causality between human exposure 
and cancer,
     Agents that should be treated as if they were known human 
carcinogens, based on a combination of epidemiologic data showing a 
plausible causal association (not demonstrating it definitively) and 
strong experimental evidence.
     Agents that are likely to produce cancer in humans due to 
the production or anticipated production of tumors by modes of action 
that are relevant or assumed to be relevant to human carcinogenicity.
    Modifying descriptors for particularly high or low ranking in the 
``known/likely'' group can be applied based on scientific judgment and 
experience and are as follows:
     Agents that are likely to produce cancer in humans based 
on data that are at the high end of the weights of evidence typical of 
this group,
     Agents that are likely to produce cancer in humans based 
on data that are at the low end of the weights of evidence typical of 
this group.
``Cannot Be Determined''
    This category of descriptors is appropriate when available tumor 
effects or other key data are suggestive or conflicting or limited in 
quantity and, thus, are not adequate to convincingly demonstrate 
carcinogenic potential for humans. In general, further agent specific 
and generic research and testing are needed to be able to describe 
human carcinogenic potential. The descriptor cannot be determined is 
used with a subdescriptor that captures the rationale:
     Agents whose carcinogenic potential cannot be determined, 
but for which there is suggestive evidence that raises concern for 
carcinogenic effects,
     Agents whose carcinogenic potential cannot be determined 
because the existing evidence is composed of conflicting data (e.g., 
some evidence is suggestive of carcinogenic effects, but other equally 
pertinent evidence does not confirm any concern),
     Agents whose carcinogenic potential cannot be determined 
because there are inadequate data to perform an assessment,
     Agents whose carcinogenic potential cannot be determined 
because no data are available to perform an assessment.

[[Page 17986]]

``Not Likely''
    This is the appropriate descriptor when experimental evidence is 
satisfactory for deciding that there is no basis for human hazard 
concern, as follows (in the absence of human data suggesting a 
potential for cancer effects):
     Agents not likely to be carcinogenic to humans because 
they have been evaluated in at least two well conducted studies in two 
appropriate animal species without demonstrating carcinogenic effects,
     Agents not likely to be carcinogenic to humans because 
they have been appropriately evaluated in animals and show only 
carcinogenic effects that have been shown not to be relevant to humans 
(e.g., showing only effects in the male rat kidney due to accumulation 
of alpha2u-globulin),
     Agents not likely to be carcinogenic to humans when 
carcinogenicity is dose or route dependent. For instance, not likely 
below a certain dose range (categorized as likely above that range) or 
not likely by a certain route of exposure (may be categorized as likely 
by another route of exposure). To qualify, agents will have been 
appropriately evaluated in animal studies and the only effects show a 
dose range or route limitation or a route limitation is otherwise shown 
by empirical data.
     Agents not likely to be carcinogenic to humans based on 
extensive human experience that demonstrates lack of effect (e.g., 
phenobarbital).
2.6.3. Case Study Examples
    This section provides examples of substances that fit the three 
broad categories described above. These examples are based on available 
information about real substances and are selected to illustrate the 
principles for weight-of-evidence evaluation and the application of the 
classification scheme.
    These case studies show the interplay of differing lines of 
evidence in making a conclusion. Some particularly illustrate the role 
that ``other key data'' can play in conclusions.

Example 1: ``Known Human Carcinogen''--Route-Dependent/Linear 
Extrapolation

Human Data

    Substance 1 is an aluminosilicate mineral that exists in nature 
with a fibrous habit. Several descriptive epidemiologic studies have 
demonstrated very high mortality from malignant mesothelioma, mainly 
of the pleura, in three villages in Turkey, where there was a 
contamination of this mineral and where exposure had occurred from 
birth. Both sexes were equally affected and at an unusually young 
age.

Animal Data

    Substance 1 has been studied in a single long-term inhalation 
study in rats at one exposure concentration that showed an extremely 
high incidence of pleural mesothelioma (98% in treated animals 
versus 0% in concurrent controls). This is a rare malignant tumor in 
the rat and the onset of tumors occurred at a very early age (as 
early as 1 year of age). Several studies involving injection into 
the body cavities of rats or mice (i.e., pleural or peritoneal 
cavities) also produced high incidences of pleural or peritoneal 
mesotheliomas. No information is available on the carcinogenic 
potential of substance 1 in laboratory animals via oral and dermal 
exposures.

Other Key Data

    Information on the physical and chemical properties of substance 
1 indicates that it is highly respirable to humans and laboratory 
rodents. It is highly insoluble and is not likely to be readily 
degraded in biological fluid.
    No information is available on the deposition, translocation, 
retention, lung clearance, and excretion of the substance after 
inhalation exposure or ingestion. Lung burden studies have shown the 
presence of elevated levels of the substance in lung tissue samples 
of human cases of pleural mesotheliomas from contaminated villages 
compared with control villages.
    No data are available on genetic or related effects in humans. 
The substance has been shown to induce unscheduled DNA synthesis in 
human cells in vitro and transformation and unscheduled DNA 
synthesis in mouse cells.
    The mechanisms by which this substance causes cancer in humans 
and animals are not understood, but appear to be related to its 
unique physical, chemical, and surface properties. Its fiber 
morphology is similar to a known group of naturally occurring 
silicate minerals that have been known to cause respiratory cancers 
(including pleural mesothelioma) from inhalation exposure and 
genetic changes in humans.

Evaluation

    Human evidence is judged to establish a causal link between 
exposure to substance 1 and human cancer. Even though the human 
evidence does not satisfy all criteria for causality, this judgment 
is based on a number of unusual observations: large magnitude of the 
association, specificity of the association, demonstration of 
environmental exposure, biological plausibility, and coherence based 
on the entire body of knowledge of the etiology of mesothelioma.
    Animal evidence demonstrates a causal relationship between 
exposure and cancer in laboratory animals. Although available data 
are not optimal in terms of design (e.g., the use of single dose, 
one sex only), the judgment is based on the unusual findings from 
the only inhalation experiment in rats (i.e., induction of an 
uncommon tumor, an extremely high incidence of malignant neoplasms, 
and onset of tumors at an early age). Additional evidence is 
provided by consistent results from several injection studies 
showing an induction of the same tumors by different modes of 
administration in more than one species.

    Other key data, while limited, support the human and animal 
evidence of carcinogenicity. It can be inferred from human and 
animal data that this substance is readily deposited in the 
respiratory airways and deep lung and is retained for extended 
periods of time since first exposure. Information on related fibrous 
substances indicates that the modes of action are likely mediated by 
the physical and chemical characteristics of the substance (e.g., 
fiber shape, high aspect ratio, a high degree of insolubility in 
lung tissues).

    Insufficient data are available to evaluate the human 
carcinogenic potential of substance 1 by oral exposure. Even though 
there is no information on its carcinogenic potential via dermal 
uptake, it is not expected to pose a carcinogenic hazard to humans 
by that route because it is very insoluble and is not likely to 
penetrate the skin.

Conclusion

    It is concluded that substance 1 is a known human carcinogen by 
inhalation exposure. The weight of evidence of human carcinogenicity 
is based on (a) exceptionally increased incidence of malignant 
mesothelioma in epidemiologic studies of environmentally exposed 
human populations; (b) significantly increased incidence of 
malignant mesothelioma in a single inhalation study in rats and in 
several injection studies in rats and mice; and (c) supporting 
information on related fibrous substances that are known to cause 
cancer via inhalation and genetic damage in exposed mammalian and 
human mesothelial cells. The human carcinogenic potential of 
substance 1 via oral exposure cannot be determined on the basis of 
insufficient data. It is not likely to pose a carcinogenic hazard to 
humans via dermal uptake because it is not anticipated to penetrate 
the skin.

    The mode of action of this substance is not understood. In 
addition to this uncertainty, dose response information is lacking 
for both human and animal data. Epidemiologic studies contain 
observations of significant excess cancer risks at relatively low 
levels of environmental exposure. The use of linear extrapolation in 
a dose response relationship assessment is appropriate as a default 
since mode of action data are not available.

Example 2: ``As If Known Human Carcinogen''--Any Exposure Conditions/
Linear Extrapolation

Human Data

    Substance 2 is an alkene oxide. Several cohort studies of 
workers using substance 2 as a sterilant have been conducted. In the 
largest and most informative study, mortality from lymphatic and 
hematopoietic cancer was marginally elevated, but a significant 
trend was found, especially for lymphatic leukemia and non-Hodgkin's 
lymphoma, in relation to estimated cumulative exposure to the 
substance. Nonsignificant excesses of lymphatic and hematopoietic 
cancer were

[[Page 17987]]

found in three other smaller studies of sterilization personnel.
    In one cohort study of chemical workers exposed to substance 2 
and other agents, mortality rate from lymphatic and hematopoietic 
cancer was elevated, but the excess was confined to a small subgroup 
with only occasional low-level exposure to substance 2. Six other 
studies of chemical workers are considered more limited due to a 
smaller number of deaths. Four studies found an excess of lymphatic 
and hematopoietic cancer (which were significant in two); no 
increase in mortality rate was observed in the other two studies.

Animal Data

    Substance 2 was studied in an oral gavage study in rats. 
Treatment of substance 2 resulted in a dose-dependent increased 
incidence in forestomach tumors that were mainly squamous-cell 
carcinomas.
    Substance 2 was also studied in two inhalation studies in mice 
and two inhalation studies in rats. In the first mouse study, dose-
dependent increases in combined benign and malignant tumors at 
several tissue sites were induced in mice of both sexes (lung tumors 
and tumors of the Harderian gland in each sex, and uterine 
adenocarcinomas, mammary carcinomas, and malignant lymphomas in 
females). In a second study--a screening study for pulmonary tumors 
in mice--inhalation exposure to substance 2 resulted in a dose-
dependent increase in lung tumors. In the two inhalation studies in 
rats, increased incidences of mononuclear-cell leukemia and brain 
tumors were induced in exposed animals of each sex; increased 
incidences of peritoneal tumors in the region of the testis and 
subcutaneous fibromas were induced in exposed male rats.
    Substance 2 induced local sarcomas in mice following 
subcutaneous injection. No tumors were found in a limited skin 
painting study in mice.

Other Key Data

    Substance 2 is a flammable gas at room temperature. The gaseous 
form is readily taken up in humans and rats, and in aqueous solution 
it can penetrate human skin. Studies in rats indicate that, once 
absorbed, substance 2 is uniformly distributed throughout the body. 
It is eliminated metabolically by hydrolysis and by conjugation with 
glutathione. The ability to form glutathione conjugate varies across 
animal species, with the rat being most active, followed by mice and 
rabbits.
    Substance 2 is a directly acting alkylating agent. It has been 
shown to form adducts with hemoglobin in both humans and animals and 
with DNA in animals. The increased frequency of hemoglobin adducts, 
which have been used as markers of internal dose, has been found to 
correlate with the level and cumulative exposure to substance 2. 
Significant increases in chromosomal aberrations and sister 
chromatid exchanges in peripheral lymphocytes and induction of 
micronuclei in the bone marrow cells have been observed in exposed 
workers.
    Substance 2 also induced chromosomal aberrations and sister 
chromatid exchanges in peripheral lymphocytes of monkeys exposed in 
vivo. It also induced gene mutation, specific locus mutation, sister 
chromatid exchanges, chromosomal aberrations, micronuclei, dominant 
lethal mutations, and heritable translocation in rodents exposed in 
vivo. In human cells in vitro, it induced sister chromatid 
exchanges, chromosomal aberrations, and unscheduled DNA synthesis. 
Similar genetic and related effects were observed in rodent cells in 
vitro and in nonmammalian systems.

Evaluation

    Available epidemiologic studies, taken together, suggest that a 
causal association between exposure to substance 2 and elevated risk 
of cancer is plausible. This judgment is based on small but 
consistent excesses of lymphatic and hematopoietic cancer in the 
studies of sterilization workers. Interpretation of studies of 
chemical workers is difficult because of possible confounding 
exposures. Nevertheless, findings of elevated risks of cancer at 
similar sites in chemical workers support the findings in studies of 
sterilization workers. Additional support is provided by 
observations of DNA damage in the same tissue in which elevated 
cancer was seen in exposed workers.
    Extensive evidence indicates that substance 2 is carcinogenic to 
laboratory animals. Positive results were consistently observed in 
all well-designed and well-conducted studies. Substance 2 causes 
dose-related increased incidences of tumors at multiple tissue sites 
in rats and mice of both sexes by two routes of exposure (oral and 
inhalation). The only dermal study that yielded a nonpositive 
finding is considered of limited quality.
    Other key data significantly add support to the potential 
carcinogenicity of substance 2. There is strong evidence of 
heritable mutations of exposed rodents and mutagenicity and 
clastogenicity both in vivo and in vitro. These findings are 
reinforced by observations of similar genetic damage in exposed 
workers. Additional support is based on SAR analysis that indicates 
that substance 2 is a highly DNA-reactive agent. Structurally 
related chemicals, i.e., low-molecular-weight epoxides, also exhibit 
carcinogenic effects in laboratory animals.

Conclusion

    Substance 2 should be considered as if it were a known human 
carcinogen by all routes of exposure. The weight of evidence of 
human carcinogenicity is based on (a) consistent evidence of 
carcinogenicity in rats and mice by oral and inhalation exposure; 
(b) epidemiologic evidence suggestive of a causal association 
between exposure and elevated risk of lymphatic and hematopoietic 
cancer; (c) evidence of genetic damage in blood lymphocytes and bone 
marrow cells of exposed workers; (d) mutagenic effects in numerous 
in vivo and in vitro test systems; (e) membership in a class of DNA-
reactive compounds that have been shown to cause carcinogenic and 
mutagenic effects in animals; and (f) ability to be absorbed by all 
routes of exposure, followed by rapid distribution throughout the 
body.
    Although the exact mechanisms of carcinogenic action of 
substance 2 are not completely understood, available data strongly 
indicate a mutagenic mode of action. Linear extrapolation should be 
assumed in dose response assessment.

Example 3: ``Likely Human Carcinogen''--Any Exposure Conditions/Linear 
Extrapolation

Human Data

    Substance 3 is a brominated alkane. Three studies have 
investigated the cancer mortality of workers exposed to this 
substance. No statistically significant increase in cancer at any 
site was found in a study of production workers exposed to substance 
3 and several other chemicals. Elevated cancer mortality was 
reported in a much smaller study of production workers. An excess of 
lymphoma was reported in grain workers who may have had exposure to 
substance 3 and other chemical compounds. These studies are 
considered inadequate due to their small cohort size; lack of, or 
poorly characterized, exposure concentrations; or concurrent 
exposure of the cohort to other potential or known carcinogens.

Animal Data

    The potential carcinogenicity of substance 3 has been 
extensively studied in an oral gavage study in rats and mice of both 
sexes, two inhalation studies of rats of different strains of both 
sexes, an inhalation study in mice of both sexes, and a skin 
painting study in female mice.
    In the oral study, increased incidences of squamous-cell 
carcinoma of the forestomach were found in rats and mice of both 
sexes. Additionally, there were increased incidences of liver 
carcinomas in female rats, hemangiosarcomas in male rats, and 
alveolar/bronchiolar adenoma of the lung of male and female mice. 
Excessive toxicity and mortality were observed in the rat study, 
especially in the high-dose groups, which resulted in early 
termination of study, and similar time-weighted average doses for 
the high- and low-treatment groups.
    In the first inhalation study in rats and mice, increased 
incidences of carcinomas and adenocarcinomas of the nasal cavity and 
hemangiosarcoma of the spleen were found in exposed animals of each 
species of both sexes. Treated female rats also showed increased 
incidences of alveolar/bronchiolar carcinoma of the lung and mammary 
gland fibroadenomas. Treated male rats showed an increased incidence 
of peritoneal mesothelioma. In the second inhalation study in rats 
(single exposure only), significantly increased incidences of 
hemangiosarcoma of the spleen and adrenal gland tumors were seen in 
exposed animals of both sexes. Additionally, increased incidences of 
subcutaneous mesenchymal tumors and mammary gland tumors were 
induced in exposed male and female rats, respectively.
    Lifetime dermal application of substance 3 to female mice 
resulted in significantly increased incidences of skin papillomas 
and lung tumors.
    Several chemicals structurally related to substance 3 are also 
carcinogenic in rodents. The spectrum of tumor responses induced by 
related substances was similar to those seen with substance 3 (e.g., 
forestomach, mammary gland, lung tumors).

[[Page 17988]]

Other Key Data

    Substance 3 exists as a liquid at room temperature and is 
readily absorbed by ingestion, inhalation, and dermal contact. It is 
widely distributed in the body and is eliminated in the urine mainly 
as metabolites (e.g., glutathione conjugate).
    Substance 3 is not itself DNA-reactive, but is biotransformed to 
reactive metabolites as inferred by findings of its covalent binding 
to DNA and induction of DNA strand breaks, both in vivo and in 
vitro. Substance 3 has been shown to induce sister chromatid 
exchanges, mutations, and unscheduled DNA synthesis in human and 
rodent cells in vitro. Reverse and forward mutations have been 
consistently produced in bacterial assays and in vitro assays using 
eukaryotic cells. Substance 3, however, did not induce dominant 
lethal mutations in mice or rats, or chromosomal aberrations or 
micronuclei in bone marrow cells of mice treated in vivo.

Evaluation

    Available epidemiologic data are considered inadequate for an 
evaluation of a causal association of exposure to the substance and 
excess of cancer mortality due to major study limitations.
    There is extensive evidence that substance 3 is carcinogenic in 
laboratory animals. Increased incidences of tumors at multiple sites 
have been observed in multiple studies in two species of both sexes 
with different routes of exposure. It induces tumors both at the 
site of entry (e.g., nasal tumors via inhalation, forestomach tumors 
by ingestion, skin tumor with dermal exposure) and at distal sites 
(e.g., mammary gland tumors). Additionally, it induced tumors at the 
same sites in both species and sexes via different routes of 
exposure (e.g., lung tumors). With the exception of the oral study 
in which the employed doses caused excessive toxicity and mortality, 
the other studies are considered adequately designed and well 
conducted. Overall, given the magnitude and extent of animal 
carcinogenic responses to substance 3, coupled with similar 
responses to structurally related substances, these animal findings 
are judged to be highly relevant and predictive of human responses.
    Other key data, while not very extensive, are judged to be 
supportive of carcinogenic potential. Substance 3 has consistently 
been shown to be mutagenic in mammalian cells, including human 
cells, and nonmammalian cells; thus, mutation is likely a mode of 
action for its carcinogenic activity. However, the possible 
involvement of other modes of action has not been fully 
investigated. Furthermore, induction of genetic changes from in vivo 
exposure to substance 3 has not been demonstrated.

Conclusion

    Substance 3 is likely to be a human carcinogen by any route of 
exposure. In comparison with other agents designated as likely human 
carcinogens, the overall weight of evidence for substance 3 puts it 
at the high end of the grouping.
    The weight of evidence of human carcinogenicity is based on 
animal evidence and other key evidence. Human data are inadequate 
for an evaluation of human carcinogenicity. The overall weight of 
evidence is based on (a) extensive animal evidence showing induction 
of increases of tumors at multiple sites in both sexes of two rodent 
species via three routes of administration relevant to human 
exposure; (b) tumor data of structural analogues exhibiting similar 
patterns of tumors in treated rodents; (c) in vitro evidence for 
mutagenic effects in mammalian cells and nonmammalian systems; and 
(d) its ability to be absorbed by all routes of exposure followed by 
rapid distribution throughout the body.
    Some uncertainties are associated with the mechanisms of 
carcinogenicity of substance 3. Although there is considerable 
evidence indicating that mutagenic events could account for 
carcinogenic effects, there is still a lack of adequate information 
on the mutagenicity of substance 3 in vivo in animals or humans. 
Moreover, alternative modes of action have not been explored. 
Nonetheless, available data indicate a likely mutagenic mode of 
action. Linear extrapolation should be assumed in dose response 
assessment.

Example 4: ``Likely Human Carcinogen''--All Routes/Linear and Nonlinear 
Extrapolation

Human Data

    Substance 4 is a chlorinated alkene solvent. Several cohort 
studies of dry cleaning and laundry workers exposed to substance 4 
and other solvents reported significant excesses of mortality due to 
cancers of the lung, cervix, esophagus, kidney, bladder, lymphatic 
and hematopoietic system, colon, or skin. No significant cancer 
risks were observed in a subcohort of one these investigations of 
dry cleaning workers exposed mainly to substance 4. Possible 
confounding factors such as smoking, alcohol consumption, or low 
socioeconomic status were not considered in the analyses of these 
studies.
    A large case-control study of bladder cancer did not show any 
clear association with dry cleaning. Several case-control studies of 
liver cancer identified an increased risk of liver cancer with 
occupational exposure to organic solvents. The specific solvents to 
which workers were exposed and exposure levels were not identified.

Animal Data

    The potential carcinogenicity of substance 4 has been 
investigated in two long-term studies in rats and mice of both sexes 
by oral administration and inhalation.
    Significant increases in hepatocellular carcinomas were induced 
in mice of both sexes treated with substance 4 by oral gavage. No 
increases in tumor incidence were observed in treated rats. 
Limitations in both experiments included control groups smaller than 
treated groups, numerous dose adjustments during the study, and 
early mortality due to treatment-related nephropathy.
    In the inhalation study, there were significantly increased 
incidences of hepatocellular adenoma and carcinoma in exposed mice 
of both sexes. In rats of both sexes, there were marginally 
significant increased incidences of mononuclear cell leukemia (MCL) 
when compared with concurrent controls. The incidences of MCL in 
control animals, however, were higher than historical controls from 
the conducting laboratory. The tumor finding was also judged to be 
biologically significant because the time to onset of tumor was 
decreased and the disease was more severe in treated than in control 
animals. Low incidences of renal tubular cell adenomas or 
adenocarcinomas were also observed in exposed male rats. The tumor 
incidences were not statistically significant but there was a 
significant trend.

Other Key Data

    Substance 4 has been shown to be readily and rapidly absorbed by 
inhalation and ingestion in humans and laboratory animals. 
Absorption by dermal exposure is slow and limited. Once absorbed, 
substance 4 is primarily distributed to and accumulated in adipose 
tissue and the brain, kidney, and liver. A large percentage of 
substance 4 is eliminated unchanged in exhaled air, with urinary 
excretion of metabolites comprising a much smaller percentage. The 
absorption and distribution profiles of substance 4 are similar 
across species including humans.
    Two major metabolites (trichloroacetic acid (TCA), and 
trichloroethanol), which are formed by a P-450-dependent mixed-
function oxidase enzyme system, have been identified in all studied 
species, including humans. There is suggestive evidence for the 
formation of an epoxide intermediate based on the detection of two 
other metabolites (oxalic acid and trichloroacetyl amide). In 
addition to oxidative metabolism, substance 4 also undergoes 
conjugation with glutathione. Further metabolism by renal beta-
lyases could lead to two minor active metabolites (trichlorovinyl 
thiol and dichlorothiokente).
    Toxicokinetic studies have shown that the enzymes responsible 
for the metabolism of substance 4 can be saturated at high 
exposures. The glutathione pathway was found to be a minor pathway 
at low doses, but more prevalent following saturation of the 
cytochrome P-450 pathway. Comparative in vitro studies indicate that 
mice have the greater capacity to metabolize to TCA than rats and 
humans. Inhalation studies also indicate saturation of oxidative 
metabolism of substance 4, which occurs at higher dose levels in 
mice than in rats and humans. Based on these findings, it has been 
postulated that the species differences in the carcinogenicity of 
substance 4 between rats and mice may be related to the differences 
in the metabolism to TCA and glutathione conjugates.
    Substance 4 is a member of the class of chlorinated organics 
that often cause liver and kidney toxicity and carcinogenesis in 
rodents. Like many chlorinated organics, substance 4 itself does not 
appear to be mutagenic. Substance 4 was generally negative in in 
vitro bacterial systems and in vivo mammalian systems. However, a 
minor metabolite formed in the kidney by the glutathione conjugation 
pathway has been found to be a strong mutagen.
    The mechanisms of induced carcinogenic effects of substance 4 in 
rats and mice are not completely understood. It has been

[[Page 17989]]

postulated that mouse liver carcinogenesis is related to liver 
peroxisomal proliferation and toxicity of the metabolite TCA. 
Information on whether or not TCA induces peroxisomal proliferation 
in humans is not definitive. The induced renal tumors in male rats 
may be related either to kidney toxicity or the activity of a 
mutagenic metabolite. The mechanisms of increases in MCL in rats are 
not known.

Evaluation

    Available epidemiologic studies, taken together, provide 
suggestive evidence of a possible causal association between 
exposure to substance 4 and cancer incidence in the laundry and dry 
cleaning industries. This is based on consistent findings of 
elevated cancer risks in several studies of different populations of 
dry cleaning and laundry workers. However, each individual study is 
compromised by a number of study deficiencies including small 
numbers of cancers, confounding exposure to other solvents, and poor 
exposure characterization. Others may interpret these findings 
collectively as inconclusive.
    There is considerable evidence that substance 4 is carcinogenic 
to laboratory animals. It induces tumors in mice of both sexes by 
oral and inhalation exposure and in rats of both sexes via 
inhalation. However, due to incomplete understanding of the mode of 
mechanism of action, the predictivity of animal responses to humans 
is uncertain.
    Animal data of structurally related compounds showing common 
target organs of toxicity and carcinogenic effects (but lack of 
mutagenic effects) provide additional support for the 
carcinogenicity of substance 4. Comparative toxicokinetic and 
metabolism information indicates that the mouse may be more 
susceptible to liver carcinogenesis than rats and humans. This may 
indicate differences of the degree and extent of carcinogenic 
responses, but does not detract from the qualitative weight of 
evidence of human carcinogenicity. The toxicokinetic information 
also indicates that oral and inhalation are the major routes of 
human exposure.

Conclusion

    Substance 4 is likely to be carcinogenic to humans by all routes 
of exposure. The weight of evidence of human carcinogenicity is 
based on: (a) Demonstrated evidence of carcinogenicity in two rodent 
species of both sexes via two relevant routes of human exposure; (b) 
the substance's similarity in structure to other chlorinated 
organics that are known to cause liver and kidney toxicity and 
carcinogenesis in rodents; (c) suggestive evidence of a possible 
association between exposure to the substance in the laundry and dry 
cleaning industries and increased cancer incidence; and (d) human 
and animal data indicating that the substance is absorbed by all 
routes of exposure.
    In comparison with other agents designated as likely 
carcinogens, the overall weight of evidence places it the lower end 
of the grouping. This is because there is a lack of good evidence 
that observed excess cancer risk in exposed workers is due solely to 
substance 4. Moreover, there is considerable scientific uncertainty 
about the human significance of certain rodent tumors associated 
with substance 4 and related compounds. In this case, the human 
relevance of the animal evidence of carcinogenicity relies on the 
default assumption.
    Overall, there is not enough evidence to give high confidence in 
a conclusion about any single mode of action; it appears that more 
than one is plausible in different rodent tissues. Nevertheless, the 
lack of mutagenicity of substance 4 and its general growth-promoting 
effect on high background tumors as well as its toxicity toward 
mouse liver and rat kidney tissue support the view that the 
predominant mode is growth-promoting rather than mutagenic. A 
mutagenic contribution to carcinogenicity due to a metabolite cannot 
be ruled out. The dose response assessment should, therefore, adopt 
both default approaches, nonlinear and linear extrapolations. The 
latter approach is very conservative since it likely overestimates 
risk at low doses in this case, and is primarily useful for 
screening analyses.

Example 5: ``Likely/Not Likely Human Carcinogen''--Range of Dose 
Limited, Margin-of-Exposure Extrapolation

Human Data

    Substance 5 is a metal-conjugated phosphonate. No human tumor or 
toxicity data exist on this chemical.

Animal Data

    Substance 5 caused a statistically significant increase in the 
incidence of urinary bladder tumors in male, but not female, rats at 
30,000 ppm (3%) in the diet in a long-term study. Some of these 
animals had accompanying urinary tract stones and toxicity. No 
bladder tumors or adverse urinary tract effects were seen in two 
lower dose groups (2,000 and 8,000 ppm) in the same study. A chronic 
dietary study in mice at doses comparable to those in the rat study 
showed no tumor response or urinary tract effects. A 2-year study in 
dogs at doses up to 40,000 ppm showed no adverse urinary tract 
effects.

Other Key Data

    Subchronic dosing of rats confirmed that there was profound 
development of stones in the male bladder at doses comparable to 
those causing cancer in the chronic study, but not at lower doses. 
Sloughing of the epithelium of the urinary tract accompanied the 
stones.
    There was a lack of mutagenicity relevant to carcinogenicity. In 
addition, there is nothing about the chemical structure of substance 
5 to indicate DNA-reactivity or carcinogenicity.
    Substance 5 is composed of a metal, ethanol, and a simple 
phosphorus-oxygen-containing component. The metal is not absorbed 
from the gut, whereas the other two components are absorbed. At high 
doses, ethanol is metabolized to carbon dioxide, which makes the 
urine more acidic; the phosphorus level in the blood is increased 
and calcium in the urine is increased. Chronic testing of the 
phosphorus-oxygen-containing component alone in rats did not show 
any tumors or adverse effects on the urinary tract.
    Because substance 5 is a metal complex, it is not likely to be 
readily absorbed from the skin.

Evaluation

    Substance 5 produced cancer of the bladder and urinary tract 
toxicity in male, but not female rats and mice, and dogs failed to 
show the toxicity noted in male rats. The mode of action developed 
from the other key data to account for the toxicity and tumors in 
the male rats is the production of bladder stones. At high but not 
lower subchronic doses in the male rat, substance 5 leads to 
elevated blood phosphorus levels; the body responds by releasing 
excess calcium into the urine. The calcium and phosphorus combine in 
the urine and precipitate into multiple stones in the bladder. The 
stones are very irritating to the bladder; the bladder lining is 
eroded, and cell proliferation occurs to compensate for the loss of 
the lining. Cell layers pile up, and finally, tumors develop. Stone 
formation does not involve the chemical per se but is secondary to 
the effects of its constituents on the blood and, ultimately, the 
urine. Bladder stones, regardless of their cause, commonly produce 
bladder tumors in rodents, especially the male rat.

Conclusion

    Substance 5, a metal aliphatic phosphonate, is likely to be 
carcinogenic to humans only under high-exposure conditions following 
oral and inhalation exposure that lead to bladder stone formation, 
but is not likely to be carcinogenic under low-exposure conditions. 
It is not likely to be a human carcinogen via the dermal route, 
given that the compound is a metal conjugate that is readily ionized 
and its dermal absorption is not anticipated. The weight of evidence 
is based on (a) bladder tumors only in male rats; (b) the absence of 
tumors at any other site in rats or mice; (c) the formation of 
calcium-phosphorus-containing bladder stones in male rats at high, 
but not low, exposures that erode bladder epithelium and result in 
profound increases in cell proliferation and cancer; and (d) the 
absence of structural alerts or mutagenic activity.
    There is a strong mode of action basis for the requirements of 
(a) high doses of substance 5, (b) which lead to excess calcium and 
increased acidity in the urine, (c) which result in the 
precipitation of stones and (d) the necessity of stones for toxic 
effects and tumor hazard potential. Lower doses fail to perturb 
urinary constituents, lead to stones, produce toxicity, or give rise 
to tumors. Therefore, dose response assessment should assume 
nonlinearity.
    A major uncertainty is whether the profound effects of substance 
5 may be unique to the rat. Even if substance 5 produced stones in 
humans, there is only limited evidence that humans with bladder 
stones develop cancer. Most often human bladder stones are either 
passed in the urine or lead to symptoms resulting in their removal. 
However, since one cannot totally dismiss the male rat findings, 
some hazard

[[Page 17990]]

potential may exist in humans following intense exposures. Only 
fundamental research could illuminate this uncertainty.

Example 6: ``Cannot Be Determined''--Suggestive Evidence

Human Data

    Substance 6 is an unsaturated aldehyde. In a cohort study of 
workers in a chemical plant exposed to a mixture of chemicals with 
substance 6 as a minor component, an elevated risk of cancer than 
was expected was reported. This study is considered inadequate 
because of multiple exposures, small cohort, and poor exposure 
characterization.

Animal Data

    Substance 6 was tested for potential carcinogenicity in a 
drinking water study in rats, an inhalation study in hamsters, and a 
skin painting study in mice. No significant increases in tumors were 
observed in male rats treated with substance 6 at three dose levels 
in drinking water. However, a significant increase of adrenal 
cortical adenomas was found in the only treated female dose group 
administered a dose equivalent to the high dose of males. This study 
used a small number of animals (20 per dose group).
    No significant finding was detected in the inhalation study in 
hamsters. This study is inadequate due to the use of too few 
animals, short duration of exposure, and inappropriate dose 
selection (use of a single exposure that was excessively toxic as 
reflected by high mortality).
    No increase in tumors was induced in the skin painting study in 
mice. This study is of inadequate design for carcinogenicity 
evaluation because of several deficiencies: small number of animals, 
short duration of exposure, lack of reporting about the sex and age 
of animals, and purity of test material.
    Substance 6 is structurally related to lowmolecularweight 
aldehydes that generally exhibit carcinogenic effects in the 
respiratory tracts of laboratory animals via inhalation exposure. 
Three skin painting studies in mice and two subcutaneous injection 
studies of rats and mice were conducted to evaluate the carcinogenic 
potential of a possible metabolite of substance 6 (identified in 
vitro). Increased incidences of either benign or combined benign and 
malignant skin tumors were found in the dermal studies. In the 
injection studies of rats and mice, increased incidences of local 
sarcomas or squamous cell carcinoma were found at the sites of 
injection. All of these studies are limited by the small number of 
test animals, the lack of characterization of test material, and the 
use of single doses.

Other Key Data

    Substance 6 is a flammable liquid at room temperature. Limited 
information on its toxicokinetics indicates that it can be absorbed 
by all routes of exposure. It is eliminated in the urine mainly as 
glutathione conjugates. Substance 6 is metabolized in vitro by rat 
liver and lung microsomal preparations to a dihydroxylated aldehyde.
    No data were available on the genetic and related effects of 
substance 6 in humans. It did not induce dominant lethal mutations 
in mice. It induced sister chromatid exchanges in rodent cells in 
vitro. The mutagenicity of substance 6 is equivocal in bacteria. It 
did not induce DNA damage or mutations in fungi.

Evaluation

    Available human data are judged inadequate for an evaluation of 
any causal relationship between exposure to substance 6 and human 
cancer.
    The carcinogenic potential of substance 6 has not been 
adequately studied in laboratory animals due to serious deficiencies 
in study design, especially the inhalation and dermal studies. There 
is some evidence of carcinogenicity in the drinking water study in 
female rats. However, the significance and predictivity of that 
study to human response are uncertain since the finding is limited 
to occurrence of benign tumors, one sex, and at the high dose only. 
Additional suggestion for animal carcinogenicity comes from 
observation that a possible metabolite is carcinogenic at the site 
of administration. This metabolite, however, has not been studied in 
vivo. Overall, the animal evidence is judged to be suggestive for 
human carcinogenicity.
    Other key data, taken together, do not add significantly to the 
overall weight of evidence of carcinogenicity. SAR analysis 
indicates that substance 6 would be DNA-reactive. However, 
mutagenicity data are inconclusive. Limited in vivo data do not 
support a mutagenic effect. While there is some evidence of DNA 
damage in rodent cells in vitro, there is either equivocal or no 
evidence of mutagenicity in nonmammalian systems.

Conclusion

    The human carcinogenicity potential of substance 6 cannot be 
determined on the basis of available information. Both human and 
animal data are judged inadequate for an evaluation. There is 
evidence suggestive of potential carcinogenicity on the basis of 
limited animal findings and SAR considerations. Data are not 
sufficient to judge whether there is a mode of carcinogenic action. 
Additional studies are needed for a full evaluation of the potential 
carcinogenicity of substance 6. Hence, dose response assessment is 
not appropriate.

Example 7: ``Not Likely Human Carcinogen''--Appropriately Studied 
Chemical in Animals Without Tumor Effects

Human Data

    Substance 7, a plant extract, has not been studied for its toxic 
or carcinogenic potential in humans.

Animal Data

    Substance 7 has been studied in four chronic studies in three 
rodent species. In a feeding study in rats, males showed a 
nonsignificant increase in benign tumors of the parathyroid gland in 
the high-dose group, where the incidence in concurrent controls 
greatly exceeded the historical control range. Females demonstrated 
a significant increase in various subcutaneous tumors in the low-
dose group, but findings were not confirmed in the high-dose group, 
and there was no dose response relationship. These effects were 
considered as not adding to the evidence of carcinogenicity. No 
tumor increases were noted in a second adequate feeding study in 
male and female rats. In a mouse feeding study, no tumor increases 
were noted in dosed animals. There was some question as to the 
adequacy of the dosing; however it was noted that in the mouse 90-d 
subchronic study, a dose of twice the high dose in the chronic study 
led to significant decrements in body weight. In a hamster study 
there were no significant increases in tumors at any site. No 
structural analogues of substance 7 have been tested for cancer.

Other Key Data

    There are no structural alerts that would suggest that substance 
7 is a DNA-reactive compound. It is negative for gene mutations in 
bacteria and yeast, but positive in cultured mouse cells. Tests for 
structural chromosome aberrations in cultured mammalian cells and in 
rats are negative; however, the animals were not tested at 
sufficiently high doses. Substance 7 binds to proteins of the cell 
division spindle; therefore, there is some likelihood for producing 
numerical chromosome aberrations, an endpoint that is sometimes 
noted in cancers. In sum, there is limited and conflicting 
information concerning the mutagenic potential of the agent.
    The compound is absorbed via oral and inhalation exposure but 
only poorly via the skin.

Evaluation

    The only indication of a carcinogenic effect comes from the 
finding of benign tumors in male rats in a single study. There is no 
confirmation of a carcinogenic potential from dosed females in that 
study, in males and females in a second rat study, or from mouse and 
hamster studies.
    There is no structural indication that substance 7 is DNA-
reactive, there is inconsistent evidence of gene mutations, and 
chromosome aberration testing is negative. The agent binds to cell 
division spindle proteins and may have the capacity to induce 
numerical chromosome anomalies. Further information on gene 
mutations and in vivo structural and numerical chromosome 
aberrations may be warranted.

Conclusion

    Substance 7 is not likely to be carcinogenic to humans via all 
relevant routes of exposure. This weight of evidence judgment is 
largely based on the absence of significant tumor increases in 
chronic rodent studies. Adequate cancer studies in rats, mice, and 
hamsters fail to show any carcinogenic effect; a second rat study 
showed an increase in benign tumors at a site in dosed males, but 
not females.

2.7. Presentation of Results

    The results of the hazard assessment are presented in the form of 
an overall technical hazard characterization. Additionally, a weight of 
evidence narrative is used when the conclusion as to carcinogenic 
potential needs to be

[[Page 17991]]

presented separately from the overall characterization.
2.7.1. Technical Hazard Characterization
    The hazard characterization has two functions. First, it presents 
results of the hazard assessment and an explanation of how the weight 
of evidence conclusion was reached. It explains the potential for human 
hazard, anticipated attributes of its expression, and mode of action 
considerations for dose response. Second, it contains the information 
needed for eventual incorporation into a risk characterization 
consistent with EPA guidance on risk characterization (U.S. EPA, 1995).
    The characterization qualitatively describes the conditions under 
which the agent's effects may be expressed in human beings. These 
qualitative hazard conditions are ones that are observable in the 
toxicity data without having done either quantitative dose response or 
exposure assessment. The description includes how expression is 
afffected by route of exposure and dose levels and durations of 
exposure.
    The discussion of limitations of dose as a qualitative aspect of 
hazard addresses the question of whether reaching a certain dose range 
appears to be a precondition for a hazard to be expressed; for example, 
when carcinogenic effects are secondary to another toxic effect that 
appears only when a certain dose level is reached. The assumption is 
made that an agent that causes internal tumors by one route of exposure 
will be carcinogenic by another route, if it is absorbed by the second 
route to give an internal dose. Conversely, if there is a route of 
exposure by which the agent is not absorbed (does not cross an 
absorption barrier; e.g., the exchange boundaries of skin, lung, and 
digestive tract through uptake processes) to any significant degree, 
hazard is not anticipated by that route. An exception to the latter 
statement would be when the site of contact is also the target tissue 
of carcinogenicity. Duration of exposure may be a precondition for 
hazard if, for example, the mode of action requires cytotoxicity or a 
physiologic change, or is mitogenicity, for which exposure must be 
sustained for a period of time before effects occur. The 
characterization could note that one would not anticipate a hazard from 
isolated, acute exposures. The above conditions are qualitative ones 
regarding preconditions for effects, not issues of relative absorption 
or potency at different dose levels. The latter are dealt with under 
dose response assessment (section 3), and their implications can only 
be assessed after human exposure data are applied in the 
characterization of risk.
    The characterization describes conclusions about mode of action 
information and its support for recommending dose response approaches.
    The hazard characterization routinely includes the following in 
support of risk characterization:
     a summary of results of the assessment,
     identification of the kinds of data available to support 
conclusions and explanation of how the data fit together, highlighting 
the quality of the data in each line of evidence, e.g., tumor effects, 
short-term studies, structure-activity relationships), and highlighting 
the coherence of inferences from the different kinds of data,
     strengths and limitations (uncertainties) of the data and 
assessment, including identification of default assumptions invoked in 
the face of missing or inadequate data,
     identification of alternative interpretations of data that 
are considered equally plausible,
     identification of any subpopulations believed to be more 
susceptible to the hazard than the general population,
     conclusions about the agent's mode of action and 
recommended dose response approaches,
     significant issues regarding interpretation of data that 
arose in the assessment. Typical ones may include:

--determining causality in human studies,
--dosing (MTD), background tumor rates, relevance of animal tumors to 
humans,
--weighing studies with positive and null results, considering the 
influence of other available kinds of evidence,
--drawing conclusions based on mode of action data versus using a 
default assumption about the mode of action.
2.7.2. Weight of Evidence Narrative
    The weight of evidence narrative summarizes the results of hazard 
assessment employing the descriptors defined in section 2.6.1. The 
narrative (about two pages in length) explains an agent's human 
carcinogenic potential and the conditions of its expression. If data do 
not allow a conclusion as to carcinogenicity, the narrative explains 
the basis of this determination. An example narrative appears below. 
More examples appear in Appendix A.
    The items regularly included in a narrative are:
     name of agent and Chemical Abstracts Services number, if 
available,
     conclusions (by route of exposure) about human 
carcinogenicity, using a standard descriptor from section 2.6.1,
     summary of human and animal tumor data on the agent or its 
structural analogues, their relevance, and biological plausibility,
     other key data (e.g., structure-activity data, 
toxicokinetics and metabolism, short-term studies, other relevant 
toxicity or clinical data),
     discussion of possible mode(s) of action and appropriate 
dose response approach(es),
     conditions of expression of carcinogenicity, including 
route, duration, and magnitude of exposure.

Example Narrative

Aromatic Compound

CAS# XXX

CANCER HAZARD SUMMARY

    Aromatic compound (AR) is known to be carcinogenic to humans by 
all routes of exposure.
    The weight of evidence of human carcinogenicity is based on (a) 
consistent evidence of elevated leukemia incidence in studies of 
exposed workers and significant increases of genetic damage in bone 
marrow cells and blood lymphocytes of exposed workers; (b) 
significantly increased incidence of cancer in both sexes of several 
strains of rats and mice; (c) genetic damage in bone marrow cells of 
exposed rodents and effects on intracellular signals that control 
cell growth.
    AR is readily absorbed by all routes of exposure and rapidly 
distributed throughout the body. The mode of action of AR is not 
understood. A dose response assessment that assumes linearity of the 
relationship is recommended as a default.

SUPPORTING INFORMATION

    Data include numerous human epidemiologic and biomonitoring 
studies, long-term bioassays, and other data on effects of AR on 
genetic material and cell growth processes. The key epidemiologic 
studies and animal studies are well conducted and reliable. The 
other data are generally of good quality also.

Human Effects

    Numerous epidemiologic and case studies have reported an 
increased incidence or a causal relationship associating exposure to 
AR and leukemia. Among the studies are five for which the design and 
performance as well as follow-up are considered adequate to 
demonstrate the causal relationship. Biomonitoring studies of 
exposed workers have found dose-related increases in chromosomal 
aberrations in bone marrow cells and blood lymphocytes.

Animal Effects

    AR caused increased incidence of tumors in various tissues in 
both sexes of several rat and mouse strains. AR also caused 
chromosomal aberrations in rabbits, mice, and rats--as it does in 
humans.

[[Page 17992]]

Other Key Data

    AR itself is not DNA-reactive and is not mutagenic in an array 
of test systems both in vitro and in vivo. Metabolism of AR yields 
several metabolites that have been separately studied for effects on 
carcinogenic processes. Some have mutagenic activity in test systems 
and some have other effects on cell growth controls inside cells.

MODE OF ACTION

    No rodent tumor precisely matches human leukemia in pathology. 
The closest parallel is a mouse cancer of blood-forming tissue. 
Studies of the effects of AR at the cell level in this model system 
are ongoing. As yet, the mode of action of AR is unclear, but most 
likely the carcinogenic activity is associated with one or a 
combination of its metabolites. It is appropriate to apply a linear 
approach to the dose response assessment pending a better 
understanding because: (a) genetic damage is a typical effect of AR 
exposure in mammals and (b) metabolites of AR produce mutagenic 
effects in addition to their other effects on cell growth controls; 
AR is a multitissue carcinogen in mammals suggesting that it is 
affecting a common controlling mechanism of cell growth.

3. Dose Response Assessment

    Dose response assessment first addresses the relationship of dose 
2 to the degree of response observed in an experiment or human 
study. When environmental exposures are outside of the range of 
observation, extrapolations are necessary in order to estimate or 
characterize the dose relationship (ILSI, 1995). In general, three 
extrapolations may be made: from high to low doses, from animal to 
human responses, and from one route of exposure to another.
---------------------------------------------------------------------------

    \2\ 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. ``Delivered dose'' for an 
organ or cell means the amount available for interaction with that 
organ or cell (U.S. EPA, 1992a).
---------------------------------------------------------------------------

    The dose response assessment proceeds in two parts. The first is 
assessment of the data in the range of empirical observation. This is 
followed by extrapolations either by modeling, if there are sufficient 
data to support a model, or by a default procedure based as much as 
possible on information about the agent's mode of action. The following 
discussion covers the assessment of observed data and extrapolation 
procedures, followed by sections on analysis of response data and 
analysis of dose data. The final section discusses dose response 
characterization.

3.1. Dose Response Relationship

    In the discussion that follows, reference to ``response'' data 
includes measures of tumorigenicity as well as other responses related 
to carcinogenicity. The other responses may include effects such as 
changes in DNA, chromosomes, or other key macromolecules, effects on 
growth signal transduction, induction of physiological or hormonal 
changes, effects on cell proliferation, or other effects that play a 
role in the process. Responses other than tumorigenicity may be 
considered part of the observed range in order either to extend the 
tumor dose response analysis or to inform it. The nontumor response or 
responses also may be used in lieu of tumor data if they are considered 
to be a more informative representation of the carcinogenic process for 
an agent (see section 3.2).
3.1.1. Analysis in the Range of Observation
    Biologically Based and Case-Specific Models. A biologically based 
model is one whose parameters are calculated independently of curve-
fitting of tumor data. If data are sufficient to support a biologically 
based model specific to the agent and the purpose of the assessment is 
such as to justify investing resources supporting use, this is the 
first choice for both the observed tumor and related response data and 
for extrapolation below the range of observed data in either animal or 
human studies. Examples are the two-stage models of initiation plus 
clonal expansion and progression developed by Moolgavkar and Knudson 
(1981) and Chen and Farland (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. Theoretical estimates 
of process parameters, such as cell proliferation rates, are not used 
to enable application of such a model (Portier, 1987).
    Similarly preferred as a first choice are dose response models 
based on general concepts of mode of action and data on the agent. For 
a case-specific model, model parameters and data are obtained from 
studies on the agent.
    In most cases, a biologically based or case-specific model will not 
be practicable, either because the necessary data do not exist or the 
decisions that the assessment are to support do not justify or permit, 
the time and resources required. In these cases, the analysis proceeds 
using curve-fitting models followed by default procedures for 
extrapolation, based, to the extent possible, on mode of action and 
other biological information about the agent. These methods and 
assumptions are described below.
    Curve-Fitting and Point of Departure for Extrapolation. Curve-
fitting models are used that are appropriate to the kind of response 
data in the observed range. Any of several models can be used; e.g., 
the models developed for benchmark dose estimation for noncancer 
endpoints may be applied (Barnes et al., 1995).
    For some data sets, particularly those with extreme curvature, the 
impact of model selection can be significant. In these cases, the 
choice is rationalized on biological grounds as possible. In other 
cases, the nature of the data or the way it is reported will suggest 
other types of models; for instance, when longitudinal data on tumor 
development are available, time to tumor or survival models may be 
necessary and appropriate to fit the data.
    A point of departure for extrapolation is estimated. This is a 
point that is either a data point or an estimated point that can be 
considered to be in the range of observation, without any significant 
extrapolation. The LED10--the lower 95% confidence limit on a dose 
associated with 10% extra risk--is such a point and is the standard 
point of departure, adopted as a matter of science policy to remain as 
consistent and comparable from case to case as possible.3 It is 
also a comparison point for noncancer endpoints (U.S. EPA, 1991f). The 
central estimate of the ED10 also may be appropriate for use in 
relative hazard and potency ranking.
---------------------------------------------------------------------------

    \3\ It is appropriate to report the central estimate of the 
ED10, the upper and lower 95% confidence limits, and a 
graphical representation of model fit.
---------------------------------------------------------------------------

    For some data sets, a choice of point of departure other than the 
LED10 may be appropriate. For example, if the observed response is 
below the LED10, then a lower point may be a better choice. 
Moreover, some forms of data may not be amenable to curve-fitting 
estimation, but to estimation of a ``low-'' or ``no-observable-adverse-
effect level'' (LOAEL, NOAEL) instead, e.g., certain continuous data.
    The rationale supporting the use of the LED10 is that a 10% 
response is at or just below the limit of sensitivity of discerning a 
significant difference in most long-term rodent studies. The lower 
confidence limit on dose is used to appropriately account for 
experimental uncertainty (Barnes et al., 1995) and for consistency with 
the ``benchmark dose'' approach for noncancer assessment; it does not 
provide information about human

[[Page 17993]]

variability. In laboratory studies of cancer or noncancer endpoints, 
the level of dose at which increased incidence of effects can be 
detected, as compared to controls, is a function of the size of the 
sample (e.g., number of animals), dose spacing, and other design 
aspects. In noncancer assessment, the dose at which significant effects 
are not observed is traditionally termed the NOAEL. This is not, in 
fact, a level of zero effect. The NOAEL in most study protocols is 
about the same as an LED5 or LED10--the lower 95% confidence 
limit on a dose associated with a 5% or 10% increased effect (Faustman 
et al., 1994; Haseman, 1983). Adopting parallel points of departure for 
cancer and noncancer assessment is intended to make discussion and 
comparison of the two kinds of assessment more comparable because of 
their similar science and science policy bases and similar analytic 
approaches.
    Analysis of human studies in the observed range is designed case by 
case, depending on the type of study and how dose and response are 
measured in the study. In some cases the agent may have discernible 
interactive effects with another agent (e.g., asbestos and smoking), 
making possible estimation of contribution of the agent and others as 
risk factors. Also, in some cases, estimation of population risk in 
addition, or in lieu of, individual risk may be appropriate.
3.1.2. Analysis in the Range of Extrapolation
    Extrapolation to lower doses is usually necessary, and in the 
absence of a biologically based or case-specific model, is based on one 
of the three default procedures described below. The Agency has adopted 
these three procedures as a matter of science policy based on current 
hypotheses of the likely shapes of dose response curves for differing 
modes of action. The choice of the procedure to be used in an 
individual case is a judgment based on the agent's modes of action.
    Linear. A default assumption of linearity is appropriate when the 
evidence supports a mode of action of gene mutation due to DNA 
reactivity or supports another mode of action that is anticipated to be 
linear. Other elements of empirical support may also support an 
inference of linearity, e.g., the background of human exposure to an 
agent might be such that added human exposure is on the linear part of 
a dose response curve that is sublinear overall. The default assumption 
of linearity is also appropriate as the ultimate science policy default 
when evidence shows no DNA reactivity or other support for linearity, 
but neither does it show sufficient evidence of a nonlinear mode of 
action to support a nonlinear procedure.
    For linear extrapolation, a straight line is drawn from the point 
of departure to the origin--zero dose, zero response (Flamm and 
Winbush, 1984; Gaylor and Kodell, 1980; Krewski et al., 1984). This 
approach is generally conservative of public health, in the absence of 
information about the extent of human variability in sensitivity to 
effects. When a linear extrapolation procedure is used, the risk 
characterization summary displays the degree of extrapolation that is 
being made from empirical data and discusses its implications for the 
interpretation of the resulting quantitative risk estimates.
    Nonlinear. A default assumption of nonlinearity is appropriate when 
there is no evidence for linearity and sufficient evidence to support 
an assumption of nonlinearity. The mode of action may lead to a dose 
response relationship that is nonlinear, with response falling much 
more quickly than linearly with dose, or being most influenced by 
individual differences in sensitivity. Alternatively, the mode of 
action may theoretically have a threshold, e.g., the carcinogenicity 
may be a secondary effect of toxicity or of an induced physiological 
change (see example 5, section 2.6.3) that is itself a threshold 
phenomenon.
    As a matter of science policy under this analysis, nonlinear 
probability functions are not fitted to the response data to 
extrapolate quantitative low-dose risk estimates because different 
models can lead to a very wide range of results, and there is currently 
no basis, generally, to choose among them. Sufficient information to 
choose leads to a biologically based or case-specific model. In cases 
of nonlinearity, the risk is not extrapolated as a probability of an 
effect at low doses. A margin of exposure analysis is used, as 
described below, to evaluate concern for levels of exposure. The margin 
of exposure is the LED10 or other point of departure divided by 
the environmental exposure of interest. The EPA does not generally try 
to distinguish between modes of action that might imply a ``true 
threshold'' from others with a nonlinear dose response relationship. 
Except in unusual cases where extensive information is available, it is 
not possible to distinguish between these empirically.
    The environmental exposures of interest, for which margins of 
exposure are estimated, may be actual or projected future levels. The 
risk manager decides whether a given margin of exposure is acceptable 
under applicable management policy criteria. The risk assessment 
provides supporting information to assist the decisionmaker.
    The EPA often conducts margin of exposure analyses to accompany 
estimates of reference doses or concentrations (RfD, RfC) for noncancer 
endpoints.4 The procedure for a margin of exposure analysis for a 
response related to carcinogenicity is operationally analogous, the 
difference being that a threshold of cancer response is not necessarily 
presumed. If, in a particular case, the evidence indicates a threshold, 
as in the case of carcinogenicity being secondary to another toxicity 
that has a threshold, the margin of exposure analysis for the toxicity 
is the same as is done for a noncancer endpoint, and an RfD or RfC for 
that toxicity also may be estimated and considered in cancer 
assessment.
---------------------------------------------------------------------------

    \4\ An RfD or RfC is an estimate with uncertainty spanning 
perhaps an order of magnitude of daily exposure to the human 
population (including sensitive subgroups) that is anticipated to be 
without appreciable deleterious effects during a lifetime. It is 
arrived at by dividing empirical data on effects by uncertainty 
factors that consider inter- and intraspecies variability, extent of 
data on all important chronic exposure toxicity endpoints, and 
availability of chronic as opposed to subchronic data.
---------------------------------------------------------------------------

    The analogy between margin of exposure analysis for noncancer and 
cancer responses begins with the analogy of points of departure; for 
both it is an effect level, either LED10 or other point (presented 
as a human equivalent dose or concentration), as data support. For 
cancer responses, when animal data are used, the point of departure is 
a human equivalent dose or concentration arrived at by interspecies 
dose adjustment or toxicokinetic analysis. It is likely that many of 
the margin of exposure analyses for cancer will be for responses other 
than tumor incidence. This is because the impetus for considering a 
carcinogenic agent to have a nonlinear dose response will be a 
conclusion that there is sufficient evidence to support that view, and 
this evidence will often be information about a response that is a 
precursor to tumors.
    To support a risk manager's consideration of the margin of 
exposure, information is provided in a risk assessment about current 
understanding of the phenomena that may be occurring as dose (exposure) 
decreases substantially below the observed data. The goal is to provide 
as much information as possible about the risk reduction that 
accompanies lowering of exposure. To this end, some important points to 
address include:

[[Page 17994]]

     The slope of the observed dose response relationship at 
the point of departure and its uncertainties and implications for risk 
reduction associated with exposure reduction (a shallow slope suggests 
less reduction than a steep slope),
     The nature of the response used for the dose response 
assessment,
     The nature and extent of human variability in sensitivity 
to the phenomena involved,
     Persistence of the agent in the body,
     Human sensitivity to the phenomena as compared with 
experimental animals.
    As a default assumption for two of these points, a factor of no 
less than 10-fold each may be employed to account for human variability 
and for interspecies differences in sensitivity when humans may be more 
sensitive than animals. When humans are found to be less sensitive than 
animals, a default factor of no smaller than a 1/10 fraction may be 
employed to account for this. If any information about human 
variability or interspecies differences is available, it is used 
instead of the default or to modify it as appropriate. In the case of 
analysis based on human studies, obviously, interspecies differences 
are not a factor. It should be noted that the dose response 
relationship and inter- or intraspecies variability in sensitivity are 
independent. That is, reduction of dose reduces risk; it does not 
change variability. To support consideration of acceptability of a 
margin of exposure by the risk manager, the assessment considers all of 
the hazard and dose response factors together; hence, the factors for 
inter- and intraspecies differences alone are not to be considered a 
default number for an acceptable margin of exposure. (See Section 
1.3.2.5.)
    It is appropriate to provide a graphical representation of the data 
and dose response modeling in the observed range, also showing exposure 
levels of interest to the decisionmaker. (See figure 3-1.) In order to 
provide a frame of reference, by way of comparison, a straight line 
extrapolation may be displayed to show what risk levels would be 
associated with decreasing dose, if the dose response were linear. If 
this is done, the clear accompanying message is that, in this case of 
nonlinearity, the response falls disproportionately with decreasing 
dose.

BILLING CODE 6560-50-P

[[Page 17995]]

[GRAPHIC] [TIFF OMITTED] TN23AP96.006



BILLING CODE 6560-50-C

[[Page 17996]]

    Linear and Nonlinear. Both linear and nonlinear procedures may be 
used in particular cases. If a mode of action analysis finds 
substantial support for differing modes of action for different tumor 
sites, an appropriate procedure is used for each. Both procedures may 
also be appropriate to discuss implications of complex dose response 
relationships. For example, if it is apparent that an agent is both DNA 
reactive and is highly active as a promotor at high doses, and there 
are insufficient data for modeling, both linear and nonlinear default 
procedures may be needed to decouple and consider the contribution of 
both phenomena.
3.1.3. Use of Toxicity Equivalence Factors and Relative Potency 
Estimates
    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 (U.S. EPA, 
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 
characteristics 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 
relationships.
    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. Criteria for constructing TEFs are 
given in U.S. EPA (1991b). The criteria call for data that are adequate 
to support summing doses of the agents in mixtures. To date, adequate 
data to support use of TEF's has been found in only one class of 
compounds (dioxins) (U.S. EPA, 1989a).
    Relative potencies can be similarly derived and used for agents 
with carcinogenicity or other supporting data. These are conceptually 
similar to TEFs, but they are less firmly based in science and do not 
have the same level of data to support them. They are used only when 
there is no better alternative.
    The uncertainties associated with both TEFs and relative potencies 
are explained whenever they are used.

3.2. Response Data

    Response data for analysis include tumor incidence data from human 
or animal studies as well as data on other responses as they relate to 
an agent's carcinogenicity, such as effects on growth control processes 
or cell macromolecules or other toxic effects. Tumor incidence data are 
ordinarily the basis of dose response assessment, but other response 
data can augment such assessment or provide separate assessments of 
carcinogenicity or other important effects.
    Data on carcinogenic processes underlying tumor effects may be used 
to support biologically based or case-specific models. Other options 
for such data exist. If confidence is high in the linkage of a 
precursor effect and the tumor effect, the assessment of tumor 
incidence may be extended to lower dose levels by linking it to the 
assessment of the precursor effect (Swenberg et al., 1987). Even if a 
quantitative link is not appropriate, the assessment for a precursor 
effect may provide a view of the likely shape of the dose response 
curve for tumor incidence below the range of tumor observation (Cohen 
and Ellwein, 1990; Choy, 1993). If responses other than tumor incidence 
are regarded as better representations of the carcinogenicity of the 
agent, they may be used in lieu of tumor responses. For example, if it 
is concluded that the carcinogenic effect is secondary to another toxic 
effect, the dose response for the other effect will likely be more 
pertinent for risk assessment. As another example, if disruption of 
hormone activity is the key mode of action of an agent, data on hormone 
activity may be used in lieu of tumor incidence data.
    If adequate positive human epidemiologic response data are 
available, they provide an advantageous basis for analysis since 
concerns about interspecies extrapolation do not arise. Adequacy of 
human exposure data for quantification is an important consideration in 
deciding whether epidemiologic data are the best basis for analysis in 
a particular case. 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 human 
risk to provide a check on plausibility of available estimates based on 
animal tumor or other responses, e.g., do confidence limits on one 
overlap the point estimate of the other?
    When animal studies are used, response data from a species that 
responds most like humans should be used if information to this effect 
exists. If this is unknown and an agent has been tested in several 
experiments involving different animal species, strains, and sexes at 
several doses and different routes of exposure, all of the data sets 
are considered and compared, and a judgment is made as to the data to 
be used to best represent the observed data and important biological 
features such as mode of action. Appropriate options for presenting 
results include:
     Use of a single data set,
     Combining data from different experiments (Stiteler et 
al., 1993; Vater et al., 1993),
     Showing a range of results from more than one data set,
     Showing results from analysis of more than one 
statistically significant tumor response based on differing modes of 
action,
     Representing total response in a single experiment by 
combining animals with statistically significant tumors at more than 
one site, or
     A combination of these options.
    The approach judged to best represent the data is presented with 
the rationale for the judgment, including the biological and 
statistical considerations involved. The following are some points to 
consider:
     Quality of study protocol and execution,
     Proportion of malignant neoplasms,
     Latency of onset of neoplasia,
     Number of data points to define the relationship of dose 
and response,
     Background incidence in test animal,
     Differences in range of response among species, sexes, 
strains,
     Most sensitive responding species, and
     Availability of data on related precursor events to tumor 
development.
    Analyses of carcinogenic effects other than tumor incidence are 
similarly presented and evaluated for their contribution to a best 
judgment on how to represent the biological data for dose response 
assessment.

3.3. Dose Data

    Whether animal experiments or epidemiologic studies are the sources 
of data, questions need to be addressed in arriving at an appropriate 
measure of dose for the anticipated environmental exposure. Among these 
are:
     Whether the dose is expressed as an environmental 
concentration, applied dose, or delivered dose to the target organ,

[[Page 17997]]

     Whether the dose is expressed in terms of a parent 
compound, one or more metabolites, or both,
     The impact of dose patterns and timing where significant,
     Conversion from animal to human doses, where animal data 
are used, and
     The conversion metric between routes of exposure where 
necessary and appropriate.
    In practice, there may be little or no information on the 
concentration or identity of the active form at a target; being able to 
compare the applied and delivered doses between routes and species is 
the ideal, but is rarely attained. Even so, the objective is to use 
available data to obtain as close to a measure of internal or delivered 
dose as possible.
    The following discussion assumes that the analyst will have data of 
varying detail in different cases about toxicokinetics and metabolism. 
Discussed below are approaches to basic data that are most frequently 
available, as well as approaches and judgments for improving the 
analysis based on additional data. The estimation of dose in human 
studies is tailored to the form of dose data available.
3.3.1. Interspecies Adjustment of Dose
    When adequate data are available, the doses used in animal studies 
can be adjusted to equivalent human doses using toxicokinetic 
information on the particular agent. The methods used should be 
tailored to the nature of the data on a case-by-case basis. In rare 
cases, it may also be possible to make adjustments based on 
toxicodynamic considerations. In most cases, however, there are 
insufficient data available to compare dose between species. In these 
cases, the estimate of human equivalent dose is based on science policy 
default assumptions. The defaults described below are modified or 
replaced whenever better comparative data on toxicokinetic or metabolic 
relationships are available. The availability and discussion of the 
latter also may permit reduction or discussion of uncertainty in the 
analysis.
    For oral exposure, the default assumption is that delivered doses 
are related to applied dose by a power of body weight. 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. To derive an equivalent human oral dose from 
animal data, the default procedure is to scale daily applied doses 
experienced for a lifetime in proportion to body weight raised to the 
0.75 power (W0.75). Equating exposure concentrations in parts per 
million units for food or water is an alternative version of the same 
default procedure because daily intakes of these are in proportion to 
W0.75. The rationale for this factor rests on the empirical 
observation that rates of physiological processes consistently tend to 
maintain proportionality with W0.75. A more extensive discussion 
of the rationale and data supporting the Agency's adoption of this 
scaling factor is in U.S. EPA, 1992b. Information such as blood levels 
or exposure biomarkers or other data that are available for 
interspecies comparison are used to improve the analysis when possible.
    The default procedure to derive an human equivalent concentration 
of inhaled particles and gases is described in U.S. EPA (1994) and 
Jarabek (1995a,b). The methodology estimates respiratory deposition of 
inhaled particles and gases and provides methods for estimating 
internal doses of gases with different absorption characteristics. The 
method is able to incorporate additional toxicokinetics and metabolism 
to improve the analysis if such data are available.
3.3.2. Toxicokinetic Analyses
    Physiologically based mathematical models are potentially the most 
comprehensive way to account for toxicokinetic 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, 1992). 
This includes considerations of model validation and sensitivity 
analysis that stress the predictive performance of the model. When a 
delivered dose measure is used in animal to human extrapolation of dose 
response data, the assessment should discuss the confidence in the 
assumption that the toxicodynamics of the target tissue(s) will be the 
same in both species. Toxicokinetic data can improve dose response 
assessment by accounting for sources of change in proportionality of 
applied to internal or 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 overestimation or underestimation of 
low dose response otherwise resulting from extrapolation from a 
sublinear or supralinear part of the experimental dose response curve 
(Gillette, 1983). Toxicokinetic processes tend to become linear at low 
doses, an expectation that is more robust than low-dose linearity of 
response (Hattis, 1990). Accounting for toxicokinetic nonlinearities 
allows better description of the shape of the curve at relatively high 
levels of dose in the range of observation, but cannot determine 
linearity or nonlinearity of response at low dose levels (Lutz, 1990a; 
Swenberg et al., 1987).
    Toxicokinetic modeling results may be presented as the preferred 
method of estimating human equivalent dose or in parallel discussion 
with default assumptions depending on relative confidence in the 
modeling.
3.3.3. Route-to-Route Extrapolation
    Judgments frequently need to be made about the carcinogenicity of 
an agent through a route of exposure different than the one in the 
underlying studies. For example, exposures of interest may be through 
inhalation of an agent tested primarily through animal feeding studies 
or through ingestion of an agent that showed positive results in human 
occupational studies from inhalation exposure.
    Route-to-route extrapolation has both qualitative and quantitative 
aspects. For the qualitative aspect, the assessor weighs the degree to 
which positive results through one route of exposure in human or animal 
studies support a judgment that similar results would have been 
observed in appropriate studies using the route of exposure of 
interest. In general, confidence in making such a judgment is 
strengthened when the tumor effects are observed at a site distant from 
the portal of entry and when absorption through the route of exposure 
of interest is similar to absorption via the tested routes. In the 
absence of contrary data, the qualitative default assumption is that, 
if the agent is absorbed by a route to give an internal dose, it may be 
carcinogenic by that route. (See section 2.7.1.)
    When a qualitative extrapolation can be supported, quantitative 
extrapolation may still be problematic in the absence of adequate data. 
The differences in biological processes among routes of

[[Page 17998]]

exposure (oral, inhalation, dermal) can be great because of, 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 relies on a case-by-case analysis of 
available data. When good data on the agent itself are limited, an 
extrapolation analysis can be based on expectations from physical and 
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 model development and 
validation is awaiting the development of more extensive data (see 
generally, Gerrity and Henry, 1990). For screening or hazard ranking, 
route-to-route extrapolation may be based on assumed quantitative 
comparability as a default, as long as it is reasonable to assume 
absorption by compared routes. When route-to-route extrapolation is 
used, the assessor's degree of confidence in both the qualitative and 
quantitative extrapolation should be discussed in the assessment and 
highlighted in the dose response characterization.
3.3.4. Dose Averaging
    The cumulative dose received over a lifetime, expressed as lifetime 
average daily dose, is generally considered an appropriate default 
measure of exposure to a carcinogen (Monro, 1992). 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. 
While this is a reasonable default assumption based on theoretical 
considerations, departures from it are expected. Another approach is 
needed in some cases, such as when dose-rate effects are noted (e.g., 
formaldehyde). Cumulative dose may be replaced, as appropriate and 
justified by the data, with other dose measures. In such cases, 
modifications to the default assumption are made to take account of 
these effects; the rationale for the selected approach is explained.
    In cases where a mode of action or other feature of the biology has 
been identified that has special dose implications for sensitive 
subpopulations (e.g., differential effects by sex or disproportionate 
impacts of early-life exposure), these are explained and are recorded 
to guide exposure assessment and risk characterization. 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. 
These issues are explored and pointed out for attention in the exposure 
assessment and risk characterization.

3.4. Discussion of Uncertainties

    The exploration of significant uncertainties in data for dose and 
response and in extrapolation procedures is part of the assessment. The 
presentation distinguishes between model uncertainty and parameter 
uncertainty. Model uncertainty is an uncertainty about a basic 
biological question. For example, a default, linear dose response 
extrapolation may have been made based on tumor and other key evidence 
supporting the view that the model for an agent's mode of action is a 
DNA-reactive process. Discussion of the confidence in the extrapolation 
is appropriately done qualitatively or by showing results for 
alternatives that are equally plausible. It is not useful, for example, 
to conduct quantitative uncertainty analysis running multiple forms of 
linear models. This would obviate the function of the policy default.
    Parameter uncertainties deal with numbers representing statistical 
or analytical measures of variance or error in data or estimates. 
Uncertainties in parameters are described quantitatively, if 
practicable, through sensitivity analysis and statistical uncertainty 
analysis. With the recent expansion of readily available computing 
capacity, computer methods are being adapted to create simulated 
biological data that are comparable with observed information. These 
simulations can be used for sensitivity analysis, for example, to 
analyze how small, plausible variations in the observed data could 
affect dose response estimates. These simulations can also provide 
information about experimental uncertainty in dose response estimates, 
including a distribution of estimates that are compatible with the 
observed data. Because these simulations are based on the observed 
data, they cannot assist in evaluating the extent to which the observed 
data as a whole are idiosyncratic rather than typical of the true 
situation. If quantitative analysis is not possible, significant 
parameter uncertainties are described qualitatively. In either case, 
the discussion highlights uncertainties that are specific to the agent 
being assessed, as distinct from those that are generic to most 
assessments.
    Estimation of the applied dose in a human study has numerous 
uncertainties such as 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 (U.S. EPA, 1992a). In a retrospective study, 
exposure may be based on monitoring data but is often 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; U.S. EPA, 1986a). 
These uncertainties can also be assessed for any confounding factor for 
which a quantitative adjustment of dose response data is made (U.S. 
EPA, 1984).

3.5. Technical Dose Response Characterization

    As with hazard characterization, the dose response characterization 
serves the dual purposes of presenting a technical characterization of 
the assessment results and supporting the risk characterization.
    The characterization presents the results of analyses of dose data, 
of response data, and of dose response. When alternative approaches are 
plausible and persuasive in selecting dose data, response data, or 
extrapolation procedures, the characterization follows the alternative 
paths of analysis and presents the results. The discussion covers the 
question of whether any should be preferred over others because it (or 
they) better represents the available data or corresponds to the view 
of the mechanism of action developed in the hazard assessment. The 
results for different tumor types by sex and species are provided along 
with the one(s) preferred. Similarly, results for responses other than 
tumor incidence are shown if appropriate.
    Numerical dose response estimates are presented to one significant 
figure. Numbers are qualified as to whether they represent central 
tendency or upper bounds and whether the method used is inherently more 
likely to overestimate or underestimate (Krewski et al., 1984).
    In cases where a mode of action or other feature of the biology has 
been identified that has special implications

[[Page 17999]]

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. Whether the lifetime average daily dose or another 
measure of dose should be considered for differing exposure scenarios 
is discussed.
    Uncertainty analyses, qualitative or quantitative if possible, are 
highlighted in the characterization.
    The dose response characterization routinely includes the 
following, as appropriate for the data available:
     Identification of the kinds of data available for analysis 
of dose and response and for dose response assessment,
     Results of assessment as above,
     Explanation of analyses in terms of quality of data 
available,
     Selection of study/response and dose metric for 
assessment,
     Discussion of implications of variability in human 
susceptibility, including for susceptible subpopulation,
     Applicability of results to varying exposure scenarios--
issues of route of exposure, dose rate, frequency, and duration,
     Discussion of strengths and limitations (uncertainties) of 
the data and analyses that are quantitative as well as qualitative, and
     Special issues of interpretation of data, such as:

--Selecting dose data, response data, and dose response approach(es),
--Use of meta-analysis,
--Uncertainty and quantitative uncertainty analysis.

4. Technical Exposure Characterization

    Guidelines for exposure assessment of carcinogenic and other agents 
are published (U.S. EPA, 1992a) and are used in conjunction with these 
cancer risk assessment guidelines. Presentation of exposure descriptors 
is a subject of discussion in EPA risk characterization guidance (U.S. 
EPA, 1995). The exposure characterization is a technical 
characterization that presents the assessment results and supports risk 
characterization.
    The characterization provides a statement of purpose, scope, level 
of detail, and approach used in the assessment, identifying the 
exposure scenario(s) covered. It estimates the distribution of 
exposures among members of the exposed population as the data permit. 
It identifies and compares the contribution of different sources and 
routes and pathways of exposure. Estimates of the magnitude, duration, 
and frequency of exposure are included as available monitoring or 
modeling results or other reasonable methods permit. The strengths and 
limitations (uncertainties) of the data and methods of estimation are 
made clear.
    The exposure characterization routinely includes the following, as 
appropriate and possible for the data available:
     Identification of the kinds of data available,
     Results of assessment as above,
     Explanation of analyses in terms of quality of data 
available,
     Uncertainty analyses as discussed in Exposure Assessment 
Guidelines, distinguishing uncertainty from variability, and
     Explanation of derivation of estimators of ``high end'' or 
central tendency of exposure and their appropriate use.

5. Risk Characterization

5.1. Purpose

    The risk characterization process includes an integrative analysis 
followed by a presentation in a Risk Characterization Summary, of the 
major results of the risk assessment. The Risk Characterization Summary 
is a nontechnical discussion that minimizes the use of technical terms. 
It is an appraisal of the science that supports the risk manager in 
making public health decisions, as do other decisionmaking analyses of 
economic, social, or technology issues. It also serves the needs of 
other interested readers. The summary is an information resource for 
preparation of risk communication information, but being somewhat 
technical, is not itself the usual vehicle for communication with every 
audience.
    The integrative analysis brings together the assessments and 
characterizations of hazard, dose response, and exposure to make risk 
estimates for the exposure scenarios of interest. This analysis is 
generally much more extensive than the Risk Characterization Summary. 
It may be peer-reviewed or subject to public comment along with the 
summary in preparation for an Agency decision. The integrative analysis 
may be titled differently by different EPA programs (e.g., ``Staff 
Paper'' for criteria air pollutants), but it typically will identify 
exposure scenarios of interest in a decisionmaking and present risk 
analyses associated with them. Some of the analyses may concern 
scenarios in several media, others may examine, for example, only 
drinking water risks. It also may be the document that contains 
quantitative analyses of uncertainty.
    The values supported by a risk characterization throughout the 
process are transparency in environmental decisionmaking, clarity in 
communication, consistency in core assumptions and science policies 
from case to case, and reasonableness. While it is appropriate to err 
on the side of protection of health and the environment in the face of 
scientific uncertainty, common sense and reasonable application of 
assumptions and policies are essential to avoid unrealistic estimates 
of risk (U.S. EPA, 1995). Both integrative analyses and the Risk 
Characterization Summary present an integrated and balanced picture of 
the analysis of the hazard, dose response, and exposure. The risk 
analyst should provide summaries of the evidence and results and 
describe the quality of available data and the degree of confidence to 
be placed in the risk estimates. Important features include the 
constraints of available data and the state of knowledge, significant 
scientific issues, and significant science and science policy choices 
that were made when alternative interpretations of data existed (U.S. 
EPA, 1995). Choices made about using default assumptions or data in the 
assessment are explicitly discussed in the course of analysis, and if a 
choice is a significant issue, it is highlighted in the summary.

5.2. Application

    Risk characterization is a necessary part of generating any Agency 
report on risk, whether the report is preliminary to support allocation 
of resources toward further study or comprehensive to support 
regulatory decisions. In the former case, the detail and sophistication 
of the characterization are appropriately small in scale; in the latter 
case, appropriately extensive. Even if a document covers only parts of 
a risk assessment (hazard and dose response analyses for instance), the 
results of these are characterized.
    Risk assessment is an iterative process that grows in depth and 
scope in stages from screening for priority-making, to preliminary 
estimation, to fuller examination in support of complex regulatory 
decisionmaking. Default assumptions are used at every stage because no 
database is ever complete, but they are predominant at screening stages 
and are used less as more data are gathered and incorporated at later 
stages. Various provisions in EPA-administered statutes require 
decisions

[[Page 18000]]

based on findings that represent all stages of iteration. There are 
close to 30 provisions within the major statutes that require decisions 
based on risk, hazard, or exposure assessment. For example, Agency 
review of premanufacture notices under section 5 of the Toxic 
Substances Control Act relies on screening analyses, while requirements 
for industry testing under section 4 of that Act rely on preliminary 
analyses of risk or simply of exposure. At the other extreme, air 
quality criteria under the Clean Air Act rest on a rich data collection 
required by statute to undergo reassessment every few years. There are 
provisions that require ranking of hazards of numerous pollutants--by 
its nature a screening level of analysis--and other provisions that 
require a full assessment of risk. Given this range in the scope and 
depth of analyses, not all risk characterizations can or should be 
equal in coverage or depth. The risk assessor must carefully decide 
which issues in a particular assessment are important to present, 
choosing those that are noteworthy in their impact on results. For 
example, health effect assessments typically rely on animal data since 
human data are rarely available. The objective of characterization of 
the use of animal data is not to recount generic issues about 
interpreting and using animal data. Agency guidance documents cover 
these. Instead, the objective is to call out any significant issues 
that arose within the particular assessment being characterized and 
inform the reader about significant uncertainties that affect 
conclusions.

5.3. Presentation of Risk Characterization Summary

    The presentation is a nontechnical discussion of important 
conclusions, issues, and uncertainties that uses the hazard, dose-
response, exposure, and integrative analyses for technical support. The 
primary technical supports within the risk assessment are the hazard 
characterization, dose response characterization, and exposure 
characterization described in this guideline. The risk characterization 
is derived from these. The presentation should fulfill the aims 
outlined in the purpose section above.

5.4. Content of Risk Characterization Summary

    Specific guidance on hazard, dose response, and exposure 
characterization appears in previous sections. Overall, the risk 
characterization routinely includes the following, capturing the 
important items covered in hazard, dose response, and exposure 
characterization.
     Primary conclusions about hazard, dose response, and 
exposure, including equally plausible alternatives,
     Nature of key supporting information and analytic methods,
     Risk estimates and their attendant uncertainties, 
including key uses of default assumptions when data are missing or 
uncertain,
     Statement of the extent of extrapolation of risk estimates 
from observed data to exposure levels of interest (i.e., margin of 
exposure) and its implications for certainty or uncertainty in 
qualtifying risk,
     Significant strengths and limitations of the data and 
analyses, including any major peer reviewers' issues,
     Appropriate comparison with similar EPA risk analyses or 
common risks with which people may be familiar, and
     Comparison with assessment of the same problem by another 
organization.

Appendix A

    This appendix contains several general illustrations of weight 
of evidence narratives. In addition, after narrative #5 is an 
example of a briefing summary format.

NARRATIVE #1 Chlorinated Alkene

CAS# XXX

CANCER HAZARD SUMMARY

    Chlorinated alkene (cl-alkene) is likely to be carcinogenic to 
humans by all routes of exposure. The weight of evidence of human 
carcinogenicity of cl-alkene is based on (a) findings of 
carcinogenicity in rats and mice of both sexes by oral and 
inhalation exposures; (b) its similarity in structure to other 
chlorinated organics that are known to cause liver and kidney 
damage, and liver and kidney tumors in rats and mice; (c) suggestive 
evidence of a possible association between cl-alkene exposure of 
workers in the laundry and dry cleaning industries and increased 
cancer risk in a number of organ systems; and (d) human and animal 
data indicating that cl-alkene is absorbed by all routes of 
exposure.
    In comparison with other agents designated as likely 
carcinogens, the overall weight of evidence for cl-alkene places it 
at the low end of the grouping. This is because one cannot attribute 
observed excess cancer risk in exposed workers solely to cl-alkene. 
Moreover, there is considerable scientific uncertainty about the 
human significance and relevance of certain rodent tumors associated 
with exposure to cl-alkene and other chlorinated organics, but 
insufficient evidence about mode of action for the animal tumors. 
Hence, the human relevance of the animal evidence of carcinogenicity 
relies on a default assumption of relevance.
    There is no clear evidence about the mode of action for each 
tumor type induced in rats and mice. Available evidence suggests 
that cl-alkene induces cancer mainly by promoting cell growth rather 
than via direct mutagenic action, although a mutagenic mode of 
action for rat kidney tumors cannot be ruled out. The dose response 
assessment should, therefore, adopt both default approaches, 
nonlinear and linear. It is recognized that the latter approach 
likely overestimates risk at low doses if the mode of action is 
primarily growth-promoting. This approach, however, may be useful 
for screening analyses.

SUPPORTING INFORMATION

Human Data

    A number of epidemiologic studies of dry cleaning and laundry 
workers that have reported elevated incidences of lung, cervix, 
esophagus, kidney, blood and lymphoid cancers. Many of these studies 
are confounded by co-exposure to other petroleum solvents, making 
them limited for determining whether the observed increased cancer 
risks are causally related to cl-alkene. The only investigation of 
dry cleaning workers with no known exposure to other chemicals did 
not evaluate other confounding factors such as smoking, alcohol 
consumption, and low socioeconomic status to exclude the possible 
contribution of these factors to cancer risks.

Animal Data

    The carcinogenic potential of cl-alkene has been adequately 
investigated in two chronic studies in two rodent species, the first 
study by gavage and the second study by inhalation. Cl-alkene is 
carcinogenic in the liver in both sexes of mice when tested by 
either route of exposure. It causes marginally increased incidences 
of mononuclear cell leukemia (MCL) in both sexes of rats and low 
incidences of a rare kidney tumor in male rats by inhalation. No 
increases in tumor incidence were found in rats treated with cl-
alkene by gavage. This rat study was considered limited because of 
high mortality of the animals.
    Although cl-alkene causes increased incidences of tumors at 
multiple sites in two rodent species, controversy surrounds each of 
the tumor endpoints concerning their relevance and/or significance 
to humans (see discussion under Mode of Action).

Other Key Data

    Cl-alkene is a member of a class of chlorinated organics that 
often cause liver and kidney toxicity and carcinogenesis in rodents. 
Like many chlorinated hydrocarbons, cl-alkene itself is tested 
negative in a battery of standard genotoxicity tests using bacterial 
and mammalian cells systems including human lymphocytes and 
fibroblast cells. There is evidence, however, that a minor 
metabolite generated by an enzyme found in rat kidney tissue is 
mutagenic. This kidney metabolite has been hypothesized to be 
related to the development of kidney tumors in the male rat. This 
metabolic pathway appears to be operative in the human kidney.
    Human data indicate that cl-alkene is readily absorbed via 
inhalation but to a much lesser extent by skin contact. Animal data 
show that cl-alkene is absorbed well by the oral route.

[[Page 18001]]

MODE OF ACTION

    The mechanisms of cl-alkene-induced mouse liver tumors are not 
completely understood. One mechanism has been hypothesized to be 
mediated by a genotoxic epoxide metabolite generated by enzymes 
found in the mouse liver, but there is a lack of direct evidence in 
support of this mechanism. A more plausible mechanism that still 
needs to be further defined is related to liver peroxisomal 
proliferation and toxicity by TCA (trichloroacetic acid), a major 
metabolite of cl-alkene. However, there are no definitive data 
indicating that TCA induces peroxisomal proliferation in humans.
    The mechanisms by which cl-alkene induces kidney tumors in male 
rats are even less well understood. The rat kidney response may be 
related to either kidney toxicity or the activity of a mutagenic 
metabolite of the parent compound.
    The human relevance of cl-alkene-induced MCL in rats is unclear. 
The biological significance of marginally increased incidences of 
MCL has been questioned by some, since this tumor occurs 
spontaneously in the tested rat strain at very high background 
rates. On the other hand, it has been considered by others to be a 
true finding because there was a decreased time to onset of the 
disease and the disease was more severe in treated as compared with 
untreated control animals. The exact mechanism by which cl-alkene 
increases incidences of MCL in rats is not known.
    Overall, there is not enough evidence to give high confidence in 
a conclusion about any single mode of action; it would appear that 
more than a single mode operates in different rodent tissues. The 
apparent lack of mutagenicity of cl-alkene itself and its general 
growth-promoting effect on high background tumors as well as its 
toxicity toward mouse liver and rat kidney tissue support the view 
that its predominant mode of action is cell growth promoting rather 
than mutagenic. A mutagenic contribution to the renal 
carcinogenicity due to a metabolite cannot be entirely ruled out.

NARRATIVE #2

Unsaturated Aldehyde

CAS# XXX

CANCER HAZARD SUMMARY

    The potential human hazard of unsaturated aldehyde (UA) cannot 
be determined, but there are suggestive data for carcinogenicity.
    The evidence on carcinogenicity consists of (a) data from an 
oral animal study showing a response only at the highest dose in 
female rats, with no response in males and (b) the fact that other 
low-molecular-weight aldehydes have shown tumorigenicity in the 
respiratory tract after inhalation. The one study of UA effects by 
the inhalation route was not adequately performed. The available 
evidence is too limited to describe human carcinogenicity potential 
or support dose response assessment.

SUPPORTING INFORMATION

Human Data

    An elevated incidence of cancer was reported in a cohort of 
workers in a chemical plant who were exposed to a mixture of 
chemicals including UA as a minor component. The study is considered 
inadequate because of the small size of the cohort studied and the 
lack of adequate exposure data.

Animal Data

    In a long-term drinking water study in rats, an increased 
incidence of adrenal cortical adenomas was found in the highest-
dosed females. No other significant finding was made. The oral rat 
study was well conducted by a standard protocol. In a 1-year study 
in hamsters at one inhalation dose, no tumors were seen. This study 
was inadequate due to high mortality and consequent short duration. 
The chemical is very irritating and is a respiratory toxicant in 
mammals. The animal data are too limited for conclusions to be 
drawn.

Structural Analogue Data

    UA's structural analogues, formaldehyde and acetaldehyde, both 
have carcinogenic effects on the rat respiratory tract.

Other Key Data

    The weight of results of mutagenicity tests in bacteria, fungi, 
fruit flies, and mice result in an overall conclusion of not 
mutagenic; UA is lethal to bacteria to a degree that makes testing 
difficult and test results difficult to interpret. The chemical is 
readily absorbed by all routes.

MODE OF ACTION

    Data are not sufficient to judge whether there is a carcinogenic 
mode of action.

NARRATIVE #3

Alkene Oxide

CAS# XXX

CANCER HAZARD SUMMARY

    Alkene oxide (AO) should be dealt with as if it were a known 
human carcinogen by all routes of exposure. Several studies in 
workers, when considered together, suggest an elevated risk of 
leukemia and lymphoma after long-term exposure to AO, even though no 
single study conclusively demonstrates that AO caused the cancer. In 
addition, animal cancer and mutagenicity studies as well as short-
term tests of mutagenicity have strongly consistent results that 
support a level of concern equal to having conclusive human studies.
    The weight of evidence of human carcinogenicity is based on (a) 
consistent evidence of carcinogenicity of AO in rats and mice by 
both oral and inhalation exposure; (b) studies in workers that taken 
together suggest elevated risk of leukemia and lymphoma to workers 
exposed to AO and show genetic damage in blood lymphocytes in 
exposed workers; (c) mutagenic effects in numerous test systems and 
heritable gene mutations in animals; and (d) membership in a class 
of DNA-reactive compounds that are regularly observed to cause 
cancer in animals.
    Due to its ready absorption by all routes of exposure and rapid 
distribution throughout the body, AO is expected to pose a risk by 
any route of exposure. The strong evidence of a mutagenic mode of 
action supports dose response assessment that assumes linearity of 
the relationship.

SUPPORTING INFORMATION

Human Data

    Elevated risks of lymphatic cancer and cancer of blood-forming 
tissue have been reported in exposed workers in several studies. The 
interpretation of the studies separately is complicated by exposures 
to other agents in each so there is no single study that 
demonstrates that AO caused the effects; nevertheless, several of 
the studies together are considered suggestive of AO carcinogenicity 
because they consistently show cancer elevation in the same tissues. 
Biomonitoring studies of exposed workers find DNA damage in blood 
lymphocytes and the degree of DNA damage correlates with the level 
and duration of AO exposure. Finding this damage in the same tissue 
in which elevated cancer was seen in workers adds further weight to 
the positive suggestion from the worker cancer studies. The human 
data are from well-conducted studies.

Animal Data

    AO causes cancer in multiple tissue sites in rats and mice of 
both sexes by oral and inhalation exposure. The database is more 
extensive than usual and the studies are good. The observation of 
multisite, multispecies carcinogenic activity by an agent is 
considered to be very strong evidence and is often the case with 
highly mutagenic agents. There are also good studies showing that AO 
causes heritable germ cell mutations in mice after inhalation 
exposure--a property that is very highly correlated with 
carcinogenicity.

Structural Analogue Data

    Organic epoxides are commonly found to have carcinogenic effects 
in animals, particularly the low-molecular-weight ones.

Other Key Data

    The structure and DNA reactivity of AO support potential 
carcinogenicity. Both properties are highly correlated with 
carcinogenicity. Positive mutagenicity tests in vitro and in vivo 
add to this support and are reinforced by observation of similar 
genetic damage in exposed workers.
    AO is experimentally observed to be readily absorbed by all 
routes and rapidly distributed through the body.

MODE OF ACTION

    All of the available data are strongly supportive of a mutagenic 
mode of action, with a particular human target in lymphatic and 
blood-forming tissue. The current scientific consensus is that there 
is virtually complete correspondence between ability of an agent to 
cause heritable germ cell mutations, as AO does, and 
carcinogenicity. All of this points to a mutagenic mode of action 
and supports assuming linearity of the dose response relationship.

NARRATIVE #4

Bis-benzenamine

CAS# XXX

CANCER HAZARD SUMMARY

    This chemical is likely to be carcinogenic to humans by all 
routes of exposure. Its

[[Page 18002]]

carcinogenic potential is indicated by (a) tumor and toxicity 
studies on structural analogues, which demonstrate the ability of 
the chemical to produce thyroid follicular cell tumors in rats and 
hepatocellular tumors in mice following ingestion and (b) metabolism 
and hormonal information on the chemical and its analogues, which 
contributes to a working mode of action and associates findings in 
animals with those in exposed humans. In comparison with other 
agents designated as likely carcinogens, the overall weight of 
evidence for this chemical places it at the lower end of the 
grouping. This is because there is a lack of tumor response data on 
this agent itself.
    Biological information on the compound is contradictory in terms 
of how to quantitate potential cancer risks. The information on 
disruption on thyroid-pituitary status argues for using a margin of 
exposure evaluation. However, the chemical is an aromatic amine, a 
class of agents that are DNA-reactive and induce gene mutation and 
chromosome aberrations, which argues for low-dose linearity. 
Additionally, there is a lack of mode of action information on the 
mouse liver tumors produced by the structural analogues, also 
pointing toward a low-dose linear default approach. In recognition 
of these uncertainties, it is recommended to quantitate tumors using 
both nonlinear (to place a lower bound on the risks) and linear (to 
place an upper bound on the risks) default approaches. Given the 
absence of tumor response data on the chemical per se, it is 
recommended that tumor data on close analogues be used to possibly 
develop toxicity equivalent factors or relative potencies.
    Overall, this chemical is an inferential case for potential 
human carcinogenicity. The uncertainties associated with this 
assessment include (1) the lack of carcinogenicity studies on the 
chemical, (2) the use of tumor data on structural analogues, (3) the 
lack of definitive information on the relevance of thyroid-pituitary 
imbalance for human carcinogenicity, and (4) the different potential 
mechanisms that may influence tumor development and potential risks.

SUPPORTING INFORMATION

Human Data

    Worker exposure has not been well characterized or quantified, 
but recent medical monitoring of workers exposed over a period of 
several years has uncovered alterations in thyroid-pituitary 
hormones (a decrease in T3 and T4 and an increase in TSH) and 
symptoms of hypothyroidism. A urinary metabolite of the chemical has 
been monitored in workers, with changes in thyroid and pituitary 
hormones noted, and the changes were similar to those seen in an 
animal study.

Animal Data

    The concentration of the urinary metabolite in rats receiving 
the chemical for 28 days was within twofold of that in exposed 
workers, a finding associated with comparable changes in thyroid 
hormones and TSH levels. In addition, the dose of the chemical given 
to rats in this study was essentially the same as that of an 
analogue that had produced thyroid and pituitary tumors in rats. The 
human thyroid responds in the same way as the rodent thyroid 
following short-term, limited exposure. Although it is not well 
established that thyroid-pituitary imbalance leads to cancer in 
humans as it does in rodents, information in animals and in exposed 
humans suggests similar mechanisms of disrupting thyroid-pituitary 
function and the potential role of altered TSH levels in leading to 
thyroid carcinogenesis.

Structural Analogue Data

    This chemical is an aromatic amine, a member of a class of 
chemicals that has regularly produced carcinogenic effects in 
rodents and gene and structural chromosome aberrations in short-term 
tests. Some aromatic amines have produced cancer in humans.
    Close structural analogues produce thyroid follicular cell 
tumors in rats and hepatocellular tumors in mice following 
ingestion. The thyroid tumors are associated with known 
perturbations in thyroid-pituitary functioning. These compounds 
inhibit the use of iodide by the thyroid gland, apparently due to 
inhibition of the enzyme that synthesizes the thyroid hormones (T3, 
T4). Accordingly, blood levels of thyroid hormones decrease, which 
induce the pituitary gland to produce more TSH, a hormone that 
stimulates the thyroid to produce more of its hormones. The thyroid 
gland becomes larger due to increases in the size of individual 
cells and their proliferation and upon chronic administration, 
tumors develop. Thus, thyroid tumor development is significantly 
influenced by disruption in the thyroid-pituitary axis.

Other Key Data

    The chemical can be absorbed by the oral, inhalation, and dermal 
routes of exposure.

MODE OF ACTION

    Data on the chemical and on structural analogues indicate the 
potential association of carcinogenesis with perturbation of 
thyroid-pituitary homeostasis. Structural analogues are genotoxic, 
thus raising the possibility of different mechanisms by which this 
chemical may influence tumor development.

NARRATIVE #5

Brominated Alkane (BA)

CAS# XXX

CANCER HAZARD SUMMARY

    Brominated alkane (BA) is likely to be a human carcinogen by all 
routes of exposure. The weight of evidence for human carcinogenicity 
is at the high end of agents in the ``likely'' group. Findings are 
based on very extensive and significant experimental findings that 
include (a) tumors at multiple sites in both sexes of two rodent 
species via three routes of administration relevant to human 
exposure, (b) close structural analogues that produce a spectrum of 
tumors like BA, (c) significant evidence for the production of 
reactive BA metabolites that readily bind to DNA and produce gene 
mutations in many systems including cultured mammalian and human 
cells, and (d) two null and one positive epidemiologic study; in the 
positive study, there may have been exposure to BA. These findings 
support a decision that BA might produce cancer in exposed humans. 
In comparison to other agents considered likely human carcinogens, 
the overall weight of evidence for BA puts it near the top of the 
grouping. Given the agent's mutagenicity, which can influence the 
carcinogenic process, a linear dose-response extrapolation is 
recommended.
    Uncertainties include the lack of adequate information on the 
mutagenicity of BA in mammals or humans in vivo, although such 
effects would be expected.

SUPPORTING INFORMATION

Human Data

    The information on the carcinogenicity of BA from human studies 
is inadequate. Two studies of production workers have not shown 
significant increases in cancer from exposure to BA and other 
chemicals. An increase in lymphatic cancer was reported in a 
mortality study of grain elevator workers who may have been exposed 
to BA (and other chemicals).

Animal Data

    BA produced tumors in four chronic rodents studies. Tumor 
increases were noted in males and females of rats and mice following 
oral dermal and inhalation exposure (rat--oral and two inhalation, 
mouse--oral and dermal). It produces tumors both at the site of 
application (e.g., skin with dermal exposure) and at sites distal to 
the portal of entry into the body (e.g., mammary gland) following 
exposure from each route. Tumors at the same site were noted in both 
sexes of a species (blood vessel), both species (forestomach) and 
via different routes of administration (lung). Some tumors developed 
after very short latency, metastasized extensively, and produced 
death, an uncommon findings in rodents. The rodent studies were well 
designed and conducted except for the oral studies, in which the 
doses employed caused excessive toxicity and mortality. However, 
given the other rodent findings, lower doses would also be 
anticipated to be carcinogenic.

Structural Analogue Data

    Several chemicals structurally related to BA are also 
carcinogenic in rodents. Among four that are closest in structure, 
tumors like those seen for BA were often noted (e.g., forestomach, 
mammary, lung), which helps to confirm the findings for BA itself. 
In sum, all of the tumor findings help to establish animal 
carcinogenicity and support potential human carcinogenicity for BA.

Other Key Data

    BA itself is not reactive, but from its structure it was 
expected to be metabolized to reactive forms. Extensive metabolism 
studies have confined this presumption and have demonstrated 
metabolites that bind to DNA and cause breaks in the DNA chain. 
These lesions are readily converted to gene mutations in bacteria, 
fungi, higher plants, insects and mammalian and human cells in 
culture. There are only a limited number of reports on the induction 
of chromosome aberrations in mammals and humans; thus far they are 
negative.

[[Page 18003]]

MODE OF ACTION

    Human carcinogens often produce cancer in multiple sites of 
multiple animal species and both sexes and are mutagenic in multiple 
test systems. BA satisfies these findings. It produces cancer in 
males and females of rats and mice. It produces gene mutations in 
cells across all life forms--plants, bacteria and animals--including 
mammals and humans. Given the mutagenicity of BA exposure and the 
multiplicity and short latency of BA tumor induction, it is 
reasonable to use a linear approach for cancer dose-response 
extrapolation.

BRIEFING SUMMARY

----------------------------------------------------------------------------------------------------------------
                                                                     Designation or                             
               Route(s)                         Class                  rationale              Dose response     
----------------------------------------------------------------------------------------------------------------
All..................................  Likely.................  High end...............  Default-linear.        
----------------------------------------------------------------------------------------------------------------

Basis for classification/dose response

    1. Human data: Two studies of production workers show no 
increase in cancer (one had a small sample size; the other had mixed 
chemical exposures). An increase in lymphatic cancer is seen among 
grain elevator workers who may have been exposed to other chemicals.
    2. Animal data: BA produces tumors at multiple sites in male and 
female rats and mice following oral, dermal, and inhalation 
exposure. Tumors are seen at the site of administration and distally 
and are often consistent across sex, species, and route of 
administration; some develop early, metastasize, and cause death.
    3. Structural analogue data: Close analogues produce some of the 
same tumors as are seen with BA.
    4. Other key data: BA is metabolized to a reactive chemical that 
binds DNA and produces gene mutations in essentially every test 
system including cultured human cells.
    5. Mode of action: Like most known human carcinogens, BA is 
mutagenic in most test systems.
    6. Hazard classification/uncertainties: There is a rich database 
on BA demonstrating its potential ability to cause tumors in humans, 
including (a) multiple animal tumors, (b) by appropriate routes of 
exposure, (c) a mode of action relevant to human carcinogenicity, 
and (d) some information in humans. Together they lead to a 
designation near the high end of the likely human carcinogen class.
    7. Dose response: Given the anticipated mode of action, a linear 
default dose response relationship should be assumed.

Appendix B

    This appendix contains responses to the National Academy of 
Sciences National Research Council report Science and Judgment in Risk 
Assessment (NRC, 1994).

Recommendations of the National Academy of Sciences National 
Research Council

    In 1994, the National Academy of Sciences published a report 
Science and Judgment in Risk Assessment. The 1994 report was written 
by a Committee on Risk Assessment of Hazardous Air Pollutants formed 
under the Academy's Board on Environmental Studies and Toxicology, 
Commission on Life Sciences, National Research Council. The report 
was called for under Section 112(o)(1)(A,B) of the Clean Air Act 
Amendments of 1990, which provided for the EPA to arrange for the 
Academy to review:
     risk assessment methodology used by the EPA to 
determine the carcinogenic risk associated with exposure to 
hazardous air pollutants from source categories and subcategories 
subject to the requirements of this section and
     improvements in such methodology.
    Under Section 112(o)(2)(A,B), the Academy was to consider the 
following in its review:
     the techniques used for estimating and describing the 
carcinogenic potency to humans of hazardous air pollutants and
     the techniques used for estimating exposure to 
hazardous air pollutants (for hypothetical and actual maximally 
exposed individuals as well as other exposed individuals).
    To the extent practicable, the Academy was also to review 
methods of assessing adverse human health effects other than cancer 
for which safe thresholds of exposure may not exist [Section 
112(o)(3)]. The Congress further provided that the EPA Administrator 
should consider, but need not adopt, the recommendations in the 
report and the views of the EPA Science Advisory Board with respect 
to the report. Prior to the promulgation of any standards under 
Section 112(f), the Administrator is to publish revised guidelines 
for carcinogenic risk assessment or a detailed explanation of the 
reasons that any recommendations contained in the report will not be 
implemented [Section 112(o)(6)].
    The following discussion addresses the recommendations of the 
1994 report that are pertinent to the EPA cancer risk assessment 
guidelines. Guidelines for assessment of exposure, of mixtures, and 
of other health effects are separate EPA publications. Many of the 
recommendations were related to practices specific to the exposure 
assessment of hazardous air pollutants, which are not covered in 
cancer assessment guidelines. Recommendations about these other 
guidelines or practices are not addressed here.

Hazard Classification

    The 1994 report contains the following recommendation about 
classifying cancer hazard:
     The EPA should develop a two-part scheme for 
classifying evidence on carcinogenicity that would incorporate both 
a simple classification and a narrative evaluation. At a minimum, 
both parts should include the strength (quality) of the evidence, 
the relevance of the animal model and results to humans, and the 
relevance of the experimental exposures (route, dose, timing, and 
duration) to those likely to be encountered by humans.
    The report also presented a possible matrix of 24 boxes that 
would array weights of evidence against low, medium, or high 
relevance, resulting in 24 codes for expressing the weight and 
relevance.
    These guidelines adopt a set of descriptors and subdescriptors 
of weight of evidence in three categories: ``known/likely,'' 
``cannot be determined,'' and ``not likely,'' and a narrative for 
presentation of the weight of evidence findings. The descriptors are 
used within the narrative. There is no matrix of alphanumerical 
weight of evidence boxes.
    The issue of an animal model that is not relevant to humans has 
been dealt with by not including an irrelevant response in the 
weighing of evidence, rather than by creating a weight of evidence 
then appending a discounting factor as the NRC scheme would do. The 
issue is more complex than the NRC matrix makes apparent. Often the 
question of relevance of the animal model applies to a single tumor 
response, but one encounters situations in which there are more 
tumor responses in animals than the questioned one. Dealing with 
this complexity is more straightforward if it is done during the 
weighing of evidence rather than after as in the NRC scheme. 
Moreover, the same experimental data are involved in deciding on the 
weight of evidence and the relevance of a response. It would be 
awkward to go over the same data twice.
    In recommending that the relevance of circumstances of human 
exposure also be taken into account, the NRC appears to assume that 
all of the actual conditions of human exposure will be known when 
the classification is done. This is not the case. More often than 
not, the hazard assessment is applied to assessment of risks 
associated with exposure to different media or environments at 
different times. In some cases, there is no priority to obtaining 
exposure data until the hazard assessment has been done. The 
approach of these guidelines is to characterize hazards as to 
whether their expression is intrinsically limited by route of 
exposure or by reaching a particular dose range based strictly on 
toxicological and other biological features of the agent. Both the 
use of descriptors and the narrative specifically capture this 
information. Other aspects of appropriate application of the hazard 
and dose response assessment to particular human exposure scenarios 
are dealt with in the characterization of the dose response 
assessment, e.g., the applicability of the dose response assessment 
to scenarios with differing frequencies and durations.
    The NRC scheme apparently intended that the evidence would be 
weighed, then given a low, medium, or high code for some combination 
of relevance of the animal response, route of exposure, timing, 
duration, or frequency. The 24 codes contain none of this specific 
information, and in fact, do not communicate what the conclusion is 
about. To make the codes communicate the information apparently 
intended would require some multiple of the 24 in the NRC scheme. As 
the number of codes increases, their utility for communication 
decreases.
    Another reason for declining to use codes is that they tend to 
become outdated as research reveals new information that was not 
contemplated when they were adopted. This has been the case with the 
classification system under the EPA, 1986 guidelines.
    Even though these guidelines do not adopt a matrix of codes, the 
method they provide

[[Page 18004]]

of using descriptors and narratives captures the information the NRC 
recommended as the most important, and in the EPA's view, in a more 
transparent manner.

Dose Response

     The 1994 report contains the following recommendations 
about dose response issues:
      EPA should continue to explore, and when 
scientifically appropriate, incorporate toxicokinetic models of the 
link between exposure and biologically effective dose (i.e., dose 
reaching the target tissue).
     Despite the advantages of developing consistent risk 
assessments between agencies by using common assumptions (e.g., 
replacing surface area with body weight to the 0.75 power), EPA 
should indicate other methods, if any, that would be more accurate.
     EPA should continue to use the linearized multistage 
model as a default option but should develop criteria for 
determining when information is sufficient to use an alternative 
extrapolation model.
     EPA should continue to use as one of its risk 
characterization metrics upper-bound potency estimates of the 
probability of developing cancer due to lifetime exposure. Whenever 
possible, this metric should be supplemented with other descriptions 
of cancer potency that might more adequately reflect the uncertainty 
associated with the estimates.
     EPA should adopt a default assumption for differences 
in susceptibility among humans in estimating individual risks.
     In the analysis of animal bioassay data on the 
occurrence of multiple tumor types, the cancer potencies should be 
estimated for each relevant tumor type that is related to exposure 
and the individual potencies should be summed for those tumors.
    The use of toxicokinetic models is encouraged in these 
guidelines with discussion of appropriate considerations for their 
use. When there are questions as to whether such a model is more 
accurate in a particular case than the default method for estimating 
the human equivalent dose, both alternatives may be used. It should 
be noted that the default method for inhalation exposure is a 
toxicokinetic model.
    The rationale for adopting the oral scaling factor of body 
weight to the 0.75 power has been discussed above in the explanation 
of major defaults. The empirical basis is further explored in U.S. 
EPA, 1992b. The more accurate approach is to use a toxicokinetic 
model when data become available or to modify the default when data 
are available as encouraged under these guidelines. As the U.S. EPA, 
1992b discussion explores in depth, data on the differences among 
animals in response to toxic agents are basically consistent with 
using a power of 1.0, 0.75, or 0.66. The Federal agencies chose the 
power of 0.75 for the scientific reasons given in the previous 
discussion of major defaults; these were not addressed specifically 
in the NRC report. It was also considered appropriate, as a matter 
of policy, for the agencies to agree on one factor. Again, the 
default for inhalation exposure is a model that is constructed to 
become better as more agent-specific data become available.
    The EPA proposes not to use a computer model such as the 
linearized multistage model as a default for extrapolation below the 
observed range. The reason is that the basis for default 
extrapolation is a theoretical projection of the likely shape of the 
curve considering mode of action. For this purpose, a computer model 
looks more sophisticated than a straight line extrapolation, but is 
not. The extrapolation will be by straight line as explained in the 
explanation of major defaults. This was also recommended by workshop 
reviewers of a previous draft of these guidelines (U.S. EPA, 1994b). 
In addition, a margin of exposure analysis is proposed to be used in 
cases in which the curve is thought to be nonlinear, based on mode 
of action. In both cases, the observed range of data will be modeled 
by curve fitting in the absence of supporting data for a 
biologically based or case-specific model.
    The result of using straight line extrapolation is thought to be 
an upper bound on low-dose potency to the human population in most 
cases, but as discussed in the major defaults section, it may not 
always be. Exploration and discussion of uncertainty of parameters 
in curve-fitting a model of the observed data or in using a 
biologically based or case-specific model is called for in the dose 
response assessment and characterization sections of these 
guidelines.
    The issue of a default assumption for human differences in 
susceptibility has been addressed under the major defaults 
discussion in section 1.3 with respect to margin of exposure 
analysis. The EPA has considered but decided not to adopt a 
quantitative default factor for human differences in susceptibility 
when a linear extrapolation is used. In general, the EPA believes 
that the linear extrapolation is sufficiently conservative to 
protect public health. Linear approaches (both LMS and straight line 
extrapolation) from animal data are consistent with linear 
extrapolation on the same agents from human data (Goodman and 
Wilson, 1991; Hoel and Portier, 1994). If actual data on human 
variability in sensitivity are available they will, of course, be 
used.
    In analyzing animal bioassay data on the occurrence of multiple 
tumor types, these guidelines outline a number of biological and 
other factors to consider. The objective is to use these factors to 
select response data (including nontumor data as appropriate) that 
best represent the biology observed. As stated in section 3 of the 
guidelines, appropriate options include use of a single data set, 
combining data from different experiments, showing a range of 
results from more than one data set, showing results from analysis 
of more than one tumor response based on differing modes of action, 
representing total response in a single experiment by combining 
animals with tumors, or a combination of these options. The approach 
judged to best represent the data is presented with the rationale 
for the judgment, including the biological and statistical 
considerations involved. The EPA has considered the approach of 
summing tumor incidences and decided not to adopt it. While multiple 
tumors may be independent, in the sense of not arising from 
metastases of a single malignancy, it is not clear that they can be 
assumed to represent different effects of the agent on cancer 
processes. In this connection, it is not clear that summing 
incidences provides a better representation of the underlying 
mode(s) of action of the agent than combining animals with tumors or 
using another of the several options noted above. Summing incidences 
would result in a higher risk estimate, a step that appears 
unnecessary without more reason.

Risk Characterization

     When EPA reports estimates of risk to decisionmakers 
and the public, it should present not only point estimates of risk, 
but also the sources and magnitudes of uncertainty associated with 
these estimates.
     Risk managers should be given characterizations of risk 
that are both qualitative and quantitative, i.e., both descriptive 
and mathematical.
     EPA should consider in its risk assessments the limits 
of scientific knowledge, the remaining uncertainties, and the desire 
to identify errors of either overestimation or underestimation.
    In part as a response to these recommendations, the 
Administrator of EPA issued guidelines for risk characterization and 
required implementation plans from all programs in EPA (U.S. EPA, 
1995). The Administrator's guidance is followed in these cancer 
guidelines. The assessments of hazard, dose response, and exposure 
will all have accompanying technical characterizations covering 
issues of strengths and limitations of data and current scientific 
understanding, identification of defaults utilized in the face of 
gaps in the former, discussions of controversial issues, and 
discussions of uncertainties in both their qualitative, and as 
practicable, their quantitative aspects.

Appendix C

Overview of Cancer Processes

    The following picture is changing as research reveals more about 
carcinogenic processes. Nevertheless, it is apparent that several 
general modes of action are being elucidated from direct reaction 
with DNA to hormonal or other growth-signaling processes. While the 
exact mechanism of action of an agent at the molecular level may not 
be clear from existing data, the available data will often provide 
support for deducing the general mode of action. Under these 
guidelines, using all of the available data to arrive at a view of 
the mode of action supports both characterization of human hazard 
potential and assessment of dose response relationships.
    Cancers are diseases of somatic mutation affecting cell growth 
and differentiation. The genes that control cell growth, programmed 
cell death, and cell differentiation are critical to normal 
development of tissues from embryo to adult metazoan organisms. 
These genes continue to be critical to maintenance of form and 
function of tissues in the adult (e.g., Meyn, 1993) and changes in 
them are essential elements of carcinogenesis (Hsu et al., 1991; 
Kakizuka et al., 1991; Bottaro et al., 1991; Sidransky et al., 1991; 
Salomon et al.,

[[Page 18005]]

1990; Srivastava et al., 1990). The genes involved are among the 
most highly conserved in evolution as evidenced by the great 
homology of many of them in DNA sequence and function in organisms 
as phylogenetically distant as worms, insects, and mammals (Auger et 
al., 1989a, b; Hollstein et al., 1991; Herschman, 1991; Strausfeld 
et al., 1991; Forsburg and Nurse, 1991).
    Mutations affecting three general categories of genes have been 
implicated in carcinogenesis. Over 100 oncogenes have been found in 
human and animal tumors that act as dominant alleles, whereas there 
are about 10 known tumor suppressor genes that are recessive in 
action. The normal alleles of these genes are involved with control 
of cell division and differentiation; mutated alleles lead to a 
disruption in these functions. The third class are mutator genes 
that predispose the genome to enhanced mutagenic events that 
contribute further to the carcinogenic process.
    Adult tissues, even those that are composed of rapidly 
replicating cells, maintain a constant size and cell number (Nunez 
et al., 1991) by balancing three cell fates: (1) continued 
replication, (2) differentiation to take on specialized functions, 
or (3) programmed cell death (apoptosis) (Raff, 1992; Maller, 1991; 
Naeve et al., 1991; Schneider et al., 1991; Harris, 1990). 
Neoplastic growth through clonal expansion can result from somatic 
mutations that inactivate control over cell fate (Kakizuka et al., 
1991; deThe et al., 1991; Sidransky et al., 1992; Nowell, 1976).
    Cancers may also be thought of as diseases of the cell cycle. 
For example, genetic diseases that cause failure of cells to repair 
DNA damage prior to cell replication predispose people to cancer. 
These changes are also frequently found in tumor cells in sporadic 
cancers. These changes appear to be particularly involved at points 
in cell replication called ``checkpoints'' where DNA synthesis or 
mitosis is normally stopped until DNA damage is repaired or cell 
death induced (Tobey, 1975). A cell that bypasses a checkpoint may 
acquire a heritable growth advantage. Similar effects on the cell 
cycle occur when mitogens such as hormones or growth factors 
stimulate cell growth. Rapid replication in response to tissue 
injury may also lead to unrepaired DNA damage that is a risk factor 
for carcinogenesis.
    Normally a cell's fate is determined by a timed sequence of 
biochemical signals. Signal transduction in the cell involves 
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). Cells are subject to growth signals from the 
same and distant tissues, e.g., 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 
et al., 1990). The cells responsive to a particular growth factor 
are those that express transmembrane receptors that specifically 
bind the growth factor.
    Solid tumors develop in stages operationally defined as 
initiation, promotion, and progression (see, for example, Pitot and 
Dragan, 1991). These terms, which were coined in the context of 
specific experimental designs, are used for convenience in 
discussing concepts, but they refer to complex events that are not 
completely understood. During initiation, the cell acquires a 
genetic change that confers a potential growth advantage. During 
promotion, clonal expansion of this altered cell occurs. Later, 
during progression, a series of genetic and other biological events 
both enhance the growth advantage of the cells and enlist normal 
host processes to support tumor development and cells develop the 
ability to invade locally and metastasize distally, taking on the 
characteristics of malignancy. Many endogenous and exogenous factors 
are known to participate in the process as a whole. These include 
specific genetic predispositions or variations in ability to 
detoxify agents, medical history (Harris, 1989; Nebreda et al., 
1991), infections, exposure to chemicals or ionizing radiation, 
hormones and growth factors, and immune suppression. Several such 
risk factors likely work together to cause individual human cancers.
    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 
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 a subject of ongoing research. For known human 
carcinogens studied thus far, there is an often decades-long latency 
between exposure to carcinogenic agents and development of tumors 
(Fidler and Radinsky, 1990; Tanaka et al., 1991; Thompson et al., 
1989). This latency is also typical of tumor development in 
individuals with genetic diseases that make them prone to cancer 
(Meyn, 1993; Srivastava et al., 1990).
    The importance of genetic mutation in the carcinogenic process 
calls for special attention to assessing agents that cause such 
mutations. Heritable genetic defects that predispose humans to 
cancer are well known and the number of identified defects is 
growing. Examples include xeroderma pigmentosum (DNA repair defect) 
and Li Fraumeni and retinoblastoma (both are tumor suppressor gene 
mutations). Much of the screening and testing of agents for 
carcinogenic potential has been driven by the idea of identifying 
this mode of action. Cognizance of and emphasis on other modes of 
action such as ones that act at the level of growth signalling 
within or between cells, through cell receptors, or that indirectly 
cause genetic change, comes from more recent research. There are not 
yet standardized tests for many modes of action, but pertinent 
information may be available in individual cases.
    Agents of differing characteristics influence cancer 
development: inorganic and organic, naturally occurring and 
synthetic, of inanimate or animate origin, endogenous or exogenous, 
dietary and nondietary. The means by which these agents act to 
influence carcinogenesis are variable, and reasoned hazard 
assessment requires consideration of the multiple ways that 
chemicals influence cells in experimental systems and in humans. 
Agents exert mutagenic effects either by interacting directly with 
DNA or by indirect means through intermediary substances (e.g., 
reactive oxygen species) or processes. Most DNA-reactive chemicals 
are electrophilic or can become electrophilic when metabolically 
activated. Electrophilic molecules may bind covalently to DNA to 
form adducts, and this may lead to depurination, depyrimidation, or 
produce DNA strand breaks; such lesions can be converted to 
mutations with a round of DNA synthesis and cell division. Other 
DNA-interactive chemicals may cause the same result by intercalation 
into the DNA helix. Still other chemicals may methylate DNA, 
changing gene expression. Non-DNA-reactive chemicals produce 
genotoxic effects by many different processes. They may affect 
spindle formation or chromosome proteins, interfere with normal 
growth control mechanisms, or affect enzymes involved with ensuring 
the fidelity of DNA synthesis (e.g., topoisomerase), recombination, 
or repair.
    The ``classical'' chemical carcinogens in laboratory rodent 
studies are agents that consistently produce gene mutations and 
structural chromosome aberrations in short-term tests. A large 
database reveals that these mutagenic substances commonly produce 
tumors at multiple sites and in multiple species (Ashby and Tennant, 
1991). Most of the carcinogens identified in human studies, aside 
from hormones, are also gene or structural chromosome mutagens 
(Tennant and Ashby, 1991). Most of these compounds or their 
metabolites contain electrophilic moieties that react with DNA.
    Numerical chromosome aberrations, gene amplification, and the 
loss of gene heterozygosity are also found in animal and human tumor 
cells and may arise from initiating events or during progression. 
There is reason to believe that accumulation of additional genetic 
changes is favored by selection in the evolution of tumor cells 
because they confer additional growth advantages (Hartwell and 
Kastan, 1994). Exogenous agents may function at any stage of 
carcinogenesis (Barrett, 1993). Some aberrations may arise as a 
consequence of genomic instability arising from tumor suppressor 
gene mutation, e.g., p53 (Harris and Hollstein, 1993). The frequent 
observation in tumor cells that both of a pair of homologous 
chromosomes have identical mutation spectra in tumor suppressor 
genes suggests an ongoing, endogenous process of gene conversion. 
Currently, there is a paucity of routine test methods to screen for 
events such as gene conversion or gene amplification and knowledge 
regarding the

[[Page 18006]]

ability of particular agents of environmental interest to induce 
them is, for the most part, wanting. Work is under way to 
characterize, measure, and evaluate their significance (Travis et 
al., 1991).
    Several kinds of mechanistic studies aid in risk assessment. 
Comparison of DNA lesions in tumor cells taken from humans with the 
lesions that a tumorigenic agent causes in experimental systems can 
permit inferences about the association of exposure to the agent and 
an observed 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 mode 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.
    Differing modes 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 carcinogenic 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 carcinogenic activity of an 
agent that acts by causing cell toxicity followed by compensatory 
growth should be a function of the toxicity.

References

Alison, R.H.; Capen, C.C.; Prentice, D.E. (1994) Neoplastic lesions 
of questionable significance to humans. Toxicol. Pathol. 22: 179-
186.
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.
Ames, B.N.; Gold, L.S. (1990) Too many rodent carcinogens: 
mitogenesis increases mutagenesis. Science 249: 970-971.
Anderson, E.; Deisler, P.F.; McCallum, D.; St. Helaire, C.; Spitzer, 
H.L.; Strauss, H.; Wilson, J.D.; Zimmerman, R. (1993) Key issues in 
carcinogen risk assessment guidelines. Society for Risk Analysis.
Ashby, J.; Tennant, R.W. (1991) Definitive relationships among 
chemical structure, carcinogenicity and mutagenicity for 301 
chemicals tested by the U.S. NTP. Mutat. Res. 257: 229-306.
Ashby, J.; Tennant, R.W. (1994) Prediction of rodent carcinogenicity 
for 44 chemicals: results. Mutagenesis 9: 7-15.
Ashby, J.; Doerrer, N.G.; Flamm, F.G.; Harris, J.E.; Hughes, D.H.; 
Johannsen, F.R.; Lewis, S.C.; Krivanek, N.D.; McCarthy, J.F.; 
Moolenaar, R.J.; Raabe, G.K.; Reynolds, R.C.; Smith, J.M.; Stevens, 
J.T.; Teta, M.J.; Wilson, J.D. (1990) A scheme for classifying 
carcinogens. Regul. Toxicol. Pharmacol. 12: 270-295.
Ashby, J.; Brady, A.; Elcombe, C.R.; Elliott, B.M.; Ishmael, J.; 
Odum, J.; Tugwood, D.; Keltle, S.; Purchase, I.F.H. (1994) 
Mechanistically based human hazard assessment of peroxisome 
proliferator-induced hepatocarcinogenesis. Hum. Exper. Toxicol. 13: 
1-117.
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: 
167-175.
Barnes, D.G.; Daston, G.P; Evans, J.S.; Jarabek, A.M.; Kavlock, 
R.J.; Kimmel, C.A.; Park, C.; Spitzer, H.L. (1995) Benchmark dose 
workshop: criteria for use of a benchmark dose to estimate a 
reference dose. Regul. Toxicol. Pharmacol. 21: 296-306.
Barrett, J.C. (1992) Mechanisms of action of known human 
carcinogens. In: Mechanisms of carcinogenesis in risk 
identification. IARC Sci. Pubs. No. 116, Lyon, France: International 
Agency for Research on Cancer; 115-134.
Barrett, J.C. (1993) Mechanisms of multistep carcinogenesis and 
carcinogen risk assessment. Environ. Health Perspect. 100: 9-20.
Barrett, J.C. (1995) Role of mutagenesis and mitogenesis in 
carcinogenesis. Environ. Mutagenesis, in press.
Barrett, J. C.; Lee, T. C. (1992) Mechanisms of arsenic-induced gene 
amplification. In: Gene amplification in mammalian cells: A 
comprehensive guide (ed. R. E. Kellems), Marcel Dekker, New York: 
441-446.
Bayly, A.C.; Roberts, R.A.; Dive, C. (1994) Suppression of liver 
cell apoptosis in vitro by the nongenotoxic hepatocarcinogen and 
peroxisome proliferator nafenopin. J. Cell. Biol. 125: 197-203.
Bellamy, C.O.C.; Malcomson, R.D.G.; Harrison, D.J.; Wyllie, A.H. 
(1995) Cell death in health and disease: The biology and regulation 
of apoptosis. Seminars in cancer biology, Apoptosis in oncogenesis 
and chemotherapy 6: 3-16.
Bianchi, A.B.; Navone, N.M.; Alda, C.M.; Conti, C.J. (1991) 
Overlapping loss of heterozygosity by mitotic recombination on more 
chromosome 7F1-ter in skin carcinogenesis. Proc. Nat. Acad. Sci. 88: 
7590-7594.
Biggs, P.J.; Warren, W.; Venitt, S.; Stratton, M.R. (1993) Does a 
genotoxic carcinogen contribute to human breast cancer? The value of 
mutational spectra in unraveling the etiology of cancer. Mutagenesis 
8: 275-283.
Birner et al. (1990) Biomonitoring of aromatic amines. III: 
Hemoglobin binding and benzidine and some benzidine congeners. Arch. 
Toxicol. 64(2): 97-102.
Blair, A.; Burg, J.; Foran, J.; Gibb, H.; Greenland, S.; Morris, R.; 
Raabe, G.; Savitz, D.; Teta, J.; Wartenberg, D.; Wong, O.; 
Zimmerman, R. (1995) Guidelines for application of meta-analysis in 
environmental epidemiology. Regul. Toxicol. Pharmacol. 22: 189-197.
Bois, F.Y.; Krowech, G.; Zeise, L. (1995) Modeling human 
interindividual variability in metabolism and risk: the example of 
4-aminobiphenyl. 15: 205-213.
Bottaro, D.P.; Rubin, J.S.; Faletto, 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. Science 251: 802-804.
Bouck, N. (1990) Tumor angiogenesis: the role of oncogenes and tumor 
suppressor genes. Cancer Cells 2: 179-183.
Burek, J.D.; Patrick, D.H.; Gerson, R.J. (1988) Weight-of-biological 
evidence approach for assessing carcinogenicity. In: Grice, H.C.; 
Cimina, J.L., eds. Carcinogenicity. New York, NY: Springer Verlag; 
pp. 83-95.
Bus, J.S.; Popp, J.A. (1987) Perspectives on the mechanism of action 
of the splenic toxicity of aniline and strucurally related 
compounds. Fd. Chem. Toxicol. 25: 619-626.
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.
Cantley, L.C.; Auger, K.R.; Carpenter, C.; Duckworth, B.; Graziani, 
A.; Kapeller, R.; Soltoff, S. (1991) Oncogenes and signal 
transduction. Cell 64: 281-302.
Caporaso, N.; Hayes, R.B.; Dosemeci, M.; Hoover, R.; Ayesh, R.; 
Hetzel, M.; Idle, J. (1989) Lung cancer risk, occupational exposure, 
and the debrisoquine metabolic phenotype. Cancer Res. 49: 3675-3679.
Cavenee, W.K.; Koufos, A.; Hansen, M. F. (1986) Recessive mutant 
genes predisposing to human cancer. Mutation Research 168: 3-14.
Chang, C.C.; Jone, C.; Trosko, J. E.; Peterson, A. R.; Sevanian, A. 
(1988) Effect of cholesterol epoxides on the inhibition of 
intercellular communication and on mutation induction in Chinese 
hamster V79 cells. Mutation Research 206: 471-478.

[[Page 18007]]

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: Implications for risk 
assessment. New York, NY: Wiley-Liss; pp. 481-499.
Choy, W.N. (1993) A review of the dose-response induction of DNA 
adducts by aflatoxin B2 and its implications to quantitative 
cancer-risk assessment. Mutat. Res. 296: 181-198.
Clayson, D.B. (1989) Can a mechanistic rationale be provided for 
non-genotoxic carcinogens identified in rodent bioassays? Mutat. 
Res. 221: 53-67.
Clayson, D.B.; Mehta, R.; Iverson, F. (1994) Oxidative DNA damage--
The effects of certain genotoxic and operationally non-genotoxic 
carcinogens. Mutat. Res. 317: 25-42.
Cogliano, V.J. (1986) The U.S. EPA's methodology for adjusting the 
reportable quantities of potential carcinogens. Proceedings of the 
7th National Conference on Management of Uncontrollable Hazardous 
Wastes (Superfund '86). Washington, DC: Hazardous Wastes Control 
Institute, 182-185.
Cohen, S.W.; Ellwein, L.B. (1990) Cell proliferation in 
carcinogenesis. Science 249: 1007-1011.
Cohen, S.M.; Ellwein, L.B. (1991) Genetic errors, cell proliferation 
and carcinogenesis. Cancer Res. 51: 6493-6505.
Cohen, S.M.; Purtilo, D.T.; Ellwein, L.B. (1991) Pivotal role of 
increased cell proliferation in human carcinogenesis. Mod. Pathol. 
4: 371-375.
Collum, R.G.; Alt, F.W. (1990) Are myc proteins transcription 
factors? Cancer Cells 2: 69-73.
Connolly, R.B.; Andersen, M.E. (1991) Biologically based 
pharmacodynamic models: tools for toxicological research and risk 
assessment. Ann. Rev. Pharmacol. Toxicol. 31: 503-523.
Cross, M.; Dexter, T. (1991) Growth factors in development, 
transformation, and tumorigenesis. Cell 64: 271-280.
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.
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.
Enterline, P.E.; Henderson, V.L.; Marsh, G.M. (1987) Exposure to 
arsenic. Amer. J. Epidemiol. 125: 929-938.
European Centre for Ecotoxicology and Toxicology of Chemicals 
(ECETOC). (1991) Early indicators of non-genotoxic carcinogenesis. 
ECETOC Monograph No. 16. Brussels: ECETOC. Printed in Mutat. Res. 
248: 211-374.
Faustman, E.M.; Allen, B.C.; Kavlock, R.J.; Kimmel, C.A. (1994) 
Dose-response assessment for developmental toxicity. Fund. Appl. 
Toxicol. 23: 478-486.
Fearon, E.; Vogelstein, B. (1990) A genetic model for colorectal 
tumorigenesis. Cell 61: 959-967.
Federated Association of Societies of Experimental Biology (FASEB) 
(1994) Evaluation of evidence for the carcinogenicity of butylated 
hydroxyanisole (BHA). Life Sciences Research Office, Bethesda, MD. 
Letter from Hamilton Brown (FDA) to John Rice (FASEB), July 28, 
1994. Letter from John Rice (FASEB) to Ed Arnold (FDA), August 4, 
1994.
Fidler, I.J.; Radinsky, R. (1990) Genetic control of cancer 
metastasis. J. Natl. Cancer Inst. 82: 166-168.
Fisher, R.A. (1950) Statistical methods for research workers. 
Edinborough, Scotland: Oliver and Boyd.
Flamm, W.G.; Winbush, J.S. (1984) Role of mathematical models in 
assessment of risk and in attempts to define management strategy. 
Fund. Appl. Toxicol. 4: S395-S401.
Flynn, G.L. (1990) Physicochemical determinants of skin absorption. 
In: Gerrity, 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.
Forsburg, S.L.; Nurse, P. (1991) Identification of a G1-type cyclin 
pug1+ 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.
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.
Gerrity, T.R.; Henry, C., eds. (1990) Principles of route to route 
extrapolation for risk assessment. New York, NY: Elsevier Science 
Publishing Co.
Gibson, D.P.; Aardema, M.J.; Kerckaert, G.A.; Carr, G.J.; 
Brauninger, R.M.; LeBoeuf, R.A. (1995) Detection of aneuploidy-
inducing carcinogens in the Syrian hamster embryo (SHE) cell 
transformation assay. Mutat. Res.; Genet. Toxicol. 343: 7-24.
Gillette, J.R. (1983) The use of pharmacokinetics in safety testing. 
In: Homburger, ed. Safety evaluation and regulation of chemicals 2. 
2nd Int. Conf., Cambridge, MA: Karger, Basel; pp. 125-133.
Goldsworthy, T.L.; Hanigan, M.H.; and Pitot, H.C. (1986) Models of 
hepatocarcinogenesis in the rat--contrasts and comparisons. CRC 
Crit. Rev. Toxicol. 17: 61-89.
Goodman, G.; Wilson, R. (1991) Predicting the carcinogenicity of 
chemicals in humans from rodent bioassay data. Environ. Health 
Perspect. 94: 195-218.
Goodman, J.I.; Counts, J.L. (1993) Hypomethylation of DNA: A 
possible nongenotoxic mechanism underlying the role of cell 
proliferation in carcinogenesis. Environ. Health Perspect. 101 Supp. 
5: 169-172.
Goodman, J.I.; Ward, J.M.; Popp, J.A.; Klaunig, J.E.; Fox, T.R. 
(1991) Mouse liver carcinogenesis: Mechanisms and relevance. Fund. 
Appl. Toxicol. 17: 651-665.
Grasso, P.; Hinton, R.H. (1991) Evidence for and possible mechanisms 
of non-genotoxic carcinogenesis in rodent liver. Mutat. Res. 248: 
271-290.
Greenland, S. (1987) Quantitative methods in the review of 
epidemiologic literature. Epidemiol. Rev. 9: 1-29.
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) Interindividual variation among humans in 
carcinogen metabolism, DNA adduct formation and DNA repair. 
Carcinogenesis 10: 1563-1566.
Harris, C.C.; Hollstein, M. (1993) Clinical implications of the p53 
tumor suppressor gene. N. Engl. J. Med. 329: 1318-1327.
Harris, H. (1990) The role of differentiation in the suppression of 
malignancy. J. Cell Sci. 97: 5-10.
Hartwell, L.H.; Kastan, M.B. (1994) Cell cycle control and cancer. 
Science 266: 1821-1828.
Haseman, J.K. (1983) Issues: a reexamination of false-positive rates 
for carcinogenesis studies. Fund. Appl. Toxicol. 3: 334-339.
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. (1985) Issues in carcinogenicity testing: dose 
selection. Fund. Appl. Toxicol. 5: 66-78.
Haseman, J.K. (1990) Use of statistical decision rules for 
evaluating laboratory animal carcinogenicity studies. Fund. Appl. 
Toxicol. 14: 637-648.
Haseman, J.K. (1995) Data analysis: Statistical analysis and use of 
historical control data. Regul. Toxicol. Pharmacol. 21: 52-59.
Haseman, J.K.; Huff, J.; Boorman, G.A. (1984) Use of historical 
control data in carcinogenicity studies in rodents. Toxicol. Pathol. 
12: 126-135.
Hattis, D. (1990) Pharmacokinetic principles for dose-rate 
extrapolation of carcinogenic risk from genetically active agents. 
Risk Anal. 10: 303-316.
Havu, N.; Mattsson, H.; Ekman, L.; Carlsson, E. (1990) 
Enterochromaffin-like cell carcinoids in the rat gastric mucosa 
following long-term administration of ranitidine. Digestion 45: 189-
195.

[[Page 18008]]

Hayward, N.K.; Walker, G.J.; Graham, W.; Cooksley, E. (1991) 
Hepatocellular carcinoma mutation. Nature 352: 764.
Hayward, J.J.; Shane, B.S.; Tindall, K.R.; Cunningham, M.L. (1995) 
Differential in vivo mutagenicity of the carcinogen-noncarcinogen 
pair 2,4- and 2,6-diaminotoluene. Carcinogenesis. In press.
Herschman, H.R. (1991) Primary response genes induced by growth 
factors or promoters. Ann. Rev. Biochem. 60: 281-319.
Hill, R.N.; Erdreich, L.S.; Paynter, O.E.; Roberts, P.A.; Rosenthal, 
S.L.; Wilkinson, C.F. (1989) Thyroid follicular cell carcinogenesis. 
Fund. Appl. Toxicol. 12: 629-697.
Hoel, D.G.; Portier, C.J. (1994) Nonlinearity of dose-response 
functions for carcinogenicity. Environ. Health Perspect. 102 Suppl 
1: 109-113.
Hoel, D.G.; Haseman, J.K.; Hogam, M.D.; Huff, J.; McConnell, E.E. 
(1988) The impact of toxicity on carcinogenicity studies: 
Implications for risk assessment. Carcinogenesis 9: 2045-2052.
Holliday, R. (1987) DNA methylation and epigenetic defects in 
carcinogenesis. Mutation Research 181: 215-217.
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. (1993) Chemicals and cancer in humans: first evidence in 
experimental animals. Environ. Health Perspect. 100: 201-210.
Huff, J.E. (1994) Chemicals causally associated with cancers in 
humans and laboratory animals. A perfect concordance. In: 
Carcinogenesis. Waalkes, M.P.; Ward, J.M., eds., New York, NY: Raven 
Press; pp. 25-37.
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 mechanisms 
of non-genotoxic carcinogenesis in mouse skin. Mutat. Res. 248: 333-
340.
International Agency for Research on Cancer (IARC). (1990) 
Ciclosporin. IARC monographs on the evaluation of carcinogenic risks 
to humans. Vol. 50. Lyon, France: IARC; pp. 77-114.
IARC. (1994) IARC monographs on the evaluation of carcinogenic risks 
to humans. Vol. 60, Some industrial chemicals. Lyon, France: IARC; 
pp. 13-33.
International Life Science Institute (ILSI). (1995) The use of 
biological data in cancer risk assessment. In: Olin, S.; Farland, 
W.; Park, C.; Rhomberg, L.; Scheuplein, R.; Starr, T.; Wilson, J., 
eds. Low-dose extrapolation of cancer risks: Issues and 
Perspectives. Washington, DC: ILSI Press; pp. 45-60.
Ito, N.; Shirai, T.; and Hasegawa, R. (1992) Medium-term bioassays 
for carcinogens. in ``Mechanisms of Carcinogenesis in Risk 
Identifications'' (eds., H. Vainio, PN. Magee, DB McGregor and AJ 
McMichael), Lyon, International Agency for Research on Cancer, pp. 
353-388.
Jack, D.; Poynter, D.; Spurling, N.W. (1983) Beta-adrenoreceptor 
stimulants and mesoovarian leiomyomas in the rat. Toxicology 2: 315-
320.
Jarabek, A.M. (1995a) The application of dosimetry models to 
identify key processes and parameters for default dose-response 
assessment approaches. Toxicol. Lett. 79:171-184.
Jarabek, A.M. (1995b) Interspecies extrapolation based on 
mechanistic determinants of chemical disposition. Human Eco. Risk 
Assess. 1(5):641-662.
Jones, P.A. (1986) DNA methylation and cancer. Cancer Res. 46: 461-
466.
Kehrer, J.P. (1993) Free radicals as mediators of tissue injury and 
disease. Crit. Rev. Toxicol. 23: 21-48.
Kelsey, J.L.; Thompson, W.D.; Evans, A.S. (1986) Methods in 
observational epidemiology. New York, NY: Oxford University Press.
Mr. Kinzler, K.W.; Nilbert, M.C.; Su, L.-K.; Vogelstein, B.; Bryan, 
T.M.; Levy, D.B.; Smith, K.J.; Preisinger, A.C.; Hedge, P.; 
McKechnie, D.; Finniear, R.; Markham, A.; Groffen, J.; Boguski, 
M.S.; Altschul, S.J.; Horii, A.; Ando, H.; Miyoshi, Y.; Miki, Y.; 
Nishisho, I.; Nakamura, Y. (1991) Identification of FAP locus genes 
from chromosome 5q21. Science 253: 661-665.
Kodell, R.L.; Park, C.N. (1995) Linear extrapolation in cancer risk 
assessment. ILSI Risk Science Institute: Washington, D.C. In press.
Kraus, A.L.; Munro, I.C.; Orr, J.C.; Binder, R.L.; LeBoeuf, R.A.; 
Williams, G.M. (1995) Benzoyl peroxide: An integrated human safety 
assessment for carcinogenicity. Regul. Toxicol. Pharmacol. 21: 87-
107.
Krewski, D.; Brown, C.; Murdoch, D. (1984) Determining ``safe'' 
levels of exposure: Safety factors of mathematical models. Fund. 
Appl. Toxicol. 4: S383-S394.
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. Vol. 
8. Washington, DC: National Academy Press; pp. 441-468.
Kripke, M.L. (1988) Immunoregulation of carcinogenesis: Past, 
present, and future. J. Natl. Cancer Inst. 80: 722-727.
Kumar, R.; Sukumar, S.; Barbacid, M. (1990) Activation of ras 
oncogenes preceding the onset of neoplasia. Science 248: 1101-1104.
Larson, J.L.; Wolf, D.C.; Butterworth, B.E. (1994) Induced 
cytotoxicity and cell proliferation in the hepatocarcinogenicity of 
chloroform in female B6C3F1 mice: comparison of administration by 
gavage in corn oil vs. ad libitum in drinking water. Fundam. Appl. 
Toxicol. 22: 90-102.
Levine, A.M. (1993) AIDs-related malignancies: The emerging 
epidemic. J. Natl. Cancer Inst. 85: 1382-1397.
Levine, P.H.; Hoover, R.N., eds. (1992) The emerging epidemic of 
non-Hodgkin's lymphoma: Current knowledge regarding etiological 
factors. Cancer Res. 52: 5426s-5574s.
Levine, A. J.; Momand, J.; Finlay, C. A. (1991) The p53 tumour 
suppressor gene. Nature 351: 453-456.
Levine, A.J.; Perry, M.E.; Chang, A., et al. (1994) The 1993 Walter 
Hubert Lecture: The role of the p53 tumor-suppressor gene in 
tumorigenesis. Brit. J. Cancer 69: 409-416.
Li, J.L.; Okada, S.; Hamazaki, S.; Ebina, Y.; Midorikawa, O. (1987) 
Subacute nephrotoxicity and induction of renal cell carcinoma in 
mice treated with ferric nitrilotriacetate. Cancer Res. 47: 1867-
1869.
Lijinsky, W. (1993) Species differences in carcinogenesis. In Vivo 
7: 65-72.
Lilienfeld, A.M.; Lilienfeld, D. (1979) Foundations of epidemiology, 
2nd ed. New York, NY: Oxford University Press.
Loeb, L.A. (1991) Mutator phenotype may be required for multistage 
carcinogenesis. Cancer Res. 51: 3075-3079.
Lutz, W.K. (1990a) Endogenous genotoxic agents and processes as a 
basis of spontaneous carcinogenesis. Mutat. Res. 238: 287-295.
Lutz, W.K. (1990b) Dose-response relationship and low dose 
extrapolation in chemical carcinogenesis. Carcinogenesis 11: 1243-
1247.
MacDonald, J.S.; Lankas, G.R.; Morrissey, R.E. (1994) Toxicokinetic 
and mechanistic considerations in the interpretation of the rodent 
bioassay. Toxicol. Pathol. 22: 124-140.
Maller, J.L. (1991) Mitotic control. Curr. Opin. Cell Biol. 3: 269-
275.
Maronpot, R.R.; Shimkin, M.B.; Witschi, H.P.; Smith, L.H.; and 
Cline, J.M. (1986) Strain A mouse pulmonary tumor test results for 
chemicals previously tested in National Cancer Institute 
carcinogenicity test. J. Natl. Cancer Inst. 76: 1101-1112.
Marsman, D.S.; Popp, J.A. (1994) Biological potential of basophilic 
hepatocellular foci and hepatic adenoma induced by the peroxisome 
proliferator, Wy-14,643. Carcinogenesis 15: 111-117.
Mausner, J.S.; Kramer, S. (1985) Epidemiology, 2nd ed. Philadelphia: 
W.B. Saunders Company.
McClain, R.M. (1994) Mechanistic considerations in the regulation 
and classification of chemical carcinogens. In: Kotsonis, F.N.; 
Mackey, M.; Hjelle, J., eds. Nutritional toxicology. New York, NY: 
Raven Press; pp. 273-304.
McConnell, E.E.; Solleveld, 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.
McMichael, A.J. (1976) Standardized mortality ratios and the 
``healthy worker effect'': scratching beneath the surface. J. Occup. 
Med. 18: 165-168.

[[Page 18009]]

Melnick, R.L.; Huff, J.E.; Barrett, J.C.; Maronpot, R.R.; Lucier, 
G.; Portier, C.J. (1993a) Cell proliferation and chemical 
carcinogenesis: A symposium overview. Molecular Carcinogenesis 7: 
135-138.
Melnick, R.L.; Huff, J.E.; Barrett, J.C.; Maronpot, R.R.; Lucier, 
G.; Portier, C.J. (1993b) Cell proliferation and chemical 
carcinogenesis. Molecular Carcinogeneis 7: 135-138.
Meyn, M.S. (1993) High spontaneous intrachromosomal recombination 
rates in ataxia-telangiectasia. Science 260: 1327-1330.
Modrich, P. (1994) Mismatch repair, genetic stability, and cancer. 
Science 266: 1959-1960.
Monro, A. (1992) What is an appropriate measure of exposure when 
testing drugs for carcinogenicity in rodents? Toxicol. Appl. 
Pharmacol. 112: 171-181.
Moolgavkar, S.H.; Knudson, A.G. (1981) Mutation and cancer: A model 
for human carcinogenesis. J. Natl. Cancer Inst. 66: 1037-1052.
Morrison, V.; Ashby, J. (1994) A preliminary evaluation of the 
performance of the mutaTM mouse (lacZ) and Big BlueTM 
(lacI) transgenic mouse mutation assays. Mutagenesis 9: 367-375.
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 Research Council (NRC). (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.
NRC. (1993a) Pesticides in the diets of infants and children. 
Committee on Pesticides in the Diets of Infants and Children, 
Commission on Life Sciences, NRC. Washington, DC: National Academy 
Press.
NRC. (1993b) Issues in risk Assessment. Committee on Risk Assessment 
Methodology, Commission on Life Sciences, NRC. Washington, DC: 
National Academy Press.
NRC. (1994) Science and judgment in risk assessment. Committee on 
Risk Assessment of Hazardous Air Pollutants, Commission on Life 
Sciences, NRC. Washington, DC: National Academy Press.
National Toxicology Program (NTP). (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.
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.
Organization for Economic Cooperation and Development (OECD). (1981) 
Guidelines for testing of chemicals. Carcinogenicity studies. No. 
451. Paris, France.
Peltomaki, P.; Aaltonen, L.A.; Sisonen, P.; Pylkkanen, L.; Mecklin, 
J.-P.; Jarvinen, H.; Green, J.S.; Jass, J.R.; Weber, J.L.; Leach, 
F.S.; Petersen, G.M.; Hamilton, S.R.; de la Chapelle, A.; 
Vogelstein, B. (1993) Genetic mapping of a locus predisposing human 
colorectal cancer. Science 260: 810-812.
Peto, J. (1992) Meta-analysis of epidemiological studies of 
carcinogenesis. In: Mechanisms of carcinogenesis in risk assessment. 
IARC Sci. Pubs. No. 116, Lyon, France: IARC; pp. 571-577.
Peto, J.; Darby, S. (1994) Radon risk reassessed. Nature 368: 97-98.
Pitot, H.; Dragan, Y.P. (1991) Facts and theories concerning the 
mechanisms of carcinogenesis. FASEB J. 5: 2280-2286.
Portier, C. (1987) Statistical properties of a two-stage model of 
carcinogenesis. Environ. Health Perspect. 76: 125-131.
Prahalada, S.; Majka, J.A.; Soper, K.A.; Nett, T.M.; Bagdon, W.J.; 
Peter, C.P.; Burek, J.D.; MacDonald, J.S.; van Zwieten, M.J. (1994) 
Leydig cell hyper plasian and adenomas in mice treated with 
finasteride, 5-reductase inhibitor: A possible mechanism. 
Fund. Appl. Toxicol. 22: 211-219.
Raff, M.C. (1992) Social controls on cell survival and cell death. 
Nature 356: 397-400.
Rall, 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 --an oncodevelopmental growth 
factor. Cancer Cells 2: 389-397.
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.
Schulte-Hermann, R.; Bursch, W.; Kraupp-Grasl, B.; Oberhammer, F.; 
Wagner, A.; Jirtle, R. (1993) Cell proliferation and apoptosis in 
normal liver and preneoplastic foci. Environ. Health Perspect. 101 
(Supp. 5): 87-90.
Shelby, M.D.; Zeiger, E. (1990) Activity of human carcinogens in the 
Salmonella and rodent bone-marrow cytogenetics tests. Mutat. Res. 
234: 257-261.
Shelby, M.D. (1994) Human germ cell mutations. Environ. Molec. 
Mutagen. 23 (Supp. 24): 30-34.
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.
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.
Sisk, S.C.; Pluta, L.J.; Bond, J.A.; Recio, L. (1994) Molecular 
analysis of lacI mutants from bone marrow of B6C3F1 transgenic mice 
following inhalation exposure to 1,3-butadiene. Carcinogenesis 
15(3): 471-477.
Snedecor, G.W.; Cochran, W.G. (1978) Statistical methods, Sixth ed. 
Ames, Iowa: Iowa State University Press; 593 pp.
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.
Stewart, B.W. (1994) Mechanisms of apoptosis: Integration of 
genetic, biochemical, and cellular indicators. J. Natl. Cancer Inst. 
86: 1286-1296.
Stiteler, W.H.; Knauf, L.A.; Hertzberg, R.C.; Schoeny, R.S. (1993) A 
statistical test of compatibility of data sets to a common dose-
response model. Reg. Tox. Pharmacol. 18: 392-402.
Stitzel, K.A.; McConnell, R.F.; Dierckman, T.A. (1989) Effects of 
nitrofurantoin on the primary and secondary reproductive organs of 
female B6C3F1 mice. Toxicol. Pathol. 17: 774-781.
Strausfeld, U.; Labbe, J.C.; Fesquet, D.; Cavadore, J.C.; Dicard, 
A.; Sadhu, K.; Russell, P.; Dor'ee, M. (1991) Identification of a 
G1-type cyclin puc1+ in the fission yeast Schizosaccharomyces pombe. 
Nature 351: 242-245.
Sukumar, S. (1989) ras oncogenes in chemical carcinogenesis. Curr. 
Top. Microbiol. Immunol. 148: 93-114.
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.
Swierenga, S.H.H.; Yamasaki, H. (1992) Performance of tests for cell 
transformation and gap junction intercellular communication for 
detecting nongenotoxic carcinogenic activity. In: Mechanisms of 
carcinogenesis in risk identification. IARC Sci. Pubs. No. 116, 
Lyon, France: International Agency for Research on Cancer; pp. 165-
193.

[[Page 18010]]

Tanaka, K.; Oshimura, M.; Kikiuchi, R.; Seki, 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.
Tarone, R.E. (1982) The use of historical control information in 
testing for a trend in proportions. Biometrics 38: 215-220.
Taylor, J.H.; Watson, M.A.; Devereux, T.R.; Michels, R.Y.; 
Saccomanno, G.; Anderson, M. (1994) Lancet 343: 86-87.
Tennant, R.W. (1993) Stratification of rodent carcinogenicity 
bioassay results to reflect relative human hazard. Mutat. Res. 286: 
111-118.
Tennant, R.W.; Ashby, J. (1991) Classification according to chemical 
structure, mutagenicity to Salmonella and level of carcinogenicity 
of a further 39 chemicals tested for carcinogenicity by the U.S. 
National Toxicology Program. Mutat. Res. 257: 209-277.
Tennant, R.W.; Elwell, M.R.; Spalding, J.W.; Griesemer, R.A. (1991) 
Evidence that toxic injury is not always associated with induction 
of chemical carcinogenesis. Molec. Carcinogen. 4: 420-440.
Tennant, R.W.; French, J.E.; Spalding, J.W. (1995) Identifying 
chemical carcinogens and assessing potential risk in short-term 
bioassays using transgenic mouse models. Environ. Health Perspect. 
103:942-950.
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.
Tinwell, H.; Ashby, J. (1991) Activity of the human carcinogen 
MeCCNU in the mouse bone marrow mironucleus test. Environ. Molec. 
Mutagen. 17: 152-154.
Tischler, A.S.; McClain, R.M.; Childers, H.; Downing, J. (1991) 
Neurogenic signals regulate chromaffin cell proliferation and 
mediate the mitogenic effect of reserpine in the adult rat adrenal 
medulla. Lab. Invest. 65: 374-376.
Tobey, R.A. (1975) Different drugs arrest cells at a number of 
distinct stages in G2. Nature 254: 245-247.
Todd, G.C. (1986) Induction of reversibility of thyroid 
proliferative changes in rats given an antithyroid compound. Vet. 
Pathol. 23: 110-117.
Tomatis, L.; Aitio, A.; Wilbourn, J.; Shuker, L. (1989) Human 
carcinogens so far identified. Jpn. J. Cancer Res. 80: 795-807.
Travis, C.C.; McClain, T.W.; Birkner, P.D. (1991) 
Diethylnitrosamine-induced hepatocarcinogenesis in rats: A 
theoretical study. Toxicol. Appl. Pharmacol. 109: 289-309.
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. Available from: NTIS, 
Springfield, VA; PB 83-153916.
U.S. Environmental Protection Agency. (1983c) Health effects test 
guidelines. Available from: NTIS, Springfield, VA; 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. Office of Health and Environmental Assessment, 
Washington, DC.
U.S. Environmental Protection Agency. (1986a) Health assessment 
document for beryllium. Office of Health and Environmental 
Assessment, Washington, DC.
U.S. Environmental Protection Agency. (1986b) Guidelines for 
carcinogen risk assessment. Federal Register 51(185):33992-34003.
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. Risk Assessment Forum, Washington, DC. EPA/625/3-89/
016.
U.S. Environmental Protection Agency. (1989b) Workshop on EPA 
guidelines for carcinogen risk assessment. Risk Assessment Forum, 
Washington, DC. EPA/625/3-89/015.
U.S. Environmental Protection Agency. (1989c) Workshop on EPA 
guidelines for carcinogen risk assessment: use of human evidence. 
Risk Assessment Forum, Washington, DC. 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, Addendum 9. Office of Pesticide 
Programs, Washington, DC. PB91-158394, 540/09-91-122.
U.S. Environmental Protection Agency. (1991b) Alpha-2u-globulin: 
Association with chemically induced renal toxicity and neoplasia in 
the male rat. Risk Assessment Forum, Washington, DC. EPA/625/3-91/
019F.
U.S. Environmental Protection Agency. (1991c) Workshop report on 
toxicity equivalency factors for polychlorinated biphenyl congeners. 
Risk Assessment Forum, Washington, DC. EPA/625/3-91/020.
U.S. Environmental Protection Agency. (1991f) Guidelines for 
developmental toxicity risk assessment. Federal Register 56(234): 
63798-63826.
U.S. Environmental Protection Agency. (1992a) Guidelines for 
exposure assessment. 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/kg\3/4\/day. Federal Register 57(109): 24152-
24173.
U.S. Environmental Protection Agency. (1992c) 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 001h.
U.S. Environmental Protection Agency. (1994) Methods for derivation 
of inhalation reference concentrations and application of inhalation 
dosimetry. Office of Health and Environmental Assessment, 
Environmental Criteria and Assessment Office, Research Triangle 
Park, NC. EPA/600/8-90/066F.
U.S. Environmental Protection Agency. (1994a) Estimating exposure to 
dioxin-like compounds. Office of Health and Environmental 
Assessment, Office of Research and Development, Washington, DC. 
External Review Draft, 3 vol. EPA/600/6-88/005Ca, Cb, Cc.
U.S. Environmental Protection Agency. (1994b) Report on the workshop 
on cancer risk assessment guidelines issues. Office of Research and 
Development, Risk Assessment Forum, Washington, DC. EPA/630/R-94/
005a.
U.S. Environmental Protection Agency. (1995) Policy for risk 
characterization. Memorandum of Carol M. Browner, Administrator, 
March 21, 1995, Washington, D.C.
U.S. Food and Drug Administration (1987) Sponsored compounds in 
food-producing animals; criteria and procedures for evaluating the 
safety of carcinogenic residues. final rule. 21 CFR Parts 70, 500, 
514 and 571.
Vahakangas, K.H.; Samet, J.M.; Metcalf, R.A.; Welsh, J.A.; Bennett, 
W.P.; Lane, D.P.; Harris, C.C. (1992) Mutation of p53 and ras genes 
in radon-associated lung cancer from uranium miners. Lancet 339: 
576-578.
Vainio, H.; Magee, P.; McGregor, D.; McMichael, A.J. (1992) 
Mechanisms of carcinogenesis in risk identification. IARC Sci. Pubs. 
No. 116. Lyon, France: IARC.
Van Sittert, N.J.; De Jong, G.; Clare, M.G.; Davies, R.; Dean, B.J.; 
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.
Vater, S.T.; McGinnis, P.M.; Schoeny, R.S.; Velazquez, S. (1993) 
Biological considerations for combining carcinogenicity data for 
quantitative risk assessment. Reg. Toxicol. Pharmacol. 18: 403-418.
Vogelstein, B.; Fearon, E. R.; Hamilton, S. R.; Kern, S. E.; 
Presinger, A. C.; Leppert, M.; Nakamura, Y.; White, R.; Smits, A. M. 
M.; Bos, J. L. (1988) Genetic alterations during colorectal-tumor 
development. New England Journal of Medicine 319: 525-532.
Weinberg, R.A. (1989) Oncogenes, antioncogenes, and the molecular 
bases of multistep carcinogenesis. Cancer Res. 49: 3713-3721.

[[Page 18011]]

Wellstein, A.; Lupu, R.; Zugmaier, G.; Flamm, 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.; Arcos, J.C. (1989) Role of structure-activity 
relationship analysis in evaluation of pesticides for potential 
carcinogenicity. In: Ragsdale, N.N.; Menzer, R.E., eds. 
Carcinogenicity and pesticides: Principles, issues, and 
relationship. ACS Symposium Series No. 414. San Diego: Academic 
Press; pp. 175-200.
Wyzga, R.E. (1988) The role of epidemiology in risk assessments of 
carcinogens. Adv. Mod. Environ. Toxicol. 15: 189-208.
Yamada, T.; Nakamura, J.; Murakami, M.; Okuno, Y.; Hosokawa, S.; 
Matsuo, M.; Yamada, H. (1994) The correlation of serum luteinizing 
hormone levels with the induction of Leydig cell tumors in rats by 
oxolinic acid. Toxicol. Appl. Pharmacol. 129: 146-154.
Yamasaki, H. (1990) Gap junctional intercellular communication and 
carcinogenesis. Carcinogenesis 11: 1051-1058.
Yamasaki, H. (1995) Non-genotoxic mechanisms of carcinogenesis: 
Studies of cell transformation and gap junctional intercellular 
communication. Toxicol. Lett. 77: 55-61.
Zhang, K.; Papageorge, A.G.; Lowry, D.R. (1992) Mechanistic aspects 
of signalling through ras in NIH 3T3 cells. Science 257: 671-674.
[FR Doc. 96-9711 Filed 4-22-96; 8:45 am]
BILLING CODE 6560-50-P