Document a4EBDROM77mEomx1ZaJ9qEXxR

- :'7 ,--iniii for enclosure DATE ~7 ~~3 -7 / Please Discuss With For Ap proval Note and Note and Forwarded Forward Return Per Your To File ToSenderf Request DUP 0811066 DU 009589 l. rUr jiTifh'*- wU Part III Consumer Product Safety Commission Environmental Protection Agency Department of Health, Education, and Welfare Food and Drug Administration Department of Agriculture Food Safety and Quality Service Scientific Bases for Identification of potential Carcinogens and Estimation of Risks______________________ Requests for Comments on Reports DU 009590 DUP 0811067 "39858 Federal Register / Vol. 44. . fc. --BM'lUnmi,I iiiiifB- jl / Friday, July 8. 1979 / Notices CONSUMER RROOUCT SAFETY COMMISSION ENVIRONMENTAL PROTECTOR AGENCY DEPARTMENT OF HEALTH EDUCATION, AND WELFARE Food and Drug Administration DEPARTMENT OF AGRICULTURE Food Safety and Quality Sarvtc* Scientific Bates for Identification of Potential Carcinogens and Estimation of Risks; Request for Comments on Report AQtNCKS: Consumer Product Safety Commission (CPSC): Environmental Protection Agency (EPA); Food and Drug Administration. Department of Health. Education and Welfare (FDA); Food Safety and Quality Service. Department of Agriculture (FSQS) ACTION: Request for public comment on scientific report.* I SOM*AMT: This notice publishes and requests comment on a scientific report entitled. "Scientific Bases for Identification of Potential Carcinogens and Estimation of Risks." The report was written by the Work Group on Risk Assessment of the Interagency Regulatory Liaison Group (IKI.C) with the assistance of senior scientists at the National Cancer Institute (NCI) and the National Institute uf Environmental I Icalih Sciences (NIEHS). The report represent* the lies) judgments of these scientists and those of the four agencies |C1>SC. EPA. FDA. and the Occupational Safety and Health Administration (OS1IA)) comprising the IRLG at the lime the report was written on the scientific concepts and methods currently in use to identify and evaluate subr'ancet that may pose a risk of cancer to humans. The FSQS has since fumed 1RI.G. Scientists at FSQS have reviewed the report and concur. The report it Iteing published by CPSC EPA. FDA. und FSQS for comment in order to give interested persons an opportunity to express their views on the validity and appropriateness of the concepts end methods described for identifying and evaluating carcinogens. After reviewing the comments received, the four agencies anticipate publishing a statement giving notice of whatever revisions lo the document are appropriate, if any. oats* Written comments on the report should be eubmitled by September 30. taro. Aooncsst Comments should be sent lo IRLG. Room MO. till lSlh Street N.W,, Washington. D.C 20207. SON FUNTMEH INFORMATION CONTACT Susan Guvnelti' at (3C-4kl4-.l3n|. SUPPIEMCNTANV MFONMATtOM Background In August. 1977. The Consume! Product Safety Commission, the Environmental Protection Agency, the Food and Drug Administration of the Department of Health. Education, and Welfare, and the Occupational Safety and Health Administration of the Department of Labor agreed to work together es the Interagency Regulatory Liaison Croup (IRLG) lo intpruve protection of the public health and the env ironment through sharing of information, avoiding duplication of effort, and developing consistent regulatory policy.- On October 11.1977. the IRLG published in the Federal Regular an Interagency Agreement relating lo the Regulation of Toxic and Hazardous Substances (42 FR 5483d). To implement this agreement, the IRI.G established work groups to develop common, consistent, or compatible practices in areas of activities common to the four activities, including risk assessments. Provisional work plans were published for the work groups in the Federal Register on February 17.197V (43 FR 7174). The work plan for the Work Group on Risk Assessment, appearing at 43 FR 7193. provides that the general goal of the work group is tu characterise the types ot health hazards that may result from human exposure' to chemicals, dev ices, consumer goods, and other uiiu let and substances. The initial task established by the work group svae to address the problems associated with health risks due to exposure to chemicals, specifically the risk of cancer. The work group tel out to examine the available scientific methods used in the assessment of carcinogenic risk and select for use by the four agencies those currently having the strongest experimental and' theoretical support The work groop explicitly restricted its task to the development ofconcepts and method# for assessing risk, without making any attempt to make statements regarding the appropriate regulatory response for particular types and levels of risk. Recently, the Food Safety end Quality Sen ice of the Department of Agriculture oined the other agencies at a participant in the IRLG. The Report The full text of the report, entitled "Scientific Bases lor Identification of Potential Carcinogens and Estimation of Risks." is set forth in the Appendix lo this notice. The report describes (1) the basis for making a qualitative evaluation of whether a particular substance presents a carcinogenic hazard and how the results of fpHJcmuiiogu'ul studun and .oiim.il Iihmss.iv s. along with othci ty pcs of information, are used in making that evaluation: and 12) the methods that are used in making quantitative estimates of the carcinogenic risk posed bv the substance, if such risk estimates are appropriate of required. It represents the licst judgment of scientists at CPSC. EPA. FDA. anti OSIIA and id the participating senior scientists at NCI and N1EI IS on the scientific principles applicable to identifying and evaluating substances that may pose A risk of cancer lo humans. Scientists at FSQS have reviewed the report and concur. The report is intended lo serve as a valuable scientific reference which may be considered by the agencies, consistent with their statutes and in assuaulHtt with other relevant information, in the evaluation of risk ond as a means uf ascertaining the adequacy of experimental and epidemiological methods used in that evaluation. The identification and evaluation of carcinogens is a fundamental step in any regulatory program. However, each of the agencies publishing this document for comment administers different laws requiring a variety of Findings precedent to regulatory action, it ii not the purpose of this notice and comment procedure to - give the principles in this document the force of law in making any of Ihose required findings. In the event any of the agencies wishes to utilize this document to develop a substantive rule of law. it will initiate appropriate proceedings under its own applicable statutes. The report does not have any regulatory status at this time ether than as a valuable scientific appraisal of scientific principles applicable to identifying and evaluating potential human carcinogens. Accordingly, this notice does not request comments related to the regulatory status of the report Scientific mi fuhlic Review ef the As stated above, the report represents the best lodgments of scientists in the BUG agencies and of the partKipeting senfor scientists at NCI and NIEIIS. In order better to enable the scientific community at large to review and comment on the report the agencies are subjecting the documents not only to review through this Federal Register publication, but also to scientific peer review through publication in the Journalof the National Cancer Institute. The report has been accepted for publication in the Journal and is expected to be published in the near future. The process of scientific peer review will proceed concurrently with this notice and comment proceeding. ' As previously stated, the report was prepared by personnel in three of the agencies publishing this notice--CPSC DUP 0811068 DU 009591 Federal Register / Vol. 44. No. 131SR* iday. July 6,1979 / Notice* * J9659 anmMBanaaBaamaBnwHBw EPA. und FDA--and tiy personnel in OSIIA. NCL and MFJ IS. The participation of FSQS in the (RUG began after Ihe report was prepared. Because OS. (A already has conducted an extensive public proceeding, including a lengthy public hearing, on its proposed rule fur the Identification. Classification, and Regulation of Toxic Substances Posing a Potential Occupational Carcinogenic Risk |42 FR 5414B. October 4. 1!T7| and will stain issue a final rule, only CISC. EPA. FDA. and FSQS ar* participating in this nulice and commenf procedure. Interested persons ire invited to Submit. on or before September 90.1979. written comments regarding the report. The comments trill be reviewed by the four agencies wilft the assistance of members of the Work Group on Risk Assessment of the IELG and scientists at NO and NIEHS. Comments and any accompanying material should be addressed to IRLC. Room 500. lilt 10th Street. N.W. Washington. D.C. 20207. Comments received after the close of the comment period wilt be considered to the extent practicable. Dated; |une 2S. 1*79. For the Container Product Safety Commission: taeoa I. King. Ckeirmaa. For the Environmental Protection Agency; Doagloe M. Costie. Admuutlralar. ^ Far the Food and Drag Administration: Doaold Koaaody,. Commufstorntr. Far *e Food Safety end Quality Sendee: Carol Tochoe Fewme. AossWanf Secretary afAgncultun. Idrodflc Bases far Identification af Poteatial Cardnogans and Eatimatioa of lift Report ofthe Interagency Regulatory Liaitoa Croup. Work Croup on Risk Assessment During the preparation of this document, the Interagency Regulatory Lisbon Croup consisted of four esoefes: the United States Consumer Product Safety Commission (CPSCh the United States Environmental Protection Agency (EPA); the Food and Drug Administration (FDA) of the United States Department of t tealth. Faiucation. and Welfare; and the Occupational Safety and Health Administration (OSHA) of the Untied States Department of Lsbnr. Work Group Musdbsn1 Eula Bingham, BtUG Principal (Assistant Secretary of loibor tor Occupational Safety and I Icalth) ' Vahtahie ewdawee m termini tr.on Arthur C ttyma lUurtw. SMimd (Unus tmiiiute) and Dand r Salt ItWoar. Nalunai In.l.lutr at Enemnismtal Iteailk 5. imcnl The W.ah Uruag aciniwrMe'a Ihr aanUaaea id Falaonl Alton IFrod Jid Und Almoualranaol: Ana Oolia |oseph V. Rodricks, Chairman (Fund and Drug Administration) Elixabelh L Anderson (Environmental Protection Agency) David W. Caylor (Food and Drug Administration. National Center for Toxicological Research) Richard A. Heller (Consumer Product Safety.Commission) Anson M. Keller (Occupational Safaty and Health Administration) Frank Kovcr (Environmental Protection Agency) |oseph McLaughlin (Consumer Product Safety Commission) Additional Participants in the Work Group Roy E. Albert (Environmental Protection Agency) Richard R. Bates (National inatituta of Envionunental Health Sciences) David G. Hoc) (National Institute of Environmental Health Sciences) Umberto Saffiotti (National Cancer Institute) Marvin A. Schneiderman (National Cancer Institute) ABSTRACT--Three types of evidence can be used to identify substances that may pose carcinogenic hazard: these types are designated in Part I of this report as 1) epidemiologic evidence derived from studies of exposed human populations. 2) experiments! evidence derived front long-term bioassays on animals, and 1) supportive or suggestive evidence derived from studies of chemical structure or from short-term or other teats that are known to correlate with carcinogenic activity. Part II delineates the scientific bases for accepting evidence from these three sources and describes their relative contributions to the determination that a aubstanca may poor a carcinogenic hazard. Further, it details the (actors that should be considered in the valuation of experimental and epidemiologic data for ascertaining the reliability and scientific merit ofeach source of evidence. It alse specifies how certain types of limitations in data may inquire qualification of coodustons. Because data on experimental animals are currently the major source of information for assessing carcinogenicity, they receive the greatest emphasis. Features of experimental design and conduct that influence the evaluation of such studies are discussed, as are the criteria for making evaluations. The report is not intended to specify how such studies should be designed and conducted: rather, it discussrs how data from Stevr* tclliaak. Rickard MOL aoS dim So-ofav lEmwwuaraul Pmiecuoa Aamcyt JUnaa looi Donald Clay and Raymond Moftman iConaumrf Undue! Safaty Fro--H.mnt Oaths C Bn--n and |awa Sonia* INaoanal Canto. Inaiifcaet Cad Ortof IOfHer af Some* and Tft hntd*s> t\du y. Eucatiis Offtca af the FisauhiMI. Nathan | Ranh Council an Earmtaai.ataf Qaahw. Rvcmraa rifttea uf ika rtfudattU. aad | Witkaa Utnd tOctuoanonal Safaty aad I Inal* AOM'ttratunl experimental animal studies uf widely varying content and quality should lie evaluated for purposes ef identifying carcinogens. Epidemiologic data and some of their limitations are discussed in lets detail. Chemical structure and die short-term tests that correlate with carcinogenic activity are briefly described, as are their roles in providing suggestive or--if coupled with positive data on animals or humana--supportive evidence of carcinogenicity. In Purl II art presented the-criteria used to ascertain the adequacy of evidence purporting to show that a substance does not pose a risk of cancer. Part I! also includes discussions of some types of experimental evidence that, if the extent and quality are adequate, may be used to show that cartaia carcinogenic responses observed in experimental animals may not be predictive of human response. Part III sets forth current methodologies for quantification of risk. Included are discussions of * aMthematical models available for extrapolation, within a biologic system, f cancer incidence data observed at experimental dose levels to estimate risks at the (usually much lower) levels that arc of concern for humans. Also presented are the factors that should be considered in attempts to identify the human population(i) at risk and to define their conditions and levels of carcinogen exposure. Part HI also deals with correlation of the magnitude of Heels observed in one human population group or m experimental annuals (under their conditions and Wval af exposure) with the magnitude of effects in the human population for which the estimate of risk Is being made. Limitations in currant risk estimation methodologies are described, as are the probtams of ensuring that human risk is net underestimated. The ioaue of Ansholds far carcinogens Is discussed in the final section sf Part 1)L Tbhtsrf Cisterns Awt . The Qualitative Oetormiaotma that a Subotawe Pates a Cardaogtaic Hazard Oefieitloa end Extent ef the Problem Nobue ef Cemaegmusii oad Coctiaogeim. Eetmutna of the Number af CareineiMic Sufaotances Enhancing Factors Variability of Effects of Carcmognts Epidenualosic Evidence Types of Epidemiologic Evidence Disease Ascertainment Evidence From Experimental Animals Criteria (or Evalualioa ef Experimental Design and Conduct Experimental Design Choice ef the Animal Model Numhee ef Animal# Route ef AdminisunUea Identity af the Substance Tseled Deee Levels Age at Treatment Conduct and Duration ef Smstsays m Annuals wwene lor Mraluolloa or Pathology IS!hstagy Exommatiaa Evaluation ef Pathologic Retails DUP 0811069 DU 009592 MOOU reueral Resitter I Vol 4*. V" 131 / Friday July 6. ,979 / Notice* Interim! Consistency of the Data Vcprodueibility of Teat Reaulta Evidence of a Positive Dose-Response Relationship Concordance of Results Evaluation of Tumor Incidence This document will provide a valuable scientific tool to be considered with other information, in the evaluation of risk and ascertainment of the adequacy of experimental and epidemiologic Evaluation of Tumor Morphology General Evaluation of Neoplastic Pathology for Carcinogenesis Bioassays Statistical Analysis of Results Short Term Test for Carcinogens Methods Based on Genetic Alterations Methods Based on Neoplastic Cell methods used in that evaluation. It is an important step in ensuring that the regulatory agencies evaluate carcinogenic risks consistently The IRLG agencies caution, however, that this document presently has no Transformation Evaluation of Short Term Test Results Molecular Structure as Supporting Evidence in Identification of Carctnogena Qualitative Judgmental Factors in Evaluation of Total Evidence Part III. The Quantitative Estimation of regulatory status. Its use will, of course, depend upon the statutory requirements of the individual agencies. The agencies have subjected this document to scientific peer review through the submission of the document Risk to the Journal of the National Cancer Mathematical Models for High-to-Low Dose Institute. In addition, a public notice and Extrapolation Within a Single Biologic comment procedure it initiated by this System The Models Procedures Chat octerization of Population Exposure Sources of Human Exposure Analytical Methods for Detection and publication in the Federal Register. Since the Occupational Safety and Health Administration {OSHAI hat already received extensive public comment on these and other issue* Measurement of Exposures Routes and Conditions of Exposure Duration. Frequency and intensity of Exposure Size and Charactenstica of Exposed Populations regarding the development of its cancer policy rulemaking and will toon promulgate its policy, only the Consumer Product Safety Commission ICPSC). the F.vironmental Protection Extrapolation From Observed Effects to Agency (EPA), the Food and Drug Estimates of Risks for Exposed Administration (FDA), and the Food Population Correlations From Observed Human Population Groups to Others Animal-to-Human Correlations Lack of Predictable Thresholds for an Exposed Population Summary of Risk Estimation References Part I. Introduction Safety and Quality Service (FSQSJ will participate in the public notice and comment procedure on this document At the conclusion of the notice and comment procedure. OSHA will consider whether.revisions to its final cancer policy art appropriate. The four agencies emphasize that the goal of this process is to articulate a consistant This document describes the best policy on the scientific principles Judgments of the scientists in tho applicable to tha identification and agencies comprising the Interagency evaluation of substances that may pose Regulatory Liaison Group (IRLGf on the a carcinogenic risk to humans. scientific concepts and methods Part II discusses the qualitative currently in use to identify and evaluate determination that a substance poses a substances that may pose a risk of carcinogenic hazard. Part IU discusses cancer to humans. These are quantitative estimation of nak. fundamental steps in any program regulating carcinogens. The document was prepared by tne Risk Assessment Work Group of the IRLG agencies and senior scientists from the National Cancer Institute (NCI) and the National Institute of Environmental Health Part 11. The Qualitative Determination That A Substance Poses A Carcinogenic Hazard' The methods used for regulatory purposes in making a qualitative determination that a substance poses a Sciences. The document describes II the basis for qualitative evaluation whether a particular substance presents a carcinogenic hazard to humans are based on a substantial scientific consensus that has emerged from expenence. research, debate, and carcinogenic hazard and how the results review Although some points need of epidemiologic studies and animal further clarification and definition, bioassays, along with other types of substantial agreement exists among the information, are used in making that Federal regulatory agencies on criteria evaluation; and 2) the methods used for for evaluating the carcinogenicity of a quantitative estimates of the . substance. carcinogenic risk posed by the In addition to determining that a substance, if such risk estimates are' substance may pose a hazard of cancer, appropriate or required. regulatory aget&tet must consider other possible health hazards, and tn some instances they are required to balance considerations of nsk with other factors (such as possible health benefits or economic costs and benefits) in reaching regulatory decisions. DEFINITION AND EXTENT OF THE PROBLEM Nature of Cardnogensis and Carcinogenic Responses The characteristic toxicologic event in carcinogenesis is a change in the regulatory mechanism of the target celts, resulting in self replicating cell lesions. The carcinogenic event so modifies the genome and/nr other molecular control mechanisms tn the target cells that these can give nse to a progeny of permanently altered ceils. This progeny of cells constitutes the basts of the neoplastic disease. The expression of tha toxic injury therefore does not derive from the same cells originally hit by tha toxic agent nor from their functional products but rather from the proliferation of a new population of altered cell*. The critical molecular injury caused by specific carcinogens may be quantitatively extremely limited--even to a few cells--and may therefore not be detectable. What will make it manifest, through the subsequent growth of a clinically detectable neoplasm, is the proliferation of the altered cell population. The intensity of the pathologic response in a subject (Le- the growth rate and spread of a cancer) depends on conditions of the host subsequent to the initial carcinogenic event and can be modified by other factors, such as enhancing agents and dietary factors. Tha continued progression of clinical manifestations of the carcinogenic process can occur in the absence of continued exposure to the carcinogen. Carcinogenic effects are therefore seif repltcshMLtoxtc effects different from the common terminal toxic effects in which the manifestations of toxicity are due to altered functional products.detenerativechanges. or death of the target cells themselves/?/. A rigorous methodology must be .followed in obtaining, reviewing, and documenting the data required for a determination of carcinogenicity from observations on humans and experimental studies. Both epidemiologic observations and experimental studies need to be correlated with information on the chemical and physical nature of the agents under consideration, their reactivity* and their fat* in the environment and in the exposed DUP 0811070 DU 009593 Federal Register / Vol. 44. No. 131 / Friday. July 6.1979 / Notices , 39661 organisms. Evidence of carcinogenicity can be obtained from three sources: 1) epidemiologic evidence from eXposedTiuman populations; 2) experimental evidence form long term bioassays in animals: 3) suggestive evidence derived from studies of chemical structure, reactivity, DNA damage and repair, mutagenicity, neoplastic transformation of cells in culture, induction of preneoplastic changes, or l orn other short-term tests that conelate with carcinogenicity. In the evaluation of the results of carcinogenesis studies, the evidence obtained from epidemiologic observations or from experimental bioassays does not necessarily fall sharply into the two categories of positive and negative: In many instances the evidence may be insufficient for a definitive assessment. Estimation of the Number of Carcinogenic Substances .Relatively few chemicals have been found to be carcinogenic. In fact, available evidence indicates that most substances do not cause cancer. The NCTs "Survey of Compounds Which Have Been Tested for Carcinogenic Activity" (2-8) and other literature surveys and reviews provide results of long-term animal bioassays on about 7,000 chemicals. Evidence of carcinogenici:y on the basis of currently accepted experimental testing methods is available for less than 1.000 chemicals and pc.r.-ibly for as few at 000-800, (9-34). Many of these substances were selected for testing because of their structural similarity to known carcinogens. Thus these data considerably overstate the true proportion of carcinogenic substances in the human environment. A critical review of the literature on carcinogenic! ly of chemicals has been undertaken by the International Agency for Research on Cancer (IARC) with the support and collaboration of NCI (9-23) Of 368 chemicals evaluated in volumes. 1-18 of the IARC monographs, some evidence of carcinogenicity was found for 247 (35) A small number of chemicals has been adequately studied by epidemiologic methods to determine whether a carcinogenic hazard exists. By one recent estimate. 20 chemical substances or processes have been idenlifie<Tai responsible for cancer-induction in bumans 19-25. 35)7Of those 20 substances. 0 were first identified as carcinogenic by tests in animals,' whereas 20 were first identified by epidemiologic evidence. Of the 368 substances for which carcinogenesis data were reviewed by the IARC. 221 showed tome evidence of in carcinogenic reactivity of individual carcinogenicity from tests in animals, agents. For example, afiatoxin Bt is but these substances had not received strongly carcinogenic in rats but is adequate epidemiologic study to ineffective in several strains of adult evaluate their effects in humans (35) In mice (53). /J-Naphthylamine is addition. IS occupational categories carcinogenic for humans, dogs, and have been reported to be associated several other species, but this compound with excess cancer incidences without has not produced tumors in rats (54). identification of a specific eliologic With some other carcinogens, there is a agent (36-50) greater concordance of results among Enhancing Factors specie*: Dimethylnilrosamine has been found to be carcinogenic in all of the Experimental and epidemiologic data strains of vertebrates tested (55). auggest that tome agents may not be carcinogenic alone but substantially Species and strain differences in susceptibility to carcinogens may be due contribute to the development of cancer to factors that affect transport and in subjects that have been exposed to metabolism, which in turn determine tbe carcinogens. Depending on experimental effective dose of the ultimate form of the circumstances, these agents have been carcinogen delivered to target cells. referred to as cocarcinogens, promoting agents, syncarcinogens. or more generally, modifying or enhancing These differences may also be due to inherent variations in susceptibility to neoplastic transformation of different factors (51,52). Research on this category of agents auggesta that they may work through a number of mechanism* of action, including (51.52): a) alteration of the organs in different species (55). Differences in the level of carcinogenic effect of individual agents can only be compared with precision wider strictly defined conditions of uptake and/or distribution of carcinogens, b) modification of the dotage and biologic end points. Frequently the level of effects, even metabolic activation of carcinogens, c) . under strictly defined conditions, will enhancement cl the susceptibility of target tissues, and d) acceleration of neoplastic progression. Current evidence suggests that some of these agents act by a mechanism that may be specific for particular organs or show saarked variability depending on (be teat aystem used. Nevertheless, in the extreme, some carcinogens are clearly more effective than others by several orders of magnitude (9-25). However, such comparative potency conditions oi exposure.'Because of the estimates must be made with caution. possible specificity of their mechanisms of actions, the activity of these agents EPIDEMIOLOGIC EVIDENCE may not be recognized by conventional bioassays. Since no common general pathway of action has been recognized, it is not expected that tests based on a single-mechanism end point will be applicable for the identification of a broad range of these substances. Evidence of carcinogenic activity of an agent can be obtained from epidemiologic studies when evaluation. of the observations shows that the test agent causes an increased incidence of neopleaau or a decrease in their latency period Enhancing mechanisms may be a major factor in the development of Evidence from studies of human populations identifies carcinogenic human cancers: therefore, their identification and control may be chemicals to which those populations were exposed In the past Many important in cancer prevention. Since no substances that have been identified as general methodology yet exists for carcinogens in humans were discovered testing and evaluation of this entire by epidemiologic studies of exposed group of substances, the special circumstances under which each may workers; this evidence dates Grom 18thcentury observations of cancer in act must be carefully evaluated, chimney sweeps to more recent interpretation of a positive effect in a observations on dye workers, asbestos carcinogenesis bioassay as being due to workers, and workers in certain one of these mechanisms would require chemicat industries (31). It was noted rigorous documentation that a full early that clinical signs of cancer are carcinogenic process is not involved. delayed for a long time after initial Variability of Effects of Cardaogeas Variability in the action of carcinogens may be due to inherent differences in susceptibility among species and strains of test animals and within populations of humans, and also to variability in the intrinsic differences exposure to carcinogens. This period of latency--often $-40 years from initial exposure until the disease appears-- * makes prompt detection of newly introduced carcinogenic substances by epidemiologic studies nearly impossible. As more substances ere introduced into the human environment and as DUP 0811071 DU 009594 aag.jigal.il ak,:.toit .jamjar., i.:yaajia.L 39862 Federal Register / Vol. 44. Ni i / Friday. July 6. 1979 / Notice! more are tested experimentally, it is expected that a latter proportion will be identified as carcinogenic: this will be ' followed by adequate control measures, so that epidemiologic confirmation may become impossible. Types of Epidemiologic Evidence Types of epidemiologic evidence of carcinogenicity in humans include neoplastic response directly related to duration and dose of exposure, incidence or mortality differences related to occupational exposure, incidence or mortality differences between geographic regions related to environmental rather than genetic differences, altered incidence in migrant populations, time trends in incidence or mortality related to either the introduction or removal of a specific agent from the environment, case* control studies, and the result of retrospective*prospective and prospective studies of the consequences of human exposure. Clinical case reports may. also provide early warning of a potential carcinogen (57). Ti.e tv.o main types of epidemiologic studies used to establish evidence of a carcinogenic hazard are cohort studies and case-control studies (50). Epidemiologic Cohort studies involve the comparison of groups differently exposed to a substance. The comparison may include a) totally unexposed versus exposed groups, b) groups having distinctly different levels of exposure, or c) rates in exposed groups versus rates prevailing in the general population. The groups need to be comparable for demographic factors such as age, sex, and race, and controlled for exposure to . known carcinogens. Epidemiologic case-control studies involve comparison of people with a given cancer type versus people without the disease but otherwise comparable with respect to appropriate demographic variables, to ascertain if they differ In exposure to the cancer hazard under investigation. Epidemiologic findings gain greater force with increasing numbers of wellconducted studied that show similar effects from a given substance under different circumstances. Absence of a positive statistical correlation does not by itself demonstrate absence of a hazard. Whereas negative epidemiologic data usually do not adequately establish the noncarcinogenicity of suspected materials, such negative data obtained for a given agent from epidemiologic studies of sufficient extent and duration may indicate the upper limits for the rate at which a specific type of exposure could affect the incidence and/or., mortality of specific human cancers under the conditions of observation. The detectability of a carcinogenic effect in a group of humans depends on several factors. Including the duration and extent of exposure, size of the exposed population, and background rate of cancer in the target organ. Evaluation of epidemiologic studies requires a knowledge of the smallest possible increase in tumor incidence detectable under the conditions of each study. Such information has rarely been Included in published reports. This information is. however, of critical importance in the evaluation of apparently negative studies. The larger the number of persons in the exposed and control groups and the greater the similarity of these groups for factors other than exposure to the suspect carcinogen, the more likely will an effect be detected. Often, only a small number of humans exposed to a substance can be studied, conditions of exposure are inadequately defined, and records are incomplete. Thus a carcinogenic effect can be easily missed by epidemiologic methods, especially when common types of cancer (such as cancer of the lung, breast colon, or rectum) are studied, inasmuch as these type* often require a large excess of risk before a causal relationship can be identified for the exposure to a particular substance. Substances distributed widely in commerce or in the environment are particularly difficult to study by epidemiologic methods unless high risk ratios are observed, because it is often impossible to identify unexposed groups as controls or to separate groups with high and low exposure. The problem of adequate controls is further compounded by the long latency of cancer, during which multiple opportunities exist for exposure to other potentially carcinogenic substances and modifying factors. The effects of such other exposures on rates of cancer are rarely known, although in some instances they were found to be more than additive (22). Disease Ascertainment Because the effect under consideration is cancer morbidity or mortality, it is important to establish the validity, consistency, and reliability of the methods used to ascertain that neoplastic disease is clinically present or that it causes death. Disease classification is also important, and uniform criteria of tumor nomenclature are needed. Some types of cancer may be classified under a generic name in such a way that changes in -atheir frequency may be missed if only the generic classification is used. Some members of a population may be "lost" to a study if their disease conditions cannot be adequately Ascertained. Specific uniform procedures are not recommended here, but careful attention needs to be given to the extent to which these problems may affect comparison of relevant characteristics between groups. In the statistical evaluation of cancer incidence or mortality differences, there has been a strong tendency for particular confidence levels (e.g.. 95%) and particular probability values (e.g., P--Q.OS or PmILOI) to be used as standard points for a Finding of statistical significance. It is recognized that probability values fall along a continuum and should be so reported. The uniform use of a standard probability value is not suggested. Regulatory needs are best served by accurate estimates of the possible rote of chance in accounting for observed differences. The most important parameter in the assessment of an epidemiologic study is the magnitude of the effect measured: its interpretation is tempered by considerations of biologic plausibility, bias, confounding factors, and chance. EVIDENCE FROM EXPERIMENTAL ANIMALS Evidence of the carcinogenic activity of an agent can be obtained from bioassays in experimental animals showing that the test substance causes either an increase in the Incidence of neoplasms or a decrease in the latency period . The experimental design and conduct should be reviewed for quality and accuracy, and the results should be evaluated statistically for significance, with the only major experimental variable between control end experimental groups being the presence of the test substance. Positive results observed in more than one group of animals or in different laboratories and the demonstration that the occurrence of neoplasms follows a dose-dependent relationship provide additional confirmation of carcinogenicity. Determination that a causal relationship exists between a test treatment and the responses observed in a bioassay is a complex judgmental activity that includes evaluation of the identity of the test agent and the biologic test system, the conditions of exposure, the methods of observation, and the qualitative and quantitative nature of the pathologic response. The assessment of carcinogenicity therefore relies upon the DUP 0811072 DU 009595 3 jstliiaaBiaaifa Federal Register / Vol. 44, No. 131 / Friday, fuly i3,1979 / Notices judgment and experience of professionals. The following discussion refers to aspects of experimental design and conduct that concern evaluation of results. They are not intended as a prescription of protocols. Criteria for Evaluation of Experimental Design and Conduct Experimental Design Commonly recommended requirements for a thorough assessment of carcinogenic potential in experimental animals generally include a] two species of rodents. Z>) both sexes of each, c) adequate controls, d) a number of animals sufficient to provide an adequate resolving power to detect a carcinogenic effect e) treatment and observation extending to most of the lifetime of the animals at a dose range .including one level likely to yield maximum expression of carcinogenic potential. /] detailed pathologic examination, and gJ statistical evaluation of results (9-25, 27,31, 32, 57, 59-73). Positive results obtained in one species only are considered evidence of carcinogenicity. Positive results in more limited tests (e.g.. when the observation period is considerably less than the animal's lifetime), but by experimentally adequate procedures, are acceptable as evidence of carcinogenicity. Negative results, on the other hand, are not considered evidence of lack of a carcinogenic effect, for operational purposes, unless minimum requirements save been met. Choice of the Animal Model The animals used most often for carcinogenesis'bioassays are mice, rats, and hamsters. These animals are used extensively because 1) their natural life spans are short- 2) they are easier to breed and handle in large numbers than larger animals; 3) they are inexpensive and easy to care for; 4) inbred strains .exist that are genetically homogeneous for such traits as "background" cancer rates, susceptibility to carcinogens at specific organ sites, longevity. and response to husbandry systems. Adequately designed and performed studies in other mammalian species may also provide useful information on' carcinogenicity. For human risk evaluation, data obtained from bioassays with the use of nonmammalian species can presently provide only suggestive evidence if positive but permit no conclusion if negative. Experience on the background incidence of tumors in the colony of animals used for testing, obtained over a to be effectively considered for the period of years by extensive observation evaluation of carcinogenesis test results of untreated animals under the same is the namber in which detection of general maintenance conditions carcinogenic effects could be expected. (historical colony controls), is useful in This number is obtained by subtracting assessing the relevance, of experimental from the number of animals started on findings, such as the appearance of rare the test the number of those lost to tumors. Rodents with different types of adequate observation (e.g., by intercurrent death followed by genetic homogeneity have been used for cannibalism or autolysis). The number carcinogenesis bioassays. These include of animals on which complete a) inbred strains, b) first-generation pathologic examination is conducted is hybrids of parents of inbred strains, c) important in the evaluation of tumor randombred animals from a closed . pathology. colony, d) noninbred animals, and e) Positive results ean be obtained in animals of unspecified strains or origins. tests with the use of a small number of As the genetic and/or environmental animals if the test ia otherwise variation increases, so does the need for adequately designed and conducted and concern about the variation of if the tumor response is significant For background tumor incidence. example, in a group of 15 animals, if 12 A particular problem is posed by the show a well-defined neoplastic lesion of use of certain strains of rodents in a kind rarely seen either in matched or which particular tumor types reach a historical controls, the finding is high frequency, often well above 50%, in positive. However, a negative finding in untreated controls. Examples of such a group of 15 animals is not adequate strains include mice of strain A for lung evidence that the test agent is not adenomas, strain AKR for lymphomas, carcinogenic. strain C3H/HcN males for liver cell Ideally, the number of animals tumors and C311 females for mammary required to provide adequate negative tumors, and females of several rat evidence would be such that an strains for mammary fibroadenomas. excessive risk would not arise if the test Although viral factors have been identified in the etiology of mouse AKR leukemia and C3H mammary tumors, no failed to detect carcinogenicity. The likelihood that such a risk would not arise increases both with the.number of such factors are known to be at work for the other types mentioned above. The effect of carcinogens has been clearly demonstrated in all of the above strains animals on test and the extent to which human exposure levels are exceeded. The probability of a false negative finding also depends on the background by detection of substantial decreases in the latency period, by definite increases in incidence or multiplicity of these tumor types, and by the induction of tumors of other histologic types in the same or other organs (2-25). Caution must be used, however, in evaluating the significance of a higher incidence of these tumors in a treated group compared with concurrent controls when the incidence in the treated animals falls within a range commonly seen in historical controls from the same colony. Background incidence rates for tumors of the lung, liver, mammary gland, and hematopoietic tissues are much lower in many other strains of mice, and for tumors of the mammary gland in other strains of rats, in these other strains, no unique biologic trait distinguishes the types of tumors mentioned above from many others, and no reason has been demonstrated for considering that they have any different significance than tumors in other organs as indicators of a carcinogenic response, under otherwise appropriate test conditions. tumor rate in the control animals. For example, if a one-sided level of statistical significance of 5% is used with 55 animals, there is an 50% chance of detecting a tumor rate of 20% in the treated animals for whom the control rate is 5%, whereas 190 animals are required to detact the same difference if the control rate is 90%. The number of animals tested may need to be increased if the nuaaber of humans exposed is large or if a small margin of safety exists between the animal dose and the human . exposure. in practice, resource limitations often require a trade-off between the number of animals used and the number of substances testedIn order to control the total cancer burden resulting from chemical carcinogens. This is particularly true with substances whose toxicity limits the test dose to a low multiplicity of human exposure levels. In those instances, it may be necessary to accept a lower than ideal degree of "negative evidence." Route ofAdministration Number ofAnimals The number of animals in each group A key factor in the comparison of an experimental result to the human 'situation is to assess whether cells DU 009596 DUP 0811073 39864 Federal Register / Vol.44. 131 / Friday. |uly6. .979 / Notice* capable of mhlignant transformation are exposed to the reactive carcinogenic Hgent(s) in both the human and the experimental animal, regardless of whether transformation occurs in identical organs and cell types. Although this comparison is most readily made from experiments with animals in which the route of adm uiitralion is the same as that in humans, other routes of administration may also be comparable and provide results useful for evaluation of the human hazard. For example, some chemicals are rapidly absorbed by inhalation, circulated through the body, and metabolised by the same pathways that occur following intravenous exposure [74). Some routes of administration in animals may fail to provide adequate metabolic activation or exposure of target tissues and therefore may lead to false-negative results. This possibility should be assessed in evaluating negative results obtained when the route of administration in animals differs from the route of human exposure. Generally, the route should be one that leads to absorption and distribution of the test substance. The induction of tumors at a remote site in the animal is evidence of absorption, distribution, and possible metabolic activation of the test substance. If exposures of both humans and animals involve absorption of the substance, any route of administration in animals may be regarded as relevant for a qualitative demonstration of human hazard unless there is evidence that the route of administration in the test species results in the production of carcinogenic substances (from degradation ar metabolism) which does not ever occur with human exposure. When tuaMtl appear only at the site of injection UT implantation, careful review is iwCMtary. If there is reason to believe that An tumors occur as a result of "solid state" carcinogenesis (75, 76). the results may be inappropriate for extrapolation to human exposure. If, however, the test material produces tumors at the site of injection or implantation as a result of its chemical reactivity, thia response is an indication of carcinogenicity. There are a number of practical reasons for studying certain substances in animals by a route of administration different from the expected route of human exposure. If a substance under test is highly volatile, accurate administration in food may be difficult because of evaporation; often feeding through a stomach tube is used so that the dose may be measured with greater accuracy. Even for nonvolatile test substances, a stomach tube may be used when it is important to know the exact amount of a substance administered to the test animals. The administration of high doses of a test substance with a disagreeable odor or taste may require the use of routes other than ingestion. Thus experimental exposures need not necessarily be by the route of human exposure in order to be meaningful, but possible physiologic and metabolic differences related to routes of absorption and distribution should be considered in the assessment of their relevance. Identity of the Substance Tested Substances to which humans are exposed through their occupations, the environment, aod the products they use vary widely both in the number and the proportion of contaminating impurities. A full assessment of the carcinogenicity of an impure mixture ideally requins that each component be tested individually at an adequata dosage and that the mixture itself be tested in order to detect cumulative or synergistic effects. Limitation of resources makes this ideal approach impractical as a routine. It is common, thereforo. simply to rely on tests either of the product to which humans ere exposed, including the impurities present, or of the purified principal chemical aubstance(s). Because the products may vary according to procedures used in manufacture and processing, tests for one commercial product may not bt applicable to another product containing a different set or level of impurities. Change in the manufacturing process of a product may require additional tests to confirm the safety of the new product if the change involves the introduction of different impurities or e substantial increase in the amount of any single component of the product Even though it is accepted practice to test mixtures, the nature of any impurities known or likely to be present as a result of the manufacturing process Is Important and may require separate examination or testing. Information on ths carcinogenicity of any single chemical in a mixture is an indication of potential hazard of the entire mixture. However, negative results obtained on a component of a mixture may not reflect the potential carcinogenicity of the entire mixture. Dose Levek `Testing should be done at doses and under experimental conditions likely to yield maximum tumor incidence." Ibis recommendation of an FDA advisory committee summarizes the issue of lest doses (65). Bioassays wl& the use of a few dozen or even a few hundred animals have relatively low sensitivity for detection of carcinogenic effects. Millions of people of varying degrees of sensitivity or exposure may be exposed to the substances under evaluation. Although a test animal cannot be strictly viewed as a "surrogate" of a large number of people without oversimplification, the role of animal tests is to provide maximum detectability of carcinogenic effects within the already narrow confines of test sensitivity. Under otherwise identical conditions, the greater the ratio of test exposure to human exposure, the greater is the safety margin provided by a negative result in a carcinogenesis bioassay. it is generally recommended that more than one dose level be tested. Most carcinogenic effects show a positive dose-response relationship, but maximum tumor incidence in test animals may not occur at the highest dote when competing toxicity, prevails. The highest test dose that can be effectively used in a carcinogenesis bioassay is limited by the conditions of . absorption, by the amount that ths animal can tolerate during lifetime administration without unwanted toxic tide effects, and by the effects on nutrition when the chemical constitutes too large a proportion of the diet Results of bioassays done at doses and under conditions permitting maximum txpression of carcinogenicity provide a sound basis for the identification of a carcinogenic hazard or its absence. It is important to estimate the highest dose level that will bs tolerated by the test animals during lifetime admistration, i.e., the estimated maximum tolerated dose (EMTD). The EMTD is defined as the highest dose that can be administered to the test animals for their lifetime and that is estimated not to produce a) clinical signs of toxicity or pathologic lesions other than those related to a neoplastic response, but which may interfere with the neoplastic response: b) alteration of the normal longevity of the animals from toxic effects other than carcinogenesis; and c) more than a relatively small percent inhibition of normal weight gain (not to exceed 10%) [71). The EMTD is determined on the basis of prechronic tests and other relevant information. If the test reveals that the EMTD is too high to meet the conditions defined herein, positive results obtained above the EMTD are acceptable as evidence of carcinogenicity unless there is convincing evidence to the contrary. Altematively..negative results obtained DUP 081107 DU 009597 --Ssrc. Federal Remitter / Vol. 44, No. 131 / Friday. July 6.1979 / Notice! i n mibjun.j.ii^niitinii--'nn--eMWlBHM^SB--ai----nan 33865 above the EMTD are considered inadequate unleu particularly strong and specific scientific reasons justify their acceptance as negative. Positive results obtained at or below the EMTD provide evidence of carcinogenicity. Age at Treatment Because of the long latency period required for induction and manifestation of tumors, treatment should be started in young cnimals, and the animals should be observed for a carcinogenic response through most of their expected life spans. The older the age at first treatment, the shorter is the remaining life-span available for tumor development: consequently, the smaller is the chance of detecting delayed carcinogenic effects. Although treatment is often started in young adult animals soon after weaning, some protocols call for treatment soon after birth (neonatal) or during fetal development (transplacental). The rationale for exposing test animals transplacentally or neonatally is based on the greater susceptibility of certain organs to carcinogens during early development. Such susceptibility has been demonsirated in several species, including those commonly used for tiioassays [77. 78). Animals first treated during the perinatal period must be also treated and observed throughout their life-spans to obtain a valid negative response. Virtually any agent that is carcinogenic in adult animals can be expected to have tome carcinogenic effect when administered to young animals including the neonate and the fetus. Unless a substance is demonstrated to be exclusively carcinogenic when administered to the fetus or neonate, enhanced perinatal susceptibility to carcinogens should be confide:ed not a separate and distinct toxicologic property; rather, it should be a means for increasing the sensitivity of conventional bioassay procedures by extension of the exposure period tothese earlier and more susceptible portions of the life-span. It should be emphasized that these protocol modifications greatly complicate dose selection and experimental design. An agent may be significantly more toxic to the fetus, the neonate, or the pregnant or lactating female animal than to the normal young adult of cither sex. This requires independent determination of the toxicity and EMTD. Furthermore. . Individuals in the litter of a treated pregnant animal cannot be considered independent units for statistical evaluation of effects. Conduct and Duration of Bioassays in Animals A long-term bioassay for carcinogenesis in animals is a complex procedure requiring control of many variables for several years. Professional experience and knowledge of the relevant biologic parameters are needed for adequate quality control. Detailed guidance on procedures is provided by reports such as the FDA's "Good Laboratory P; actice Regulations'* (9) and the NCI's "Guidelines for Carcinogen Bioassays in Small Rodents'* (71). Review at the observations made during the bioassay (on food intake, weight, clinical course, and pathologic conditions of the animals) provides a basis for determining whether these experimental variables are recorded in sufficient detail and are internally consistent to permit independent assessment of their validity. The purpose of these bioassays is primarily to provide maximal opportunity for detection of a neoplastic response; therefore, the longer the period of observation the better is the chance of detecting delayed effects. A "point of diminishing return" can be reached when Intercurrent disease and/ or survival considerations make the observation or evaluation of old animals particularly difficult. It is expected that the animals will be observed for most of their life-spans. The best negative evidence for the carcinogenicity of a substance is obtained from tests in which both exposure and observation last through all or nearly all of the expected life-spans of the animals under study. Negative results decrease in value as the exposure and observation periods are shortened, and they become practically meaningless if these periods are shorter than half the life-spans of the animals. When some animals die early in the course of a test, the value of the test is reduced as a function of the percentage of animals dying without tumors at periods markedly shorter than the life-span of the species. Sometimes, a positive carcinogenic response may be definitely demonstrated in a shorter period of observation if the experiment is adequately controlled; in such cases the test is considered valid even if it is shorter than usual (40). Accepted procedures include o) the observation of all animals in the study (treated and control groups) until their spontaneous death, b\ the sacrifice of animals that show clinical signs of severe illness or impending death (sacrifice of moribund animals prevents losses due to autolysis and provides better observation of tissue pathology), and c) terminal sacrifice at a scheduled date near the end of the life-span (e.g., after 24 months on test). Criteria for Evaluation of Pathology Pathology Examination The evaluation of carcinogenesis bioassay results rests on the extent and accuracy with which organs and tissues of both treated and control animals are examined for morphologic changes. After the termination of a bioassay, the only physical evidence that can be used to permit reevaluation of results, even years afterwards, is represented by the written descriptive and diagnostic records, the graphic or photographic records of grow or microscopic observations, and most importantly, the original slides of tissue sections for microscopic examination. The histologic slides are of critical importance as a lasting direct documentation of the conditions of normal and abnormal tissues and organs, both for scientific and regulatory purposes. Quality and extent of pathologic documentation are therefore major factors in establishing the validity of bioassays in animats (71, 73). Although a well-conducted pathologic examination cannot generally rescue a poorly designed or badly conducted bioassay, inadequate pathologic examination can significantly reduce or eliminate the value of an otherwise well-conducted experiment. Among the factors to be considered in evaluation of the pathologic examination are: 1) the care and thoroughness, of gross tissue examination and the " qualifications of the persons conducting this examination to recognize abnormalities; 2) the quality of preservation, sectioning, and staining of tissues; 3) the accuracy of the record-keeping system used for labeling tissues as they -are moved from the animal through slide-processing to final diagnosis and reporting: 4) the extent of selection of normal and abnormal tissues for microscopic examination; and 5) the qualifications of the pathologist making the microscopic examination. The numbers of tumors or other lesions diagnosed by the pathologist are not a thorough assessment of incidence unless each factor is adequately considered, controlled, and documented. The strength of evidence provided by a bioassay also depends on the number of tissues examined. Failure to observe excess tumors in treated animals cannot be considered evidence of the absence of a carcinogenic hazard unless all DUP 0811075 DU 009598 . nKuii i *01. *H. WO.131 I Friday. ]uly 6.1979 / Notice* 11111 a, meraM--------a--a--iMs/wnamur mmmmmm----------m----mag--a organs have been examined grossly and all grussly4rUible suspect lesions have been examined microscopically. In s' large organ, the taking of a single^ random section for histologic examination can result in failure tc detect small tumors. Thus multiple cuts through such orgnns should be made. It is also important to open and search the entire cavity of all hollow organs for abnormalities. For example, the entire length of the gastrointestinal tract should be opened and inspected. Crossly visible lesions should be selected for histologic examination, and if they are not subsequently observed on tissue slides, preparation of additional sections may be necessary until the gross lesion is verified histologically. Furthermore, histopathologic examination should be made of major organs in the treated groups and matched controls, and specific organs should be studied in detail in all dose groups and controls in which there is either gross or microscopic evidence of lesions. Major organs are defined in the NCI's "Guidelines for Carcinogen Bicassays *n Small Rodents" (77). Positive evidence of carcinogenicity may be valid for a particular organ if it has been adequately examined in both treated and control groups. Negative reports are inadequate for any organ that has not received careful gross examination in all animals and histologic examination of suspect lesions. The mote limited the number of organs examined grossly and microscopically, the less the value of the c.-.prriment in providing evidence of a negative result. Evaluation ofPathologic Results The evaluation of bioissay results and their quality requires a detailed review and expert judgment of all the experimental conditions and observations, including the identity of the test substance; the conditions of administration; the identity, source, and characteristics of the test animals; the accuracy and systematic recording of observations: the extent of pathologic examination; and the competence of the investigator. Meticulous and detailed documentation is of great importance. Several criteria are applied in the evaluation of bioassay results. t) Internal consistency of the data is important in reviewing the conduct of the test. Apparent inconsistencies should be investigated by analysis of records. 2) Hepraducibiiity of test results can . be demonstrated within a single experiment (in different groups of - milady treated animals or in different dose-level groups) or in separata bioassays conducted with the same experimental design in the same or in different lultoralories. Evidence of reproducibility adds greater confidence to the evaluation of results. Statistical considerations provide an estimate of the level of detectability of an effect and the consequent level of probability that the effect may be missed in a repetition of the test in a given number of animals. Apparent contrary results in any two tests may be simply an effect of chance variation and may be fully compatible with an identical mechanism and level of activity of the test compound. 3) Evidence of a positive doeeresponse relationship adds further confidence to the evaluation of a positive test, but lack of U may be due to testing in a portion of the dose-reponse curve with a shallow slope or even with a declining slope due to competing risks. In the presence of positive results in * well-designed, well-conducted teats, evidence of reproducibility and positive dose-response relationships it not necessary to reach a conclusion of carcinogenicity. 4) Concordance ofresults obtained under differing test conditions (e.g* different species, different routes of . administration, or markedly different . basal diets) provides greater confidence in the evaluation of both positive and negative studies, but it has a different meaning from "reproducibility" within the same tests or under the same conditions. Lack of concordance from tests performed under different conditions does not in itself, detract from the validity of the positive test Reasons for a discordance in observation may be identified by evidence obtained during a test or may be sought through further research. The response to carcinogens in different animal apedea and even strains is known tc vary greatly because of genetic, metabolic, nutritional, and other factors that effect susceptibility in a given test animal. Present knowledge indicates that a substance that is clearly carcinogenic in one test species is likely to be carcinogenic in other species, (hat it may take extensive tests in several species to demonstrate this correlation, and that the responsive target tissues or organs and the types of tumors induced in different species may vary greatly. Therefore, although concordance of positive results (even if different tumor types nrc involved) adds support to an evaluation of carcinogenicity, the finding of negative results in tome other species generally does not detract from the validity of a positive result as evidence of carcinogenicity for the test substance. In this respect positive results supersede negative ones. The assessment of such apparent discrepancies in results requires consideration of all experimental variables, since apparently negative results may derive from limitations in the sensitivity of the test (e.g.. early scheduled sacrifice, limited extent of pathologic examination, and atatisticai probability). If the positive result is itself not fully conclusive or if reasons exist for questioning its validity as evidence of carcinogenicity, the result is generally classified as "inconclusive" or "only suggestive" even in the absence of other negative test results. S) Evaluation of iuator incidence is made on the basis of the pathologic findings end therefore depends on professional diagnostic judgment Tumor incidence is evaluated by consideration of ail tumors of specific organ sites or anatomically or physiologically related systems. At present there is considerable uncertainty about the interpretation of carcinogenic responses in terms of the total tumor yield in contrast to the.response in terms of a statistically significant increase of tumors in specific target organs or tissues.- Traditionally, carcinogens have been recognized in studies on humans and animals by a decisive increase in tumors of target organs. However, it is conceivable that a genera! increase in total tumor yield, in the absence of an excess incidence in one or more target tissues, could occur--for example, by a promoting effect that generally increases the spontaneous incidence of tumors in test animals or by the action of a multipotent carcinogen whose response did not reach statistical significance in any one organ even at the maximum tolerated dose. In some instances, however, control animals may have a high frequency of tumors at certain sites (e.g testicular tumors in F344 male rats), in such instances, a simple cumulative count of tumor-bearing versus tumor-free animals may fail to reveal carcinogenic effects in the treated groups. Prudent judgment is needed on the appropriate categorization of tumors used to evaluate induced effects. A positive result in a carcinogenesis bioassay can be based on evidence of the induction of an increased incidence or a substantially decreased latency period. The latter is often difficult to establish. Determination of the latency period can be made by various ' techniques of observation during a bioassay. If both test and control animals are sacrificed at a fixed time, only the early part of a temporal distribution curve may be observable; consequently, the estimate of the DUP 0811076 DU 009599 Federal H lister / avenge latency period for all tumors or tumor-bearing animals may be artificially altered, tf the test and control groups are allowed to live out their life spans. the comparison of latency periods must take into account the relative survival and the number of animals at risk, particularly in the case of competing risks. The methods used in estimating the latency period must be defined in the context of each bioassay, it is always difficult to determine the exact onset of a neoplasm. Morphometric criteria may be used for tumors (e.g.. skin or subcutaneous tumors) detectable during clinical observation of the animals and a minimum size may be established as a criterion for identification. For neoplasms of the internal organs it is practically impossible to determine an adequate time of onset Methods such as palpation of the abdomen are highly subjective and generally unreliable. Serial sacrifice studies provide excellent data on time to tumor induction, but they should not be substituted for adequate numbers of animals under lifetime observation. In most instances, what is referred to as latency period is the time between the beginning of the exposure and the observation of a tumor at death. This parameter is obviously influenced by all the factors that determine time of death, e.g, intercurrent diseases, other tumors, or .growth rate of individual tumors. Here too, the judgment of experienced pathologists may provide critical evaluation of such aspects as tumor sise, location, cell differentiation, and invasion: these factors may pontribute to an estimate of temporal sequence. The observation in treated groups of tumors that are considered rare in antreated and historical controls may raise considerable sispicion even when their incidence is below the required level of statistical significance. Careful review and cautious judgment are necessary in their evaluation; often the rarity of a tumor type is estimated on the basis of a small Control population. The occurt ence of one or a few neoplasms of a kind, however rare, is not necessarily evidence that a substance is carcinogenic in the absence of other supporting evidence. 6) Evaluation of tumor morphology in the final analysis of bioassay results is highly dependent on the way in which pathology data are categorized. It is incorrect, for example, to subdivide diagnoses into so many individual .categories based on different stages of disease different morphologic features that no single category is large enough to be statistically significant. At the other extreme, it Is incorrect to group unrelated end points in a way that maximizes the opportunity to find statistical significance, whether or not such groupings are biologically meaningful. Carcinogenic and chronic toxic effects of a chemical on an organ, tissue, or cell develop through a series of stages from minimal changes to advanced and possibly fatal end points (5/). The stage reached at any particular time is related to the dose of the substance, the conditions of exposure, the time elapsed since beginning of exposure, and host susceptibility factors. Early lesions that are pathognomonic of a disease process resulting from toxic chemicals should be grouped with more advanced lesions, whether or not the animal has survived long enough for the process to develop to the latest stages. The carcinogenic process may go through early stages Including atypical hyperplasia, carcinoma in situ, and/or historically benign tumor before progressing to a clearly malignant stage. Although the stage of development is of critical importance in clinical oncology for assessing the prognosis of a patient at the time of therapy, it is not relevant in deciding whether a chemical is capable of inducing cancer as long as the induction of lesions recognized as neoplastic is conclusively demonstrated. The induction of preneoplastic lesions in the process of cancer development is an indication that the test substance is capable of inducing cancer in a susceptible host given sufficient exposure and time for cancer to arise. Care must be taken, however, to distinguish atypical hyperplasias that are pathognomonic of neoplastic progression from other nonspecific or reactive hyperplasias. In the evaluation of bioassays, the concern is with the capability of a test substance to react with a biologic system to give rise to a neoplastic response that may develop through all stages to malignancy. One issue is whether or not the response Is the kind that stops at the benign stage and never evolves further to the invasive and metastasizing stage. Few if any tumor types are presently known to belong to this category, which could be called "permanently benign" tumors. For benign tumors, no specific mechanism of induction is known that can be distinguished from the mechanisms of induction of other neoplasms. Moreover, no established body of evidence exists showing that certain substances or groups of substances are capable of inducing exclusively permanently benign tumors without ever inducing any more malignant ones. The mammary fibroadenoma is generally considered to be a benign tumor In both the human (2) and the rat (3), and it has been suggested that ita experimental induction provides little evidence that the inducing agent can cause cancer. X- rays or carcinogenic polycyclic aromatic hydrocarbons, however, which principally induce fibroadenomas in some rat strains, induce mostly malignant adenocarcinomas in other strains; the genetic characteristics of the animal rather than the inducing agent determine whether benign or malignant tumors develop (M). Thus the induction of benign tumors, even of a type that rarely progresses to a malignant stage, must be considered a warning that the inducing chemical may be capable of causing cancer in some humans. The induction of benign neoplasms, even if they were demonstrated to be of a permanently benign type, would therefore be considered evidence of carcinogenic activity unless definitive evidence is provided that the test chemical is incapable of inducing malignant neoplasms.' Neoplasms at a benign stage may jeopardize the health and life of the host Furthermore, it is extremely difficult to rule out the presence of malignant changes simply on the basis of a limited histologic examination of the primary tumor, because focal malignant change or local invasion may have occurred in other areas of the tumor that were not examined microscopically. Similarly, it is very difficult to tula out the metastatic spread of a neoplasm that may be biologically capable of aietastasizing without an extremely detailed search for metastases, which can begin as small foci of one or a few cells lodged in the arteriolar walls of peripheral organs (5). The frequency of observation of such BMtastases depends directly on the amount of peripheral tissue that is examined (6). Another case to be considered is the combination of neoplasms diagnosed as benign and malignant. This may include instances in which the incidence of histologically malignant tumors is only a relatively small fraction of the total tumor incidence but represents the most advanced stages of the neoplastic response. Although thd number of tumors diagnosed as malignant may not reach statistical significance as such in the number of animals at risk, the total neoplastic response (benign and malignant) may ba clearly significant. Some common types of neoplasms found in carcinogenesis bioassays in laboratory rodents arv among those DUP 0811077 DU 009600 UJitUO federal Register / Vol. 44. No. 13t / FrUl .y 6. 1979 I Notice* often diagnosed a* being nt a benign o) The induction of sarcomas around stage When observed in lest animals. a "solid state" implant of the test Examples include lung adenomas, skin substance into a connective tissue is not and bladder papillomas, liver cell considered an indication of the adenomas (hepatomas), and carcinogenicity of that substance when hemangiomas in various organs. All of it is administered in another physical these tumor types are known to progress ) form (7 76). to frank malignant stages. No b) The induction of a carcinogenic pathogenetic mechanisms have been response by asbestos and other fibrous identif cd that could demonstrate that materials by a mechanism linked to the induction of such tumors, whether in certain physical characteristics such as a benign or malignant stage, in fiber length and diameter is recognized r'herwise appropriate, comparable, and as a basis for categorizing the exposure well-controlled experimental conditions, to such fibrous materials as a provides any different kind of evidence carcinogenic hazard [22). for carcinogenesis than the induction of c) The effect of particulate materials other tumor types. In the evaluation of in the induction of respiratory tumor incidence, therefore, neoplasms in neoplasms, when they are administered different stages of progression are jointly with certain carcinogens counted together. (probably through their capacity to 7) General evaluation of neoplastic pathoi- >gy for carcinogenesis bioassays absorb and retain carcinogens, to penetrate the respiratory tract tissues, includes consideration of the total and to stimulate early cellular number of animals with tumors in each responses) is not recognized as evidence group, the total number of individual tumors, and the index of tumor of carcinogenicity of these substances but rather as an indication of their rola multiplicity in tumor-bearing animals. as cofactors in earcinogensis, particulate The tumor response can be further characterized by a detailed observation materials require careful but separate consideration as a potential hazard (Ml of the tumor morphology and related 90). preneoplastic changes. The extent of d) The induction of a neoplastic tumor growth and spread and special morphologic characteristics may give useful indications of the time of response by a substance because of its radioactivity is recognized as a cancer hazard. development of the neoplastic response. Other factors are sometimes The quality of the pathologic response is determined by a comprehensive evaluation of all the pathologic changes observed in both treated and control animals. Special attention is required in suggested to be sufficient to refute the presumption of positive evidence of a carcinogenic effect These factors must be critically examined to avoid falsenegative judgments based on the evaluation of toxic effects other than unsubstantiated hypothetical carcinogenicity, because their pathologic explanations of the circumstances of manifestations have to be distinguished from.those due to die neoplastic response. The organs and times that are the targets of carcinogens may vary greatly in different species and even under different exposure conditions; therefore, no direct analogy ef morphologic response can be expected from a carcinogen in animals of different species and in humans. Examples are known both of widely different target sites [e.g., benzidine induces bladder carcinoma in humans and cholangiomas and liver cell carcinomas in hamsters and rats (67)) and of similar responses (e.g., vinyl chloride induces the same type of angiosarcomas of the liver in humans, rats, and mice (0)]. tumor induction. The following factors are considered in this respect: a) Indirect mechanism* of action requiring special exposure level* or conditions. An example has been suggested in the case of substances that may induce bladder neoplasms only in the presence of bladder stones resulting from high levels of intake and urinary excretion of the test substance (PJ). Support for such a mechanism as an explanation for development of bladder tumors is provided by determination of a specific association of tumors with stones, a dose-response correlation between stones and tumors, and the absence of other chemical or biological indications that the substance might be carcinogenic by other mechanisms. In Special conditions of tissue exposure evaluation of the relevance of such or reaction may result in a tumor experimental observations to the response by mechanisms that appear assessment of human hazard, special due to physical rather than chemical consideration is needed for mechanisms properties of the test material. The. by which exposures or intercurrent following conditions are evaluated diseases in the human may act as the differently in this respect: cofactor (e.g., in bladder stone induction), thus producing a susceptible state for the possible carcinogenic activity of the test substance. b) The action of promoting agents only on tissues previously initiated by carcinogens [St, 52). Few examples are well documented, such as the phorbol esters in epidermal carcinogenesis in mice. Criteria of risk evaluation need to be defined and dose-response relationships considered. Any claim that a substance acts only by .this mechanism and thus is ofless concern to humans needs to be supported by experiments showing the mechanism of action and demonstrating that the effect does not occur at human exposure levels. c) Metabolic pathways of carcinogen activation [92) which are suggested as occurring exclusively under certain test conditions in experimental animals but not under other test conditions or in other species. This situation would be important if thorough studies demonstrate that the metabolic pathways for carcinogenic activation of a substance in animals do not occur in humans. Another important situation would be the demonstration that the metabolic pathways of activation of a particular carcinogen identified by studies at high levels of exposure are exclusively formed at such high levels but are absent at lower dose levels. Statistical Analysis of Results Statistical hypothesis testing provides an estimate of the likelihood that an experimental observation may or may not be a result of chance alone. The 95% confidence level is widely accepted as a reasonable assurance that the observed effect is real but confidence that an * increased incidence of tumors is a true indication of the carcinogenicity of a substance increases with increasing statistical significance of the results. ~ Thus the level of statistical significance should be reported rather than the fact that a result is statistically significant or not significant at a single preassigned level of confidence. Failure to detect an increase of tumors in a bioassay may be due to an insufficient number of animals tested and does not unequivocally prove that a substance does not pose a risk of cancer. Tumors rarely seen in experimental animals may raise considerable suspicion even if the statistical significance is well below the 95% confidence leveL Because of the frequent use in chronic studies of both sexes, more than one species or strain, and more than one dosage level, and because many different tissues arc examined, a large DUP 0811078 DU 009601 Federal I 'ister / Vol. 44. No. 131 / Friday. July 6.1979 ! Noticee .-Ml-. 39869 IMini'MlMn-- in munber of statisticalcofflparisons are possible between control and treated animals. Thus the results from a chronic study must be interpreted cautiously to control the rate of false positives arising from the large number of possible statistical comparisons (93). Lifetime animal experiments are often difficult to interpret because of competing causes of death, which may alter the of the observation period of the tumor type under study. A common but inadequate form of presenting tumor data is a report only of the proportion of animals in whi :h particular tumor tynes were observed during the study. This proportion may contain a mixture of three types of observations: 1) The tumor causes the death of an animal and is subsequently observed upon necropsy; 2) an animal i dies due to some cause other than a particular tumor and the tumor is observed upon necropsy; or 3} the tumor is observed when an animal is - necropsied at the time of a scheduled sacrifice, generally at the termination of an experiment. Simply combining tumors observed under these three - situations makes interpretation difficult, and in fact the data may be misleading if the mortality pattern is altered by the toxicity of the substance. Serial or terminal sacrifices provide an opportunity to compare the prevalence of tumors in various groups of animals unperturbed by mortality. Hov.-v\ or, sacrifice data do not provide an opportunity to study the effect of a substance on survival or on causes of death. . The analysis of a bioassay is limited by the quantity and quality of data. Such studies must include the age of " each animal at the beaming of the experiment, its age at time of removal from the experiment, reason for removal (death, moribund condition, scheduled sacrifice, or others), and all clinical and pathologic observations, including gross and microscopic examination. When survival curves of control and treated animals differ due to competing causes of death, adjustment of the number of animals at risk may be necessary. For a tumor type generally leading to the death of an animal statistical analyses of survival experiments should incorporate life* table or competing risk techniques in order to estimate and test tumor incidence. This approach requires assumptions concerning the independence of the competing causes of death, if all the animals are utilized from a survival study, including sacrificed animals, the net probability of death due to a tumor type can be estimated as though that were the only cause of death of a group of animals. Statistical tests for differences between control and treated groups can be performed on the adjusted tumor incidence rates (94-96). For a tumor type that is unlikely to kill the animals, methods of analysis based on life-table techniques are not appropriate for adjusting the number of animals at risk. These tumors are observed conditionally as a result of other events occurring first: death of the animal or a scheduled sacrifice. To estimate the prevalence rate of these tumors, mortality is assumed to be unrelated to the presence of the tumor. Statistical methods for the analysis of tumors that are not generally lifethreatening are discussed by Hoel and Walburg (94) and Peto (95). SHORT-TERM TESTS FOR CARCINOGENS Carcinogenesis tests have traditionally been based on the experimental induction of tumors in laboratory animals. Such tests usually involve the observation of treated animals for most of their life-spans. Recently, short-term methods have been developed to provide more rapid markers for the tentative identification of carcinogenic effects. These methods are directed toward the atudy of mechanisms underlying neoplastic transformation as well as toward provision of reproducible and rapid methods for testing chemicals and physical agents for potential carcinogenic activity. The use of short term methods for the evaluation of carcinogens was the subject of a recent review (97) from which die following discussion is largely derived. Methods Based on Genetic Alterations The analysis of mutagenic effects has been developed mainly to assess the ability of a substance to induce genetic alterations. The resulting information can be used for estimating the genetic hazard of chemical agents for man. Because of the similarities ofbasic molecular mechanisms by which chemical mutagens and most chemical carcinogens appear to induce genetic effects (i.e_ molecular alterations of DNA), it has been postulated that mutagenic effects can be used to predict carrihogcnicity. The use of a battery of short-term genetic tests is usually recommended in order to minimize false-negative and false-positive results and to select compounds that require.further long term investigations. This battery of testa may include: ) tests for nutations in'bacteria and eukaryotic microorganisms; ) tests for mutations in aomatic mammalian cells;' c) tests for effects on chromosomes in higher eukaryotes, including mammals; d\ evaluation of DNA repair synthesis. For screening purposes, preference has usually bean given to tests that have already been validated with a large ample of compounds belonging to different chemical groups. Among the mutagenicity tests on microorganisms, the one most widely used and validated is the Ames (aversion test in Salmonella. Tests in Escherichia coll Saccharomyccs. Neorospora. end Aspergillus an also being used. Mutagenicity testing is also being conducted in Drosophila. Several other methods currently being evaluated may be seed to monitor genetic damage in mammalian cells by carcinogens in vivo and in vitro. These methods include the production of sister chromatid exchanges as well as measurement of the induction of direct damage to DNA and its subsequent repair. Various short-term mutagenesis tests, some of which are used to provide supportiva evidence of carcinogenicity, are discussed in (96). Methods Based on Neoplastic Cell Transformation Several systems are now available at the mammalian cell level for the identification and study of substances that represent a possible cancer hazard im In recent years a number of systems have been developed to test for neoplastic call transformation by chemical and physical carcinogenic agents. Some of these systems are being ased in several laboratories with good reproducibility; other systems an still being developed. Those that have been most widely studied are a) the golden hamster embryo cell system and 6) the mouse embryo fibroblast cell line systems. In the golden hamster embryo cell system, primary and/or secondary cultures of normal embryo cells are used. Transformation is determined 7-10 days after treatment of cells seeded for colony formation. Quantitation is based on the frequency of morphologically altered colonies. In the mouse embyo fibroblast systems, established homogeneous ceil lines are used. Thus cloned populations of cells can be grown in Urge quantities and used by many Uboralorics. Transformants are identifiabia 4-6 DUP 0811079 DU 009602 39870 Federal gister / Vol. 44. No. i / Friday. (uly fl. 1979 / Notice* weeks after exposure to the carcinogen. They may be scored quantitatively by morphologic criteria (focus assay), which correlate highly with tumnrigcnicity in animals. Among these established lines, the C3H 10TV4 Clone 8 cell system has been the most widely studied. In these tests for neoplastic transformation, the cells derived from transformed colonies or foci, when inoculated into syngeneic or immunosuppressed animals, can grow as,malignant tumors. Although the definitive evidence for neoplastic transformation of cells in culture remains their tumorigenicity In animals, a number of phenotypic changes of the cultured target cells are commonly used as indicators. Other in vitro systems are being developed with the use of specialized cell types such as epithelial cells from liver, epidermis, and other organs. Neoplastic transformation of well* defined epithelial cells by chemicals has been ichieved in vitro: conditions for quantitative studies are under development Sedt systems may be needed to Identify critical target cell populations within target tissues closely correlated with carcinogenesis in vivo. To be effective, most chemical carcinogens require metabolic activation by cell enzymes to an ultimate reactive metabolite. In mammals metabolic activation of carcinogens takes place in many organs and tissues. Cells in .cultvre can retain enzyme activities, but specific culture systems or preparations may lack or lose the enzyme activity necessary to activate certain chemicals. Therefore, adequate consideration should be given to the effectiveness of metabolic activation functions in sack test system used. Evaluation of Short-Term Test Results The study of carcinogenesis at the cell level presently offers an effective metns to identify carcinogenic effect! and mechanisms. In vitro mammalian cell transformation systems are simple models for the study of the mechanisms of chemical and physical carcinogenesis. As these systems become more widely used as test methods, they will lead not only to better development and definition of screening techniques but also to better understanding of the underlying mechanisms of carcinogenesis. Short-term tests for chemical carcinogens presently do not, in the absence of animal bioassays and. epidemiology data, constitute definitive evidence that a substance does (or does not) pose a carcinogenic hazard to humsns. However, positive responses in these tests are considered suggestive evidence of a carcinogenic hazard. ' Such positive results also supply supporting evidence to positive animal bioassays or epidemiology results. In some instances results from short-term tests may conflict with animal bioassay data. If an animal bioassay shows a positive response, it cannot be dismissed because a negative response was observed in these testa. However, positive responses in such short-term tests are ordinarily sufficient to provide suggestive evidence of carcinogenicity, even if the substance tested has shown only negative responses in some animal bioassays. As the degree of certainty attached to the negative responses in animal bioassays increases because the observation is reproduced in other animal species and strains or under more rigorous test conditions, the suspicion about the chemical as a result of shortterm tests may be reduced and eventually eliminated. These conclusions are in accord with those of the National Cancer Advisory Board's Subcommittee on Environmental Carcinogenesis (57): At the present, none of the short-terra tests can be used to establish whether a compound will or will not be carcinogenic in humans or experimental animals. Positive results obtained in these systems suggest extensive testing of the agent in long-term animal bioassays, particularly if there are other reasons for testing. Negative results in a short-term teat. however. do not establish the safety of the agent MOLECULAR STRUCTURE AS SUPPORTING EVIDENCE IN IDENTIFICATION OF CARCINOGENS Information useful in identifying possible carcinogens is provided by their molecular structures. It is well established that certain groupings of atoms (functional groups) in tone molecules may impart carcinogenic properties--e.g, some polynuclear, aromatic systems, hydrazine group*. Nnitroso groups, and a, -unsatitrated lactones. There is s moderately aubstantia! base of empirical data that permits conclusions about carcinogenic potential on the basis of molecular structure [33.100). Similarly, some functional groups have never been shown to impart carcinogenic properties to molecules, although the data bass for such negative correlations is much smaller and probably inconsequential The reason for the absence of a strong empirical data base for noncarcinogens is that structure has frequently been used as a guide to testing chemicals for carcinogenicity, and priorities for testing have often been based on the suspected cancer-inducing properties of chemicals. In some instances, the predictive power of molecular structure of functional groups known to be correlated with carcinogenic properties has proved unsatisfactory. Therefore, the general consensus of the scientific community appears to be that chemical structure has limited value in identifying carcinogens and is to be used in carcinogenesis hazard assessment only as corroborative supporting evidence. In the absence of other data, however, there are instances in which structure may provide suggestive evidence that a risk of carcinogenesis exists. When strucure is to be used as suggestive evidence, well-documented support should be presented end qualified where necessary by complete notation of substancea of similar structure that have been adequately studied for carcinogenic activity. QUALITATIVE JUDGMENTAL FACTORS IN EVALUATION OF TOTAL EVIDENCE Evidence that'a substance poses a carcinogenic hazard is contributed by each source discussed in the preceding sections of this report epidemiologic studies, studies on experimental animals, and studies based on short term and other tests that have been shown to correlate with carcinogenicity, this includes studies of biochemical . pathways and chemical structure. For some substances data may be available from all three sources: for others, there may be data from only one or two sources. Each source of relevant data needs to be critically evaluated by consideration of the many aspects discussed in this document The judgment that a substance poses e carcinogenic hazard derives from the evaluation of.the total evidence . provided by all of the sources. Different data sources may not contribute equally to the cumulative evaluation, depending on the specific nature end extent of the data, the scientific quality of the studies, and the adequacy of their documentation. Conclusions on the carcinogenicity of a substance may be reached on the basis of evidence provided by epidemiologic studies, bioassays in animals, or both. Suggestive evidence is provided by the other types of studies. In the absence of adequate epidemiologic or animal evidence, a positive response in any of the short term in-vitro tests that correlate with carcinogenicity is considered suggestive of a carcinogenic hazard. Suggestive evidence may also derive from DUP 0811080 DU 009603 Federal Register / Vol. 44, No. 131 / >. .day, July g.1979 / Notices arm 39871 considerations of chemical structure or biochemical pathways. Ordinarity. if a substance has produced positive results in a single adequately designed and conducted animal bioassay and no other data are available, the conclusion is that the substance is likely to pose a risk of cancer to humans. These results may be further confirmed by data on chemical structure, in vitro testing, or relevant biochemical studies that suggest a carcinogenic potential. However, negative data from the latter three sources do not override adequate positive data from an animal bioassay. Further confirmation that the substance poses a carcinogenic hazard to humans is obtained from bieaesay data showing reproducibility of results. positive doseresponse relationships, and concordance of results (see "Evaluation of Pathologic Results"). Because of biologic variability among species, the conclusion that the evidence is positive on the basis of results obtained in one animal species is not altered by negative data obtained in other species or strains of test animals. Moreover, negative epidemiologic data, questionable because of limitations In die power of detection of such studies, do not deny the conclusion of carcinogenicity on the basis of animal bioassays. Negative evidence from properly designed and conducted epidemiologic studies may.-however. be used to set an upper limit on.human risk to comparable populations under analogous conditions of exposure. It should be stressed that the qualitative judgment whether a substance poses a carcinogenic hazard is based on the evaluation of cumulative evidence from all pertinent data sources. The reasons for specific conclusions need to be clearly detailed. The terns "strong** and "weak" have been used in the literature to describe both the nature of the hazard or risk and the exent and quality 6f the evidence. A certain confusion may have ensued, since one could refer fe weak evidence of a strong effect or to Strong evidence of a weak effect. The two categories are clearly not equivalent and should not be confused. PART HI. THE QUANTITATIVE ESTIMATION OF RISK The previous section of this document dealt with the issue of the likelihood that a substance poses a carcinogenic hazard humans. In some instances a regulatory agency may be required, or may find it useful, to estimate quantitatively the cunccr risk of such a substance in exposed humans if the compound is assumed to be a human carcinogen. Quantitative assessment of human cancer risk may be based on epidemiologic or animal data. In either instance, methodologic problems arise because of the need to extrapolate from effects observed under one condition and level of exposure and in one population group or biologic system to arrive at an estimate of the effects expected in the human group or individual. Because extrapolations are involved, uncertainties are necessarily attached to the cancer risk estimates that can be made with current methodologies. Furthermore, uncertainties arise from other sources, particularly from attempts to identify accurately conditions and levels of exposure of the human group or individual. Despite the uncertainties, risk estimates can be and are being made, not only by some regulatory agencies but by other scientific bodies. Because of the uncertainties, however, and because of the serious public health consequences if the estimated risk were understated, it has become common - practice to make cautious and prudent assumptions wherever they are needed to conduct a risk assessment. This approach has a precedent in other areas of public health protection where similar problems arise because of gaps in knowledge {101.103). Thus current methodologies, which permit only crude estimates of human risk, are designed to avoid understatement of the risk. It must be recognized, however, that in some circumstances this cannot be guaranteed because of other factors that may enhance human response, such as synergistic effects. Thus risk * assessments should be used with caution in the regulatory process. If data on animals are used as the basis for estimating human risk, data obtained from the most sensitive animal species or strain tested are commonly recommended as the starting point for extrapolation. Of the available data, these are clearly the least likely to understate human risk. Use of data from less sensitive species or strains is justifiable only if there are strong reasons to believe that the most sensitive animal model is-completely irrelevant to any segment of the exposed human population. A limited comparison of human and animal data for carcinogens is contained in a report of the National Academy of Sciences (103). Data were compared for benzidine, chlomaphazine, dlethylstiibcstrol. aflatoxin B(. vinyl chloride, and cigarette smoke. The authors stated that *** * * as a working hypothesis, in the absence of countervailing evidence for the specific agent in question, it appears reasonable to assume that the lifetime cancer incidence induced by chronic exposure in man can be approximated by the lifetime incidence induced by similar exposure in laboratory animals at the same total dose per body weight." These preliminary observations suggest that current methodologies may not lead to serious errors. Whether quantitative risk assessment is based on data from animals or humans, there is uncertainty about the shape of the dose-response relationship at the (usually low) levels of actual human exposure. Mathematical extrapolation models are discussed in detail later in this section. The linear nonthreshold dose-response model is most commonly used at the present time. Of the various models, it appears to have the soundest scientific basis and is less likely to understate risk than other plausible models. It has, for many of the same reasons, a long history of ).use in protection against radiation {101, 102 The most favorable foundation for quantitative risk assessment is based on well-characterized responses in human populations with well-defined exposures. Unfortunately, the exposure estimates are often unavailable or crude. Negative epidemiologic studies on populations for which usable exposure estimates are available can be valuable in conjunction with animal data; the studies on animals provide evidence for carcinogenic haiard. and die epidemiologic data may provide upper limits of response for-crosscomparison with the animal data. Although extrapolation from the observed human population group to other groups carries leas uncertainty than extrapolations from animals to humans, the possibility of significant differences in the characteristics and conditions of exposure of the two population groups must be recognized. Any such differences that may affect the estimate of risk should be noted, although information is rarely available that will Srmit specific integration of these :tors into the risk assessment methodology. To the extent currently possible, the methods described in the following section permit a crude order-of- magnitude estimate of risk for substances that may pose a cancer hazard to humans. As more knowledge develops, risk assessment methodologies should be improved. Some of the kinds of information and DUP 0811081 DU 009604 nuMW tvegister ( v ill. . I' f Friday, Iulytt*t979 / Notice* knowledge that will likely prove useful in the future^ire discussed in the sections to follow. At present, most such information is not available and thus cunnol ordinarily be used in risk assessment without the imposition of numerous assumptions. Caution is needed in risk assessment as long as thes< gaps in knowledge exist. Much has been written about threshold dosea for carcinogenic effect, but unfortunately, there is no recognized method for determining their existence. A model recently proposed by Cornfield (7W) permits tha inclusion of thresholds. However, as Cornfield stipulated originally and again recently (705), a threshold could be derived from this model only if then were instantaneous and complete deactivation of the material before any carcinogenic effect occurs--an improbable event Since threshold doses for carcinogenesis have not been established, a prudent approach from a safety standpoint it to assume that any dose may induce or promote carcinogenesis. Some of the mathematical models proposed to describe the dose-response relationship for carcinogenesis are discussed in the following section. With the present state of knowledge, the quantitative assessment of cancer risks provides only a rough estimate of the magnitude of the cancer risks; this estimate may be useful in setting . priorities for control of carcinogens and in obtaining a very rough idea of the magnitude of the public health problem posed by a given carcinogen. MATHEMATICAL MODELS FOR H1GH-TO-LOW DOSE EXTRAPOLATION WITHIN A SINGLE BIOLOGIC SYSTEM Mathematical Models were developed u. the last two decides for estimating the effects of ttyfttun levels well below levels for which toot data were available, with thf goal of ensuring that the risk will not be underestimated. These models of dose-response relationships make use of data obtained in a given biologic system to extrapolate from high to low doses. Consideration must be given to the many biologic variables that influence the level of response in different species or under different exposure conditions. The Models In order to extrapolate outside the experimental range of exposure levels, some mathematical formulation relating response to dose must be available. The two categories of mathematical models commonly used to depict the relationship between response and dose are dichutomous-responsa models and time-to-response models. In tha dichotomous-response situation the response of interest is the presence or absence of some specified condition. Time-to-response models attempt to relate dose level to distribution of the time until the occurrence of a given event, such as tumor observation or death. (Both categories of models an completely specified except for a few unknow n parameters, which an typically estimated from a given sat of experimental data.) A variety of different approaches have been proposed to deal with the problem of low-dose extrapolation involving a dichotomous nsponse. included an the Mantel-Bryan procedure, the one-hit model linear extrapolation, and various extensions of the multistage model developed by Armitage and Doll [106). Mantel and Bryen [107,106) proposed an extrapolation technique based on the log-probit model, which had long been used for bioassays to estimate median lethal doses. They selected this model because it seemed to provide a reasonable fit to a large body of experimental carcinogenesis data and not because of any mechanistic arguments in its support Under this procedure, extrapolation is conducted from the upper confidence limit on the observed experimental response along e probit tog-dose fine with a pnassigned ' slope of one tasome specified low level of risk. By using the upper confidence limit and fixing the slope at one (a shallower slope than they had typically teen with their experimental data lets). Mantel and Bryan hoped to generate an upper bound on the estimated dose associated with the predeterminated risk level, regardless of the true form of the underlying and unknown doseresponse curve. However, subsequent theoretical and applied nsearch has demonstrated that the Mantel-Bryan procedure is not as conservative as once thought and may underestimate risk in some situations [t09.110), The one-hit model is based on the concept that a tumor can be induced after a single susceptible target or receptor has been exposed to a single effective dose unit of a substance [109, 110). Thus, unlike the Mantel-Bryan procedure, then is an assumed biologic mechanism of action for the carcinogen underlying the one-hit model. This action implies that the probability [P\ that a tumor will be induced by exposure to a chemical at doae d Is given by the equation P(d)*\-cxp[-\d). when X is an unknown non-negative constant. When --......... y--=---------------------------- W is small (i.e' in the low-dose ngion). it can readily be shown that PfJlsKd, i.e.. for low dose levels the one-hit model is well approximated by a simple linear model in which the probability of tumor observation is directly proportional to dose. If the unknown (true) dosc-reponse curve is assumed to have a sigmoidal shape--an assumption supported by a wealth of toxicologic data--then the nsponse will curve upward in the low (or. typically, environmental) dose region. Thus a linear model will provide an upper bound to curves of this shape and. it is hoped, e conservative estimate of the dose associated with any specified lavel of risk (117). A line connecting zero with a point on the dose-response curve for the excess tumor nte above background will always lie above the true dose-nsponse curve for the convex portion of the curve. An additional degne of conservatism is introduced by extrapolating back to zero from an upper confidence limit (UCL) for the net excess tumor nte above the background nte. In the linear model the tumor rate is assumed to be proportional to dose: PfdJ^M. Ths upper confidence limit for the slope X is UCL-^experimental dose. Thus the maximal risk for a given dose d may be estimated by the equation maximal risk(UCL/t/,)x<f. when d, is the experimental dose. Conversely, the equation for a predicted dose for a maximal level of risk is: predicted dose(riskx</,)/UCL. A number of investigators have published papers [112-1IS) based on the Armitage and Doll (775) formulation of the multistage model of esreinogensis. Under the multistage model it is assumed that the cancer originates as a "malignant" ceil which is initiated by a series of somatic-like mutations occurring in finite steps. It is also assumed that each mutational stage can be depicted as a Poisson process in which the transition rate is approximately linear in dose nte. Then the lifetime probability of tumor induction can be expressed approximately as P(d)*l-exp{-X-Xid------- X^d*). when X*>0 for all values of /. and k corresponds to the number of transitions or mutational stages. (Highly sophisticated computer algorithms have been developed for fitting the multistage model to laboratory data with the use of a restricted maximum likelihood approach which does not require that the value of A be pn-specified.) Both the total incidence of tumors and the time at which tumors occur an importanL Tumors leading to early DUP 0811082 DU 009605 Federal Register / Voi. 44. No. 131 / l. -ay, (uly 6,1979 / Notices 39873 death and life-shorterning need to be considered. Time-to-tumor it the time at which a .tumor is detected or observed by palpation or by gross or microscope examination of an animal at the time of death or sacrifice. Time-to-tumor is not used here to indicate the instant at which a pretumorous condition becomes a tumor. Time-to-observance is better terminoic.y. Some hope for improving risk estimates has been based on use of the time-to-observance of tumors in addition to use of the proportion of anir.als possessing tumors. On the basis of Druckrey's work {117), the median time to tumors appeared to increase as the dose decreased. It was hoped that low doses could be found that would result in median times-to-tumor observation well beyond the expected lifetime; this might result in the identification of "practical thresholds." Albert and Altshuler [11B) expanded on the use of median time-to-tumor observance by employing distributions of time-totumors for individual animals. Chand and Hoel {119) showed that use of a log normal time-to-tumor distribution leads to a probit-log dose relationship, and - use of a Weibull time-to-tumor distribution leads to an extreme value model for the proportion of animals with tumors: P(d)=1 - expj - exp(a+/3 log .</)), where alpha and beta are constants. Schneiderman et al. (120) demonstrated th*: even though tha median time-totumor may be well beyond the expected lifetime, a significant proportion of animals or humans may still develop . tumors within the normal life-span. Peto {121) examined human data and questioned the concept that lower doses result in longer latency. Whittemore and Altshuler [122), analysing data on cigarette smoking, concluded that it was hot possible to distinguish between the log-normal and the Weibull models. The available data do not permit a conclusion as to whether lower doses lengthen the latency periods. Animal experiments at high doses may induce more tumors resulting in easier and therefore earlier detection, and this may not be due to an actual decrease of latency period. Time-to-observance response models have not received the same degree of attention as dichotomous-response models in carcinogenesis risk extrapolation. One of the major factors ` underlying this relative lack of emphasis may be that studies in which animals were given the substance in their feed have not generated sufficient information to determine the relationship between age and cumulative cancer incidence. Procedures In the preceding section it was noted that the Mantcl-Dryan procedure is essentially empirical and lacks biologic relevance with respect to current knowledge about carcinogenesis. Since risk extrapolations developed by-the Mantel-Bryan technique tend to zero much more rapidly in the low-dose region than do extrapolations based on somatic mutation models, the MantelBryan procedure would certainly not be appropriate if th e carcinogen under study were thought to act directly on cellular DNA [109). Initially, extrapolation based on a multistage model appeaps to offer significant advantages over linear extrapolation procedures. Under the multistage approach, no assumptions are made a priori about the exact form for the mathematical extrapolation. Instead, the experimental data are used to estimate the shape of the doseresponse curve. However. Crumpet al. (709.114) and Guess et al. (110) have shown that the upper confidence limit on estimated risk becomes essentially linear for generalized polynomial extrapolation in the low-dose region. This approximate linearity holds even when the maximum likelihood estimate of excess risk does not contain a linear component (estimated). Therefore, there is some question whether the mathematical refinements of generalized polynomial extrapolation are justified for appl.cation to animal bioassays, which may be only crude approximations to the human situation (M9). As an interim procedure, it has generally been recommended (70S) that whenever quantitative risk analysis is deemed necessary, linear extrapolation thould always be included among any methods used unless there is reason to believe that the experimental (observed) response does not fall in the convex portion of the dose-response curve. If the response is in the concave portion of the curve, the one-hit model is suitable. At low observed responses the linear and one-hit models yield nearly identical results. An added degree of protection can be achieved by starting the extrapolation from the upper confidence limit of the response. The mathematical procedures per se are intended to provide upper limit estimates of risk from a statistical standpoint. However, the risk estimates as applied to humans should not be regarded as upper limit estimates because of large biologic uncertainties (see "Extrapolation From Observed Effects to Estimates of Risk for the Observed Population"). CHARACTERIZATION OF POPULATION EXPOSURE The estimation of total population exposure to a given substance (and/or to its decomposition and metabolic products) requires consideration of the following aspects: o) sources of human exposure (occurrence, production, uses, and environmental distribution); b) analytical methods for detecting and measuring exposures in the environment and in the population; c) routes and conditions of exposure; d) duration, frequency, and intensity of exposure; and *) ate and characteristics of the exposed populations. During examination of exposure data, important qualitative and quantitative factors beyond definable numerical values of dose level and population size will emerge; although such information may net be usable directly in a matheoretical calculation of risk estimate, it will frequently provide additional perspective and insight during risk evaluation. Because of the great diversity in sources and estimating procedures available in various situations, it does not seem practicable at this point to set minimum detailed specifications for the reliable estimation of exposures or to identify recomamnded or approved methods and procedures for producing exposure estimates. The following general considerations indicate the kind of data useful for assessment of population exposures. The better defined these data are, the higher will be the confidence that a realistic estimate of risk for the exposed populations has been made [123). Sources of Human Exposure Two types of exposure sources are considered: primary sources and human contact sources. Primary sources of exposure to a chemical are those that determine its release into the human environment and they include natural occurrence, extraction from natural products, mining chemical synthesis, manufacturer or production, and specific uses. Human contact sources are those that bring about the contact of the substance with tbs human body, and they include items, ar preparations containing the chemical (such as foodstuffs or consumer products), vehicles, or a medium in which the chemical is present (such as ambient air or drinking water). Some substances may originate from a single primary source and be present in a wide range of human contact sources; conversely, a specific-human contact source may be traced to several DUP 0811083 jiSiMI . 39074 Federal Register / Vol. 44. No. 131 / F.ii..y. July 6. 1979- ( Notices different primary sources. It is important that for each substance the entire range of sources and environmental distribution be examined. frequently, there is more than one source of human exposure, and an individual may be exposed to a substance of concern from an array of sources depending upon the circumstances. Analysis of environmental distribution and exposure pathways allows idenification of the most significant sources, so that both the size of the population exposed and the intensity of exposure can be established. In some instances, it la possible to estimate combined exposures to the same substance from different sources, primarily where the populations affected by these different sources are the same. Frequently, however, differences in the populations exposed from various sources are so large that any attempt to combine the estimates may produce an unrcalis'ic or unclear description of the acutal human exposure conditions. Then it is preferable to constder each source separately and subsequently use whatever knowledge is available on multiple sources of exposures to interpret these observations. . Estimates of the total level of production of a substance can be useful indicators of the extent of exposure, particularly over time. Dates of first synthesis and commercial production of a subsi .nee are useful in the evaluation of delayed toxic effects and allow an estimate of the time before which no human exposure could have occurred. The accuracy of data on national production and foreign trade of individual substances (which are often difficult to obtain) needs to be ascertained. Uses of a substancs are important descriptors of its environmental distribution and the extent of human exposure. Whenever possible, all uses of carcinogenic substances should be identified. An important distinction is that between uses for which human exposure is intended (intentional exposures) and that for which it is not intended (unintentional exposures). Individual exposure or consumption of a substance may be voluntary or involuntary. The sociologic bases and implications of these definition# are beyond the scope of this report. Analytical Methods for Detection and Measurement of Exposures 1 tie specificity and limit of detection of analytical procedures for the identification of many carcinogenic substances, both in the environment and in exposed orgunisms. have been remarkably improved in recent years. Progress in analytical chemistry is expected to undergo further refinement and improvement in the near future. The limit of detection of analytical methods v-Hrics considerably for different substances and different conditions of analysis, and this is a critical factor in assessing a source of exposure, it is important to consider that the ugent may not be measurable but may still be present below the minimum detectable level. The minimum detectable level of a substance may vary depending on different vehicles, media, and conditions of exposure. Quantitative determinations of the level of a substance in various exposure sources should consider time and space distribution and variations, and ranges of values may be useful to estimate the conditions of exposure. The chemical and physical properties of the substance should be identified. Such characteristics as particle size distribution for aerosols and dust should be determined insofar as possible. Analytical determination of the levels of a substance in exposed organisms, particularly in the exposed population, is of great value but not always obtainable. Available data on the levels of substance (or its metabolites) in tha target tissues or body fluids should be considered. The dose of an ultimata carcinogen at the site of action in tha tissues or ceils, which is measured at all time* after its introduction ("target tissue dost") is ideally the dots that should ha estimated and corralatad with expectsd effects. This targat tissus dote usually cannot be closely estimated because of many variables and uncertainties [J02). The relationship between target tissue dose and exposure doss may vary considerably under different conditions. To the extent practicable, documentation of the analytical methods, the sampling conditions, the limits of detectability, and the range of observed values is desirable. Routes and Conditions of Exposure All possible routes of exposures associated with each source should be identified. If any routes of exposure are considered irrelevant for estimation of effective doses, the circumstances should be specified. Careful consideration of sources of exposure-- e.g.. product use patterns, environmental or occupational situations, and background--may suggest or reveal routes of exposure not immediately apparent. For example, a chemical may also be absorbed through the akin or by ingestion when inhalation is apparently the primary route. For estimation of anitaai-to-human correlations in the evaluation of test data on animals, it is necessary to obtain the human dose level in units consistent with those used to describe the effective dose in the animal bioasaay being used for comparison. In some instances, any necessary conversion from the actual measurement at the source to the needed units describing exposure dose can be straightforward (e^. by simple application of observed or estimated food ingestion rates to a chemical's concentration in food). In other instances, complex calculations or modeling procedures may be necessary, as in the estintation of effective exposure distributions from ambient airon the basis of monitoring data or emission inventories for point sources. This conversion or translation step, often necessary in the estimation of human exposure, should always be explicity identified and reported. When available data show substantial differences between the rout* and conditions of exposure in test animals and in humans, it is necessary to rely on estimates of comparability and to attempt to establish an acceptable equivalent dose. In the absence of satisfactory equivalent dose data, only defensible conservative assumptions should be used in such a way that the possible risk is not underestimated. Duration. Frequency, and Intensity of Exposure An important factor in tha quantitative evaluation of population exposure is the length of time during which exposures occur. Although the time of exposure may vary considerably within a population, there are cates when it can be reasonably well-defined. Theta include case# of specified duration of exposure (e.g.. to certain drugs or certain occupational carcinogen*) or continuous lifetime exposure to widely disseminated environmental carcinogens (e.g- polycyclic aromatic hydrocarbons). Effective exposure rates corresponding to typical patterns of individual exposure, whether short-term or long-term temporal trends, must be reported wherever significantly different patterns exist.The two components of the estimated level or amount of exposure-- the effective rate per unit time or per incident of exposure and tha frequencyduration pattern--should be explidty identified for each exposure pattern considered. DUP 0811084 DU 009607 Federal Register / Vol. 44. No. 131 / Frida" luly 6.1879 / Notices 39875 Size and Characteristic* of Exposed Populations The total number of people exposed to any level of a carcinogenic substance represents a major indicator of the extent of risk related to that substance. Because combinations of exposures to 'different carcinogens may contribute to the cancer risk in the same population or individual, and because no threshold level for exposure to a carcinogen can presently be reliably determined for a population, a contributory risk level from any exposure level, however small, must be assumed. Age of exposure should be considered, i.e., 'whether exposure is essentially lifelong (et more or less constant rates) or ie concentrated in certain age ranges. The relationship between total lifetime exposure in each exposure pattern end the amount of this exposure that may be concentrated in any specific age ranges should be identified. Wherever feasible, the degree of stratification of exposed populations Should be identified to permit distinctions between effective exposure amounts by age (e.g., childhood, working age. and elderly age groups) and by sex. As noted above, populations having high-risk age groups should be identified. Attention should be given to exceptional exposure groups of special concern, such as infants, children, and pregnant women, as well as to groups with special genetic conditions or concurrent disease. In addition, in descriptions of certain population subgroups, the smoking habits, dietary and alcohol consumption patterns, and other cultural and environmental characteristics should be considered if possible. -- EXTRAPOLATION FROM OBSERVED EFFECTS TO ESTIMATES OF RISKS FOR EXPOSED POPULATION The quantitative animation, of risk from a carcinogeifttftbatance for the entire exposed or ffetenttally exposed population may be Conducted with the use of observation* OH the effects of the substance in 1) a 44fll*ed human population group Mil I) experimental animal tests. In both situatiMN dm extrapolation will take into accmtttt the factors that characterize and distinguish the groups observed and the factors to which the extrapolation applies. Correlations From Observed Human Population Groups 1o Others The luoblem to be considered here is the estimation of present or potentiul risks for all people exposed to a given substance by means of data obtained from observations in s defined population group. The observed group may be small and its exposure conditions may be well defined, as for certain studies of drugs or for occupational exposures. In other situations the observed group may be poorly defined even if larger. In analyzing the correlation between observed and estimated population effects, it is desirable wbere feasible to review the critical differences between the two conditions, such as age and sex distribution of the population: genetic, racial, and ethnic differences: environmental differences and migration patterns: dietary and cultural habits: smoking patterns; alcohol consumption: patterns of intercurrent disease: and particular susceptibility states including pregnancy and fetal and neonatal exposures. Many of these complex variables are considered under "Epidemiologic Evidence" in Part II and "Characterization of Population Exposure" in Part UL * Animal-to-Human Correlations Although a dost qualitative similarity has been established in the nature of the response of laboratory animals and humans to carcinogenic substances, a quantitative correlation ie more uncertain because of the marked variation of susceptibility in different animal species and among individuals in the human population. It is not possible to reduce the variable* to e single safety factor for general use (70S). Several species-conversion factors should be considered in estimating risk levels for humans from data obtained in another species. Species-conversion factors are affected by many variables, such as body surface, body weight, metabolic pathways, nutritional conditions, genetic variability, and bacterial flora as well as tissue distribution and the retention and fata of the chemical. In evaluating exposures to the general population, one should consider all ages, transplacental ` exposures, concurrent disease conditions, and special susceptibility states. Other conversion factors should also be considered when observations are obtained for test species under expocure conditions markedly different from those in the population (e.g.. different routes or modes of exposures, vehicles, modifying factors, variations in age, sex. perinatal exposures, disease states, end single vs. multiple exposures). The limits of uncertainty should be stated whenever possible [10Z 106). Different carcinogens tested under comparable experimental condition* show a wide range of response; if extreme cases ere included, the range of variation is more than one millionfold. Changes in experimental conditions, particularly ones that alter the effective dose, can markedly affect the observed level of effect of a carcinogen within the seme genetic strain of animal Exposure of experimental animals to certain other chemicals in addition to a carcinogen under test may change the observed effect in either direction end at the extremes up to one hundredfold or even one thousandfold [121). Differences between species can be even greater. On the other aide of the correlation, the human response to carcinogens as well as to many other chemical* and drugs may also show great quantitative variations among individuals. Studies on the metabolic activation and chemical interaction of carcinogens in human tissues in vitro have shown interindividual quantitative variations of about one hundredfold in relatively small population samples [125-127). Individuals resistant or sensitive to one carcinogen may not be equally resistant or sensitive to another carcinogen or to combined effect* of several exposures. Such wide interindividual variations are also well known from many pharmacokinetic studies. A number of variables are relevant to the correlation of animal and human conditions. Some problems inherent in the use of animals must be kept in mind when animal studies are used for estimation of the quantitative carcinogenic potential of a substance for human*. A concise statement of tome of these factors is contained in "Drinking Water end Health," prepared by the Safe Drinking Water Committee. Advisory Center on Toxicology. National Research Council National Academy of Sciences (56). Factors discussed in this document include the rate of chemical absorption, distribution within the body, metabolic differences among exposed animals, effect of intestinal bacteria, rates of excretion end reabsorption, differences in molecular receptor sites for the carcinogen, environmental and genetic differences, end number of exposed animals and susceptible cells. Metabolism end pharmacokinetics account for major differences in sensitivity to chemical carcinogens between species. In principle, this information could be used in estimating the relative sensitivity of humans compared to experimental animals. In practice, detailed metabolic pathways in humans are not known for many carcinogens: moreover, the marked variation in metabolism and sensitivity DUP 0811085 DU 009608 39876 Federal Register / Vol. 44, No. 131 / FriJay. July 6. 1979 / Notices .I among individual! of different ages, states of health, and other biologic conditions require more information on the heterogeneity of human metabolic and pharmacokinetic responses than is usually a\ ailable. It is hoped that future research will clarify these important correlations in much greater depth. Such information, if available, should be used to correct for an underestimate of hj.r. .n risk, but it should be used to correct for an overestimate of human risk only when there is substantial information on diversity of humen response. The contribution of animal test data to the estimation of the risk level for humar.s should be based on experiments with the most sensitive species available. Confidence that this procedure will not underestimate the human risk increases with the number of experiments and the number of species and strains studied. * LACK OF PREDICtABLE THRESHOLDS FOR AN EXPOSED POPULATION The self-replicating nature of cancer, the multiplicity of causative factor* la which individuals can be exposed, the additive and possibly synergistic combination of effects, and the wide range of individual susceptibilities work together in making it currently unreliable to predict a threshold below which human population exposure to carcinogen has no effect on cancer risk. Observation of tha marked individual differences in the response of human subjects to carcinogens shows that same individuals do not develop cancer in their lifetime, whereas others develop it readily after the sttne exposure to a carcinogen. Although Bwae observations are compatible wfB (lA existence of different "thresholds** for individual subjects in certain traditions, they are not a basis for predicting a no-effect level of a carcinogen in other individuals or under different conditions. There is no presently acceptable way to determine reliably a threshold for a carcinogen for an entire population. Individual human subjects in the population are exposed throughout life to a number of carcinogens, which may be considered to provide a background of carcinogenic risk; exposure to any amount of a single carcinogen, however small, is regarded as capable of adding to the total carcinogenic riik(I09). Cancer susceptibility varies greatly among individual members of human populations due to genetic, racial, and ethnic factors; to environmental and dietary exposure: and to other modifiers. Variability among individuals makes it very difficult to have confidence that an observed no-cfTcct level of exposure in animals or even in a specific human population (for which individual variation may be small in comparison to the total population) will be applicable to the total human population at risk. A large number of factors (c.g.. age. sex. race, nutritional status, immunologic status, general state of health, previous exposure to the substance in question or to other substances) could affect individual susceptibility. Even if thresholds for carcinogens could be demonstrated for certain individuals or for a defined population, no reliable method is known for establishing a threshold that could apply to the total human population [67). SUMMARY OF RISK ESTIMATION For a given substance, the usefulness of dose-response data obtained from a specific human population group or from animal tests for estimation of riak in the general population is limited by the consideration that general population exposures to one substance are usually only a component of the total carcinogenic burden derived from multiple sources, with their possible interactions. Recognition of these limitations, however, does not imply that no attempt should be made to develop reasonable risk estimates for different conditions of human exposure. The several components of quantitative risk asses iment include, the following: o) definition and quantification of exposures; b) characterization of the exposed populations in quantitative terms; c) chemical and physical properties of the substance and its chemical reactivity in relation to exposure; d) prudent quantitative mathematical extrapolation of the responses from observed to estimated exposure ranges within the observed biologic system; and e) qualification of the estimated risk in light of identifiable biologic and toxicologic differences that may be present in the exposed human population. REFERENCES (7) Saffiotti U: Scientific bases of environmental carcinogenesis end cancer prevention: Developing an interdisciplinary science and facing its ethical implications. J Toxicol Environ Health 2:1434-1447,1977 (2) Hartwell |L Survey of Compounds Which Have Been Tested for Carcinogenic Activity. Natl Cancer Inst. Public Health Serv Pub! No. 149. Washington. D.C.: U.S. Covt Print Off. 1951 (J) Shubik P. Hartwell JL: Survey of Compounds Which Have Been Tested for Carcinogenic Activity. Netl Cancer Inst, Public Health Serv Publ No. 149 (suppl 1). Washington. D.Cj U.S. Covt Print Off. 1957 (4) Shubik P, Hartwell JL Peter* (A. eda: Survey of Compounds Which Have Been Tested for Carcinogenic Activity. Natl Cancer Inst. Public Health Serv Publ No. 149 (suppl 2). Washington. D.C: U.S. Covt Print Off. 19G9 (5) National Cancer Institute: Survey of Compounds Which Have Been Tested for Carcinogenic Activity. 1901-1997 vol. Public Health Serv Publ No. 149. Washington, D.C: U.S. Covt Print Oft 1973 (6) --: Survey of Compound* Which Have Bden Tested for Carcinogenic Activity. 1960-1909 vol. Public Health Serv Publ No. 149. Washington, D.C: U.S. Covt Print Off, 1971 (7) ----: Survey of Compounds Which Have Been Tested far Carcinogenic Activity. 1970-1*71 vol Public Health Serv Publ No. 149. Washington. D.C: U.S. Govt Print Off. 1974 (0)------: Survey of Compounds Which Have been Tested for Carcinogenic Activity. 1972-1973 vol Public Health Serv Publ No. 149. Washington. D.C: US. Covt Print Off, 1975 (9) International Agency for Research on Cancer Inorganic substances, chlorinated hydrocarbon*, aromatic amines. N-nitroso compound*, natural products, miscellaneous. 1ARC Monogr Eval Cardnog Risk Chem Man 1:1-194.1972 (70)------: Some Inorganic end OrganometaUic Compounds. 1ARC Monogr Eval Cardnog Risk Chem Man 2:1-181.1973 (//) --: Certain Polycyclic Aromatic Hydrocarbons and Heterocyclic Compounds. IARC Monogr Eval Cardnog Risk Chem Man 3:1-271.1973 (72)------: Some Aromatic Amines. Hydrazin* and Ralatad Substances. NNilroso Compound* and Miscellaneous Alkylating Agent*. IARC Monogr Eval Cardnog Riak Chem Man 4:1-286.1974 (73) ----: Some Organochlorine Pesticides. IARC Monogr Eval Cardnog Risk Chem Man 5:1-241.1974 [14)------: Sex Hormone*. IARC Monogr Eval Cardnog Risk Chem Man 9:1-243.1974 (73) --: Some Anti-thyroid and Related Substance*. Nitrefur*ns and Industrial Chemicals. IARC Monogr Eval Cardnog Risk Chem Man 7:1-326.1974 (73) ----< Some 'Aromatic Azo Compounds. IARC Monogr Eval Cardnog Risk Chem Man 9:1-357,197$ (77) --: Soma Aziridines. N-, S-. and OMustards and Selenium. IARC Monogr Eval Cardnog Risk Chem Man 9:1-209,1975 (73) ----: Some Naturally Occurring Substances. IARC Monogr Eval Cardnog Risk Chem Man 10:1-353,1975 (15)------ Cadmium, Nickel Some Epoxides. Miscellaneous Industrial Chemical*, and General Considerations on Volatila Anesthetics. IARC Monogr Eval Cardnog Risk Chem Man 11:1-300.1979 (30} ----: Some Carbamates. Thiocarbametea and Carbazides. IARC Monogr Eval Cardnog Risk Chem Men 12:1282.1*79 (27) : Some Miscellaneous Pharmaceutical Subetances. IARC Monogr Eval Cardnog Risk Chem Man 13:1-255.1977 DUP 0811086 DU 009609 SKHliJ I-MIHII E*[ J IMI Federal Register / Vol. 44. No. 131 / Friday. July 6. 1979 / Notices 39877 [22]------: Asbestos. 1ARC Monogr Evl Carcinog Risk Chem Man 14:1-106.1977 [23]----- : Some Fumigants. the ilcrbicidc-s 2.4-D and 2.4.S-T. Chlorinated Dibcnzodioxins and Miscellaneous Industrial Chemicals. IAKC Mor.ogr Eval Carcinog Risk Chem Man 15:1-345.1977 (24j------: Some Aromatic Amines and Related Nitro Compounds--Hair Dyes. Colouring Agents and Miscellaneous Industrial Chemicals. (ARC Monogr Eval Carcinog *isk Chem Man 16:1-400.1978 (251------' Some N-Nitroso Compounds, 1ARC Monogr Eval Carcinog Risk Chem Man 17:1-365,1978 (76] National Institute for Occupational Safety and Health: Suspected Carcinogens. A subfile of the KIOSH Toxic Substances List DMEW Publ No. (N10SH) 75-188. Rockville. Md.: U.S. Dept Health. Educ. Welfare. 1975 (77) Arcos jC. Argus MF. Wolf C: Chemical Induction of Cancer: Structural Oases and Biological Mechanisms, vol I. New York and London: Academic Press. 1968 (28) Arcos |C Argus MF: Chemical Induction of Cancer. Structural Bases and Biological Mechanisms, vol BA. New York and London: Academic Press. 1974 (79)------: Chemical Induction of Cancer Structural Bases snd Biological Mechanisms, vol UB. New York and London: Academic Press. 1974 [30] Derenblum L Carcinogenesis ts s Biological Problem. Vol 34. Frontiers of Biology. Amsterdam: North-Hoiland. 1969 (Jl) lfucper V.C, Conway WD: Chemical Carcinogenesis and Cancers. Springfield. 111.: Thomas, 1964 (32) Clayson DC: Chemical Carcinogenesis. Boston: Little. Brown ft Co.. 1962 (33) Searle CE. ed: Chemical Carcinogens. Amrr.can Chemical Sectaty Monograph 173. Washington. D.C: Am Chem Soc.,1976 (34) Teichraann B. Schramm T: Substanzen tail kanxetogener Wfcfcwg. Berlin-Bueh: Zentralinstitut fur Krehtforschung der Akademie der Wissenechaften der DDR. 1973 (35) Tomatis L Agthe C Bartsch IL et al: Evaluation of the carchMgencity of chemicals: A review of the monograph program of the Internatienal Agency for Reaearch on Cancer (1971 to 1977). Cancer Rea 38:877-685.1979 (36) National Cancer Institute. National Institute of Environmental Health Sciences. National Institute for Occupational Safety and Health: Estimates of the Fraction of Cancer in the United States Related to Occupational Factors. US. Dept Labor. Occupational Safety and Health Admin. Docket No. 11-090. Washington. D.C: 1978 (37) Rockette HE: Cause specific mortality ef cost miners.| Occup Med 19:795-801.1977 (38} Li FP. Frsumeni ]F )r. Mantel N. et al Cancer mortality among chemists.) Natl Cancer Inst 43:1159-1164. I960 (39) Koskela R-S, liemberg S. Karava R. et al A mortality study of foundry workers. Scand | Work Environ Health 2 (suppl 1): 7389.19G9 [40) Cibson ES. Martin R! L Lockingtpn |N: Lung cancer mortality in a ateel foundry.) Occup Mod 19 807-8111977 (41) Moss F. Lit WR: Occurrence of oral and pharyngeal cancers in textile workers. Be ) Ind Med 31224-232.1974 (42) Lloyd |W. Decoufle P. Salvin LG: Unusual mortality experience of printing pressmen. | Occup Med 19:543-550.1977 (43) Wagoner |K. Miller RW, Lundin FE Jr, et al: Unusual cancer mortality among a group of underground metal miners. N Engl) Med 209:284-289.1963 (44) Redmond CK. Strobino BY. Cypess Rll: Cancer experience among coke by product workers. Ann NY Acad Sci 271:102- 115.1976 (45) Lemon RA. lee JS, Wagoner JK. et al* Cancer mortality among cadmium production workers. Ann NY Acad Sci 271: 274-279.197ft (46) Monson RR. Nakano KK: Mortality among rubber workers. L White male union employees in Akron. Ohio. Am) Epidemiol 103:284-296,1978 (47) Aehcsoa ED: Nasal cancer in the furniture and boot and shoe manufacturing industries. Prev Med 8295-315,1978 (48) Brinton LA: A death certificate analysts of nasal cancer among furniture workers in North Carolina. Cancer Res 37: 3473-3474. 1978 (49) Aksoy M. Erdem & Din^o) C: Leukemia in shoe-workers exposed chronically to benzene. Blood 44:837-841.1974 (56) Cole P. Coldman MB: Occupation. In Persons at High Risk of Cancer (Fraumeni JF Jr, ed). New York: Academic Press. 1975. pp 167-184 (57) Slugs T). Sivak A. Boutweii RK. eds: Mechanisms of Tumor Promotion and Cocarcinogenesis. Carcinogenesis, a Comprehensive Survey, vol 2. New York: Raven Press. 1976 (52) Colbum NH: Tumor promotion and prencoplastic progression. In Modifers of Carcinogenesis. Carcinogenesis, a Comprehensive Survey (Slaga TJ. ed). vol 7. New York: Raven Press, in press (531 international Agency for Research on Cancer. Aflatoxins. (ARC Monogr Eval Carcinog Risk Chem Man 10:51-72.1978 (54) --: 2-Naphthylamine. IARC Monogr Eval Carcinog Risk Chem Man 437-111,1974 (55) Magee PN. Montesano R. Preusaman R: N-Nitroso compounds and related carcinogens. In Chemical Carcinogens (Sesrte CE. ed). American Chemical Society Monograph 173. Washington. D.C: Am Cham Soc. 1976. pp 491-825 (56) Safe Drinkinf Water Committee. Advisory Center on Toxicology, National _ Research Council National Academy of Sciences: Drinking Water and Health. Washington. D.C: Natl Acad Sci. 1977 (57) National Cancer Advisory Board: General criteria for assessing the evidence for carcinogenicity of chemical substances: Report of the Subcommittee oo Environmental Carciiwsenesis. National Cancer Advisory Board.) Natl Cancer Inst 58:461-405.1544.1977 (58) Fraummi |F |r. ed: Persons at High Risk of Cancer: Aa approach to Cancer Etiology and Control Proceedings of a conference sponsored by the National Cancer Institute and the American Cancer Society, Key Bixcayne. 11a- Dec. 18-12 1974. New York: Academic Press. 197S (59) Shubik P. Sice |: Chemical carcinogenesis as a chronic toxicity test A review. Cancer Res 18:728-742.1958 (69) International Union Aftainst Cancer. Report of symposium on potential cancer hazards from chemical additives and contaminants to foodstuffs. Acta Un Ini Contra Cancr 13.170-193.1957 (67) Subcommittee on Carcinogenesis. Food Protection Committee. Food and Nutrition Board. National Academy of Sciences- National Research Council: Problems in the evaluation of carcinogenic hazard from use of food additives- Cancer Kes 21:429-458.19G1 (62) Joint FAO/WHO expert committee on food additives: Fifth report of ths Joint FAO/ WHO expert committee an food additives. Evaluation of Carcinogenic Hazards of Food Additives. WHO Tech Rep Ser 220:1-32.1901 (63) WHO Expert Committee on the Prevention of Cancer Report of the WHO Expert Committee on the Prevention of Cancer. WHO Tech Rep Ser 278:1-53.1984 (64) Berenblum L ed: Carcinogenicity Testing. Union Internationale Contra ie Cancer Tech Rep Ser. vol 2. Ceneva. Switzerland: U1CC. 1968 (65) WHO Scientific Croup: Report of the WHO Scientific Croup on Principles for the Testing and Evaluation of Drags for Carcinogenicity. WHO Tech Rep Ser 428:1- 28.1969 (65) Advisory Pane) on Carcinogenicity of Pesticides: Carcinogenicity of pesticides. In Report of the Secretary's Commission on Pesticides and Their Relationship to Environmental Health. U.S. Dept Health. Educ. Wetfara. Washington. D.C: U.S. Govt Print OfT. 1969. pp 459-508 (67) Ad Hoc Committee on the Evaluation of Low Levels of Environmental Carcinogens: Evaluation of environmental carcinogens-- report to the Surgeon General. In Chemicals and the Future of Men. Hearings before the Subcommittee on Executive Reorganization and Government Research of the Committee on Government Operation* of the U.S. Senate, 92d Congress. 1st session. Washington. D.Cu U.S. Govt Print Oft 1971, pp 171-183 -- (68) Food and Drag Administration Advisory Committee on Protocols for Safety Evaluation: Pane) on carcinogenesis report on cancer testing in the safety evaluation of food additives and pesticides. Toxicol App( Pharmacol 20:419-436.1971 (89) Health snd Welfare, Canada: The Testing of Chemicals for Carcinogenicity. Mutagenicity, and Teratogenicity. Ottawa. Canada: Ministry Health Welfare, 1973 [70] WHO Scientific Croup: Report of the W1)0 Scientific Croup for the Assessment of the Carcinogenicity and Mutagenicity of Chemicals. WHO Tech Rep Ser 548:1-19,1974 (77) Sontag JM. Page KP. Saffiotti U: Guidelines for carcinogen bioassays in small rodents. Natl Cancer Inal Carcinogenesis Tsch Rep Ser No. 1. Nad Inst Health. DilEW Publ No (NIH) 78-801. Washington. D.C: U.S. Covt Print Off. lftTft [72] Albert RE. Train RE. Anderson E: Rationale developed by the Environmental Protection Agency for the assessment of carcinogenic risks.) Natl Cancer Inst 56:15371541.1977 (73) Hiatt 101 Watson )D. Winsten |A. eds: Origins of Human Cancer. Cold Spring Harbor Conferences oo Cell Proliferation, vol DU 009610 DUP 0811087 39078 Federal Register / Vol. 44. No. 131 / Friday. July 6.1979 / Notices 4. Cold Spring Harbor. N.Yj Cold Spring Harbor Laboratory, 1977 [74) Klaasscn CD: Absorption, distribution and excretion of toxicants. In Toxicology, the Basic Science of Poisons (Casarctt!.(. Doll J, eds). New York: Macmillan. 1975. pp 2b--44 (75] Bischoff F. Bryson G: Carcinogenesis through solid state surfaces. Prog Exp Tumor Res 5:85-97.1964' (78) Brand KG: Foreign body induced sarcomas. In Cancer, a Comprehensive Treatise (Becker FF. ed). vol 1. New York and London: Plenum Press. 1975. pp 485-511 (77) Rice )M: Carcinogenesis: A late effect of irreversible toxic damage during development. Environ Health Perspcct 18:133-139.1978 (78) Rice |M, ed: Perinatal Carcinogenesis. Natl Cancer Inst Monogr 51:1-282.1979 (79) Food and Drug Administration: Kondinical Laboratory Studies. Good Laboratory Practice Regulations. Fed Register 43: No. 247. 59908-80025.1978 (SO) Saffiotti U. Page NP: Releasing carcinogenesis teat results: Timing and extent of reporting. Med Pediatr Oncol 3:159-167, 1977 (87) Farber E Sporn MB. cochairmen: Symposium: Early lesions end the development of epithelial cancer. Cancer Res 36:2475-2708, 1878 (82) Stewart FW: Temors of the breast. In Atlas of Tumor Pathology, sect IX, fasc 34. Washington, D.G: Armed Forces Inst PathoL 19SO. pp 7-10 (82) Young S. Hettowes RC: Tumours of the mammary gland. In Pathology of Tumors in Laboratory Animala (Turusov VS. ed). vol 1. part 1.1ARC Sci Publ 5:31-74,1973 (84) Shcllabarger C|: Mammary neoptastic response of Lewis and Sprague-Dawley fem.t'.e rats to 7,12-dimethylbenz(a)anthracone or X-ray. Cancer Res 32:583-685.1972 (85) Baserga R. SefTiotti U: Experimental studies on histogenesis of blood-borne metastases. Arch Pathol (Chicago) S9-.26-34. 1955 (88) Kyriazis AP, Koka M. Vestelinovilch SO: Metastatic rate of liver tumors induced by diethylnitrosamine in mice. Cancer Res 34:2681-2886,1974 (87) International Agency for Research on Cancer. Benzidine. IARC Monogr Evil Carcinog Risk Chem Man 1:80-66.1972 (88)----- : Vinyl chloride. IARC Monogr Evil Carcinog Risk Chem Man 7:291-318. 1974 (89) Hanna MG Jr, Neltesheim P, Cilbert JR. eds: Inhalation Carcinogenesis. Atomic Energy Commission Symp Ser No. 18 (CONF691001). Oak Ridge. Tcnn.: U.S. Atomic Energy Comm. Div Tech Information Extension. 1970 (90) riarbo E. Park JF. eds: Experimental Lung Cancer. Carcinogenesis and Bioassays. Berlin, Heidelberg, New York: SpringerVerlag. 1974 (97) Clayson D6: Bladder carcinogenesis in rats and mice: Possibility of artifacts. J Natl Cancer Inst 52:16*5-1689.1974 . (92) Miller EC Milter JA: The metabolism of chemical carcinogens to reactive electrophiles and their possible mechanisms . of action in carcinogenesis. In Chemical Carcinogens (Searle CE ed). American Chemical Society Monograph 173. Washington. D.C.: Am Chem Soc, 1978. pp 737-782 (92) Fears TR. Tamone RE Chu KC: False positive and false negative rales for carcinogenicity screens. Cancer Res 37:1941- 1945.1977 (97| Hoel DC. Walburg HE Jr. Statistical analysis of survival experiments. J Natl Cancer Inst 49:381-372,1972 (95) Peto R: Guidelines on the analyses of tumour rates and death rates in experimental animals. Br J Cancer 29:191-105.1974 IPS) Thomas DC. Breslow N. Cart JJ: Trend and homogeneity analyses of proportions and life table data. Comput Biomed Res 10:373381.1977 (97) Methods for Carcinogenesis Tests at the Cellular Level and Their Evaluation for the Assesment of Occupational Cancer Hazards. Proceedings of the meeting of the Scientific Committee. Milan. Italy. Dec. 4-8. 1977. Milan: Fondazione Carlo Erba. 1977 (98) Working Group on Mutagenicity Testing. Subcommittee on Environmental Mutagenesis. U.S. Department of Health, Education, and Welfare Committee to Coordinate Toxicology and Related Programs: Approaches to determining the mutagenic properties of chemicals: Risk to future generations. J Environ Pathol Toxicol 1:301-352.1977 (99) Saffiotti U, Autrup ii eds: In Vitro Carcinogenesis. Guide to the Literature, Recent Advances and Laboratory Procedures. Natl Cancer Inst Carcinogenesis Tech Rep Ser No. 44. Nad Inst Health. DHEW Publ No. (NIH) 78-844. Washington, D.G: U.S. Covt Print Off. 1978 (700) Asher 1M. Zervos G Eds: Structural Correlates of Carcinogenesis and ' Mutagenesis: A Guide to Testing Priorities? Proceedings of the Second Food and Drug Administration Office of Science Summer Symposium. Annapolis. Md. Aug. 31-Sept 2. 1977. Rockville. Md.: Food Drug Admin. 1978 (707) International Commission on Radiological Protection: Radiation Protection--Recommendations of the1 .International Commission on Radiological Protection. 1CRP Publ 8 Oxford: Perganvon. Press. 1968 (102) Task Croup: Air pollution end cancer: Risk assessment methodology end epidemiological evidence. Environ Health Perspect 22:1-12.1978 (702) Environenta) Studies Board. National Research Council. Nationat Academy of Sciences: Carcinogenesis in man end laboratory animals. In Pest Control: An Assessment of Present end Alternative Technologies. Vol 1. Contemporary Pest Control Practices and Prospects: The Report of the Executive Committee. Washington, D C.: Natl Acad Sci. 1975. pp 66-82 (104) Cornfield |: Carcinogenic risk assessment. Science 198:893-609,1977 (705) --: Models lor carcinogenic risk assessment. Science 202:1107-1109.1978 (708) Hoel DC. Caylor DW. Kirschstein RE et xl: Estimation of risks of irreversible, delayed toxicity. ] Toxicol Environ Health 1.133-151.1975 (707) Mantel N. Bryan WR: "Safety" testing of carcinogenic agents. J. Natl Cancer Inst 27:455-470.1961 (!0g) Mantel N. Bohidar NR. Brown CC. et jjl: improved Mantel-Bryan procedure for "safely" testing of carcinogens. Cancer Res 3S:8AS-872.1*75 (100) Crump KS. Hoel DC. Langley CH. et el: Fundamental carcinogenic processes and their implications for low dose risk assessment. Cancer Res 36.2973-2979.197# (770) Cuess HA. Crump KS. Peto R: Uncertainty estimates for low-dose-rste extrapolations of animal carcinogenicity data. Cancer Res 37:3475-3483.1977 (777) Gross MA. Fitzhugh OG. Mantel N: Evaluation of safety for food additives: An Illustration involving the Influence of methyl ealieytate on ret reproduction. Biometric* 26:111-1*4.1978 (112) Cox DR: Regression models and lifetables. J R Stal Soc (B) 34:187-202.1972 (113) Goess HA, Crump KS: Low-dose-rate extrapolation of date from animal carcinogenicity experiments--analysis of a new statistical technique. Math Biosci 32:1536,1978 (774) Crump KS. Cuess HA. Deal KG Confidence intervals end lest of hypotheses concerning dose response relations inferred from animal carcinogenicity data. Biometrics 33:437-451.1977 (775) Hartley HO. Sielken RE Estimation of "safe doses" in carcinogenic experiments. Biometrics 33:1-30,1977 (775) Armitege P. Doll R:.Stochastie models for carcinogenesis. In Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability. Berkeley. Calif., June 20-July 30.1960 (Neyman J. ed). vol 4. Berkeley. Calif.: Univ California Press. 1961. pp 19-3* (777) Druckrey H: Quantitative aspects of chemical carcinogenesis. In Potential Carcinogenic Hazards From Drugs (Evaluation of Risks) (Truhart RT. ed), Unio Internationale Centre la Cancer Monograph Ser. vol 7. Berlin: Springer-Veriag. 1967. pp 80-78 (778) Albert RE Altshuler 8: Considerations relating to the formulation of limits for unavoidable population exposures to environmental carcinogens. In Radionuclide Carcinogenesis (Senders CE Busch RH, Ballou )E et el. eds). Atomic Energy Commission Symp Ser No. 29 (CONF- 720805). Springfield. Va.: U.S. Atomic Energy Comm Office of information Services. 1973, pp 233-253 (779) Chand N, Hoel DC: A comparison of models for determining safe levels of environmental agents. In Reliability and Biometry: Statistical Analysis of Lifelength (Proschen F, Serfling RJ, eds). Philadelphia: Soc indust Appl Math. 1973. pp 681-700 (120) Schneiderman MA. Decoufle P, Brown CC: Thresholds for environmental cancer-- biological and statieticel considerations. NY Acad Sci. In pres* (121) Peto R: Epidemiology, multistage models, snd short-term mutagenicity tests. In Origin* of Human Cancer. Cold Spring Harbor Conferences on Cell Proliferation (Hiett HR Wilson JD, Winsten JA. eds). vol 4. book G Cold Spring Harbor. N.Y.: Cold Spring Harbor Laboratory, 1977 pp 1403-1428 (122) Whittcmare A. Altschuler B: Lung cancer incidence in cigarette smokers: Further analysis of Doll and Hill's data for DUP 0811088 DU 009611 federal Register / Vol. 44. No. 131 f Friday. |uly i. 1#T9 / Notice* British physicians. Biometric* 3240S-tft, (f2J) Second Task Fore* for Research PlMtal to Environmental Health Science; Eavkamaental measurement* of chemicals iar aeeesement oi human exposure. chapt 7. dr Human Health and the Environment Soma BaMamh Needs. Nad Inst Health. DHEW Ml Mm (NIH) 77-1277. Wwfcingtoa. D.C: D*ft Health. Educ. Welfare. 1*77, pp 217-242 (2M| Bingham E, Falk H: Environmental : The modifytaa effect of i on the thraalaM luapoc**. Arch Health lfc77a-7M, ww (124 Harris CC. Autnap H. Btonar C. et ah Matahollim of ben*o(*4p|Wma and 7,12dtmethytbenzfa|anthraaMB In cultured human bronchus and paMMMc duct Cancer Baa artiiin nil. if77 (120) Harris CC Autrup ft Oanncr R. et ah htarindividual variatiau in htadin* of banoaMpyrane to DMA h unbared human breach!. Science lM:IMMMh 107S (227) Harris CC. Autrup It Btoner C; llstahollim of benm(a|pyMM in cultured human Maaoes and cdls. la Mjrcjrdic Hydrocarbon* and Canesr (Tso PO. Calboin HV, ads), vol 2. New York and London: Aoadamic Press. 187*. pp 221-242 fu u.ie munuir^nsa i DU 009612 DUP 0811089