8104

8/14/2019 8104

http://slidepdf.com/reader/full/8104 1/240

8/14/2019 8104


Library of Congress Catalog Card Number 81-600081

For sale by the Sup erintendent of Documents,

U.S. Governm ent P rinting Office, Washington , D.C. 20402

8/14/2019 8104


Foreword

Congressional interest in cancer is long standing and continuing. Programs in basic cancerresearch, and in treatment and prevention of the disease are now complemented by some twodozen laws directed at reducing exposures to cancer-causing substances. This report examinesthe technologies used to gather and analyze information about cancer in our society, as well asthe ways in which those technologies affect and are affected by the public health and en-vironmental legislative mandates.

The report discusses the strengths and weaknesses of data sources used for determining

trends in cancer occurrence and mortality, and reviews estimates of the contribution of variousfactors—behaviors and exposures—associated with cancer in th is country. Evidence linking to-day’s cancers with past carcinogenic influences has come mainly from epidemiology, whichcontinues to scrutinize aspects of the American lifestyle, for p ossible associations w ith cancer.

Congressional mandates intended to shield people from new and already-present car-cinogens have heightened the need for methods to identify such harm ful agents before theyhave an impact on human health. Laboratory testing technologies currently used to determinethe carcinogencity of substances, and technologies that may become important in the near fu-ture are d iscussed and evaluated, The assessment examines the use of extrapolation techniquesfor estimating human carcinogenic risks from test-derived data; the advantages and disadvan-tages of the available extrapolation models; and the ultimate use of these techniques in settingstandards for controlling exposures und er diverse legislation. The report then looks at the prob-lems of decisionmaking in the face of the often-great u ncertainties accompanying scientific find-

ings and the proposals for regulatory reform that have grown ou t of concern for these issues.In preparing the full report, OTA staff consulted with members of the advisory panel for

the study, with contractors who prepared material for the assessment, and with otherknowledgeable persons in environm ental organizations, Government, indu stry, labor orga-nizations, research institutions, and universities.

A draft of the final report was reviewed by the advisory panel, chaired by Dr. NortonNelson, the OTA Health Program Advisory Committee, chaired by Dr. Sidney S. Lee, and byapp roximately 80 other individuals and grou ps. We are grateful for their assistance and that of many other people who assisted and advised in the p reparation of this report.

#

/f+//q

JOH N H. GIBBON S Director

. . .111

8/14/2019 8104


Assessment of Technologies for Determining Cancer

Risks From the Environment

Advisory Panel

Norton N elson, Panel Chairman

Department of Environmental Medicine, New York University Medical School

David AxelrodCommissioner of HealthState of New York

Peter A. A. Berle Berle, Butzel, Kass, and Case

Theodore L. Cairns E. I. DuPont de Nemours & Co.,

Inc. (retired)

Paul F. Deisler, Jr.Vice President, Health, Safety

and Environment Shell Oil Co.

George S. Dominguez Director of Government

RelationsCIBA-Geigy

David Doniger Natural Resources Defense

Council

A. Myrick FreemanProfessor of Economics

Bowdoin CollegeRobert Harris

Environmental Defense Fund

Priscilla W. LawsProfessor of Physics

Dickinson College

Mark Lepp erVice President for Evaluation

Rush-Presbyterian MedicalSchool, St Lukes Medical Center

Brian MacMahonChairman, Epidemiology

Department Harvard University School of Public Health

Robert A. N ealPresident Chemical Industry Institute

of Toxicology

Vaun A. New ill Director, Research and

Environmental Health Division

Exxon Corp.

William J. Nicholson Department of Community

Medicine Mt. Sinai School of Medicine

R . T a l b o t P a g e

California Institute of Technology

Margaret Seminario Department of Occupational

Safety and Health

American Federation of Labor/ Congress of IndustrialOrganizations

Alice S. Whittemore Division of EpidemiologyStanford University School of

Medicine

Michael WrightSafety and Health Department United Steelworkers of America

8/14/2019 8104


Assessment of Technologies for Determining Cancer

Risks From the Environment

OTA Project Staff

Joyce C. Lashof, Assistant Director, OTA Health and Life Sciences Division

H. David Banta, Health Program Manager

Michael Gough , Project Director

Robert J. Fensterheim , * Research Associate

Hellen Gelband, * Research Associate

Virginia Cwalina, Administrative Assistant Shirley Ann Gayheart, Secretary

Nancy L. Kenney, SecretaryMartha Ingram Finney, * Editor

Other Contributing Staff

Barbara Lausche, Senior Analyst Elizabeth Williams, Senior Analyst

Contractors

Richard Doll, Oxford University

Richard Pete, Oxford UniversityMichael Baram, Bracken and Baram, Boston, Mass.

Roy Albert, New York University Medical SchoolWilliam Rowe, American UniversityClement Associates, Inc., Washington, D.C.

OTA Publishing Staff

John C. Holmes, Publishing Officer

John Bergling Kathie S. Boss Debra M. Datcher Joe Hen son

q OTA contract personnel,

8/14/2019 8104


Contents

1.

2.

3.

4.

5.

6.

Summary . . . . . . . . . . . . . . . . .

Cancer Incidence and Mortality

Factors Associated With Cancer

Page

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Methods for Determining and Identifying Carcinogens . . . . . . . . . . . . . .113

Extrapolation From Study-Generated Data to Estimates of Human Risk .157

Approaches to Regulating Carcinogens. . . . . . . . . . . . . . . . . . . . . .. ...175

Appendixes

A. A Comparison of the National Institute’s and the

International Agency for

of Bioassay Results . . . .

B. Unreasonable Risk . . . .

Research on Cancer’s Evaluation

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

C. Abbreviations and Glossary of Terms. . . . . . . . . . . . . . . . . . . . . .. ....221

References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ........227

vi i

8/14/2019 8104


1 .

Summary

8/14/2019 8104


8/14/2019 8104


1

Summary

Cancer occupies center stage in Americanconcern about disease because of its toll in lives,

suffering, and dollars. It strikes one ou t of fourAmericans, kills one out of five, and as thesecond -leading cause of death, following h eartdisease, killed over 400,000 people in the Un itedStates in 1979. According to estimates from theNational Center for Health Statistics (NCHS),cancer accounted for about 10 percent of theNation’s total cost of illness in 1977. These nu m-bers are distressing, but the impacts of cancerextend beyond the numbers of lives taken anddollars spent. The human suffering it causestouches almost everyone.

CANCER AND “ENVIRONMENT”

Studies over the last two decades yielded avariety of statements that 60 to 90 percent of cancer is associated with the environment andtherefore is theoretically preventable. As it wasused in those statements and is used in thisreport, “environment” encompasses anythingthat interacts with humans, including sub-stances eaten, drunk, and smoked, natural andmedical radiation, workplace exposures, drugs,

aspects of sexual behavior, and substances pres-ent in the air, water, and soil. Unfortunately,the statements were sometimes repeated w ith“environment” used to mean only air, water,and soil pollution.

Relating exposures and behaviors to canceroccurr ence is a first step in cancer prevention.Once carcinogenic influences are identified, ef-forts to control them can be undertaken toward

Cancer is a collection of about 200 diseasesgrouped together because of their similar

growth processes. Each cancer, regardless of thepart of the body it affects, is believed tooriginate from a single “transformed” cell. Atransformed cell is unrespon sive to normal con-trols over growth, and its progeny may growand multiply to produce a tumor. Studies inhuman populations and in laboratory animalshave linked exposures to certain su bstances w ithcancer. This knowledge of cancer’s origins hasled to the conclusion that preventing interac-tions between cancer-causing substances andhumans can reduce cancer’s toll.

CANCER MORTALITY AND INCIDENCE

the goal of reducing cancer. This study is in-tended to illuminate the debates about the im-porta nce of environmental factors in cancer oc-currence, the laws that requ ire actions to redu ceexposures to cancer-causing substances (car-cinogens), and describes:

q what is known about the occurrence of

cancer and death from cancer in the Un itedStates;

. methods to identify cancer-causing sub-stances, exposures, and behaviors;

q methods to estimate th e amount of cancer

which may result from a particular be-havior or exposure;

q Federal laws that provide for regulatorycontrol of carcinogenic exposures; and

q options for Congress.

Nationwide mortality data are used to an- doubt, the number of Americans dying fromswer questions about the number of deaths cancer has increased during the last century.caused by cancer in the United States. Withou t Paradoxically, a major part of this increase re-

3

8/14/2019 8104


4 qTechnologies for Determining Cancer Risks From the Environment

suited from improvements in public health andmedical care. In yearn past, infectious diseaseskilled large numbers of people in infancy andduring childhood. Now that improved healthcare has softened the imp act of those diseases,many more people live to old ages when cancercauses significant mortality.

Cancer deaths are not evenly distributedamong all body sites, the lung, colon, andbreast accounting for over 40 percent of thetotal (see table 1). Changes in cancer rates overtime also vary by body site. For this reason,discussion of cancer rates at particular bodysites is more revealing than discussion of overalltrends wh ich mask changes at ind ividual sites.Moreover, because some cancer-causing sub-stances act at specific’ sites, more informationabout opportunities for prevention is obtainedfrom the analysis of particular sites.

To permit the examination of cancer ratesover time, standardization, a statistical tech-nique, is applied to make allowances for achanging population structure. Standardizationallows the direct comparison of single, sum-mary statistics, e.g., the mortality rates fromlung cancer for the entire population in 1950and 1981. In this report, mortality rates arestandardized to the age and racial structure of the 1970 U.S. census, u nless otherwise sp ecified.

Age-specific rates are also used extensivelyfor examining trends. These rates measure the

prop ortion of people in d efined age classes whohave developed or d ied from cancer, and are un-affected by changes in the age structure of thepopulation. Of greatest importance in detectingand identifying carcinogens, changes over timein younger age groups often presage future,larger changes in that group of people as theyenter older age groups.

In general, cancer mortality rates are higheramong nonwhite males than among whitemales. Differences between nonw hite and wh itefemales are less pronounced. The observedgreater fluctuations in rates from year to yearfor nonw hites is consistent with the conclusionthat reporting of vital statistics is poorer fornonw hites than for wh ites.

Greatest concern is expressed about the in-creasing trends. The largest increases since 1950are in respiratory cancers (mainly of the lung,

larynx, pharynx, trachea), wh ich ar e largely as-cribed to the effects of smoking. Male respira-tory cancer rates began to rise about 25 yearsearlier th an female rates, w hich reflects the d if-ference in time when the two sexes adoptedsmoking. Further evidence for the imp ortance of smoking in lung cancer is the recent decrease inlung cancer mortality among males youngerthan 50. The percentage of males wh o smoke isknown to have decreased during the last 20years, and studies have shown that smokingcessation reduces lung cancer occurrence. Addi-

Table 1.– Mortality From Major Cancer Sites in th e United States, 1978, All Races

Number of deaths Percentage of total

Anatomic site Male Female Total Male Female Total

All malignant neoplasms . . . . . . . . . . . . . . . . . . . . . 215,997 180,995 396,992 100% 100% 100%

Lung, trachea, and bronchus . . . . . . . . . . . . . . . . . . 71,006 24,080 95,086 32.9 13.3 24.0Colon. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20,694 23,484 44,178 9.6 13.0 11.1Breast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280 34,329 34,609 0.13 19.0 8.7Prostate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21,674 — 21,674 10.0 5.5Pancreas. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

—

11,010 9,767 20,777 5.1 5.4 5.2Blood (leukemia) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8,683 6,708 15,391 4.0 3.7 3.13Uterus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . — 10,872 10,872 — 6.0 2.7Ovary, fallopian tubes, and broad ligament. . . . . . . — 10,803 10,803 — 6.0 2.7Bladder. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6,771 3,078 9,849 3.1 1.7 2.5Brain and other parts of nervous system. . . . . . . . . 5,373 4,362 9,735 2.5 2.4 2.5Rectum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5,002 4,089 9,091 2.3 2.3 2.3Oral: Buccal cavity and pharynx. . . . . . . . . . . . . . . . 5,821 2,520 8,341 2.7 1.4 2.1Kidney and other urinary organs . . . . . . . . . . . . . . . 4,809 2,916 7,725 2.2 1.6 1.9Esophagus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5,552 2,030 7,582 2.6 1.1 1.9Skin. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3,537 2,511 6,048 1.6 1.4 1 .5All other . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45,785 39,446 85,231 21.2 21.8 21 .5

SOURCE: Office of Technology Assessment.

8/14/2019 8104


Ch. l—Summary q 5

tionally, changes in cigarette composition arethought to contribute to a reduced risk of lungcancer. Decreases among m en now over 50 ar enot expected because those populations includea large prop ortion of long-time smokers w ho re-main at h igh risk.

Death rates from prostate and kidney cancersamong males have risen somewhat, and mortal-ity rates from malignant skin tumors (melano-mas) have increased in wh ite males and females.Mortality from breast cancer, the number onecancer killer of women, has remained relativelyconstant. Overall mortality from nonrespira-tory cancers (i.e., excluding most cancers gener-ally associated w ith smoking) has d ecreased infemales and remained constant in males duringthe last 30 years.

The more satisfying trend s are those that aredecreasing. The most striking, among both men

and women, has been the great decrease instomach cancer since 1930. Although generallyascribed to changes in diet, the reasons for thedecrease are not known with any certainty. Adecrease in uterine cancer within the last fewdecades is attributed to higher living stand ards,better screening tests for early cancer, and an in-crease in hysterectomies, which reduces thenum ber of wom en at risk.

In general, mortality data (numbers of deaths) are considered more reliable for de-ciding about trends in cancer occurrence than

are data about cancer incidence (numbers of new cases). This is largely because nationw idemortality data have been collected on a regu larbasis for almost 50 years. In contrast, incidencedata for a sample of the entire country havebeen collected systematically only since 1973 bythe N ationa l Cancer Institu te’s (NCI’s) Sur veil-lance, Epidemiology, and End Results (SEER)progr am. Before that, incidence dat a are avail-able only for three points in time since 1937.The lo-percent samp le of the popu lation includ-ed in the SEER areas is not representative of theentire population. Some groups—orientals—are

overrepresented in the data collected, and somegroups—rural blacks—are underrepresented.Incidence rates for nonwhites, at least duringthe first 4 years of the SEER program, were con-sidered too un reliable for m eaningful analysis.

Incidence data are important because theyprovide information not captured in mortalitydata. They record each new case of cancerwhether the person dies from cancer, is cured,or d ies from other causes.

Followup studies of SEER program partici-pants have provided information about survivalfrom the var ious typ es and stages of cancer. Aproblem encountered in such stud ies was thatpeople who move from the registration areaafter treatment are sometimes lost to furtherstudy, making it difficult to ascertain whetherthey eventually succumb to cancer or if treat-ment cur ed them. Use of the new ly established(1981) National Death Index, by which deathscan be identified through a single query toNCHS rather than through a request to everyState, is expected to facilitate SEER program

followup studies. If this expectation is realized,information from the “End Results” componentof SEER should be improved.

Data collected in the SEER program (1973-76), in combination with data from the ThirdNational Cancer Survey (TNCS), carried outfrom 1969 throu gh 1971, have been interpretedas showing an increase of more than 10 percentin cancer incidence du ring the last decade. Themajor changes seen in the incidence data p arallelthose seen in mortality data—increases in lun gcancer and decreases in stomach and uterine

cancers. However, publication of this analysissparked a controversy about th e true natu re of incidence trends, since only 2 years earlier ananalysis of data from the three national cancersurveys had shown an overall decrease of about4 percent between 1947 and 1970. Some ob-servers are concerned about th e p ossibility that,after at least half a century of stable or decliningrates, cancer incidence has gone up and that theincrease might result from newly introducedchemical carcinogens. Those who dispute theimportance of the observed increase contendthat it reflects changes in the reporting of cancer

incidence between TNCS and SEER (1973through 1976), and not real changes in cancerinciden ce. As more d ata are collected du ring thenext few years, a clearer picture of incidencetrends may emerge.

8/14/2019 8104


8/14/2019 8104


Ch. 1—Summary q 7

The magnitu de of associations between air andwater p ollution and cancer are argued and stu-dies to examine the associations are difficult todesign and execute. The same is true of asso-ciations between consumer p rodu cts and cancer.

There is no disputing that occupational ex-posures to asbestos and some chemicals have

caused human cancer, and table 2 presents es-timates both for asbestos-caused cancer andtotal occupationally associated cancer. As thedata in the table show, there is significant dis-agreement about how much current cancer andcancer in the near futu re is to be associated w ithoccup ational exposur es.

Associating a high or low percentage of can-cer with a factor does not reflect the present-dayopp ortunities for prevention. For instance, dietis considered very important, but because asso-ciations with specific elements and cancer are

poorly und erstood, there are few p ractical pre-ventive measures now available.

The opportu nities for prevention of occup a-tion-related cancers at this time are better. Iden-tification of a cancer-causing substance in theworkplace can lead to reductions in exposureeither by regulation or through voluntary ac-

tivities on the pa rt of indu stry. While redu cingor eliminating occupational exposures to car-cinogens m ight only slightly red uce the overallcancer toll, it could have a profound effect onthe amount of cancer among workers who maynow be at risk. A redu ction of only 1 percent incancer mortality means 4,000 fewer cancerdeaths each year, so that even small redu ctionstranslate into relatively large nu mbers.

IDENTIFICATION OF CARCINOGENS

The Federal Government has centered effortsto control cancer on reducing exposures tochemical and physical carcinogens.

Carcinogens can be identified through epide-miology—the study of diseases and their de-terminants in human populations—and throughvarious laboratory tests. Currently 18 chemicalsand chemical processes are listed as humancarcinogens and an ad ditional 18 listed as p rob-

able human carcinogens by the InternationalAgency for Research on Cancer (IARC), aWorld H ealth Organization agency. IARC con-clusions, based on reviews of the worldwideliterature, are accepted as authoritative byg o v e r n m e n t a g e n c i e s a n d m a n y o t h e rorganizations.

In the United States, Congress has directedthe National Toxicology Program (NTP) to pro-duce an annual list of carcinogens, The first list,published in 1980, was composed of the sub-stances identified as human carcinogens byIARC. The next publication is to be consider-ably expanded and will include usage and ex-posure d ata and information on the regulatorystatus of over 100 chemicals either considered tobe carcinogens or regu lated by th e Federal Gov-ernment because of carcinogenicity.

Cancer epidemiology established the associa-tions between the 36 substances and human can-cer listed by IARC as w ell as the carcinogenicityof smoking, alcohol consumption, and radia-tion. However, epidemiology is limited as atechnique for identifying carcinogens becausecancers typically ap pear year s or d ecades afterexposure. If a carcinogen were identified 20years after its widespread u se began, many peo-ple might d evelop cancer from it even thou gh itsuse is then immed iately d iscontinued . Certainly,those people who w ere identified in the study ashaving had their cancer caused by the substancewould have been irreparably harmed. Epidemi-ology is complicated because people are difficultto study; people move from place to place,change their type of work, change their habits,and it is hard to locate them and to estimatetheir past exposures to suspect agents.

Laboratory tests, which do not depend onhum an illness and d eath to produ ce data, havebeen developed to identify carcinogens. Cur-rently, the testing of suspect chemicals inlaboratory animals, generally rats and mice, isthe backbone of carcinogen identification. Thesuspect chemical is administered to the animals

8/14/2019 8104



Table 2.-Summary of Cancer-Associated Environmental Factorsa

Range of estimatesFactor b Sites considered in drawing the estimates associated with factor

Diet Digestive tract, breast, endometrium, ovary 35-50 percent

Associations between diet and cancer are suggested by epidemiologic and experimental laboratory studies. Significant dif-ferences in cancer rates are observed between different population groups with varying eating habits. Dietary components,such as high-fat and low-fiber content, and nutritional habits that affect hormonal and metabolic balances are believed moreimportant than additives and contaminants. The magnitude of the estimates reflect observed relationships between diet and

prominent cancer sites, e.g., breast and colon.Tobacco Upper respiratory tract, bladder, esophagus, kidney, pancreas 22-30 percent

Tobacco is associated with cancer at many anatomical sites, principally the lung. Many estimates of the proportion of overallcancer mortality associated with tobacco smoking are firmly based on epidemiologic studies that compared cancer mortalityamong individuals with varying smoking habits. Several carcinogens act synergistically with tobacco, e.g., asbestos, alcohol,radiation.

Occupation, asbestos Upper respiratory tract, others 3-18 percent

Several occupational exposures are firmly linked to cancer occurrence, the most important of these is asbestos. Estimates forthe contribution of asbestos to current cancer deaths and cancers in the near future range from 3 percent (1.4-4.4 percent) toan upper estimate of 13-18 percent. Most estimates Iie toward the lower end of the range. The exposures responsible for thesecancers occurred primarily in the 1940’s and 1950’s and the resultant cancers are expected to peak in the early to mid-1980’s.

Occupation, all exposures Upper respiratory tract, others 4-38 percent

Estimates of the proportion of cancer associated with all occupational exposures range from 4 percent (2-10 percent) to ahigh of 23-38 percent. The higher estimates are from a paper that estimated that asbestos is associated with 13-18 percent ofall cancer and added to that estimates of cancer associated with five other occupational exposures. Almost all otherestimates are near the lower end of the range.

Alcohol Upper digestive tract, larynx, liver 3-5 percent

Alcohol consumption is associated with cancer in the upper digestive tract and in the liver. The digestive tract cancers occurmore frequently in smokers than nonsmokers, and therefore many of these cancers could be prevented if either tobacco oralcohol were discontinued. The majority of reliable estimates are based on apportioning a percentage of the cancers at thealcohol-related sites to alcohol, and the numerical estimates are very similar.

Infection Uterine cervix, prostate, and other sites 1-15 percent

Epidemiologic data strongly suggest an association between a virus and cervical cancer, and cancer at that site accounts forthe lower numerical estimate. The higher estimate is much more tentative and associates all urogenital cancers in both sexeswith infections of venereal origin. Some other cancers which occur commonly in other parts of the world are stronglyassociated with viral infection. They are rare in the United States.

Sexual development,reproductive patterns, andsexual practices Breast, endometrium, ovary, cervix, testis 1-13 percenfc

All of the hormonally related cancers in women, breast, endometrial, and ovarian are believed associated with sexual develop-ment and reproductive patients. The important characteristics are: 1) age at sexual maturity; 2) age at birth of first child; 3) age

at menopause. The higher numerical estimate includes the large number of breast cancers. Testicular cancers are associatedwith developmental and hormonal abnormalities.

Pollution Lung, b/adder, rectum Less than 5 percent

Air pollution: Several epidemiologic studies of the effects of air pollution demonstrate an increased risk of lung cancer inheavily polluted areas, but these conclusions are weakened because smoking and occupational exposures were not alwaystaken into account. The most important carcinogens are believed to be combustion products of fossil fuels. There is con-tinued concern that chlorofluorocarbons introduced into the atmosphere may deplete the ozone layer. This would result inmore ultraviolet light reaching the surface of the Earth and increase the number of cases of skin cancer.

Drinking water pollution: Many carcinogenic chemicals have been identified in drinking water but the extent to which past andpresent levels contribute to the overall cancer rate is uncertain. Several descriptive epidemiologic studies have suggested anassociation with an increased risk of cancer but the studies are plagued by confounding variables. A soon to be released NCIepidemiologic study is expected to provide more definitive evidence regarding the association between quality of drinkingwater and bladder cancer.

Medical drugs and radiation Breast, endometrium, ovary, thyroid, bone, lung, blood (leukemia) 1-4 percent

Drugs known to be carcinogenic are used in the treatment of diseases, including some cancers. In addition, hormonal ther-

apies, particularly the estrogens, are firmly linked to an increased cancer risk. Medical radiation exposures are known to havecaused cancer and while dosage levels can be estimated, the level of risk from present day exposures is uncertain.

8/14/2019 8104


Ch. l—Summary q 9

Table 2.—Summary of Cancer-Associated Environmental Factorsa—Continued

Range of estimatesFactor b

Sites considered in drawing the estimates associated with factor

Natural radiation Skin, breast, thyroid, lung, bone, blood (leukemia) Less than 7-3 percent

There is no doubt that natural radiation, consisting of ionizing radiation from cosmic rays and radioactive materials, can causecancer. While disagreements persist regarding the amount of risk associated with low-level ionizing radiation, the estimatesgenerally agree within one order of magnitude. Ultraviolet radiation from the Sun is believed responsible for most of the

400,000 nonmelanoma skin cancers. These tumors are not usually included in quantitative estimates of cancer rates becausethey are poorly recorded and generally curable. They are not included here.

Consumer products Possibly all sites Less than 1-2 percent

Substances known to be carcinogenic are present in consumer products at usually very low levels. The extent to which theycontribute to the overall cancer rate is uncertain.

Unknown associations All sites (?)

Many substances have not been tested for carcinogenicity and associations between some of those substances and cancermay exist. Furthermore, substances newly introduced into the environment may have an impact in the future. In particular,there is concern that point sources of pollutants, such as dumps, may be contributing to cancer. Because the associations areunknown, the estimate is uncertain but it is certainly not zero. Additionally, stress, which may be manifested by overeating,smoking, or in other ways, probably plays a role in cancer causation.

aManY cancers ma y be associated with more than one factor, Factors are not mutually excluswe, and the total, If al I associationswere known, would add to much more

than 100 percentbEst

lmat

es are Ilsted under th

efactors that most closely approx,rnate the description published With them The estimates are detailed and their sources referenced In

ch. 3CRange of single estimate

SOURCE” Off Ice of Technology Assessment

either in their food, water, air, or (less frequent-ly) by force feeding, skin p ainting, or injection.As the animals die, or when the survivors arekilled at the end of the exposure p eriod (which isgenerally the lifespan of the anim al), a pathol-ogist examines them for tumors. The number of tumors in the exposed an imals is then comparedwith the number in a group of “control” ani-mals. The controls are treated exactly as the ex-

perimentals except that they are not exposed tothe chemical under test. The finding of a sig-nificant excess of tumors in the exposed animalscompared with the number found in controls ina w ell-designed, w ell-executed animal test forcarcinogenicity leads to a conclusion that thechemical is a carcinogen in that species.

IARC has reviewed the literature concerning362 substances which have been tested in ani-mals and considers the d ata “su fficient” to con-clude that 121 are carcinogens. For about 100others, there was “limited” evidence of carcino-

genicity, indicating that further information isdesirable, but that the available evidence pro-duces a strong warning about carcinogenicity.Data were “insufficient” to make decisionsabout the carcinogenicity of the remaining sub-

stances. The IARC review program is active andcontinuing and updates it findings periodically.

The reliability of animal tests, bioassays, de-pends on their design and execution. NCI pub-lished guidelines for bioassays in 1976. Bio-assays now cost between $400,000 and $1 mil-lion and require up to 5 years to complete.Clearly such expensive tools should be usedonly to test highly suspect chemicals, and mucheffort is devoted to selecting chemicals fortesting.

Molecular structure analysis an d examinationof basic chemical and physical properties areused to make preliminary decisions about thelikelihood of a chemical being a carcinogen andwhether or not to test it. For instance, greatersuspicion is attached to chemicals that sharecommon features with identified carcinogens.Unfortunately, not all members of a structuralclass behave similarly, which places limits onthis approach. In making decisions about

whether chemicals should be tested further, sci-entists consider other data, including any avail-able toxicological information. These prelimi-nary decisions may be critical, because if a de-cision is made not to test a substance, nothing

8/14/2019 8104



more may be learned about its toxicity. Thewrong decision might result in a carcinogen en-tering the environment and being ignored untilit causes disease in a large number of people.

The most exciting new d evelopments in test-ing are the short-term tests, which cost from afew h und red to a few th ousand d ollars and re-quire a few days to months to complete. Suchtests have been u nd er developmen t for about 15years, and most depend on biologically measur-ing interactions between the suspect chemicaland the genetic material, deoxyribonucleic acid(DNA). The best-know n test, the “Ames test,”measures mutagenicity (capacity to cause genet-ic changes) in bacteria. Other short-term testsuse m icro-organisms, nonmammalian labora-tory animals, and cultured hu man and animalcells. Some measure mutagenicity and some thecapacity of a chemical to alter DNA metabolismor to transform a normal cell into a cell ex-hibiting abnormal growth characteristics.

Many chemicals that have already been iden-tified as carcinogens or noncarcinogens in bio-assays have also been assayed in short-termtests to measure congruence between the tw otypes of tests. Results from these “validation”studies vary, but u p to 90 percent of both car-cinogens and noncarcinogens were correctlyclassified by short-term tests. These figures aresometimes questioned because they were de-rived from studies that excluded classes of chemicals known to be difficult to classify by

the short-term tests being evaluated. However,the International Program for the Evaluation of Short-Term Tests for Carcinogenicity con-clud ed that th e Ames test, in combination withother tests, correctly identified about 80 percentof the tested carcinogens and noncarcinogens.That study purposefully included some chem-

icals known to be difficult to classify by short-term tests, and it further demonstrates thepromise of short-term tests.

Short-term tests now play an important rolein “screening” substances to aid in making deci-sions about w hether or not to test them in an i-mals. The role of short-term tests is expected toincrease in the future as more such tests are de-veloped and validated. How ever, the eventualreplacement of animal tests by short-term testsis probably some time aw ay.

One factor likely to retard replacement of ani-

mal tests by short-term tests is the poor quan-titative agreement between the two kinds of tests. Qualitative agreement, as measured invalidation studies, is good—i. e., a mu tagen isvery likely to be a carcinogen—but poor quan -titative agreement means th at a p owerful mu-tagen may be a weak carcinogen or the otherway around. Additionally, because there issome evidence to support the idea that thepotency of a carcinogen in animals is predictiveof its potency in humans, the poor agreementabout p otency between animal and short-termtests may inh ibit wider u se of the latter tests.

PROGRAMS TO IDENTIFY CARCINOGENS

Government Programs Consumer Product Safety Commission (CPSC),

The most important recent development ingovernmental management of test developmentand implementation is the establishment of NTPby the Department of Health, Education, andWelfare in 1978. The program encompasses theshort-term and bioassa y testing activities of theDepartment of Health and Human Services

and the Occupational Safety and Health Ad-ministration (OSHA), participate in the selec-tion of substances to be tested by NTP. Each of these agencies retains responsibility for develop-ment of policies and guid elines for testing an dinterpretation of results und er the laws that theyadminister.

(DHHS) but n ot the testing prog rams that exist NTP has assumed the management of the car-in other executive branch departments. Other cinogen bioassay program that was formerlyagencies w ith a stake in carcinogen testing, the located at N CI. This is the largest single test pro-Environmental Protection Agency (EPA), the gram , and began th e testing of about 50 chemi-

8/14/2019 8104


Ch. I–Summary q 1 1

cals in fiscal year 1980; the number w ill d rop toabout 30 in fiscal year 1981 because of budget-ary limitations.

Government-sponsored cancer epidemiologyis supported principally by the National In-stitutes of Health, with the National Institute of Occupational Safety and Health, and otheragencies carrying out some research. Epidemi-ologic research is marked by flexibility in ex-perimental design, and it has not been placedunder an umbrella organization like NTP.

Nongovernment

Many chemical,panics have large,and epidemiology

Programs

drug, and petroleum com-active, inhouse toxicologyunits. These resources are

employed to develop information about sub-

ANALYSES OF TEST RESULTS

Results from tests are conveniently discussedas being “positive,” “negative,” or “inconclu-sive.” A “positive” test is sufficient to conv inceall (or most) experts that the tested substancecauses the measured effect —e. g., cancer in bio-assay. Similarly, a “negative” result is one thatconvinces all (or most) experts that the testedsubstance does not exert the effect measu red inthe test. “Inconclusive” means that no conclu-sion can be d rawn from the test. Test results are

analyzed initially by the scientists who conductthe tests. Their conclusions may be reviewed byother experts later on, and such peer review isimpor tant for th e acceptan ce or rejection of theconclusions.

Positive epidem iologic results show an asso-ciation between an exposure or behavior andhuman cancer. When they are available andbased on a valid study, they tend to dominateany d ecision to be mad e about carcinogenicity.When no or limited epidemiologic data areavailable, bioassays which measure carcino-genicity in intact animals are the most impor-

tant source of information. The last decade sawGovernment organizations, Congress, executiveagencies, and the courts, as w ell as private sec-tor organizations endorse bioassays and agreethat they can be used to identify potential hu-

stances of concern to th e compan ies and also tosupply data to Federal regulatory agencies. Oneof the most modern toxicology laboratories isthat of the Chemical Ind ustry Institute of Tox-icology (CIIT). This laboratory recently com-pleted extensive testing of formald ehyde, w hichdemonstrated that the chemical causes nasal

cancer in rats. CPSC and other agencies haveproposed regulations to curtail exposures toformaldehyde based on information from CIITstudies.

Many epidemiologic studies and much of thedevelopment of test procedures take place inacademic institutions. Funding for these ac-tivities comes from both Federal and non-Fed-eral sources, and these institutions have beenimportant in gaining knowledge and improvingtechniques.

man carcinogens. For instance, IARC con-cluded:

. . . it is reasonable, for practical purposes, toregard chemicals for which there is sufficient evi-den ce of carcinogenicity . . . in an imals as if they presented a carcinogenic risk for humans.

The use of short-term test results varies de-pend ing on wh ether the substance being tested isin use or new. When making decisions aboutcurrently used chemicals, short-term test results

are used to d ecide wh ether or not to proceed to abioassay, and they are accorded a supportingrole in making decisions about carcinogenicity.

In industry, short-term tests play a role inmaking decisions about whether or not to pro-ceed with development of a new chemical. Apositive result, indicating that the cost of devel-oping th e chemical for m arket might h ave to in-clude extensive and expensive toxicological test-ing, may be factored into a m anu facturer’s deci-sion to develop or n ot to develop a chemical. Arisky chemical may be dropped from considera-

tion for further developm ent.A problem that bedevils decisionmaking is

the existence of both “p ositive” and “negative”results from tests of the same substance. Carefulanalysis of the d esign an d execution of the “pos-

8/14/2019 8104


8/14/2019 8104


Ch. l–Summary q 1 3

q

q

q

Some individuals object to any use of nu meric extrapolation. For th em, identifi-cation of a substance as a carcinogen isenough to justify efforts to reduce or toeliminate exposu re.

Other people see extrapolation as useful toseparate more risky from less risky sub-

stances.The most extensive use of extrapolation isrecommended by people who urge thatextrapolation method s be used to estimatequantitatively the amount of human cancerlikely to result from exposures. Such es-timates are seen as necessary by those whowish to compare quantitatively the risksand benefits from carcinogens.

The disagreements among the groups whohold d ifferent opinions about u se of extrapola-tion are vocal and current. A particular problem

in qu antitative extrapolation arises from the factthat different extrapolation models produceestimates of cancer incidence that differ by fac-tors of 1,000 or more at levels of human ex-posure. Given such uncertainty, some labor andenvironmental organizations and many indi-viduals refuse to choose one model or anotherfor estimating the impact of a carcinogen onhumans, and oppose the use of quantitative ex-trapolation. Fewer objections are raised againstchoosing a model to order carcinogens on thebasis of their likelihood of causing cancer.Regard less of wh ich p articular m odel is chosen,

it should produce approximately the same rel-ative ranking as any other.

Proponents of quantitative extrapolationargu e that careful attention to the available dataaids in choosing the correct model and redu ceschances for error. Arguments about the applica-bility of these techniques will continue, especial-

ly because efforts to app ly cost-benefit ana lysisto making d ecisions about carcinogens will re-quire quantitative estimates of cancer incidence.

There are now no convincing data to dictatewh ich extrapolation m odel is best for estimatinghuman cancer incidence, whether from epidemi-ologic data or animal data, or even that onemod el will be consistently better th an all others.However, one particular model for estimatinghuman incidence from animal data (linear, nothreshold extrapolation and relating animal andhumans on the basis of total lifetime exposure

divided by body w eight) has been reported toestimate h um an cancer inciden ce within a factorof 10 to 100 when compared to incidence meas-ured by epid emiologic studies. While this agree-ment is gratifyingly good, data exist to makethese comparisons for fewer than 20 substances.

STATUTORY AND REGULATORY DEFINITIONS OF “CARCINOGEN”

Regulation of carcinogens has been markedby repeated arguments about the amount andkind of evidence necessary to m ake d ecisions toregulate substances as carcinogens. Several Fed-eral documents describe the types of tests agen-cies will consider an d the criteria they w ill app lyto make such decisions. Statements of regula-tory agency policy are found in EPA’s Interim

Guideline for Carcinogenic Risk Assessment,EPA’s air carcinogen policy statement, OSHA’sgeneric cancer policy, and the Regulatory Cou n-cil’s policy statement, which drew heavily on

recommendations of the Interagency RegulatoryLiaison Group (IRLG). IRLG now coordinatesFederal regulatory agency discussion aboutidentifying and characterizing toxic substances,includ ing carcinogens.

All Federal agencies accept positive epide-miologic studies as strong evidence for carcino-

genicity, and a positive bioassay result in asingle species as evidence that the substance is apotential human carcinogen. All relegate short-term tests to a supporting role. Trade associa-

8/14/2019 8104


8/14/2019 8104


Ch. 1—Summary q 1 5

these suggestions. They see the regulatory agen-cies as the approp riate and lawful locations formaking decisions about risk. In general, they seea science panel as anoth er layer of bureaucracythat might hinder regulatory activities, andworry that a single panel might be more sen-sitive to pressure from interested parties. Fur-

thermore, they see the division between “sci-ence” and “policy” in d ecisions abou t cancer asillusionary. They argue that such a panel mighthave the p ower to d elay decisions by imp osing ahigher standard of proof that a substance is acarcinogen than is required by law. This, theysay, would stymie preventive “precautionary”governmental action tha t they view as necessary

REGULATED CARCINOGENS

App roximately 96 substances have been reg-ulated as carcinogens or suspect carcinogens,and an add itional 49 toxic chemicals wh ich h avebeen identified as carcinogens by EPA are re-quired to be considered for regulation un der the1977 Clean Water Act Amendments. Whenoverlaps between the list of already regulatedsubstances and those required for regulation aretaken into account, there is a total of approx-

to protect lives and health when certainty can-not be achieved.

The number of proposals for risk-determi-nation panels almost guarantees that the panelswill remain an issue in Federal policy abou t car-cinogens. Establishment of such a p anel wou ld

represent a significant change in how the FederalGovernment makes decisions about health risksand wou ld p robably require sp ecific legislation.In November 1980, Congress provided $0.5 mil-lion to the Food and Drug Administration(FDA) to place a contract to investigate the fea-sibility of a panel. The report from the stud y isexpected by the en d o f 1982.

imately 102 substances. Fifty-seven of thosesubstances are regulated un der m ore than onelaw. This is expected because exposures to a car-cinogen may occur in air, in water, from solidwaste on land, and in the workplace so a car-cinogen may be regulated under several stat-utes. Twenty-one of the substances that are reg-ulated under a single law are FDA-regulatedcomponents of food.

COLLECTION A N D C O O R D I N A TI O N O F

EXPOSURE AND HEALTH DATACongress has enacted several pieces of legisla-

tion that require Federal agencies to controlcarcinogenic chemicals. OSHA is responsiblefor the occupational setting; CPSC, consumerprod ucts; FDA, some food, drugs, and cosmet-ics; EPA, the “environment” (air, water, andsoil); and the Departmen t of Agricultu re, food.In ord er to meet th eir responsibilities, agenciesmust collect information and assess risks.

Data about exposure histories and health

status are useful in assessing associations be-tween the environment and cancer. Both typesof information are collected by Federal agenciesbut often in separate data systems in differentagencies. Because of privacy and confidential-ity restrictions, these records can seldom be

brought together to “link” information pertain-ing to an individual. In general, these recordseither cannot be made available to researchersor can be made available only without personalidentifiers, which makes linkage impossible.Efforts to ease these restrictions are beingpursued.

Federal, State, and local groups collect en-vironmental data for a multitude of reasons,and individual programs periodically review

their monitoring capabilities and directions.However, there is no federally coordinatedfocus to review the quality and quantity of datathat are collected. Thus, there is no assurancethat adequate exposure data are collected foridentifying and estimating carcinogenic risks.

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


. . .


Congress, reflecting; the difficulties inherent inregulating when risks and benefits are uncertainand difficult to qu antify, delegated to th e agen-cies the task of operationally defining certainkey balancing term s. Words in the law s, for in-stance, those requiring EPA to regulate “unrea-sonable risks” und er TSCA, were pu rposefully

left und efined. The agencies and the courts, bytheir decisions, are now defining those terms.

Some regulatory agency lawyers have beenasked abou t the ability of the agencies to work within the confines of the present laws. They ex-pressed confidence that the agencies can ad-minister the balancing laws (such as TSCA) andapparently appreciate the flexibility of the lawsas they are now w ritten.

Other observers are of the opinion thatgreater attention to the balancing terms in thelaw w ould imp rove the regulatory processes. Tosome extent, defining a balance may mean ac-cepting a specified risk, and Congress, com-posed of elected representatives, is most oftenseen as the bod y to d ecide on such a level. Con-

gress already avails itself of opportun ities dur-ing oversight and reauthorization hearings toquestion agencies and other organizations aboutdifficulties in implementing the laws. Continua-tion of these activities may be sufficient tosatisfy Congress that the language of the lawsdoes or does not p resent a p roblem that can berectified by congressional action.

BALANCING RISKS, COSTS, AND BENEFITS IN DECISIONS

ABOUT REGULATING CARCINOGENS

Beyond the technical level of deciding wheth-er or not a substance is a carcinogen and es-timating the amount of human cancer it maycause is any decision that requires weighingrisks against the benefits of its continued use.The decision is complicated by equity consider-ations. The people who most directly bear risksfrom exposures to carcinogens are not necessari-ly the people w ho m ost directly benefit from theactivities that produce the risk. Depending on

conditions, either th e risks or the benefits maybe accorded greater quantitative importance.

REGULATORY REFORM

Concern about increasing regulatory costsand burdens in recent years has produced apush for changes in regulatory decisionmaking.Charges of overregulation or untimely regula-tion have been leveled at many programs, in-cluding those that regulate exposures to carcino-gens.

Current p rocedu ral reform p roposals includ e:

. more emphasis on regulatory benefit-cost

analysis;

Numerous surveys show that society wants pro-tection from health risks and is willing to payfor it. At the same time, economic and otherconsiderations cause society to attempt to spendno more than is necessary. Uncertainties at-tached to estimates of health risks and economicbenefits complicate regulatory decisionmaking.Improvements in cancer risk identification andmeasurement will reduce the uncertainty, butbalancing health risks against costs of control

and the benefits of the regulated substance willremain a difficult value judgm ent.

q

q

q

q

more systematic regulatory review;more flexibility in rulemaking;appointment of add itional adm inistrativelaw jud ges and greater involvement of theAdministrative Conference; andprovid ing financial assistance to pu blic in-tervenors who are seen to be at a disadvan-tage when opposed by resources of ind us-try or agencies.

Each of these pr oposals is directed at improv-ing regulatory decisionmaking, but not all

8/14/2019 8104


8/14/2019 8104


20 . Technologies for Determining Cancer Risks From the Environment

rence and distribution of cancer. This infor-mation would allow more precise estimates tobe made about the incidence of cancer andtherefore allow accurate monitoring of cancertrend s. Results from the specific studies in op -tion 3 might clarify many qu estions about rela-tionships between particular exposures and be-

haviors and cancer. They would be immediatelyuseful for prevention programs.

OPTION 1

Expan d the op eration of N CI’s SEER pro-gram to collect cancer incidence and sur-vival data rep resentative of the entire coun-try and design programs to assess validityof collected data.

The SEER program, which started in 1973, isthe first continuous cancer incidence reportingprogram in the United States and p rovides an

approximation of cancer incidence and survivalrates for the coun try as a wh ole. The SEER pro-gram has been and should continue to be a u se-ful source of identified cancer cases which great-ly facilitates studies about the disease. Detaileddiagnostic information is av ailable on each caserecorded by the SEER program, and patients,family members, and friends can be quer ied tolearn more about exposures, behaviors, and oc-currence of cancer.

SEER program data are collected from abou t10 percent of the total population, but thegeograph ical regions covered by the SEER pro-gram do not closely represent the demographicmakeup of the entire country. A slightly ex-pand ed SEER program could encompass m oreof the country an d be constru cted so as to col-lect data representative of the entire population.Expand ed coverage would generate data for amore careful and detailed examination of cancerrates over time than what is now possible. Animportant component of an expanded SEERprogram could be rigorous examination of thevalidity of the collected data. For instance, theaccuracy of diagnoses and transfer of the diag-

nosis information to the SEER program datacould be monitored to reduce uncertaintiesabout data. Such attention to data reliabilitywould make conclusions draw n from the d atamore convincing and accepted .

The current SEER program costs about $10million ann ually out of the total NCI bud get of about $1 billion. Expanding the program wouldcost more m oney and wou ld also require coop-eration of additional local medical organiza-tions to establish new SEER data collectionareas. Balanced against these costs are oppor-

tunities to gather incidence and survival datarepresentative of the whole country and to learnmore about cancer in the U.S. population.

OPTION 2

Establish a National Cancer Registry(NCR) to record al l new cases of cancer inthe Un ited States.

An NCR would p rovide the most compr ehen-sive data possible on cancer incidence in thiscountry. With a n ational registry, cancer wou ldbecome a rep ortable disease, as are some infec-

tious diseases. This data base would be usefulfor trend analysis and for identifying cancercases for ep idemiologic stud y. The NCR wouldrecord the date on which cancer was diagnosedand could be used in conjunction with the Na-tional Death Index to generate informationabout survival.

An N CR would collect less detailed informa-tion on each case than is now collected by theSEER program, but would record on the orderof ten times more cases. Establishing an NCRwould not reduce the need for the SEER pro-gram, but rather would add to the cancer in-ciden ce data base.

About 30 States presently have enacted regu-lations or laws requiring that cancer cases bereported to a central authority, but most of those States have not yet initiated p rogram s toimplement the laws. Establishment of an NCRwhich would invite States to participate mightprovide incentive for States to imp lement th eirown programs. At the Federal level, the Centersfor Disease Control, or another ap propriate or-ganization, might serve as the agen cy for receiv-ing, storing, and d isseminating registry data.

The establishment of a well-structured NCRmight become a seed project for a comprehen-sive registry for many chronic diseases of na-tional importance. The interrelationships and

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


——


scientists continue to discuss and refine them,and in the United States, NTP has improved themanagement of the Government test program.Despite all this progress, no improvements areexpected in two aspects of the tests: they are ex-pensive (up to $1.0 million for each substancetested) and they require a great length of time

(from 3 to 5 years).In its first annu al plan (1979), NTP iden tified

the development and validation of less expen-sive, quicker tests as a priority goal. NTP hasoutlined a testing scheme involving both short-term and long-term tests and is working to de-cide w hich short-term tests work best for iden -tifying a number of toxics, including carcino-gens. The attention paid to short-term tests byNTP promises that progress will be made. Theconcentration of DHHS toxicological expertisein NTP and the development of NTP’s workingrelationships with agencies outside DHHS as-sure that the p rogram can call on the ap propri-ate people in pursuing the goal of new tests.

Congress might encourage short-term test de-velopment and validation in its oversight ac-tivities, and it might consider ad ditional fun dingfor the programs. There is currently a great dealof interest in the short-term tests and add itionalcongressional support might have a profoundeffect on their development.

A potential disadvantage of relying on NTPfor guiding and directing this research and de-

velopm ent effort is that NTP has m any other re-sponsibilities. As discussed in this assessmentand in option 7 below, NTP also is responsiblefor the management of large animal test pro-grams. As a part of a multipurpose program,short-term test development has to compete forresources with other parts of the NTP. If it weredecided that short-term tests are sufficiently im-portant to be set apart from other NTP activ-ities, the following option might be considered.

OPTION 6

Establish a commission to advise the Feder-al Government about optimal methods fordevelopment of short-term tests.

A commission, composed of experts fromacademe, industry, public interest groups, and

Government agencies could be established tomake recommendations about short-term tests.This wou ld have the ad vantage of concentratingthe talents of diverse people on test developmentand bringing increased attention to the tests.

The existence of a commission would prob-ably result in short-term tests being given higher

priority in NTP. The exact tasks of th e commis-sion w ould be decided by NTP and other p artieswith interest in the tests. However, one task might be the serious consideration of w hich, if any, tests offer p romise as su bstitutes for long-term animal carcinogenicity bioassays whenmaking regulatory decisions. The establishmentof criteria that a single test or a combination of tests would have to meet to be considered forregulatory decisionmaking would be a spur anda gu ide to test development. The possible disad-vantage of a comm ission is that it may p rovidenothing different from what NTP (as in the pre-vious option) might provide.

The comm ission could focus attention on thetests, the most likely ways for their employmentand what criteria they must meet. Adoption of this option would reinforce the conclusionsalready reached by many experts that the short-term tests show great p romise. In a major way,the commission might answer the question“promising for what?”

OPTION 7

Expan d sup port of the National ToxicologyProgram.

NTP has made a promising start at managingDHH S’ short-term testing an d animal toxicolo-

gy. Programs. AS mentioned above, it has iden-tified the d esirability of obtaining a lternativesto the current carcinogenicity bioassay, andwhether the direction of that effort remainswithin N TP (option 5) or is shared w ith a com-mission (option 6), NTP personnel will continueto be involved in test development. In add itionto short-term test research, NTP administers thelargest animal test program for carcinogenicity.

Those expensive and time-consuming tests areused for two general purposes: to test sub-stances that are of interest to regulatory agen-cies and to provide information useful in devel-oping and validating p ossible new tests.

8/14/2019 8104


Ch. 1–Summary q 2 5

NTP has been successful in organizing expertadvice for its programs. It has assembled aboard of non-Government scientists to advisethe overall program, a panel of regulatory agen-cy representatives to aid in selecting chemicalsfor animal carcinogencity bioassay, and, mostrecently, a panel of nongovernmental experts to

review the results of animal tests for carcino-genicity. Each of these efforts increases thesources of advice for NTP, and assures NTPhigher visibility.

Arguments for encouraging and expandingNTP center on its prom ising start, its attentionto immediate testing needs through the bioassayprogram, and from the possibility of futurepayoffs from new test development. Its estab-lishment of advisory committees of Govern-ment and private sector representatives helpsassure that it will remain responsive to nationalneeds. A possible disadvantage of NTP is that

its wide pu rview and efforts both to develop testmethod s and to serve the needs of the regulatoryagencies may stretch it too thin. Continued con-gressional interest and oversight can help avoidthis possibility.

Options concerned with EPA’s implementa-tion o f TSCA:

The following three options discuss collectionof sufficient information to protect the publicfrom u nreasonable risks, as required by TSCA.The first option is to provide additional support

to EPA, The second and third op tions considerchanges in TSCA.

OPTION 8

Increase the resources available to EPA toenable it to assess more effectively potentialrisks from substances before they are intro-duced into commerce and from substancesalready p resent in comm erce.

One of the most important tools for protec-ting the p ublic from toxic substances is pr ovidedin the sections of TSCA which enable EPA togather information about substances beforethey are in troduced in to commerce. Themechanism for obtaining this information is thePMN w hich must be subm itted to EPA 90 day sbefore a manufacturer or processor can produce

or import a substan ce. EPA must then evalua tethe PMN to d etermine whether or not the sub-mitted information supports a conclusion thatthe substance “may present an unreasonablerisk. ” If the decision is made that a substancemay present such a risk, EPA can then requiresubmission of additional information before

allowing introduction of the substance intocommerce.

EPA, which bears responsibility for evalua-tion of PMNs, is overburd ened. EPA estimatedthat 1,500 people were necessary for the pro-gram in fiscal year 1979; 382 permanent posi-tions w ere au thorized; 313 were filled.

A GAO rep ort characterized EPA’s progressin implementing TSCA as “disappointing,” anddrew attention to too few staff members as partof the problem. If more resources are not mad eavailable, review of PMNs will likely become

less complete, because more are being sub-mitted. EPA estimates that 800 PMNs, almosttwice the number in fiscal year 1981, will besubmitted in fiscal year 1982.

The premanufacturing review program is de-signed to screen ou t risky substances before theyenter commerce. Making decisions at that pointis more protective of pu blic health, and has theadditional advantage of identifying hazardsbefore industry had tied up large amounts of money in manufacture and distribution. Reali-zation of these objectives apparently will re-quire more people as the burden to review

PMNs increases at EPA.

Other sections of TSCA specify tha t EPA canrequire that industry test substances alreadypresent in commerce that “may present an u n-reasonable risk. ” Congress, throu gh TSCA, es-tablished the Interagency Testing Committee(ITC) to recommend chemicals for testing, andto date EPA has considered only ITC-recom-mended substances for testing requirements.Even so, EPA has fallen behind schedule inmeeting its requirements to develop test rules.

One criticism leveled at EPA’s process of de-veloping a test ru le is that it spend s more effortthan necessary to m ake the case that a substan ce“may present an unreasonable risk. ” If thecriticism is correct, EPA may be able to improve

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


2

Cancer Incidence and Mortality

INTRODUCTION

In 1979, cancer killed more than 400,000Americans (254) and 765,000 new serious can-cers’ were diagnosed (6) (see table 3). Over 3million Americans alive today have had a diag-nosed cancer. Cancer accoun ts for abou t 20 per-cent of total U.S. mortality, second only toheart d iseases, which are respon sible for about38 percent of deaths. Moreover, cancer is thenu mber one killer of adu lt Americans, ages 25to 44 (see table 4).

Of equal or greater importance than knowingthe nu mber of cancers and cancer deaths is the

matter of wh ether age-specific cancer rates ar eincreasing, decreasing, or remaining constant.Are people of a given age at greater risk of developing cancer today than were people of that age in th e past? This question of changingrates bears on wheth er aspects of the modern en-

‘Not including nonmelanoma skin cancers, estimated at 400,000

per year.

vironment, largely introduced within the lasttwo to four decades, might be causing today’scancers. If so, preventive efforts shou ld star t byidentifying these elements in the environmentand modifying them. If most of the now com-mon cancers have been common for a long time,it might suggest th at the causes of cancer havenot changed greatly. In that case, preventionmight require changes in long-establishedaspects of the American lifestyle.

When seeking means to prevent cancer, spe-

cial attention m ust be given to increases in can-cers at particular sites. This attention is war-ranted because the increase indicates that thecause (or causes) might have been introducedrecently and can presumably be identified andeliminated . How ever, to concentrate on a searchfor new agents to the exclusion of other causesmay ignore the possibility of preventing pres-

Table 3.—Estimated New Cancer Cases and Deaths by Sex for Major Sites, 1981

Females Males

Total cases Total deaths Cases Deaths Cases Deaths

Percent Percent Percent Percent Percent Percent

Site Number of total Number of total Number of total Number of total Number of total Number of total

L u n g . . . . . . . . . 1 2 2 , 0 0 0

Colon-rectum .. 120,000

B r e a s t . . . . . . . . 1 1 0 , 9 0 0

Prostate . . . . . . 70,000Uterus . . . . . . . . 54,000Urinary . . . . . . . 54,600Oral (Buccal

cavity andpharynx) 26,600

Pancreas. . . . . . 24,200Leukemia . . . . . 23,400Ovary. . . . . . . . . 18,000b

Skin . . . . . . . . . . 14,300All others . ....177,000

15.0

14.7

13.6

8.6

6.6

6.7

3.33.02.92.21.8

21.7

105,00054,90037,10022,70010,30018,700

9,150

22,000

15,900

11,400

6,700

106,150

25.013.18.85.42.54.5

2.25.23.82.71.6

25.3

34,00062,000

110,000 —

54,00016,600

8,20011,50010,400

18,0007,300

80,000

8.315.026.7

—

13.14.0

2.02.82.54.41.8

19.4

28,000 14.528,700 14.936,800 19.1

— —

10,300 5.46,500 3.4

2,850 1.510,5007,000 : :;

11,400 5.92,700 1.4

47,750 24.8

88,00058,000

90070,000

—

38,000

18,40012,70013,000

—

7,00097,000

21.8

14.4

0.2

17.4 —

9.4

4.63.23.2 —1.7

24.1

77,000 33.826,200 11.5

300 0.122,700 10.0

— —

12,200 5.4

6,300 2.811,500 5.18,900 3.9

— —

4,000 1.858,400 25.7

T o t a l . . . . . . . 8 1 5 , 0 0 0 420,000 412,000 192,500 403,000 227,500

alnvasive cancer only.bMelanoma only.

NOTE: Estimates of new cancer cases and deaths are offered as a roQgh guide and should not be regarded as deflnitwe

SOURCE: American Cancer Society, 1980.

31

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


Ch. 2—Cancer Incidence and Mortality q37

not ava ilable until several years after the latestcensus is completed. Thus, there may be largediscontinuities between th e later postcensal pop-ulation estimates and the actual census count—the adjusted estimate for the number of malesage 85 or older between 1959 (postcensal esti-mate based on th e 1950 census) and 1960 (census

enumeration) shows a 50-percent increase.These discrepancies are not present when in-tercensal estimates for 1959 are compared to1960 figures.

The censuses have been characterized by un -derenumerations which vary from census tocensus. The undercounts are thought to besmall, near zero, for some demographic groups,such as white females 40 to 45 years old (in1970), but are considerably greater for othergroup s, such as black males 25 to 45 years old,for w hom the estimated und ercounts are on theord er of 10 percen t or more (see fig. 1) (349). Anum ber of stud ies have shown that serious un-dercount of the population exists for the veryelderly, those age 85 years and over (350).

Mortality Data

Information on deaths has been collectedthrough the national vital registration systemsince the beginning of this century and is themost reliable basis for calculating U.S. cancerrates. Since 1933, data hav e been collected con-tinuou sly for the entire Un ited States. Nationalvital statistics functions are centered in the Divi-sion of Vital Statistics of the National Center forHealth Statistics (NCHS). U.S. mortality statis-tics are based on information obtained directlyfrom copies of original death certificates re-ceived from the registration offices and fromdata provided to NCHS through the Coopera-tive Health Statistics System. A number of States now provide their data—medical anddemographic—entirely on computer tapes.

NCHS is not a repository of original certifi-cates. Those are available only from the States.U.S. mortality statistics for all years except 1972

are based on inform ation for all death s. In 1972,they were based on a 50-percent sample, be-cause of unusual budgetary and personnel con-straints.

The mortality figures used in this report arebased on vital statistics information fromNCHS, which (except for the most recent years)have been published in the annual volumes of Vital Statistics of the United States, Volume II,

Mortality. The data refer to the aggregate pop-ulation of the 50 States and the District of

Columbia.

Incidence Data

While mortality data are extremely useful forstud ying cancer, incidence data are necessary toadvance the state of knowledge about when andwhere cancers occur, irrespective of outcome.Population-based cancer registries are attemptsto identify all incident cases in a defined p opu -lation. In practical terms, registries coverdiscrete political or geograph ic areas. There area nu mber of countr ywid e cancer registries (e. g.,

Canada, Norway, and Sweden). These registrieshave generally taken ad vantage of preexisting,centralized data -collection mechanisms. Thereis no nationwide cancer registry in the UnitedStates. By law, cancer is a reportable disease inabout 30 States and the District of Columbia(58), but m ost States have n ot pu t in p lace themechanisms necessary to handle reported data.The most p rominen t efforts by States are Con-necticut’s statewide registry, operating since1935, and the New York State registry, in opera-tion since 1940 (New York City was not in-cluded until 1973).

The first major cancer incidence surveys inthe United States were th e Ten Cities Sur veys of 1937 and 1947’, (now referred to as the First andSecond National Cancer Surveys, FNCS andSNCS, respectively; see table 6), and the Iowastud y of 1950. Metropolitan ar eas were chosenfor FNCS and SNCS to assure a h igh p ercentageof correctly diagnosed cases. The areas sur-veyed w ere selected to p rovide reliable data an dincluded about 10 percent of the U.S. popula-tion. The sample population was representativeof the geographic distribution of Northern,

Southern, and Western cities with populationsgreater th an 100,000, but w as not en tirely dem -ograph ically representative of the U.S. pop ula-tion.

8/14/2019 8104


8/14/2019 8104


Ch. 2–Cancer Incidence and Mortality q39

Table 6.—Areas Covered by the Three National Cancer Surveys and the Surveillance, Epidemiology, and EndResults Program of NCI

FNCS SNCS TNCS SEER SEER +

Area

SEER +

(1937-39) (1947-48) (1969-71) 1973-present 1-3 NCSa

TNCSb

Atlanta Cherokee, Clayton, Cobb, Cherokee, Clayton, Cobb, Clayton, Cobb, DeKalb, Fulton,

DeKalb, Douglas, Fayette, DeKalb, Douglas, Fayette, GwinnettForsythe, Fulton, Gwinnett Forsythe, Fulton, Gwinnett x x x

Birmingham Jefferson Jefferson Jefferson, Shelby, Walker

Dallas Dallas Dallas Collin, Dallas, Denton, EIIis,

Kaufman, RockwellDenver Denver Adams, Arapahoe, Denver, Adams, Arapahoe, Denver,

Jefferson Jefferson, Boulder ——

Detroit Wayne Wayne Macomb, Oakland, Wayne x x x

Pittsburgh Allegheny Allegheny Allegheny, Beaver, Washington,Westmoreland

San Francisco- Alameda, San Francisco Alameda, San Francisco Alameda, Contra Costa, Marin,

Oakland San Francisco, San Mateo x x x

Chicago Cook Cook

New Orleans Orleans Orleans x c

Philadelphia Philadelphia — —

Philadelphia —Fort Worth Tarrant Johnson, Tarrant

Colorado Entire State.

Iowa Entire State x x.-

Minneapolis- Anoka, Dakota, Hennepin,

St. Paul Ramsey, Washington —

Seattle - PugetSound Area

(13 counties) xConnecticut

(entire State) x

Hawaii(entire State) x

Utah(entire State) x

Puerto Rico x

New Mexico(entire State)

Abbreviations: FNCS — First National Cancer Survey.SNCS — Second National Cancer Survey.TNCS — Third National Cancer Survey.

SEER — Surveillance, Epidemiology and End Results.NCS — National Cancer Survey.

~“X” Indicates area was covered by all 3 NCSS and SEER.“X” mdlcates area was covered by TNCS and SEER.

cNew Orleans IS being dropped from the SEER program in 1981.

SOURCE Devesa and Silverman (84) and Off Ice of Technology Assessment

Several changes took place in the next majoreffort, the Third National Cancer Survey(TNCS) of 1969 through 1971 (84). The report-ing period was 3 years instead of 1. Thetimeframe was used to improve data for rarecancers and for smaller population groups.Working under contract to NCI, local, non-profit medical organizations (e.g., schools of public health, medical schools, health depart-ment offices) condu cted th e project at each site.

Ten percent of the cancer patients were includedin a more intensive survey to determine meth-ods of treatment used , dura tion of hospitaliza-

X

tion, cost of medical care, and economic impacton the family. There was considerable geo-graphical overlap between FNCS, SNCS, andTNCS. Seven of the original 10 cities were in-cluded in TNCS, although coverage was ex-panded from only central city areas to theirStandard Metropolitan Statistical Areas. Theentire state of Iowa was added. The Common-wealth of Puerto Rico cancer registry also sup-plied d ata (see table 6).

The first nationally coordinated , continu ous-registration system, the Surveillance, Epidemi-ology, and End Results (SEER) program, was

8/14/2019 8104


40 Technologies for Determining Cancer Risks From the Environment

begun in 1973 by NCI. NCI derives the SEERprogram mandate from the National Cancer Actof 1971 which directs that the Director of NCI:

Collect, analyze, and disseminate all data use-ful in the p revention, diagnosis, and treatmentof cancer . . . . (sec. 407(b)(4)).

In SEER areas, attempts are made to ascertainevery primary cancer, excluding nonmelanomaskin cancer. All information pertaining to a caseis consolidated into one record, to facilitatefollowup and to correlate survival data withtreatment, age, and other variables.

Eight geograph ical locations, four in comm onwith TNCS were chosen originally, and threewere add ed subsequently (see table 6). Recently,one area, New Orleans, was dropp ed from theprogram. Approximately 10 percent of the U.S.population resides in SEER areas. A SEER re-port for the first 4 years of operation compared

the d emograp hic characteristics of the popu la-tion with the total U.S. population (248):

. . . the p articipants . . . are fairly representa-tive with respect to age. Blacks are somewhatund errepresented while other nonwhite popula-tions (Chinese, Japanese, Hawaiians, and Amer-ican Indians) are somewhat overrepresented.Rural populations (especially rural blacks) arealso underrepresented.

The SEER program has considered addingregistration areas toward the goal of improvingthe reliability of data for all segments of society.

Exact proportional representation is not, how-ever, the ideal situation, since for small dem o-graphic groups overrepresentation may be nec-essary to prod uce reliable data.

As of September 1980, data were availablefrom the first 6 years of SEER operation. NCIplans to publish complete incidence and mor-tality data every 5 years, and to make ava ilabledata for interim years. A small amount of treat-ment and survival data have been pu blished asthe “Cancer Patient Survival” reports (247,251)following the earlier “End Results in Can cer”series. As more years of followup data are ac-

cumu lated, NCI surv ival analyses will increas-ingly draw on SEER progr am information.

Error and Bias in Mortality andIncidence Data

There are various sources of error and bias inboth mortality and incidence data. Reasoned in-terpretation of trends depend s on un derstanding

the forces, other than true chan ges in inciden ceand mortality, which impact on certified mor-tality (cause of death as reported on d eath cer-tificates) and registered incidence. The im por-tant imp acts are outlined below:

1.

2.

Improper diagnosis. -Individuals maycontract cancer and die but the correctdiagnosis may never be made, affectingboth incidence and mortality rates. It ispossible, for reasons of inadequate med-ical care or simple oversight, for lung can-cer to be diagnosed as pneumonia; braincancer as stroke or senility, and leukemiasor lymp hom as as infectious d iseases. Con-versely, cancer may be reported as thecause of death for people dying of othercauses. For instance, in a 1970 autopsyseries in Atlanta, Engel et al. (99) foundthat 86 percent of cancers found at autop-sy were listed as the underlying cause of death on the death certificate. A misseddiagnosis of cancer in a dying patient ispresumably more likely to occur amongold than among young cancer patients, if only because the old ar e less likely to behospitalized. These errors are likely tohave become less numerous over the pastfew d ecades, par ticularly in older peop leafter the introd uction of med icare in 1966.

Improvements in ascertainment. —It i slikely that fewer cancers are missed todaythan in the past, affectin g both incidenceand mor tality rates, probably to a greaterdegree in the older age groups. In addi-tion, incidence rates may be influenced bya progressive improvement in the readi-ness of physicians to collaborate with acancer registry. Data from the Connect-icut cancer registr y suggest that since its

inception in 1935, the completeness of coverage may have imp roved so as to in-

8/14/2019 8104


Ch. 2–Cancer lncidence arid Mortality q 41

troduce substantial upward biases into therates. Better ascertainment of incidentcases is also seen in comparing SNCS toTNCS. The proportions of cases that wereascertained by death certificate only, andfor which no clinical record was everfound, were 11.8 and 2.2 percent, respec-

tively (84). This suggests that the earliersurvey may have underestimated total in-cidence rates. Overall, there is a tendencyfor better recording of cases over time,which causes an apparent increase in in-cidence.

3. Primary site not specified. —The site of the pr imary cancer in p atients w ith metas-tatic disease may never be d etermined. Sixto eight percent of American cancer deathcertificates are for cancer of an un specifiedprimary site (255). This percentage is alittle lower among whites than among

nonwhites, and among middle-aged thanamong older people, but it has not ma-terially changed for decades, and may notseriously bias the assessment of trends incancers at specified sites. However, themore th an 20,000 cancer deaths per yearnow classified as unspecified represent anuncomfortably large amount of missinginformation.

4 . I n c o r r e c t pr imary s i te or cel l type. —

Misdiagnosis of the primary site or celltype of fatal cancer is probably th e mostimportan t bias affecting cancer d eath cer-

tification rates. Patients with cancer of one primary site (e. g., lung) may be mis-diagnosed as h aving a cancer originatingfrom another site (e.g. , pancreas orbrain), if the cancer has either extendeditself to other nearby organs, or m etasta-sized to distant organs. Boyd et al. (93),concluded that at ages over 65 most bonetumor deaths were in fact misdiagnosedsecondaries from other sites. This mayalso have been true in the past for livercancer, since bone and liver are not siteswh ere cancers commonly arise but, along

with brain and lung, are sites to whichcancers commonly spread. Likewise, can-cers of one particular cell type may bemisdiagnosed as cancers of another cell

5.

6.

7.

type. This problem can som etimes be cir-cumvented by group ing together particu-lar types of cancers which are often m is-diagnosed as each other, e.g., all benignand malignant brain tumors or all colonand rectal cancers. Colon and rectum can-cers, which together account for 18 per-cent of cancer deaths, are often lumpedtogether to imp rove the reliability of thestatistic. However, they are different dis-eases and their individu al characteristicsare obliterated by this procedu re.

Incomplete transfer of information todeath certificates. —Even if the cancer iscorrectly diagnosed while the patient isalive, the correct information may neverreach the death certificates. Percy,Stanek, and Gloeckler (291) tabulated cor-respondence between the primary site of cancer, as diagnosed in the hospital, and

the primary site as it eventually appearedon the death certificate for 48,826 patientsin TNCS. Many discrepancies emerged.Cancers of the colon were overreportedwh ile cancers of the rectum were u nd erre-ported on d eath certificates as comparedto hospital records. In add ition, in over 50percent of all cases where a more specificcancer site (cecum, ascending colon andapp endix, transverse colon, etc. ) appearsin hospital records, it was recorded dif-ferently—most often as “colon, not other-wise specified’’—on the dea th certificate.

Cancers of specific uterine sites (cervixand corpus) suffer from the same prob-lem. Cancers of the buccal cavity areunderreported on death certificates, whilebone cancer is overreported, and othersites are misreported to v arying d egrees.

Inclusion of prevalent cases. —This bias islimited to incidence data. In a stu dy wh ichrun s for a relatively short period, “preva-lent” cases, which were actually diagnosedbefore the start of the study period, mayinadvertently be includ ed. This was m orelikely a problem in FNCS and SNCS,

somewhat less so in TNCS, and less still inthe ongoing SEER program.Changes in the definitions for some can-cers. —The definition of what constitutes a

90-4 B 1 9 - B 1 - 4

8/14/2019 8104


42 q Technologies for Determining Cancer Risks From the Environment

cancer changes with increasing knowl-edge. For instance, all salivary gland tu-mors, whether malignant or of mixed cel-lularity, were classified as cancer up to1967, but the mixed tumors were droppedthereafter. Likewise, all brain tum ors w erecounted in SNCS, while only those speci-fied as m alignant were included in TNCS,causing a su bstantial artifactual d ecreasein brain tumor incidence between the twosurveys (84).

8. Increased access to medical care and changes in diagnosis. —An even moreserious bias stems not from classificationchanges but from the more vigoroussearch for lumps, and imp rovements in di-agnostic procedures which affect mainlyincidence rates. By old age the humanbody may contain various lumps which, if examined histologically, would be clas-sified as cancer, yet many are biologicallybenign and cause no reductions in life-

span. For instance, by age 70, 2.5 percentof males in the areas covered by TNCSwere diagnosed as having prostatic can-cer, wh ile at autop sy 25 percent of malesaged 70 or over who died of unrelatedcauses were found to have cancerous pr os-tates (33). Similarly, among women un-dergoing mastectomy for cancer of onebreast, and in whom cancer is not clinical-ly evident in the opposite breast, biopsyand histological examination of the oppo-site breast finds 15 to 20 percent of themcancerous (93). Normally only 0.5 percentof these women will develop clinicalevidence of cancer in the op posite breast.The scope for biased trends in incidencewhich are due to either more completeregistration or the identification of cancerwh ich w ould n ot cause a serious d isease isprobably large but unknowable with cur-rent procedures.

INCIDENCE AND MORTALITY TREND ANALYSIS

The magnitude and rate of change in inci-dence and mortality at any p articular cancer siteare seldom equal. Highly fatal cancers are theexception. In fact, comparable incidence andmortality rates are seen for pancreatic and lungcancers wh ich h ave the p oorest survival rates of the leading cancer sites. The same d oes not h old

true for cancers of the endom etrium, breast, orbladder, for which survival rates are muchhigher and improved over the recent past.

Major problems confront analysis of existingU.S. cancer incidence data, while such problemsare less severe for m ortality data. Incidence datarepresentative of the U.S. pop ulation w ere col-lected at only three points in time over a periodof more than 30 years before 1973. During thattime there were changes in survey m ethods, def-initions of disease, diagn ostic patterns, classifi-cation of disease, and in the rules for assigningcause of death, as well as imp rovements in ac-cess to medical care. All of these factors mayhave affected registered incidence. The more re-cent SEER data, while collected according to the

same basic procedures since the p rogram ’s in-ception, may reflect different degrees of ascer-tainment in startup years as compared with sub-sequent years. More importantly, the programhas been operating for too short a time fortrends which are real but small in magnitude toemerge. Comparing SEER data with informa-

tion from one or more of the earlier surveysraises questions about whether data from suchdifferent sources can be analyzed together withsufficient validity. Despite these drawbacks,data from the n ational cancer surveys and SEERhave been analyzed for trends, and providesome useful indicators, though authors of theanalyses acknow ledge inherent p roblems.

Devesa and Silverman (84) analyzed inci-dence data from the national cancer surveyswhen TNCS was completed, along with mortal-ity data for corresponding years. They reportedthat between the two survey points, 1947 and1970, the “overall age-adjusted incidence ratefor all sites combined decreased 3.9 percent.”This summary figure is the result of ups and

8/14/2019 8104


Ch. 2–Cancer incidence and Mortality q43

downs in race-sex-site groups, including anoverall striking decrease for women and anoverall increase for men. Lung cancer was by farthe most active site, increasing more than 350percent between FNCS, wh en it ranked eighth inincidence as a p rimary site, and TNCS, when ittook first place. Uterine cancer, wh ich h ad the

highest incidence rate in FNCS, had decreased40 percent. Even m ore d ramatic was th e overall72 percent decrease in stomach cancer by thetime of TNCS. Incidence trends for the most fre-quent sites and for all sites combined, as re-ported by Devesa and Silverman are displayedin figures 2 and 3.

Devesa and Silverman believe that some of the apparent changes may be artifactual, no-tably part of the dramatic increase in cancers of the lung and prostate in nonwhite males. In ad-dition, the lower rates for nonwhites in theearlier periods may reflect underdiagnosis.However, they conclude (84):

Changes are likely to hav e occurr ed in th eprevalence of carcinogenic agents either in thegeneral or personal environm ent, since the shiftsin trends, especially when considered by raceand sex, are greater than those that could be ex-plained by the problems discussed earlier.

Doll and Peto (93) conclud e that the m ost re-liable estimates of trends in cancer incidence areprobably those derived by direct comparison of SNCS and TNCS, though even in this compari-son substantial artifacts are possible. Figures 4

and 5 display the age-standardized male andfemale incidence data from SNCS and TNCS,respectively. Figures 6 and 7 display age-stand-ardized mortality for the period 1955 through1975. The important changes are increases inlung cancers and melanomas, a decrease instomach cancers for both sexes, increasing ratesof prostate, bladder, and kidney cancers inmales, and a sharp decrease in cervical cancer inwomen.

Pollack and Horm (296) presented the firstanalysis of cancer incidence rates from SEERdata. The pap er, according to the auth ors, had

three objectives:

1. to assess the comparability of the cancerincidence ra tes across the total SEER pro-gram over the period 1973-76;

2. to assess the validity of use of TNCS in-cidence rates for 1969-71 and SEER ratesfor 1973-76 to analyze incidence trendsover the p eriod 1969-76; and

3. to present trends in cancer incidence andmortality for 1969-76, where data are suf-ficiently comparable, for some of the major forms of cancer and to summarizethese trend s by use of a convenient set of measures.

For whites, rates for all SEER areas combined ,for individual cancer sites and all cancer sitescombined were found to be “reasonably com-parable. ” How ever, rates for blacks over the 4-year period are not comparable, and thereforeanalysis of incidence data was carried out forwhites only; mortality data were analyzed forwhites and blacks. The authors concluded, re-garding the second stated purpose, that “the useof incidence rates for TNCS areas for 1969-71and for SEER areas for 1973-76 app ears to p ro-vide a good approximation to trends over thattotal time period for the white population”(296). Pollack (295) reached the same conclu-sion after comparing Connecticut tu mor regis-try d ata, the data from th e three national cancersurveys and the SEER program .

The auth ors (296) calculated the av erage an-

nu al percent change in incidence at each site andall sites combined for each sex (see table 7).However, they caution that “it can be mis-leading to focus on the p icture for all sites” be-cause these overall figures are affected by manydynamic rates for different sites. For all sitescombined, the incidence rate increased an aver-age of 1.3 percent/ yr for wh ite males, and 2.0percent/ yr for wh ite females. Mortality ratesfor all sites combined increased an average of 0.9 and 0.2 percent/ yr for white males andwhite females, respectively. Mortality rates forblacks increased an average of 2.1 and 0.6 per-

cent/ yr for males and females, respectively.

8/14/2019 8104


8/14/2019 8104


Ch. 2–Cancer Incidence and Mortality q 45

Figure 3.—Age-Adjusted Cancer Incidence Trends Among Females for All Sites-- -— . . . .

400 I

Combined Showing the Most Frequent Sites

White female incidence trends

300

000-00

z: 200zK

Breast

Uterus

Intestines

Cervix

Corpus

Rectum

Lung100

All other

sites

o

400

1940 1950 1960 1970

Year

Nonwhite female incidence trends

Breast

Cervix

CorpusUterus

Rectum

Lung100

Intestines

All othersites

o

SOURCE From Devesa and Silverman (84)

1940 1950 1960 1970

Year

8/14/2019 8104



70C

60C

5oc

400

300

200

10 0

0Mouth,

esophagus,pharynx,and larynx

.

Figure 4.—Site-Specific Cancer Incidence Rates per Million Males, All Agesa

miver,

d ’ Incidence

LA @:~~::derl ‘one I “adde r I ‘rosta’e I ‘eukemia ~!%~lial l

11Iucts

I I I I I sites

ILung Intestines Pancreas Melanoma Kidney Brain and Hod~kin’s Other and

(including central dise&e unspecified solid,rectum) nervous except

system skin

a Age.standardized to 1950 U.S. Census population, assuming 90 percent of the population is white.

SOURCE: Doll and Peto (93).

8/14/2019 8104


Ch, 2–Cancer Incidence and Mortality q 47

700

600

500

400

300

200

100

0

esophagus,

pharynx,

and larynx

Figure 5.—Site-Specific Cancer Incidence Rates per Million Females, All Agesa

QIncidence

1

KeySNCS, 1947/8

TNCS, 1969-71

Breast

IKidney

I IEndometrium Brain and

IHodgkin’s

IOther

L,”=, ,

I

““0, ”

~allhladder I and other central disease and unspec.L - - - - - -and

bile ducts I ‘terus F::s I Iified solid,except skin

Lu;g Intestines Panc!reas Mela;oma Blabder Cervi~ uteri Ovary Leukemia Other

(including reticulo-

rectum)endothel ialsites

a Age.~ tandardlzed to 1950 U.S Census population, assuming 90 percent of the population IS white

SOURCE Doll and Peto (93)

8/14/2019 8104



Figure 6.--Site-Specific Cancer Mortality Rates per Million Males at Ages aO-64 Years:1953-57; 1963=67; and 1973-77

30,00C

25,00C

20,000

15,000

10,000

6000

2000

0 dl-Mouth,esophagus,pharynxand.larynx

d’Mortality

c!!!!!!

1953-7

Key 1963-71973-7

Bone Skin Hodgkin’s [ Other ancStomach ] Liver, Igallbladderand bileducts

I I c e n t ra l I disease I (chiefly)

I Inervous

I Iunspecified

system sites

Lung Intestines Pancreas Bladder Prostate Leukemia Other(including reticulo-rectum) endothel ial

sitesa Age.standardized to the 197’O U.S. ceIISUS population.

SOURCE: Doll and Peto (33)

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104



Table 8.—U.S. Age-Standardized Cancer Death Rates for Males and Females Under 651953-78

Average annual rates/100 million aged under 65a

OnlySite Sex 1953-57 1958-62 1963-67 1968-72 1973-77 1978

Mouth, pharynx,larynx, or esophagus

Remaining respiratory:

Trachea, bronchus,or Iungc

Pleura, nasal sinuses,and all other

Both

Stomach

Intestines (chiefly largeintestine, i.e., colon andrectumd)

Liver e

Gallbladder andducts f

Both

Pancreas

Remaining digestive (chieflyperitoneum*

Bone

Connective and soft tissuesarcomas

Skin (chiefly melanomah)

Breast

Bladder i

Kidney i

Cerv ix u te r iEndometrium J

OvaryProstateOther genital:

MalignantMalignantBenign and unspecified

Brain or nerves, malignant,or benignk

Eye

Thyroid

Leukemia

MF

MFMFMF

MF

MF

MFMFM

FMF

MF

MFMF

MF

MF

MF

MFFFFM

MFFMF

MF

MF

MF

5,9361,213

18,2752,714

6,8083,293

8,9549,014

1,2031,520

3,9842,210

465414

936656355276

1,325916

13815,880

2,066760

2,0121,0087,5504,2185,6922,785

854356835

4,9083,675

127116

236340

4,7543,562

6,4851,478

21,2903,378

5,5392,717

8,7398,576

1,3961,425

4,3362,363

427370

795552419338

1,410935

12716,158

1,919676

2,0511,0066,6513,2825,7362,602

852326444

4,8223,663

120106

207304

4,8433,477

6,8581,700

25,3904,734

4,4782,216

8,6247,977

1,3621,219

4,5362,459

404330

747483492373

1,6591,040

12117,053

1,810655

2,134991

5,6732,6505,6802,549

837291302

4,8313,653

102100

186255

4,7053,338

— --- — -- — ---7,059

2,000

28,7997,133475215

29,2747,348

3,7531,815

8,5217,486

807354535712

1,342

1,0664,4642,482

351265

680444516421

1,5471,022

13017,358

1,658547

2,2061,0184,4232,1935,6212,555

811274173

4,6933,520

9282

182210

4,3443,049

7,123

2,143

30,9119,803447192

31,3589,995

3,2701,551

8,2987,130

795347488644

1,283

9914,2672,598

283212

600380464414

1,8281,086

12317,260

1,538500

2,2361,0163,3651,9665,3042,612

72924695

4,4753,364

7766

153177

4,0362,753

7,200

2,111

32,08011,598408178

32,48811 ,776

2,9831,403

8,276

6,807

789

383487625

1,277

1,0084,1482,618

256193

575360473426

1,9961,161

11017,229

1,386455

2,3001,0112,9111,8155,0422,590

540224

534,2933,246

7058

146181

3,8452,622

8/14/2019 8104


8/14/2019 8104



Table 9.—U.S. Age-Standardized Cancer Death Rates for Males and Females Over 65,1953-78

Average annual rates/10 million aged 65 or overa

OnlySite Sex 1953-57 1958-62 1963-67 1968-72 1973-77 1978

Mouth, pharynx,larynx, or esophagus

Remaining respiratory:

Trachea, bronchusor Iungc

Pleura, nasal sinuses,and all other

Both

Stomach

Intestines (chiefly largeintestine, i.e. colon andrectumd):

Liver e

Gallbladder andducts

Both

Pancreas

Remaining digestive (chieflyperitoneum

Bone

Connective and soft tissuesarcomas

Skin (chiefly melanomah)

Breast

Bladder i

Kidney j

Cervix uteriEndometrium j

OvaryProstateOther genital:

MalignantMalignantBenign and unspecified

Brain or nerves, malignantor benignk

Eye

Thyroid

Leukemia

MF

MFMFMF

MF

MF

MFMFM

FMF

MF

MF

MF

MF

MF

MF

MF

FFFM

MFFMF

MF

M

FMF

8,0271,786

— — — —

14,2772,937

14,3687,547

17,91615,502

7,5801,654

— —

19,0162,981

11,8275,930

17,74914,672

7,2141,551

24,8233,442

9,5524,635

17,76114,024

1,921

2,6515,8163,842

729649

837468

228156

1,8671,118

20111,356

5,4162,258

1,7351,047

3,1274,0683,195

19,300

435645240935596

131106

276

5243,9242,273

2,106

2,5616,4264,074

668610

617359

273184

1,739961

16610,633

5,4962,042

1,9691,066

2,8843,5123,344

18,584

359604147

1,068692

12391

251

4504,5122,474

2,208

2,3576,8994,226

628540

521308

323209

1,679863

17510,351

5,5011,876

2,1661,105

2,5133,1753,460

18,488

320545115

1,375857

10679

234

4174,8552,612

7,3241,643

31,5394,692483205

32,0224,897

7,7083,667

17,95813,497

926364

1,2671,7492,193

2,1137,0904,390

500431

502286

357243

1,448780

19210,603

5,6261,673

2,4881,160

2,0212,8613,680

18,591

295514

721,7311,187

11279

232

3725,0152,704

7,4781,787

37,4246,550492195

37,9166,745

6,5193,047

18,26513,256

957376

1,1931,4792,150

1,8557,1694,463

483365

465261

353244

1,527789

18811,087

5,7811,615

2,5431,222

1,6422,6623,743

19,465

252495

512,1631,522

9970

210

3385,0532,609

7,4871,933

40,8888,296

468203

41,:3568,499

5,8922,870

18,83913,437

1,067

3951,1921,4592,259

1,8547,2474,637

417317

436244

355273

1,608840

18111 ,070

5,7321,623

2,6701,252

1,4032,593

3,79620,392

224470

392,5811,862

10264

217

3105,1422,627

8/14/2019 8104


8/14/2019 8104


56 qTechnologiesfor Determining Cancer Risks From the Environment

Figure 9.–Respiratory Cancer (lCD 8: 160-163)Mortality Rates per 100,000, Age-Standardized to

U.S. 1970 Population: 1950-77

1950 1955 1960 1965 1970 1975 1980

SOURCE Office of Technology Assessment

so of the large intestine, and to describe the re-maind er, includ ing u nspecified intestinal sites,as “colon.” Inspection of the data gives themisleading impression that rectal cancer ratesare decreasing and colon cancer rates are in-creasing, wh ile in fact the decreases in the d eathcertification rates for rectum are if anythingslightly less extreme than for other specifiedparts of the intestines. In view of the fact thathalf of all fatal cancers diagn osed in h ospital as“rectum” in TNCS, were eventually certified as“colon,” the most plausible interpretation of thedata is that there have been no material trends

in either colon or rectal cancer mortality duringthe past 25 years among males, although boththe incidence and mortality data do suggest aslight d ecrease in onset rates below 65 years of

Figure 10.—Trends Since 1950 in U.S. Male‘Lung Cancer Mortality at Young Ages

1950 1955 1960 1965 1970 1975 1980

Y e a r


age (93). Similar difficulties of classification af-fect data for females, and when all intestinalsites are combined, total female intestinal cancerdeath rates appear to be decreasing steadilysince 1950.

Liver cancer currently accounts for 0.8 per-cent of cancer deaths among Americans under65 and no statistically significant trends in livercancer mortality are eviden t du ring the past d ec-ade. Incidence trends show a decrease in liver

cancer, which is probably artifactual and due toimproving differential diagnosis between pri-mary liver cancer and cancers which have me-tastasized from other sites to the liver. This

8/14/2019 8104


Ch. 2—Cancer Incidence and Mortality q 57

Figure 11 .—Combined Mouth, Pharynx, Larynx, andEsophagus Cancer(ICD8:140-150~ 161) MortalityRates per 100,000, Age-Standardized to U.S. 1970

27

24

21

19

16

13

10

7

4

2.5

.Population: 1950-77

1950 1955 1960 1965 1970 1975 1980

SOURCE Off Ice of Technology Assessment

decrease may be somewha t surprising in that theliver is intimately exposed to much of what is in-gested, and it is composed of cells which arecapable of rapid proliferation w hen n ecessary.Moreover, many of the chemicals that h ave thu sfar been found to be carcinogenic in animalfeeding experiments cause liver cancer inanimals.

Gallbladd er and bile du ct cancers are unfor-tunately not reported separately either in mor-tality data or in the data from SNCS, thoughthey are in TNCS. Cancers at these sites havedifferent causes, how ever. Gallstones are a risk factor for gallbladd er but not bile du ct cancer.According to TNCS data, females develop

Figure 12.—Stomach Cancer (ICD 8: 151) MortalityRates per 100,000, Age-Standardized to U.S.

1970 Population: 1950-77

1950 1955 1960 1965 1970 1975 1980


Figure 13.—Combined Small Intestine, Colon, andRectum Cancer (ICD 8: 152-154) Mortality Rates per100,000, Age-Standardized to U.S. 1970 Population:

1950-77

n I I I I 1

1950 1955 1960 1965 1970 1975 1980


8/14/2019 8104



cancer of the gallbladd er more frequen tly thancancer of the bile ducts, while for males the ratiois the inverse.

Decreases have occurred an d ar e continu ingto occur in the aggregate of the two cancers, andthese decreases are larger am ong females than

among males. The figures suggest that it iscancer of the gallbladder that is chiefly decreas-ing, rather than cancer of the bile ducts.

Pancreatic cancer (fig. 14) is now decreasingin ma les at ages und er 65, the d ecreases in earlymiddle age being particularly rapid. Thisdecrease is especially encouraging because itcomes after d ecades of gradua lly but steadily in-creasing rates. Pancreatic cancer is so uniformlyfatal that treatment cannot have affected thesetrend s. If the correlation of smoking w ith pan -creatic cancer represents a cause-and-effect rela-tionship, one might expect the ratio of rates

among smokers to nonsmokers to be increasing,as has been the case in recent years for lungcancer. If the association is causal, then amongmidd le-aged non smokers the trend in pancreaticcancer mortality must be even more steeplydow nw ards than these national data suggest.

Bone cancer death certification (and inci-dence) have shown apparent decreases, which

Figure 14.—Pancreas Cancer (ICD 8: 157) MortalityRates per 100,000, Age-Standardized to U.S.

1970 Population: 1950.77

15

10

5

0

1950 1955 1960 1965 1970 1975 1980

SOURCE. Off Ice of Technology Assessment

may be due largely to the progressive elimina-tion of misdiagnosed secondaries (93). How-ever, it is now impossible to determine whetheror not actual d ecreases have occurred.

Skin cancer mortality increases from 1950through 1978 are a result of an increasing death

rate from melanomas, offset partially by adecreasing death rate from other skin tumors.The increases are most rapid in mid dle age, sothe rates in old age will probably increase evenmore rapidly in future decades than is now th ecase. The causes of melanoma are not well-un derstood; exposure to sunlight seems to be in-volved, and people with a genetic deficienc

yin

their ability to repair the dam age done to DNAby sunlight are at extraordinarily high risk of melanoma (310). How ever, people whose work involves regular outdoor exposure seem para-doxically to be at lower risk of melanom a than

otherwise similar people who work indoors(93), perhaps because a permanent suntan isprotective. This may be at least in part a resultof self-selection for outdoor work, or perhapsthe conditions that maximize risk are thosewhich involve sudden exposure of untannedskin to sunlight. It is possible that theworldwide increases in melanoma are due mere-ly to some change in the pattern of human ex-posure to sunlight, e.g., changes in clothing andincreases in sunbathing, particularly since thechief increases seem to be in melanoma of thetrunk and legs rather than face (93). Alter-

natively since melanocytes are subject to hor-monal influences, it could be that oth er causesare also imp ortant.

Breast cancer (fig. 15) incidence and mortalityat ages under 65 show no substantial changes,but that overall rate conceals smaller fluctua-tions in mortality in particular age groups. Bas-ed on the accepted association of age at firstchildbirth and breast cancer risk (219), Blot (27)has argued that the reprod uctive patterns of dif-ferent cohorts of American women can accountfor some or all of the small fluctuations in breast

cancer death rates in particular cohorts of women. Women who were young during theGreat Depression of the 1930’s somewhatdelayed having their children, and their breastcancer mortality now is slightly increased.

8/14/2019 8104


8/14/2019 8104


hygiene may be relevant. It is not known whateffect cervical cancer screening p rogram s hav ehad on cervical cancer mortality. Also, manycervical deaths between 1933 and 1978 were cer-tified merely as being due to “cancer of theuterus” (with the exact site not otherwise speci-fied). If these d eaths could be tran sferred from

endometrium, where they now are, to cervix,the downward trend in cervical cancer wouldpresumably be much steeper and that fromcancer of the endometrium much shallower,which is supported by the trends in incidence.Finally, an increasing p ercentage of Am ericanwomen in middle and old age, when cancer ismost common, have already undergone hyster-ectomy for various reasons, thereby removingboth uterine cervix and endometrium (and,sometimes, both ovaries) from risk. A betterstatistic for these cancers might be the d eath rateper uterus or per ovary and not per woman.

Prostate cancer (fig. 17) becomes increasinglycomm on w ith age, more so than for m ost othercancers. Incidence rates for prostate cancerprobably are not reliable, being influenced by

Figure 17. —Prostate Cancer(lCD 8: 185) MortalityRates per 100,000, Age-Standardized to U.S.

1970 Population: 1950.77

40

35

30

25

20

15

01950 1955 1960 1965 1970 1975 1980

SOURCE. Off Ice of Technology Assessment.

poorer census data for older age groups, andmost importantly, the more thorough search forlumps, which would not have come to clinicalattention in earlier years.

Mortality data appear to be reliable, how-ever, and ind icate a marked divergence of ratesfor whites and nonwhites. The rate for whites

has remained more or less constant since ‘1950,wh ile the rate for nonwhites has risen steadily,and has not leveled off. The reasons for thesepatterns are not known.

Brain tumors, whether malignant, benign, orof unsp ecified typ es are not r eliably distinguish-able, therefore the three typ es are combined toexamine trends. Under age 65, a l-percent peryear decrease in brain tumor death rates is seen.Over 65, the op posite is true, and there is a veryrap id increase in brain cancer death rates, pos-sibly due to a steady improvement of diagnostic

standards.

Thyroid cancer is not common, accountingfor less than 1 percent of all cancer deaths. “Thy-roid cancer d eath rates h ave fallen steadily from1950 to the present, while a large increase in in-cidence occurred through at least 1970. The dis-crepancy between incidence and mortality canbe at least partly explained by the m any cases,mostly nonfatal, induced by medical radiationof the thyroid, head, and neck. These X-raypractices have largely been discontinued. Thedecrease in mortality may be a result of im-proved treatment, as well as possible real de-creases in the incidence of serious cases.

Hodgkin’s disease and certain forms of leu-kemia are mu ch more treatable now than a dec-ad e or tw o ago. This fact alone accoun ts for theobserved substantial down ward trends in mort-ality, especially among younger people. Theavailability of successful treatments may alsohave encouraged more thorough efforts at cor-rect diagnosis. Reliable estimation of trends inthe various completely different types of leu-kemia is, unfortunately, not possible from theavailable data. The lack of any net trend in

either direction in leukemia mortality amongolder people may represent a balance betweenincreasingly thorough diagnosis among elderlypatients who are dying of leukemia and slightly

8/14/2019 8104


Ch. 2—Cancer lncidence arid Mortality q 61

better treatment of the disease. The incidencedata suggest that some decreases in real onsetrates for leukemia are in progress, at leastamong females (93).

Myelomatosis death rates have been risingsteadily from 1968 through 1978, more so forolder than for m iddle-aged p eople, for u nknownreasons. The apparent increase may be real, butmay be largely or wh olly du e merely to be im-proved case-finding, for imp roved case-finding

SUMMARY

The ability to analyze cancer incidence andmortality dep end s on the available data. Quali-fications ar e attached to the reliability of bothkinds of data, more to incidence data becausethey hav e been collected over sm aller geograp h-

ical regions for shorter p eriods of time, but cer-tain conclusions can be d raw n.

Respiratory cancer incidence and mortalityhave increased dramatically in both sexes, butthe last few years have seen a decline in 1ungcancer rates among young men, The increasesand the more encouraging decreases are asso-ciated with changes in smoking habits. Thedecrease in cancer death rates for females du ringthe last 50 years is partly explained by dramaticdecreases in deaths from cancer of the uterinecervix. Stomach cancer mortality has decreased

substantially in both sexes.When cancer mortality from all nonrespira-

tory sites are considered togeth er, a decrease isseen in females since 1950 (due to decreased cer-

must have occurred during 1968-78 and mightbe expected to hav e its greatest effect amon g theold.

Unspecified sites accoun t for 6 to 8 percent of all cancers at ages below 65. The exact per cent-age var ies irregularly from 1950. Rather surp ris-ingly, given better diagnostic criteria in recentyears, slight increases in cancer at unspecifiedsites have been seen during the past decade,

vical and stomach cancers). Male death rateshave remained abou t level over the same peri-od. While stomach cancer death rates have de-creased in men increases at other sites havebalanced th ose decreases.

A controversy is swirling around the inter-pretation of incidence rates. Data collected inTNCS (1969-71) and NCI’s SEER program(1973-78) show that overall cancer incidence in-creased m ore than 1 percent per year over thatdecade. The large changes, major increases inlung cancer and decreases in cervical andstomach cancers, are the same as th ose observedin mortality trend s. How ever, increases seen atother sites are not universally accepted as re-flecting actual change, because of differences inmethodology between TNCS and SEER. Con-

tinuation of the SEER program may providedata to better answer questions about cancer in-cidence.

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


Ch. 3—Factors Associated With Cancer q67

prod uce cancer of the lung; and w ith alcohol toproduce cancer of the esophagus. One of thebest known synergisms is the interaction of tobacco an d asbestos.

Asbestos and Smoking

In 1979, the Surgeon General st ated (286):

Asbestos provides one of the most dramaticexamples of adverse health effects resulting frominteraction betw een the smoking of tobaccoproducts and an agent used in the workplace.

Exposure to asbestos is one of the most exten-sively studied occupational health hazards, andone which continues to attract the efforts of epi-demiologists, in part because of its interactionwith smoking, Indisputable evidence links as-bestos to lun g cancers and to rar e cancer typ es,mesotheliomas of the pleura and the perito-neum. Smokers exposed to asbestos, especially

those exposed at high levels, have a m uch great-er pr obability of developing lung cancers thaneither smokers not exposed to asbestos, or non-smokers exposed to asbestos. Rates for meso-theliomas, on the other hand, are similar forsmokers and nonsmokers exposed to asbestos.

Hammond, Selikoff, and Seidman (156) haveanalyzed mortality data from a large group of U.S. male asbestos insulation workers. Com-putations were carried out comparing the in-sulation workers to a control group of malesfrom a large cohort assembled by the AmericanCancer Society (ACS). Table 10 presen ts the re-sults of the analysis, supporting the existence of strong synergism between smoking and expo-sure to asbestos in the prod uction of lung can-cer. As the au thors explain, if the tw o exposures

were acting independently in this cohort, onewou ld p redict the following (156):

. . . the lung cancer d eath rate of asbestosworkers with a history of cigarette smokingshould be very close to the sum of the followingthree numbers: 11.3 (the rate for the “no,no”group), 47.1 (the mortality difference for the

“yes, no” grou p), and 111.3 (the mortality d if-ference for the “no, yes” group ). The sum comesto 169.7 lung cancer deaths per 100,000 man-year s . . . . In contrast, the observed lun g can-cer death rate of the “yes, yes” group was 601.6

per 100,000 man-years. The difference (601.6 –169.)7 = 431.9 lung cancer deaths per 100,000man-years, was presumably due to a synergisticeffect in men w ith both of the tw o types of ex-posu re . . , .

Another measure, in addition to the mortalitydifference, is the mortality ratio, which in thiscase is the lung cancer d eath rate in each of thefour exposure groups, divided by the lungcancer death rate in the group of nonsmokerswho were not exposed to asbestos (group 1).The mortality ratio for exposure to both agents,53.24, is much higher than would be expectedfrom the additive effects of cigarette smokingalone (mortality ratio = 10.85) and asbestos ex-posure alone (mortality ratio = 5.17). The ef-fects appear to be multiplicative.

Attribution of Risk

In the case of smoking and asbestos, and inother cases where two or a variety of factors in-teract to produce some cases of cancers, thedisease may be prevented by interventions inany of the factors. Shared responsibility, how-ever, complicates the attribution of risk to a

Table 10.—Age-Standardized Lung Cancer Death Ratesafor Cigarette Smoking and/or Occupational Exposure

to Asbestos Dust Compared With No Smoking and NO Occupational Exposure to Asbestos Dust

Exposure to History of Mortality MortalityGroup asbestos? cigarette smoking Death rate difference ratio

Control No No 11.3 0.0 1.00Asbestos workers Yes No 58.4 + 47.1 5.17Control No Yes 122.6 + 111.3 10.85

Asbestos workers Yes Yes 601.6 + 590.3 53.24

aRat,s ~~r 100000 man. YearS standardized for age on tfle distribution Of tfle man-years of all the asbestos workers. Number Of lUn9 cancer deaths b=ed on death Certlfl-

cate information.

SOURCE: Hammond, Sellkoff, and Seldman (156)

8/14/2019 8104



particular factor. Adding up the numbers of cases that might be prevented by various indi-vidual measures taken separately may producea total num ber of “preventable” cancers largerthan the number that actually occurs. An exam-ple of this concept, presented in table 11, isbased on Lloyd’s (214) analysis of the asbestosinsulation worker mortality data describedabove. He estimated the percentage redu ction inlung cancer mortality that would accrue byeliminating smoking cigarettes alone to be 88.5percent, and by eliminating exposure toasbestos alone to be 79.6 percent. By eliminat-ing both exposures, the total redu ction w ould be97.8 percent, and not the su m of the ind ividualreductions, which w ould amou nt to a w hopping168.1 percent. The multifactorial nature of

Table 1 1.—Estimates of Percentage Reduction inLung Cancer Mortality in Asbestos Workers by

Elimination of Exposure to Cigarettesand to Asbestos

Percentage reduction fromStatus current rate

Current. . . . . . . . . . . . . . . . . . . . 0.0Eliminate smoking only . . . . . . 88.5

Eliminate asbestos only. . . . . . 79.6Eliminate smoking andasbestos. . . . . . . . . . . . . . . . . . 97.8

SOURCE: Lloyd (214).

cancer means that it is impossible a present aneat balance sheet add ing up to 100 percent thatindicates the prop ortion of all cancers that canbe attributed to factors X, Y, and Z.

ESTIMATES OF THE AMOUNTS OF CANCER ASSOCIATED WITH

ENVIRONMENTAL FACTORS

Diverse methods have been used to produceestimates of the amounts of cancer associatedwith various exposures and behaviors. Themethods are variously described as rangingfrom “seat of the pants, ” to “top of the head ,” tofigures based on more quantified inputs. Evi-dence for associations between various “factors”and cancer ranges from very strong to veryweak. There is no relationship between thestrength of the association and the estimatedmagnitude of the amount of cancer associatedwith a factor. For instance, the strongest asso-ciations include those ‘between smoking tobaccoand respiratory cancers, between asbestos andcancer of the lung an d oth er sites, and betw eenionizing radiation and cancer at many sites.While each of the three associations is strong,the percentage of cancer associated with each isdifferent. Smoking is associated with more than20 percent of cancer, asbestos with between 3and 18 percent, and natural background radia-tion, with less than 1 to 3 percent.

The importance of a factor, as measured bythe magnitude of the proportion of cancer withwhich it is associated, is one starting point fordeciding on the developm ent of preventive strat-

egies. However, a large proportion, by itself does n ot give any indication of the practicalityor availability of strategies. If all factors wereequally well understood, and preventive strat-egies equally available, one wou ld choose to gofor big redu ctions. Under real conditions, strat-egies that can be implemented receive pref-erence.

Most nu merical estimates of associations arebased on cancers occurr ing today or in the last

few years. Therefore, they are n ot predictors of carcinogenic activity today, but reflections of carcinogenic activity in the past, possibly 20 to50 years ago. A few au thors have attem pted torelate today’s carcinogenic risk to future cancersand these are also cited.

All factors discussed here h ave not been” in-vestigated equally in the scientific community,either because of perceived differences in rela-tive importance, or because of difficulty in ob-taining meaningful results. Therefore, it has notbeen possible for “equal” evidence to be pre-sented for each factor in this assessment. Evi-dence for carcinogenicity in humans from goodepidemiologic stud ies is given more w eight thanare animal data.

8/14/2019 8104


Ch. 3—Factors Associated With Cancer q 69

The remainder of this chapter discusses whatis known about th e contributions of the follow-ing factors to cancer incidence and mortality:tobacco, alcohol, diet, occupation, pollution,consumer products, medical drugs and radia-tion, sexual development, reproductive patternsand sexual practices, natural radiation, infec-

tion, and other or unknown associations. Cur-rent thinking, with some historical background,is presented for the m ajor elements of each fac-tor, and attempts are mad e to mention studiesgiving quan tified estimates of the importan ce of various factors.

1

I Discussion of the factors draws heavily on a contract report bySir Richard Doll and Richard Pete, who were charged with review-ing existing literature about the quantified estimates of cancer cau-sation. They also made their own estimates, which are cited in thisassessment. Doll and Pete’s repor t, in its entirety, is pu blished inth e ]ournul of National Cancer lm-titute (93).

TOBACCO

Diseases related to the smoking of tobacco in-clud e lung cancer and cancer at oth er sites, cor-onary h eart d isease and stroke, chronic bronchi-tis and emp hysema, and m any other diseases,includ ing p eptic u lcers (157). Tobacco smoking“is the single most important preventable en-vironmental factor contributing to illness,disability, and death in th e United States” (286).WHO (365) states:

Smoking-related diseases are such importantcauses of disability and premature death in de-veloping countr ies that the control of cigarettesmoking could d o more to improve h ealth andprolong life in these countries than an y single ac-tion in the whole field of preventive medicine.

The harmful effects of tobacco are greaterwh en it is smoked as cigarettes than w hen con-sumed in other forms. This may be because acidcigarette smoke is less irritating than thealkaline smoke from p ipes and cigars and , there-fore, more easily inhaled. However, tobaccoconsump tion in any form ap pears to be accom-

pan ied by adverse effects, most recently demon-strated in a study show ing that long-time snuff dippers experience a highly increased risk of oral cancer (363).

The categories used ar e not necessarily “natu -ral” assemblages, nor are they the only possiblegroupings of the components. The discussion of each factor includes a description of importantinclusions and exclusions. For instance, the“diet” section looks at all “foodstuffs,” includingnaturally occurring and added contaminants.

Drinking water, which is discussed und er “pol-lution, ” and “alcohol,” which is treated as a fac-tor u nto itself, are exclud ed from diet.

All of the estimates considered in preparingthe following discussion of factors are listed intable 19, at the end of this chapter. Only “bestestimates, ” either point estimates or intervals, aspresented by each author are included in the ta-ble. The p rimary references should be consultedfor acceptable ranges and / or confidence limits,exact data sou rces and method s, and caveats.

Tobacco is known to contribute more heavilyto the number of cancer deaths than any othersingle substan ce.. The relationship of cigarettesmoking and cancer was first suggested in the1920’s. During th e 1950’s, results from man y ep -idemiologic studies confirmed this association(287). Many carcinogens have been identified incigarette smoke, and the differences consistentlyobserved between rates of lung cancer amongregular cigarette smokers and lifetime non-

smokers is so extreme th at it is not likely to bean artifact of the epidemiologic method (287):

The 1964 Surgeon General’s Report reachedthe following conclusion: ‘Cigarette smoking iscausally related to lung cancer in men . . . Thedata for women, though less extensive, point inthe same d irection. ’

Today, cigarette smoking is regarded as the major cause of lung cancer in both males and fe-males and is largely responsible for the recentrapid rise in female lung cancer rates. The 1980Surgeon General’s report, The Health Conse-

quences of Smoking For Women, states (287):

. . . the first signs of an epidemic of smoking-relateddisease among women are now appearing.

8/14/2019 8104



It has been predicted that by the early 1980’s,the age-adjusted female lun g cancer death ratewill surpass the breast cancer rate, which istoday’s leading cause of cancer death in w omen.The Surgeon General’s 1980 report estimates(287):

. . . smoking will contribute to 43 percent of themale and 18 percent of the female newly diag-nosed cancer cases in the United States in 1980and to 51 percent of the male and 26 percent of the female cancer deaths.

The principal impact of tobacco smoking ison the incidence of cancer of the lun g althoughcancer at many other body sites, includinglarynx, oral cavity, esophagus, urinary bladder,kidney, and pancreas are also associated (287).By late mid dle age, the r isk of developing lungcancer is more than 10 to 15 times greater incigarette smokers th an in lifelong non smokers

(152). Present evid ence ind icates that th e effectof smoking on the d evelopment of lung cancer isaffected by n um ber of years smoked, age whensmoking began, the number of cigarettes smok-ed per day, the degree of inhalation, and thecomposition of the cigarette.

Studies have shown that discontinuing smok-ing d ecreases the risk of developing lung cancer.Ten to fifteen years following cessation of smoking, an ex-smoker’s risk of dying of lungcancer d ecreases to a level only abou t tw o timeshigher than the risk for lifelong nonsmokers(286). This phenom enon is nicely illustrated by

data in table 12 from an ep idemiologic stud y of British doctors (92).

Higginson and Muir (166) attributed 30 per-cent of male and 7 percent of female cancersfrom 1968 through 1972 in the Birmingham andWest Midland region of England to tobacco.They specifically ascribed 80 to 85 percent of lung cancers to smoking. Wynder and Gori(368) estimated that 28 percent of male and 8percent of female 1976 cancer deaths in theUnited States can be attributed to smoking.Their estimate is based on calculating the per-

cent difference between U.S. mortality rates andthe lowest reported worldw ide mortality ratesfor each site and by considering specific case-control stud ies.

Table 12.—Lung Cancer Mortality Ratiosin Ex.Cigarette Smokers, by Number of

Years Stopped Smoking

MortalityYears stopped smoking ratio

Still smoking. . . . . . . . . . . . . . . . . . . . . . . . . . . 15.81-4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . “16.@

5-9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.910-14 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.315+ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.0Nonsmokers . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.0

aThe higher mortality ratios observed in the 1-4 year Category after quittin9

compared to those continuing smoking is believed to be due 10 thoseindividuals who quit smoking because of illness,


Enstrom (100) reported:

If all Americans did not smoke, the mortalityreduction that would occur has been estimatedto be 80,000 lung cancer deaths plus 22,000other cancer deaths of the “1978 total of 390,000cancer deaths—a redu ction of 26 percent.

Data from a representative sample of non-smokers derived from a 1966–68 National Mor-tality Survey indicated that this group had atotal cancer rate which was 24 percent less thanthat of all U.S. whites (100).

Nu merous estimates have been mad e of thecontribution of tobacco to the overall cancerrate. Most have taken advantage of concurrentepidemiologic cohort studies, in which largenumbers of people are queried about theirsmoking habits at the initiation of the stud y andthen followed to determine their causes of

deaths. The largest such stud y involved 1 mil-lion Americans who were self-identified assmokers or nonsmokers in 1959 and whose sub-sequent m ortality was mon itored by ACS. Themagnitude of the excess risk observed forwomen in the ACS population was less than formen. The difference is thought to be due towom en having smoked fewer cigarettes per day,inhaling less deeply, and being more likely tosmoke cigarettes with redu ced tar and nicotine(153). In addition, women are less frequently ex-posed to occupational hazard s, including thosethat m ay act ad ditively or synergistically with

tobacco to cause cancer.

Several researchers have evaluated data fromthe ACS study population. Doll and Peto (93)

8/14/2019 8104



computed mortality rates from the ACS studygroup and estimated that between 25 and 40percent of 1978 American cancer deaths, with abest estimate of 30 percent (94,782 or 43 percentmale; 27,266 or 15 percent fema le), could be at -tributed to tobacco smoking. Their computationmethod assum es that the male and female age-

specific cancer death rates observed am ong theACS nonsmokers, between 1960 and 1972,would have applied to the entire country had noone smoked.

Hammond and Seidman (155) estimated thatfrom 25 to 35 percent of cancer mor tality in theU.S. male pop ulation and 5 to 10 percent in th efemale population are “mainly due to smokingof tobacco products and cigarettes. ” Enstrom(100) calculated a 38-percent reduction in thetotal cancer rate in the “never sm oked regu lar-ly” group as compared with all U.S. whites.These estimates from ACS data d o not greatly

differ and the reported differences can be ex-plained by the different methodological ap-proaches and assumptions used. Hammond andSeidman (155) assumed that the age-specificdistribution of smoking habits in the ACS groupwas similar to the country as a w hole and reliedon mortality rates computed for individualsclassified as smokers in 1959, many of whomare know n to h ave quit th e habit by 1967. In theapproach taken by Doll and Peto (93), it wasunnecessary to speculate on the smoking habitsof the general population and instead of using

mortality rates for smokers, they relied on can-cer rates in the nonsmoking group. The specificsof Enstrom’s (100) calculations were not given .

Data derived from a uniqu e study cohort suchas the ACS popu lation ar e not free of bias, butthe risks estimated from the study are com-parable with those found in a study of U.S.veterans. Rogot and Murray (312) reportedmortality ratios for cigarette smokers versusnonsmokers and ex-cigarette smokers versusnonsmokers in a cohort of 250,000 AmericanWorld War I veterans. After a 16-year followupperiod, lun g cancer d eaths occurred 11.3 timesmore frequently among smokers; laryngeal can-cer 11.5 times, buccal cavity cancer, 4.2 times;pancreas and bladder cancer, approximately 2times; and death s from cancer at all other sites1.38 times more often (see table 13). Studiesfrom Gr eat Britain (94,192) and other count ries(286,287) show similar elevated cancer deathrates among smokers.

Many of these estimates have been criticizedfor overstating the impact of tobacco on cancer.Objections are expressed because the studiesmeasured mortality rather than incidence, andbecause they did not take into account im-provements in survival, changes in smokinghabits, and changes in cigarette composition.None of the estimates attempt to account forany effects of “passive smoking” and it is onlyrecently that evidence of a carcinogenic effect

Table 13.–Mortality in a Population of U.S. Veterans Classified as Smokers Compared to Mortality Expectedfor Nonsmokers

Cigarette smokers Ex-cigarette smokersa

Cause of death (7th Revision lnter- Observed Expected Observed Expected

national Classification of Disease) deaths deaths 0 - Eb

deaths deaths 0- Eb

All cancers (140-207). . . . . . . . . . . . . . . . . . . . . . . . 7,608 3,590’ 2.12 2,816 2,025C 1.39

Cancer of buccal cavity (140-144) . . . . . . . . . . . . . 110 26 4.22 24 14 1.67

Cancer of pancreas (157) . . . . . . . . . . . . . . . . . . . . 459 256 1.79 170 145 1.17

Cancer of larynx (161) . . . . . . . . . . . . . . . . . . . . . . . 94 8 11.49 22 5 4.78Cancer of lung and bronchus

(162.1, 162.8, 163) . . . . . . . . . . . . . . . . . . . . . . . . . 2,609 231 11.28 517 130 3.97Cancer of bladder and other urinary

organs (181) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326 151 2.16 126 90 1.41

All other cancers . . . . . . . . . . . . . . . . . . . . . . . . . . . 4,010 2,916 1.38 1,957 1,642 1.19

aonly ex.cigarette Smokers who stopped smoking cigarettes for reasons other than physicians’ orders.% + E values are based on expected numbers to two decimal places.cvalues d. not exactly total due to rounding.

SOURCE Rogot and Murray (312)

8/14/2019 8104


.


from such exposure has been documented.Hirayama (167) reported an approximately two-fold increased risk of developing lung canceramong nonsmoking wives of cigarette smokers.The effects may even be greater; a fourfold ex-cess was estimated by Doll and Peto (93), wh oconsidered lifelong exposure, as would be thecase w ith children.

A more p recise estimate of the pr oportion of cancer associated with tobacco smoking re-quires a m ore exhaustive study of lung cancer.Such an effort might be desirable because of theimportance of lung cancer to overall cancerrates. As American lung cancer death rates con-tinue to rise, the estimate of the p ercentage of cancer deaths caused by tobacco will likewiseincrease. The futu re course of lung cancer ratesdepends largely on patterns of cigarette con-sumption and possibly on changes in tar andnicotine yields. The latter point may be par-

ticularly importan t in light of the large num bersof smokers who have switched to lower tar andnicotine cigarettes.

There is some evidence to indicate that low-tar/ nicotine cigarettes may h ave a lesser car-cinogenic risk (6,153,157,287). The tar and nic-otine yield is believed to be a function both of the tar/ nicotine content of the tobacco and thenumber of puffs taken (202). Therefore, a reduc-ed risk is depend ent on th e smoker’s behavior aswell as the cigarette itself. The decreased risk from a “less hazardous” cigarette may besomewhat offset by an increase in the num ber of

cigarettes smoked. Reports indicate that withthe increased produ ction of low-tar/ nicotinecigarettes, there has been an increase in thenumber of cigarettes consumed per currentsmoker (286). It is also unknown whetherchemicals newly added to cigarettes andchanges in the composition of the gaseous p hasewill have health consequences.

The extent to w hich the increase in m ale andfemale lung cancer rates over the last few d ec-ades can be accounted for by tobacco is de-bated. Doll and Peto (93) argue that the increase

within the last century can almost totally be ex-plained by cigarette smoking, while others,

Schneiderman (321) and the Toxic SubstancesStrategy Committee (345) included, contendthat additional factors are responsible. (For amore complete discussion, see ch. 2.)

Examination of cigarette-consumption pat-terns in this country leads to speculations aboutfuture cancer statistics at tobacco-related sites.

The proportion of adult men smoking cigaretteshas declined from 51 to 37 percent during theperiod 1965 to 1978 (287). There has also been aslowdown in the rate of initiation of smokingamong adolescent m ales. The decreases began atthe time of release of the first Surgeon General’sreport in 1964, and the w idespread discussion of the dangers of smoking that followed (287).This information demonstrates that worthwhiledecreases in cigarette consumption can takeplace even without radical Government inter-vention , chiefly by increasing public awarenessof the hazards of smoking. The proportion of

adult women who smoke remained virtuallyconstant at aroun d 32 percent between 1965 and1976, and has since started declining slightly.Unfortunately, the rate of smoking initiationamong young wom en has n ot declined (287).

Changes in smoking habits that should lowerrisks are believed to have already influencedlung cancer rates. The rate of increase in lungcancer among men under 65 years of age hasslowed du ring the last decade (see table 8 andfig. 10). Recent female mortality statistics arealso prom ising, for they indicate that femalelung cancer rates in the 30- to 4&year age group

are steady or decreasing. There is reason tohope that w ith continued r edu ctions in exposur eto the harmful comp onents of cigarettes, the d e-creases will follow throu gh to older age grou ps.

Public health laws exclude tobacco from reg-ulatory action because smoking tobacco isviewed as a p ersonal decision, and one in w hichCongress has decided not to intervene. TheGovernment limits its responsibility to inform-ing smokers and potential smokers of thehazards of cigarettes, conducting behavioralstudies on ways of affecting smoking habits,

and supporting research on low-tar/ nicotinecigarettes.

8/14/2019 8104



ALCOHOL

Alcohol is considered next because of the in-teraction betw een tobacco smoking and alcohol-related cancers. The National Institute on Alco-hol Abu se and Alcoholism (NIAAA) (257) esti-mated that approximately one-third of adult

Americans drink alcoholic beverages at leastonce a week and another third do not drink soregularly but drink primarily on special occa-sions. In ad dition, NIAAA estimated that thereare 9.3 to 10 million “problem drinkers, ” in-cluding alcoholics, in the adult population.Alcohol consumption is influenced by num erouspersonal characteristics including sex, age,education, socioeconomic status, occupation,residence, ethnicity, and religious affiliation.

Alcohol’s association with cancer was firstsuspected around the turn of the century. To-day, as is the case with tobacco, there is in-

disputable evidence that alcohol consumptionincreases the risk of developing cancer atvarious body sites. The 1978 Department of Health, Education, and Welfare (DHEW) re-port, Alcohol and Health (257), states:

In comparison to the general population,heavy consumers of alcohol always show amarked excess of mortality from cancers of themouth and pharynx, larynx, esophagus, liver,and lung. In the United States, these cancersrange from 6.1 to 27.9 percent of the total in-cidence of all cancer recorded in those locationswhere there are cancer registries.

Epidemiologic stud ies have d emonstrated th atcancers are more common in m en emp loyed intrades that encourage the consumption of largeamou nts of alcohol. A recent stud y showed thatDanish brewery workers have a higher risk of developing cancer then th e general popu lation(193). Clinical evidence suggests that con-sum ption o f alcohol at levels sufficient to causecirrhosis of the liver also increases the incidenceof liver cancer (257).

Researchers have tried to determine whetherthe association between alcoholic drink and

cancer is due to the alcohol itself or to otherchemicals found in spirits, wines, and beers(348). Pure gra in alcohol (ethanol) has n ot been

shown to be carcinogenic in animal bioassays(93,277); however, both pure ethanol and meth-anol, a contaminant of many alcoholic bever-ages, are mutagenic in bacteria (160,277). Addi-tionally, many alcoholic drinks ar e found to be

mu tagenic. The resuIts from such experimentsdo n ot yet lead to firm conclusions, but it m aybe that components of alcoholic beverages aswell as alcohol itself are related to increasedcancer risk. Some evidence suggests that the risk of cancer is greatest when alcohol is consumedas spirits and that the apple-based drinks con-sumed in Northwest France are particularlyharmful (93). However, the carcinogenic effectof alcoholic beverages is largely independent of the form in which it is drunk. In addition to adirect effect, alcohol may also exert a carcino-genic effect by facilitating contact between ex-

trinsic carcinogenic chemicals and the contentsof the stem cells that line the upper digestivetract and larynx.

Epidemiologic evidence supports the viewthat excessive alcohol consumption increasesthe risk of developing cancers of the mouth (ex-cluding lip), larynx, and esophagus and thatalcohol acts additively and even synergisticallywith tobacco in th e pa thogenesis of cancers of the u pp er d igestive tract (257,367).

Most estimates of the percentage of cancerassociated with alcohol fall in the 3- to 5-percent

range. From data presented by Schottenfeld(233), alcohol appears to be associated with 4 to5 percent of 1978 U.S. cancer deaths; Wynderand Gori (368) estimated 4 percent male and 1percent fema le 1976 U.S. cancer inciden ce; andHigginson and Muir (166) estimated 5 percentmale and 3 percent female 1968–1972 cancers inthe Birmingham and West Midland region of England.

Rothman (313) (see table 14) and Doll and

Peto (93) estimated that 3 percent of U.S. cancermortality is related to alcohol. Rothman’s

calculations are based on attributing a p ropor-tion of 1974 cancer deaths at each alcohol-related bod y site to alcohol consum ption. Doll

!30-4’31 o - 81 - 6

8/14/2019 8104



Table 14.—Praportion of Cancer Deaths Attributable to Alcohol Consumption by Site and Sex

Males Females

Number of Number ofdeathsa

Percent ascribedb Number of deaths deathsaPercent ascribedb

Number of deathsBody site from cancer to alcohol ascribed to alcohol from cancer to alcohol ascribed to alcohol

Buccalcavity andpharynx . . . 5,686 50 2,843 2,282 40 913

Esophagus . . 4,917 75 3,688 1,735 75 1,301Liver. . . . . . . . 1,600 30 480 865 30Larynx . . . . . . 2,826 50 1,413 436 40 174 — —

Total. . . . . . 15,029 8,424 5,318 2,648

Total cancerdeaths . . . . 199,194 4.2 166,338 1.6

aln 1974.

bpropo~lon of alcohol. cause~ cancer at each site from (313).

SOURCE: Adapted from Rothman (313).

and Peto assum e that cancers at sites related toalcohol consumption account for 7 percent of allcancer deaths in men and 3 percent in wom en.

They attribute to alcohol about two-thirds of these cancer deaths in men, about one-third inwomen, plus a small proportion of the livercancer d eaths (wh ich constitute less than 1 per-cent of all cancer deaths) and derive overallestimates for the combined sexes of 3 percent(range: 2 to 4 percent:). It should be emphasizedthat most of the cancer risk posed by alcohol

consumption is also related to cigarette smok-ing. Therefore, most of the cancer deaths causedby alcohol would be avoided in the absence of

smoking even if alcohol consumption remainedunchanged.

Table 15 shows ann ual incidence rates for al-cohol-related cancer sites for several regions inthe Un ited States wh ere cancer registries exist

(348). Among the white population, the propor-tion of total cancer incidence associated with

Table 15.—Age-Adjusted Annual incidence Rates for Selected Cancer Sites in Various Population Groups inthe United States

Total for Proportionthe 7 01 all

Place Population Tongue Mouth Oropharynx Hypopharynx Esophagus Liver Larynx cancer sites cancers (O/.)

California:Alameda White 3.0

Black 2.23.6

13.23.74.1

2.22.2

1.11.5

2.24.3

7.912.9

23.740.4

8.512.3

13.612.5

Bay Area White 3.2Black 2.1

4.24.8

2.63.3

2.1

1.51.5

4.0

15.22.84.2

7.511.8

25.842.9

Connecticut 2.8 4.3 1.5 5.7 2.0 7.8 26.2 9.2

Iowa 1.4 2.6 1.1 1.2 3.0 1.6 5.8 16.7 6.7

Detroit White 2.7Black 3.3

3.33.3

2.02.1

1.21.1

4.0

14.12.64.5

7.57.7

23.336.1

&711.3

New Mexico Spanish 0.4Othe r wh i t e 2 .2

0.72.8

0.41.4

0.23.0

2.23.0

3.03.1

2.75.8

9.618.6

6.1

6.7

New York State 2.2 3.2 1.3 0.8 4.5 1.9 5.9 19.8 8.0

Puerto Rico 7.5 7.8 4.3 4.4 14.8 3.3

0.9

6.4 48.5 27.9

Utah 2.1 2.5 0.9 0.4 1.8 4.4 13.0 6.1

SOURCE: Tuyns (348)

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104



Table 16.—Some Currently Attractive Hypothetical or Actual Waysin Which Diet May Affect the Incidence of Cancer

Possible ways or means Example a

Ingest ion of powerful, direct-acting carcinogens or t heir Carcinogens in natural foodstuffs (plant products)precursors Carcinogens produced in cooking

Carcinogens produced in stored food by microorganisms(bacterial and fungal)

Affecting the formation of carcinogens in t he body Providing substrates for the formation of carcinogens in thebody (e.g., nit rites, nitrates, secondary amines)

Altering intake or excretion of cholesterol and bile acids (andhence the production of carcinogenic metabolizes in thebowel)

Altering the bacterial flora of the bowel (and hence the capaci-ty to form carcinogenic metabolizes)

Affecting transport, activation, or deactivationof carcinogens

Altering concentration in, or duration of contract with, fecesAltering transport of carcinogens to stem cellsInduction or inhibition of enzymes (which affect carcinogenmetabolism or catabolism)

Deactivation, or prevention of formation, of short-lived in-tracellular species (e.g., by use of selenium, vitamin E, orotherwise trapping free radicals; by use of b-carotene orotherwise quenching singlet oxygen; by use of other antiox-idant)

Affecting “promotion” of cells (that are already

initiated)b

Vitamin A deficiency (clinical or subclinical)

Retinol [Binding Protein] (hormonal and other factors deter-mine blood RBP, though vitamin A intake may not affect itmuch)

Otherwise affecting stemcell differentiation (carotenoids?determinants of lipid “profile”?)

Overnutrition Age at menarcheAdipose-t issue-derived estrogensOther effects

aThere may be considerable overlap between many of the efltrleS In this tablebor, more ~enerally, affecting th

e~robablllty that a partially transformed stem cell will become fully transformed and Will proliferate successfully Into Cancer

SOURCE Doll and Peto (93)

while in 1977 the rate w as 8.8/ 100,000 for m alesand 4.0/ 100,000 for females. This decrease hasoccurred in many other countries, including

those w ith high initial rates, such as Japan andIceland, and those w ith lower initial rates, suchas Canada and New Zealand. The factors be-lieved to have contributed to these decreases in-clude: reduction in use of salt and pickling,lower consumption of smoked foods, and in-creased use of refrigeration, increased con-sumption of milk, green vegetables, fruit, andantioxidants (237).

In general, as a result of increased intake of calories, proteins, and certain other nutrients,Americans have been growing taller and reach-ing sexual maturation earlier. A great improve-

ment in the Nation’s health has resulted fromthis change, but increased risk for certain cancersites may accompany the improvement. For ex-

ample, earlier sexual maturation in women isassociated with higher risk of breast cancer laterin life, though as yet, no increases in breast

cancer has been attributed directly to improvednutrition (see Sexual Development, Reproduc-tive Patterns, and Sexual Practices for a morecomplete d iscussion of this topic).

Many estimates of the importance of diet tocancer hav e been p ut forth. Doll and Peto (93)estimated that altering dietary practices mayredu ce cancer by as mu ch as 35 percent (stom-ach and large bowel cancer by 90 percent; en-dometrium, gallbladder, pancreas, and breastby 50 percent; lung, larynx, bladder, cervix,mouth, p harynx, and esophagu s by 20 percent;others by 10 percent). The great u ncertainty of

this estimate is ind icated by the ra nge of 10 to 70percent which they attach to their estimate.Wynd er an d Gori (368) estimate, by calculating

8/14/2019 8104



the percent difference between U.S. mortalityrates and the lowest rates reported worldw ideand by considering specific case-control studies,that an even larger proportion of cancer, ap-proximately 40 percent in males and 60 percentin females, could be attributable to d iet.

In testimony before the Subcommittee on

Nutrition of the Senate Committee on Agri-culture, Nutrition, and Forestry, Dr. ArthurUpton (353), then-Director of the National Can-cer Institute (NCI), succinctly discussed th e ex-tent of involvement of diet and cancer w hen hestated:

Despite the impression that cancers are linked withdietary patterns and the inability to pinpoint specificdietary carcinogens, scientists generally agree thatfactors in d iet and nutr ition-includ ing dr inkingwater contaminants—appear to be related to a largenumber of human cancers, perhaps app roaching 50percent.

Dietary Intake

Fat/Meat Intake

Examination of cancer rates in different coun-tries show positive correlations between coloncancer and consumption of meat and animalprotein and between cancer of the breast and en-dometrium with total fat consumption (17).These cancers are common in the United States,Canada, and W est ern Eu rop e and r ar erthroughout the developing world.

The most widely held theory is that fat in thediet has a promotional effect on the devel-opment of cancer. One suggested mechanism isthat fats affect hormone levels. Several cancers,includ ing breast, ovarian, and endom etrial are,in turn, influenced by hormones. This associa-tion of dietary fat with higher cancer rates is,however, not found uniformly. Breast and col-orectal cancers are not uncommon among vege-tarians, and the observed incidence in SeventhDay Adventists who are largely vegetarian, ismatched by the same incidence in Mormonswho eat meat (293). This finding and the fact

that meat intake among Morm ons is not mark-edly lower than among th e general pop ulation isoften cited as a rebuttal to the meat/ fat cancercausation hypothesis. These interpretations

should be viewed cautiously for the stud ies maynot have been adequate to reflect any promo-tional effect in these low risk groups who areless exposed to many other types of carcino-genic stimuli, such as, cigarettes and alcohol,than is the general population. Additionalstudies in this area are needed.

More and more, studies have focused onspecific types of fat. Evidence suggests that dietswith a high ratio of polyunsaturated fats (main-ly of vegetable origin) to saturated fats (mainlyof animal origin) may increase the risk of can-cer. Paradoxically, this type of diet is recom-mended to lower the risk of heart disease. Re-sults from epid emiological stud ies (98,290) andanimal trials (53,305) have been inconsistentand confusing, and have show n different effectson different cancer sites. In rats, polyun-saturated fats need be only a small proportionof a total high fat intake to promote breasttumor incidence. If the animal model is ap-plicable to humans, virtually all high-fat dietswill exert a p romotional effect (52).

Serum cholesterol has been investigated inmany studies as a risk factor for variouscancers. Serum levels are directly affected by in-take of cholesterol and fat. A diet with a lowsaturated fat to polyunsaturated fat ratiodecreases the amount of ingested cholesterolthat will appear in the serum. The discoverythat the stools of colorectal cancer patients con-tain an abnormally high proportion of acidsteroids, derived from bile salts, and cho-

lesterol, supports the hypothesis that certaintypes of fat play a r ole in th e produ ction of col-orectal cancer.

A recent epidemiologic study found an asso-ciation between high-density lipoproteins andcancer risk (199). On the other hand, in a pro-spective epid emiologic study of heart d isease inFramingham, Mass., serum cholesterol levelswere inversely associated with the incidence of colon cancer and other sites in men. Men w iththe lowest serum cholesterol levels experienceda colon cancer rate which was three times higher

than men with the higher cholesterol levels(361). A similar negative association was re-ported in data from the Paris Prospective Studyof Coronary Heart Disease (45).

8/14/2019 8104



The hypoth esis that dietary cholesterol playsa d irect role in the p rodu ction of colon cancer issupp orted by som e animal studies. For instance,the ad dition of cholesterol to the d iet of rats on acholesterol-free liquid diet, had a p romotionaleffect and cholesterol enhanced the carcinogeniceffect of a known carcinogen (1,2 dimethyl-hydrazine).

Fiber Intake

Burkitt (40) observed that several intestinaldiseases that are common in developed coun-tries are rare in rural Africa where unprocessedfood is consum ed, and the stools tend to be soft,bulky, and frequent. It is suggested that dietaryfiber m ay r edu ce colorectal cancer by d ecreas-ing the time stools remain in th e bowel, by in-creasing bulk (thereby decreasing the concentra-tion of carcinogens in stool), or by perhapsaltering the distribution of bacteria, some of

wh ich m ay pr odu ce or d estroy carcinogenic me-tabolizes.

The effects of fiber on cancer incidence haveattracted mu ch interest but remain problemati-cal. As in th e case of fats, dietary fiber is a termwhich covers a multitude of different sub-stances, each of which may have different influ-ences on carcinogenesis. The methods by whichtypes of fiber are chemically characterized arestill in a primitive state, and analyses of thecomposition of dietary fiber in many foods islacking. Despite the drawbacks, some associa-

tions have emerged. A close inverse correlationhas been found between th e pentose fiber con-tent of the diet and mortality from colon cancerin pa rt of the Un ited Kingdom (93). No correla-tion was shown with any other fiber types, orwith d ietary fiber as a w hole.

Elements and Vitamins

Several metals and vitam ins are linked withcancer d evelopment. They can either act as car-cinogens themselves or they may biologicallycompete with other dietary constituents toenhance or sup press a cancerous response.

An iron-deficient syndrome was shown to beassociated with a high risk of developing can-cers of the pharyn geal and esophageal mucosa

in northern Scandinavia. The incidence of gastric cancer is foun d to be 4 to 5 times high erin countries where iron d eficiency is prevalentthan in the United States (356).

Selenium, an effective antioxidant, is oftenfound at higher levels in plants, milk, or humanblood in sections of the United States with low

cancer rates. Selenium deficiency in rats con-sistently increases the carcinogenic effect of known chemical carcinogens, particularly inanimals fed on high polyunsaturated fat diets(190). Supplementation with selenium abovedietary requirements decreases tumor yield inanimals on both low- and high-polyunsaturatedfat diets (188).

Stocks and Davies (24) found that a high zinc-copper ratio in soils was associated withelevated rates of human stomach cancer whileStrain et al. (24) reported an association of elevated cancer rates with low zinc-copperratios. Marginal zinc deficiency is associatedwith increased esophageal cancer in animals(122) and esophageal cancer patients had lowerlevels of zinc in their blood, hair, and tumortissue than controls in a stu dy of Chinese men(211). Zinc deficiency may interact synergisti-cally with alcohol to enhance esophageal car-cinogenesis (135).

The overall relationship between cancer andvitamins is not well und erstood. Vitamin C hasbeen shown to redu ce carcinogen formation inexperimental animals (359), and this activity

may be important in reducing cancer occur-rence. Vitamin A (retinol) has been more ex-haustively studied than any other vitamin (forreview see 292) and it has been suggested th atvitamin A, or , more particularly, its vegetableprecursor, beta-carotene, may decrease the su s-ceptibility of a variety of epithelial tissues to thedevelopment of cancer. Retinol (or its analogsretinoic acid and various retinoids) has beenrepeatedly demonstrated to diminish the risk of experimentally induced cancer in laboratoryanimals (93). These results are particularly in-trigu ing because the pro tective effect is observed

even when these substances are fed long afterthe animal is treated with an initiating car-cinogen, and the vitamin and its analogs appearto be effective at a w ide va riety of sites.

8/14/2019 8104


8/14/2019 8104



and mutagens are of plant origin, but an impor-tant class, nitrates and nitrites, occurs in bothplants and animals.

Nitrates and N itrites

Nitrate and nitrite salts naturally occur invegetables, fish, and m eat, are added to food fortheir preservative properties (see Additives p .82), and are present in pesticide and drugresidues in food (127). They react with otherchemicals in the body to produce N-nitro-samines and N-nitrosamides. N-nitrosaminesare among the most powerful chemical car-cinogens in laboratory animals, producing tu-mors at a variety of sites including liver, kidney,lungs, esophagus, bladd er, pancreas, trachea,nasal cavities, and peripheral nervous system(221). Experimentally, vitamin C has beenshow n to inhibit the formation of nitrosamines(359).

Epidemiological evidence linking nitrites andnitrates to cancer is fragmentary. Evidence for arelationship b etween gastric cancer and the ni-trite content of the diet is not wholly consistent.It is notable that vegetables, which are u suallythe main source of dietary nitrates, appear topr otect against the developm ent of the disease,possibly because of their vitamin C content (93).On present knowledge, the possible contribu-tion of dietary nitrates and secondary am ines tothe production of cancer is uncertain.

Other Substances

Cycasin in the cycad nut, safrole in sassafras,pyr rolizidine alkaloids in som e plants, and ex-tracts of coltsfoot and bracken fern are a fewnaturally occurring compounds which exhibitcarcinogenic propert ies in animals. Onlybracken fern has been demonstrated to causecancer in man . Bracken fern is comm only eatenin Japan, and Japanese who eat it daily havethree times the r isk of developing cancer of theesophagus as Japanese who d o not eat it at all(93). It has been p ostulated th at the high bowelcancer rates in Scotland m ay be d ue to the inges-tion of cattle fed on bracken fern or the leaching

of the carcinogenic components into water orfood (24). Range-fed cattle and sheep may pass

pyrrolizidine alkaloids along to hu man s in theirmeat a nd milk (215).

Certain plant-derived preparations are asso-ciated with cancer. Extracts of some plants usedto make herba l infusions, for use as hom e rem-edies, tonics or beverages, are carcinogenic inanimals, and their tannin-containing fractionsare p articularly active. Some p opu lation grou pswho use these products have high rates of esophag eal cancer, suggesting an association(297). Perhaps of greater concern, because of itswidespread use, coffee has been associated withhuman bladder (24) and pancreatic cancer (220),but whether the associations are causal is notyet know n. Stud ies in an imal cell cultures haveshown caffeine, a constituent of both coffee andtea, to potentate the effect of carcinogenicsubstances (96).

Mutagenic substances have also been iden-tified in cruciferous plants (cabbage, broccoli,

etc.), from cereal grains, and some grazingrange p lants in the South western United States(MacGregor, 1980). Other not-yet-identifiedcarcinogens and mutagens may occur naturallyin food, but on pr esent evidence, natu rally oc-curring carcinogens are not regarded as an im-portant cause of cancer in the United States (93).

Carcinogens and Precursors Producedby Cooking

Another possible source of carcinogens istheir produ ction in cooking. Hum ans are the on-ly animals that cook their food, and it has beenknow n for man y years that carcinogenic chem-icals such as benzo(a)pyrene (B(a)P) and otherpolycyclic hydrocarbons are produced whenmeat or fish is broiled or sm oked or w hen foodis fried in fat which has been used repeatedly,Sugimura et al. (336) demonstrated that broilingalso prod uces powerful mu tagens that cannot beaccounted for by the production of B(a)P alone.Few people eat more broiled foods than Ameri-cans and while the declining stomach cancerrate provides some assu rance that cancer at thatsite is not related to broiled food, the possibility

remains that colorectal cancer, which has notdecreased m aterially, might be r elated, Recent

8/14/2019 8104


evidence has show n that m utagens can in fact beproduced by cooking at relatively low tempera-tures (100-200C) (93).

Contaminants

Natural Contaminants

A less obvious source of carcinogenic activityand one that was overlooked altogether until theearly 1960’s, is the p rodu ction of carcinogens bymicro-organisms in stored food. Aflatoxin, aproduct of the fungus Aspergillus flavus, is themost pow erful liver carcinogen know n for someanimal species. In addition, human liver cellscontain the enzymes necessary to produce themetabolic products that appear to be responsi-ble for its activity. l-here is eviden ce for believ-ing that aflatoxin is a major factor in the pro-duction of liver cancer in certain tropical coun-tries where it is a contaminan t of carbohyd ratefoods, particularly grains and nuts. The in-cidence rates of primary liver cancer in differentpar ts of Africa are app roximately proportionalto the amount of aflatoxin in the diet (212).

Liver cancer occurs more comm only in p eoplewho are chronically infected with hepatitis Bvirus and it seems p robable that wh ere aflatoxinis present in the diet, both aflatoxin and hepa-titis B virus contribu te to the risk of liver cancer,each mu ltiplying the other’s effects.

In the United States, primary cancer of theliver is a rare d isease accoun ting for less than 1percent of cancer d eaths (2,796 death s in 1978).The amount of aflatoxin in the American diet issmall and is only one am ong several other possi-ble causes of liver cancer. In the Am erican con-text, the chief importance of the discovery of thedangers of aflatoxin is that it raises the possibili-ty that other as yet unrecognized mycotoxins infood may be carcinogenic. Recently, there hasbeen some evidence to suggest that a fungusmay contribute to th e high incidence of esoph-ageal cancer in p arts of China by increasing thenitrite content of contaminated food (209).

Environmental ContaminantsIn late 1979, OTA (284) examined both reg-

ulatory approaches and monitoring strategies

for coping with contaminated food. A widevariety of industrial products ranging fromheavy metals and pesticide residues, to sub-stances that leach out of packaging, such aspolyinyl chloride, can pollute food. Organiccompoun ds are felt to pose the greatest potentialfor environmental food contamination based onthe number, volume, and toxicity of organics

manu factured in the United States.

Polycyclic aromatic hydrocarbons (PAH) arefound wid ely in all types of foods, sometimes atthe same level which can be present in charcoal-broiled meats and smoked h am, i.e., up to 15 ugB(a)P/ kg net w eight. Some shellfish and finfishfrom polluted waters have been reported to con-tain up to 1,000 ug B(a) P/ kg. The importance of ingested B(a)P in cancer ind uction is uncertain .Epidemiologically, no relationship has been es-tablished, and when cancer is produced experi-mentally in laboratory anim als, it either affects

an organ not represented in man or occurs by amechanism that does not app ear relevant (93).

Some chlorinated hydrocarbons that wereused as pesticides (e.g., DDT, aldrin, dieldrin)prod uce liver tum ors (hepatomas) in mice, andsome have been shown to be carcinogenic inother species. These comp ound s accum ulate inhuman fat, but no increase in liver tumors ap-pears to have accompanied their introductionand use. However, the latent period is notknown, and the possibility exists that effectsmay be seen some years from now. Pesticideresidues in the form of second ary am ines may

constitute a hazard if the formation of nitros-amines in the stomach proves to be a cause of cancer in man .

Many other chemicals fall into this categoryof environmental contaminants. The extent towhich these have contributed to the productionof human cancers is difficult to evaluate, butbelieved to be small (328).

Additives

Chemicals are used to preserve food and giveit color, flavor, and consistency. Food dyes

were among the first chemicals investigated be-cause of their structural similarities to acceptedchemical carcinogens (24). Consumption of

8/14/2019 8104


8/14/2019 8104


taken to minimize human exposure to N-nitrosocompounds, including the restriction of theamounts of nitrate and nitrite added to meatproducts.

There is great un certainty regard ing the con-tribution of the comp oun ds d iscussed abov e tothe formation of cancer. The possibility also ex-

ists that other additives might have detrimentaleffects.

Diet Summary

Dietary components discussed above, andmany others, are currently the subjects of inten-sive research, from which some results shouldbe known within the next few years. The out-comes may show diet to be a factor in determin-ing cancer occurrence at many sites, particularlythe stomach, large bowel, endometrium, gall-bladder, and in tropical countries, the liver.Diet may also be show n to affect the incidence

OCCUPATIONAL EXPOSURES

The last several years have seen a h eated d is-cussion concerning the contribution of occu-pational exposures to cancer in the UnitedStates. Since the form ation of th e Occup ationalSafety and Health Administration (OSHA) in1970, 20 regulations have been prom ulgated andtwo more proposed relating to suspect carci-nogenic factors in the w orkp lace. Labor u nions

and pu blic interest group s have criticized OSHAfor moving slowly, while industry and theirtrad e associations claim u nn ecessary irrationalregulation and undue expense. In an effort toimplement a comprehensive and rational policyfor the regulation of carcinogens in the work-place, OSHA held 2 mon ths of hearings in 1978which attracted the participation of laborunions, environmental groups and spawned amajor new trade association, the American In-du strial Health Cou ncil (AIHC). Subsequ ently,OSHA promulgated “a general policy for theidentification and regulation of physical and

chemical substances that pose a potential oc-cupa tiona l carcinogenic r i sk to humans”(278,279).

of cancers of the breast and pancreas, and,throu gh th e antiprom oting effects of retinoids,to reduce incidence of epithelial cancers in manyother tissues. If these or other hypotheses arepr oven, practicable means of d ietary mod ifica-tions may eventu ally be identified to r edu ce can-cer rates.

Cairns (43) draws attention to the probabledifficulty of changing cancer incidence even if adirect link is shown between a diet constituentand cancer incidence:

Cancer of the lung is d ue to a p leasant andhighly ad dictive habit, cancer of the large in-testine and breast are most common in affluentcountries and so are p resumably associated withsome desirable habit, such as a diet high in ani-mal fats, that the rich nations can afford and theothers cann ot . . . . These are signs, therefore,that the campaign to prevent cancer may comeinto some conflict with people’s immediatedesires.

The first recognized industrial cancer wasidentified by Percival Pott, a British surgeon,wh o observed that scrotal cancer occurred m orefrequently in m en wh o had been employed aschimney sweeps as boys. This led to the iden-tification of soot as the first chemical and oc-cupational carcinogen. In the ensuing years,many other groups of workers have been found

to suffer from occupationally induced cancer.For the p urp ose of this assessment, occup ationalexposure is defined as exposure to a substanceor physical agent through any route d uring thecourse of emp loyment.

The workplace setting has proved to be thesingle most pr odu ctive sour ce of information inthe discovery of carcinogenic substances. This isbecause of the defined p opu lations involved andhigher exposures which can be more easilymonitored and identified. Table 17 lists carcino-gens and processes found in the workplace

which are associated with increased cancer risk.Occupations known to produce an elevated risk of cancer, though the specific agents responsible

8/14/2019 8104



Table 17.—Occupational Cancer Hazards

Agent

Acrylonitrile

4-aminobiphenyl

Arsenic and certain arsenic

compounds

Asbestos

Auramine and the manufacture

of auramine

Benzene

Benzidine

Beryllium and certain berylliumcompounds

Bis(chloromethyl) ether (BCME)

Cadmium and certain cadmiumcompounds

Carbon tetrachlorideChloromethyl methyl ether

(CM ME)Chromium and certain

chromium compoundsCoal tar pitch volatilesCoke oven emissionsDimethyl sulphateEpichlorohydrinEthylene oxideHematite and underground

hematite mining

Isopropyl oils and themanufacture of isopropyl oils

Mustard gas2-naphthylamine

Nickel (certain compounds)and nickel refining

Polychlorinated biphenyls(PCBs)

Radiation, ionizing

Radiation, ultravioletSoots, tars, mineral oilsThorium dioxide

Vinyl chloride

Agent(s) not identified

Cancer site or type —

Lung, colon

Bladder

Lung, skin, scrotum, lymphatic

system, hemangisarcoma of theliver

Lung, larynx, GI tract, pleural andperitoneal mesothelioma

Bladder

Leukemia

Bladder, pancreas

Lung

Lung

Lung, prostate

LiverLung

Lung, nasal sinuses

Lung, scrotumLung, kidney, prostateLungLung, leukemiaLeukemia, stomachLung

Paranasal sinuses

Respiratory tractBladder, pancreas

Nasal cavity, lung, larynx

Melanoma

Skin. pancreas, brain, stomach,

breast, salivary glands, thyroid,GI tract, bronchus, Iymphoid tis-

———. —. — A .

— .

Type of workers exposed — — — — —

Manufacturers of apparel, carpeting, blanket% draperies. syntheticfurs, and wigs

Chemical workers

Workers in the metallurgical industries, sheep-dip workers, pesticide

production workers, copper smelter workers, vineyard workers, in-

secticide makers and sprayers, tanners, miners (gold miners)

Asbes tos fac tory workers , tex t i l e workers , rubber- t i re manufac tur ing

industry workers, miners, insuIation workers, shipyard workers

Dyes tuf fs manufac turers , rubber workers, texti le dyers, paint

manufacturersRubber-tire manufacturing industry workers, painters, shoe manufac -

turing workers, rubber cement workers, glue and varnish workers,

distillers, shoemakers, plastics workers, chemical workers

Dyeworkers, chemical workersBeryll ium workers, electronics workers, missile Parts producers

Workers in plants producing anion-exchange resins (chemicalworkers)

Cadmium production workers, metallurgical workers, electroplating

industry workers, chemical workers, jewelry workers, nuclear

workers, pigment workers, battery workers

Plastic workers, dry cleanersChemical workers, workers in plants producing ion exchange resin

Chromate-producing industry workers, acetylene and aniline worker%

bleachers, glass, pottery, pigment, and linoleum workersSteel industry workers, aluminum potroom workers, foundry workers

Steel industry workers, coke plant workersChemical workers, drug makers, dyemakersChemical workersHospital workers, research lab workers, beekeepers fumigators

Miners

Isopropyl oil workers

production workersDyeworkers , rubber- t i re manufac tur ing indus t ry workers , chemica l

workers, manufacturers of coal gas,nickel refiners, copper smelters,

electrolysis workersNickel refiners

PCBs workers

Uranium miners , rad io log is ts , rad iographer% luminous d ia l Pa in ters

SOUFiCES; Institute of Med!clne (179),

sue, leukemia, multiple myelomaSkin Farmers, sailors, arc welders

Skin, lung, bladder, GI tract Construction workers, roofers, chimney sweeps, machinists

Liver, kidney, larynx, IeukemiaChemical workers, steelworkers, ceramic makers, incandescent, lamp

makers, nuclear reactor workers, gas mantle makers, metal refiners,

vacuum tube makers

Liver. brain, lung, hematolympho-Plastics factory workers, vinyl chloride polymerization plant workers

‘poietic system; breastPancreasStomachBrain, stomach

Hematolymphopoietic systemBladderEye, kidney, lungLeukemia, brainColon, brainEsophagus stomach, lung

ChemistsCoal miners

Petrochemical industryRubber industry workersprinting pressmenChemical workers

FarmersPattern and model makersOil refinery workers

— — ——-.—— ———‘International Agency for Res;a~cF~~ C-anc~r (185), and the Occupational SafetY and

————

Health Administration(279a)

8/14/2019 8104



have n ot yet been identified, are also included .Many of these exposures represent substantialhazards, and it is likely that other human car-cinogens exist which have not yet been detectedin the workplace. This may be because theadded cancer risk is small in comparison toother causes of cancer, because only a fewwor kers have been exposed, or simply because a

hazard has not been suspected and th erefore notlooked for. Besides the su bstances listed, man yother industrial chemicals have been found tocause cancer in laboratory animals. It must alsobe borne in mind that human cancer seldom de-velops until one or more d ecades after exposu reto a carcinogen, and thus it is too early to ascer-tain wh ether chemicals introd uced into indu stryduring the last few decades will affect cancerrates.

Epidemiologically demonstrating an in-creased risk of developing cancer from an oc-

cupational exposure generally involves resultsfrom examining large numbers of exposed peo-ple, the existence of relatively large risks, orsometimes a rare cancer, which facilitates draw-ing associations. Even then, years may p assbefore a hazard is iden tified. As the work placechanges, workers w ill be exposed to ad ditionalsubstances, some of wh ich m ay turn out to becarcinogenic.

Estimating the p ropor tion of today’s cancersthat can be attributed to occupational hazards isdifficult. Determining how many future cancersmay arise from past an d present occup ational

exposures is even more speculative. As with allstudies of carcinogenesis, the latent periodmakes it extremely problematic to associate aparticular occupational exposure with an in-creased cancer risk. Individuals frequentlychange jobs which can result in exposure tonu merou s substances, several of which may in-teract w ith each other either to p romote or to in-hibit tumor formation. With rapidly changingtechnologies, new exposures emerge whileothers cease. Often, the available information isbased on studies which have inadequate ex-posu re information. Exposure d ata are generally

poor because they are u sually not collected u ntila hazard is recognized, or at least suspected.OSHA requ ires monitoring only for some of thesubstances which it regulates. Oftentimes em-

ployees, and even in som e cases emp loyers, areunaware of specific chemical exposures.

Certain occup ational d iseases are difficult todiagnose and this can lead to underreporting, asexemplified by recent letters to the New England

Journal of Medicine concerning the inability of board -certified p athologists to iden tify cases of

asbestosis (1,318). Studies can be designed toobtain exposure data and continuously monitorworkers’ health, but they are expensive. Eventhough no system has yet been adopted togather these types of information, such an ap-proach m ay be feasible in som e indu stries.

Many estimates have been m ade of the p er-centage of human cancers associated with occu-pational exposures, most ranging from 1 to 10percent. In most instances the bases for theseestimates are not presented and the estimatesthemselves appear little more than informedguesses. Since epidemiologic studies do not

always follow individuals for their full lifetimes,calculated risks may be underestimates. Themagnitude of a detected risk is dependent uponthe time elapsed between exposure and the timeindividuals are studied.

Higginson and Muir (165) estimated the im-pact of environm ental factors in hu man cancer.They stated:

Although occupational cancers recognized sofar provide some of the most satisfactory d atafor identifying external agents, the absolutenumber of cancers due to occupational expo-

sures wou ld appear to be relatively small, prob-ably 1 to 3 percent of all cancers.

Wynd er and Gori (368) estimated the p ercentof 1976 cancer incidence in the United States at-tributable to occupation, From their summaryfigure, it appears that they estimated that thefraction of cancers attributable to occupa tionalfactors was 4 percent for men and 2 percent forwomen. Their explanation for these estimates isas follows:

The data presently available are, at best, edu-cated estimates of the relationship betweenspecific cancers and specific occupationalgroups. Cole et al.2 (1972) suggested that 20%

2P. Hoover, R. Cole, and G. H. Friedell, “Occupation and Can-cer of the Lower Urinary Tract, ” Cancer 29: 1250-60, 1972.

8/14/2019 8104


Ch, 3–Factors Associated With Cancer 87

of bladder cancers occurring in males in theBoston ar ea are r elated to occup ational ex-posure. In certain counties of New Jersey, the in-creased risk for this cancer appeared to be highamong workers in certain chemical industries.General estimates of the percentage of all humancancers related to occupational exposure rangebetween 1 and 10 percent.

Cole (62) estimated slightly higher figures forworkplace-caused cancer:

My estimates are less than 15 percent for menand less than 5 percent for women. These ratherlow figures emp hasize that we are not d ealingwith a single major p ublic health p roblem.Rather, we are d ealing w ith a series of modestproblems which are, however, of great impor-tance in the specific industries where they occur,

Cole does not describe the data used to derivethe estimate.

Higginson and Muir (166), in a later study

estimated that 6 percent of m ale cancers, skincancer included, in the Birmingham and WestMidland region of England in 1968 through1972, could be attributed to occupational ex-posu res. Their pap er d escribes some of the in-formation considered.

Fox and Adelstein (126) reviewed 1970-72British vital statistics and found that 12 percentof the differences in cancer mortality among dif-ferent occupational classifications could be ex-plained by variations associated with work andthe remaining 88 percent by lifestyle factors.They describe their estimate as an “over-

simp lified calculation , ignorin g interactions be-tween direct and indirect influences . . . clearlyvery crude. ”

Recently, a different approach was used togenerate an estimate of workplace cancers by 10distinguished r esearch w orkers of NCI, the Na-tional Institute of Environmental Health Sci-ences, and the National Institute for Occupa-tional Safety and Health (82). The report calcu-lated the proportion of cancers “for the nearterm and future, ” from quantitative estimates of risk and the nu mbers of workers exposed to sixknown carcinogens—asbestos, arsenic , ben-

zene, chrom ium, n ickel oxides, and petroleumfractions. This paper (82) concludes:

Reasonable projections of the future conse-quences of past exposure to established carcino-gens suggest that . . . occupationally relatedcancers may compr ise as much as 20% or moreof total cancer m ortality. Asbestos alone w illprobably contribute up to 13-18% and the data

(relating to five other carcinogens) su ggest atleast 10-2070 more.

The total of occupationally related cancers fromthose six exposures is then 23 to 38 percent of the overall cancer total, to which must pre-sumably be added the effects of other occupa-tional carcinogens not includ ed in the calcula-tion. This report h as been widely misqu oted asan assessment of the contribution of occupa-tional exposures to present cancer rates and isfrequently discussed and criticized because of the magn itud e of the estimates and their diver-gence from previous estimates.

While this report has not been su bmitted forpublication, it has been widely circulated andreviewed (c f., 9,93,332,345). The two reviewsthat have been published are discussed below.

Toxic Substances Strategy Committee(TSSC) Review

TSSC (345) identified several methodologicalaspects which might have caused the HEW (82)pap er to overestimate or to un derestimate thecontribution of occupational factors to the over-all cancer rate.

Overestimating

1.

2.

3.

The number of cases of lung cancer asso-ciated with asbestos may have been over-estimated by attributing all cases of lung can-cer among asbestos workers to asbestos ex-posure.Exposures to the examined carcinogens wereprobably higher in the past than they havebeen recently.Estimates of the number of workers exposedmay have been too high since: a) workerswh o were employed in more than one of thesix industries, a likely possibility in thechromium and nickel industries, might havebeen counted twice, and b) estimates of the

8/14/2019 8104


8/14/2019 8104



demonstrate mutagenicity or other toxic effects,animal bioassays, and epidemiologic data de-rived from grou ps exposed to h igh levels of pol-lutant chemicals. The latter are often obtainedin special situations, usually occupational,where people are exposed to much largeramou nts than the general public.

Air Pollution

Air pollution is primarily a problem of majorcities but it can move in and out of differentareas an d affect everyone. The Censu s Bureau(35) estimates that over 70 percent of the U.S.population lives in urban centers and that thisnumber is increasing with each passing year.The amount and type of air pollution in a givenregion changes not only with the time of day,but from day to d ay and year to year, makingmeaningful m easurement difficult. Air p ollutioncan be generally classified as a secondary en-

vironmental stressor, aggravating existing d is-ease conditions or increasing th e risk of diseasein those pr edisposed to ill health (326).

A variety of carcinogenic substances has beeniden tified in air—e. g., asbestos, arsenic, PAH—and EPA has promulgated several regulationsunder section 112 of the Clean Air Act to limitexposu re to these substances. Table 17 (see Oc-cupational Exposures) lists several recognizedoccupational carcinogens which may escapeinto the atmosph ere in the course of indu strialactivity and constitute a potential cancerhazard.

Morbidity and mortality due to chronic ob-structive lung d isease—asthma, bronchitis, andemphysema —are probably the most importantlong-term health effects of air pollution. Urbanareas generally have a higher death rate fromchronic obstructive lung d isease and lung cancerthan more rural environments (54,205). Thishas been used as the basis for suggesting that at-mosp heric pollution is an etiologic factor in thedevelopm ent of cancer. How ever, like many ap -parently simple observations, this one isplagued by conflicting and confound ing factors.

Individuals residing in u rban centers generallyhave smoking, drinking, and eating habits dif-ferent from their rural counterparts, and their

risk of infection and exposures to indu strial en-vironments also varies. In addition, urban areastend to have more accurate death certificationand disease diagnosis which m ight bias collectedstatistics. Epidemiologic studies that examinethe h ealth effects of air pollution mu st deal w iththese factors.

Many airborne carcinogens are believed to in-teract synergistically w ith tobacco an d occup a-tional exposu res. Since cigarette sm oking is thepredominant cause of lung cancer, it is a pri-mary suspect for explaining the observed ur-ban/ rural gradient. There have been nu merousattempts to disentangle the health effects of smoking from environm ental pollution. Unfor-tunately, many studies have compared broadcategories of smoking habits, such as “smokers”versus “nonsmokers” or “never smoked” versus“ever smoked” w hich redu ces the probability of discerning any effects. More specific catego-

ries —e.g., taking into account number of yearssmoked, age at which smoking began, numberof cigarettes smoked p er d ay—are necessary toevaluate any urban/ rural variation. In addition,other imp ortant d ifferences may exist in smok-ing habits between urban an d ru ral dw ellers, forinstance, in the tar and nicotine content, inhala-tion habits, and other factors contributing todifferences in effective dose of carcinogenreceived (see Tobacco). Most studies concen-trating on nonsmokers have failed to demon-strate, or found only a slight d ifference, in lun gcancer rates between nonsmokers in urban an d

rur al environs, suggesting th at airborne carcino-gens by themselves can be responsible for only asmall portion of th e d isease. Several have sug-gested that the effects of smoking a givenamount may be greater in urban than ruralareas (for r eview, see 91).

One of th e m ost extensive investigations intothe effects of air pollution is a cohort study con-ducted by ACS. Data concerning men withoutknown carcinogenic occupational exposure pro-du ced “little or no sup port to the hyp othesis thaturban air pollution has an important effect on

lung cancer death rates” (154). When lungcancer rates for men with occupational exposurewere comp ared to those for men without occu-

2 ,_+ ~ 1-- 1? 1 -

-

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


92 q Technololgies for Determining Cancer Risks From the Environment

phasis was on detecting a class of organics, thetrihalomethanes, formed during water treat-ment w ith chlorine and believed to be the mostubiquitous group of organics found in drinkingwater. Levels have been found to range frombelow 1 ug/ liter to above 300 ug/ liter. TheFederal Government has recently sought toreduce exposure to trihalomethanes by pro-

mulgating a regulation establishing a “max-imum contaminant level” of 100 ug/ liter totaltrihalomethanes for all community water sys-tems serving 10,000 persons or more th at ad d adisinfectant in the treatment of their water(107).

Much attention has been centered on th e rel-atively high concentrations of synthetic organ-ics found in untreated drinking water fromground water sources in certain areas of theUnited States. The extent to w hich these water -borne pollutants contribute to human carcino-

genesis has been evaluated by the examinationof both laboratory and epidemiologic data.Many of these studies were reviewed by theNAS Safe Drink ing Water Com mittee (266,270)as mandated by the Congress. (These reportsstand as references for man y issues concerningthe health effects of drinking water con-taminants. )

The numerous epidemiologic studies thathave examined th e association between d rinkingwater and cancer have been of varying r eliabili-ty and have taken many forms. These studiessuggest an association betw een constituents indrinking water and cancer mortality (169).How ever, the degree and extent of the associa-tion is disputed. NAS discussed and evaluated10 of the stud ies which focused on th e relation-ship of trihalomethanes and cancer mortality.All but one of the studies were found to demon-strate a statistical association between exposureand excess cancer rate, but excess cancer at nosingle anatomical site w as associated with ex-posu re. The epid emiologic analyses do n ot havethe extrapolation difficulties of animal ex-periments, but they are plagued by confoundingvariables such as exposures to other sources of carcinogenic stimuli, including smoking, whichwere gen erally not accounted for. Many of thestudies utilized the cancer deaths rates of a

region and related them to the water quality of that r egion. The d ifficulty with this app roach istwofold: 1) an individua l’s place of death maydiffer considerably from where that personresided d uring m ost of his water-drinking life,and 2) the data on water quality are extremelylimited, often reflecting the water quality of only one water sou rce in a region.

Prior to 1970, about 100 different organiccompoun ds h as been identified in water (169).In the past, organic pollutants were measuredby crude, nonspecific methods such as biochem-ical oxygen dem and and total organic content.Today, analytical techniques permit detectionof chemicals in minu te quan tities, and many arefound in concentrations of parts per trillion.NAS (270) found that over 700 volatile organiccompounds contaminate U.S. drinking water.Volatile substances m ake u p only 10 percent of the total organic content of water and approx-

imately 90 percent of this comp onent h as beenidentified and quantified. On the other hand,only 5 to 10 percent of the nonvolatile constit-uents of dr inking water h ave been characterized(266). As technology improves and detectionmethods become more sensitive, an ever-in-creasing num ber of comp oun ds may be d etectedin drinking water.

In a report for the Council on EnvironmentalQuality, Crump and Guess (74) reviewed five of the recent case-control studies on cancer risk associated with drinking water quality in thiscountry. While inadequacies were identified ineach of the stud ies (74):

. . . increased risks . . . are large enough to beof concern yet small enough to be very difficultto separate from confounding risks associatedwith other environmental factors.

Rectal cancer risk ratios for chlorinated v. un-chlorinated water were found to range from1.13 to 1.93. In three of the studies the risk ratios wer e statistically significant . Statisticallysignificant r isk ratios were also found in three of the stud ies for colon cancer and in two stu diesfor bladder cancer. Although the studies did notindicate a consistently increasing cancer risk with increasing exposure to organic contam-inants, one stud y did show su ch a relationship

8/14/2019 8104



for rectal cancer and another was suggestive forcolon cancer.

Crump and Guess made crude estimates of the possible range of carcinogenic risks fromconsuming grou nd water (from wells) contain-ing various levels of synthetic organic chem-

icals. They comp uted an u pp er risk limit of 7.5x 10

-4or about one case per 1,300 persons

from consuming water from a hypothetical wellcontaining th e highest d etected concentrationsof commonly found contaminants. Levels towh ich m ost Americans are exposed are typicallymuch lower. The estimate is based on summingthe individu al upper limits on human risk de-rived by a pr ocedure w hich tends to produ ce ex-aggerated estimates of human risk. On the otherhand, only risks from chemicals known to bepresent in th e wells and for which positive car-cinogenicity d ata exist, were considered in the

estimate (74). Crump and Guess also computedan u pp er limit on the lifetime total cancer risk associated with an average concentration of chloroform to be 1.8 x 10

-4or about one case

per 5,000 persons. The actual risk is probablymuch lower.

The NAS Safe Drinking Water Committeelists 19 carcinogens and 3 suspected carcinogensidentified in or thou ght to be in drinking w ater(266). NAS estimated the 95-percent upper con-fidence limit of cancer risks to humans fromlifetime exposure to w ater containing 1 ug/ 1 of

each of these compounds. These estimates aredisplayed in table 18 and are compared with risk estimates computed by EPA’s Carcinogen As-sessment Group . It is interesting to n ote that inman y cases a tenfold to hu nd redfold d ifferenceexists between the tw o estimates. These differ-ences are du e to the use of different extrapola-tion models and the use of point rather than in-terval estimates.

The problem of evaluating the effects of mul-tiple exposures is acutely relevant to water-borne pollutants because of the multitude of

compounds found in drinking water. Bioassayshave rarely attempted evaluation of synergisticor antagonistic effects. Therefore, these esti-mates may not accurately reflect the hu man car-cinogenic risk from these comp oun ds.

Table 18.—Concentration of Drinking WaterContaminants and Calculated Excess Cancer Risk

NAS aCAGb

1 0 -6

1 0 -6

Ug/lc Ug/lc

Acrylonitrile . . . . . . . . . . . . . . . .Arsenic . . . . . . . . . . . . . . . . . . . .

Benzene . . . . . . . . . . . . . . . . . . .Benzo (a) pyrene. . . . . . . . . . . . .Beryllium . . . . . . . . . . . . . . . . . .Bis (2-chloroethyl) ether . . . . . .Carbon tetrachloride. . . . . . . . .Chlordane. . . . . . . . . . . . . . . . . .Chloroform . . . . . . . . . . . . . . . . .DDT . . . . . . . . . . . . . . . . . . . . . . .1,2-Dichloroethane . . . . . . . . . .1,1-Dichloroethylene . . . . . . . . .Dieldrin . . . . . . . . . . . . . . . . . . . .Ethylenedibromide . . . . . . . . . .ETU . . . . . . . . . . . . . . . . . . . . . . .Heptachlor . . . . . . . . . . . . . . . . .Hexachlorobutadiene . . . . . . . .Hexachlorobenzene . . . . . . . . .N-nitrosodimethylamine. . . . . .

Kepone . . . . . . . . . . . . . . . . . . . .Lindane. . . . . . . . . . . . . . . . . . . .PCB . . . . . . . . . . . . . . . . . . . . . . .PCNB . . . . . . . . . . . . . . . . . . . . .TCDD . . . . . . . . . . . . . . . . . . . . . .Tetrachloroethylene . . . . . . . . .Trichloroethylene . . . . . . . . . . .Vinyl chloride . . . . . . . . . . . . . . .

0 .77—

— — —

0 .83

9 .09

0 .056

0 .59

0 .083

1 .4—

0 .004

0.11

0 .46

0 .024—

0 .034—

0 .0230 .108

0 .32

7.14—

0.71

9.09

2 .13

0 .034

0 .004

3. 0—

0.02—

0 .086

0 .012

0 .48—

1.46

.2 8—

0 .0022—

2 .4

1 .4—

0 .0052—

— — —

5 .0 x 10 - ’

0 .82

5 .8

106 .0

astandardized to 1O-S risks from Natfonal Academy of Sciences, Drinking Waterand Health (266) for consumption of 1 I/water/daybReca[culated to exclude aquatic food Intake from Cancer Assessment Group,

Arnbent Water (Jua//ty Cr/ter/a (104) Standardized to 1 I/water/day IntakecAverage adult water consumption IS 2 I/day.


As more chemicals are tested, additional sub-

stances found in drinking water will be shownto be carcinogenic. Two substances identified indrinking water, toxaphene and 1,2 dichloro-ethane, not considered carcinogenic by NAS in1977, have since been show n by NCI to be car-cinogenic in animals. Two additional com-pound s, bromodichloromethane and chlorodi-bromomethane, found in most drinking watersystems surveyed by EPA, have not been appro-priately tested for carcinogenic properties buthave been shown to be mu tagenic in the Am estest (169). In 1978, NCI (203) published a list of 23 recognized or suspected carcinogens, 30

mu tagens and 11 tumor p romoters which havebeen foun d in American drinking water.

Turning from the organ ics to inorganic, sev-eral stud ies have attempted to correlate levels of

8/14/2019 8104



carcinogenic trace metals in water supplies, withcancer rates. Associations of high levels of arse-nic in d rinking water w ith cancer of the skin andother sites were found in Taiwan and Argentina(203,370). Berg and Burbank (25) comparedconcentrations of eight carcinogenic metals inwater sup plies with State cancer mortality rates.

They found strong correlations with cadmiumconcentrations and cancer mortality; no signifi-cant correlations with iron, cobalt, and chro-mium ; a low bu t not biologically interpretablelevel of significance with nickel and arsenic; andsome correlations with lead and berylliumwhich correspond to known biological ac-tivities.

Asbestos, an established hu man carcinogen,is also of some concern since it has been foundto contam inate water sup ply systems. No rela-tionship between asbestos in drinking w ater and

cancer has been found but available data arevery limited.

Experimental and epidemiologic data providesome evidence that a carcinogenic hazard existsfrom drinking water contaminants, but does notperm it adequ ate quan tification. Most attention

CONSUMER PRODUCTS

Consum er prod ucts are a class of agents thatare so nu merou s that it is only possible to echo

the uncertainty with which pollutants were dis-cussed in the previous section. Detergents andother surfactants, hair dyes and other cos-metics, solid or foam plastics, paints, dyes,polishes, solvents, fabrics, and even the proc-essed paper and the printer’s ink in this reportare but a few. It is likely that some of theseproducts are already causing, unnoticed, a num-ber of today’s cancers, and it is quite possiblethat after prolonged exposu re substantial risksmight be detected in the future.

The d ifficulty in assessing cancers associatedwith exposure to consumer p rodu cts is that thedose is usually low, and the exposed pop ulationtends to be very large. An adequate assessmentof the carcinogenic effect of a consumer productmight require a study of a million people,

on quantifying risk estimates from waterbornecarcinogens has focused on chloroform, since italone has consistently generated dose-responsedata sufficient for extrapolating hum an r isk. Anad h oc stud y group of the EPA’s Hazard ous Ma-terial’s Advisory Committee (158) estimated therisk as follows:

The level of risk, estimated from considera-tion of the worst case [Miami, 311 pp b] and forthe expected cancer site for chloroform (theliver), might be extrapolated to account for upto 40% of the observed liver cancer incidence.

The results of a large NCI case-control studywhich is examining possible relationships be-tween w ater quality and bladd er cancer are ex-pected in 1981. Information was obtained inthat study on a wide range of demographic andexposure factors and a data bank on water qual-ity was created for more than 1,000 utilities

serving many of the people in the study. Thatinformation will hopefully permit linking an in-dividu al’s risk factors w ith d isease state. In ad-dition, a case-control interview stud y to investi-gate relationships between drinking water andcolorectal cancer is being conducted (74).

wh ereas a stud y of only 1,000 workers might beadequate in an occupational setting. Further,

wh ile occup ational exposure to a carcinogen isfrequently limited to a group of relativelyhealthy workers, consumer exposure includessubpop ulation groups usu ally considered to be“at high r isk, ” such as infants, small children,the elderly, and th e infirm,

For most consumer products, available lab-oratory and human evidence is insufficient todeterm ine wheth er they pose a cancer risk. Themagnitude of the potential health hazard be-comes apparent when one begins to examinepopulation exposure to different consumerproducts. In 1973, U.S. production of aerosolproducts was estimated at 2.9 billion units. Arecent Consumer Product Safety Commission

(CPSC) study indicated that manufacturers re-moved trichloroethylene (TCE), a widely used

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


Ch, 3—Factors Associated With Cancer q 97

that the benefit will prove to outweigh theharm. Many of the known hazard ous dru gs areused in the treatment of relatively uncommonserious conditions and the sum of the cancerscaused by them can amou nt to no more than afew score throughout the entire country eachyear. However, some drugs, including oral con-traceptives, and estrogens are or have been u sed

extensively. These and the medical use of ioniz-ing radiation, also known to be carcinogenic,are discussed below.

Oral Contraceptives

Oral contraceptives are taken by over 80 mil-lion healthy wom en the world over (366), spur-ring international concern about possible healtheffects, Not surprisingly, the possibility of in-creased cancer risk after long-term use and along latent period has been th e focus of m ajorepidemiologic efforts.

The evidence to date has clearly implicatedone type of sequential oral contraceptive (anestrogen and a progestin taken separately dur-ing the first and second halves of the monthlycycle), the use of which has been abandoned,with some cases of endom etrial cancer in youn gwom en. Conversely, there is some evidence thatusers of combination type oral contraceptives(an estrogen and a progestin taken together inthe same pill) may be at slightly lower risk forendometrial cancer (360).

In animal tests, both breast an d liver cancers

have been ind uced by comp onents of oral con-traceptives. A small number of benign livertumors in humans have been associated withcertain typ es of oral contraceptives, and thoughnot cancers in the usual sense, they can causefatal internal hem orrhage. No controlled stu dyhas yet evaluated the possible risk of malignantliver tumors (173).

Studies of breast cancer in oral contraceptiveusers have thu s far yielded no clear-cut evidencefor either increased or decreased risk, exceptperhaps some increase for women already athigh risk (see p. 99, Sexual Development,

Reproductive Patterns, and Sexual Practices fora full discussion of risk factors). Cervical cancer

also has been stud ied extensively with no con-sistent findings.

There is some evidence to suggest that oralcontraceptives may reduce the risk of ovariancancer. The red uction may be related to loweredlevels of ovulatory activity, since high levelshave been associated with an increased r isk of

ovarian cancer.

Because oral contraceptives have been part of the American lifestyle for a relatively shortperiod, large num bers of users have not reachedold age, the period of greatest cancer risk. Untilsuch time, perhaps w ithin the decade, the full ef-fects cannot be known. Even then, because thereare m any typ es of oral contraceptives, some of wh ich already have been shown to have differ-ent effects, and because formulations anddosage levels have changed significantly overtime, risks identified for that first cohort may

not ap ply to tod ay’s users. Overall, evaluatingthe evidence available at this time, oral con-traceptives do n ot app ear to be a major cause of cancer, but d o require continued epidem iologicattention.

Menopausal Estrogens

The use of “replacement estrogens” to relievemenopausal symptoms became widespread inthe 1960’s. A sharp rise in the incidence of en-dometrial cancer followed through the mid-1970’s in what has been termed “one of thelargest epidemics of iatrogenic disease , . . inthis country” (194). A cause-and-effect relation-ship was established through epidemiologicstud ies, precipitating a d ecline in the use of theseagents. Shortly after the levels of use drop ped ,incidence began to fall toward previous levels,and the “epidem ic” app ears to be largely over.Fortunately, these endometrial cancers have arelatively good prognosis, and while morbidityincreased sharp ly, mortality attributable to es-trogen therap y has not p aralleled the incidencetrend. Estrogen therapy is still prescribed forsome women, generally for shorter periods of time than previously. Used in this way, the risk of endometrial cancer appears much lower,though some cases may still occur, but the ben -

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


Ch. 3–Factors Associated With Cancer 101

Project on Breast Cancer of the InternationalUnion against Cancer, speculate:

A popu lation that achieved a 5-year redu ctionin age at first delivery might achieve a 30-per-cent reduction in incidence of breast cancer.

An an alysis of popu lation-based cancer mor-tality rates (27) supports the correlation be-tween nulliparity and higher rates. The datasuggest that, if the current pattern continues, thedecreasing fertility trends of the 1960’s and1970’s may foretell increased breast cancer ratesfor women in th is decade.

Age at menarche has also been shown to h avesome degree of predictive value in nearly allbreast cancer studies, with a lower age indi-cating a higher risk. Tulinius et al., (346) in theirstudy of Icelandic women, found a “slight in-fluence” rema ining after adjustment for par ityand age at first pregnancy. Henderson et al.

(162) found that w omen w ith menarche beforeage 13 had 1-1/ 2 times the breast cancer risk of awoman with later menarche.

Age at menarche, itself, maybe influenced byenvironmental factors, notably nutrition. Pop-ulation-based d ata reveal a steady decline in theage at menarche for American women, thoughtto be attributable to improved nutritional status(238). The effect of this on futu re breast cancerrates is uncertain. Early studies in rats cor-related body size, more than age, with onset of menarche (121). Similarly, observations in hu-man s, including a r ecent look at menarche anddisturbances in the menstrual cycle of balletdancers (134), provide evidence that lean bodymass is related to later m enarche. Later age atmenop ause brings increased risk. Women withnatural menopause after age 55 have abouttwice the risk of developing breast cancer as dowomen with natu ral menopause before age 45(219).

Breast cancer risk has a strong familial com-ponent th at has not been entirely explained bylifestyle similarities among relatives. First-de-gree (sisters, mothers and daughters) and sec-

ond-degree (nieces and aunts) relatives of women with breast cancer have two to threetimes the risk of the general population. Rel-atives of women with bilateral breast cancer are

at higher risk, and are more likely to be diag-nosed at an earlier age and to d evelop bilateraldisease than are relatives of women with uni-lateral d isease (15).

Gray, Henderson, and Pike (144) looked atthe higher rate of breast cancer in U.S. whitewomen compared with U.S. black women.They found that the relationship between black and white rates is not constant. Below age 40,black women have higher rates than whites,while after age 45, black rates were 20 to 30 per-cent lower than those for whites. This can beonly partially explained by differences in age atmenarche, but a full explanation has not beendemonstrated.

Cancers of the endometrium and ovary shareat least two important risk factors with breastcancer. They are all associated with high-fatdiets and women who have not borne childrenare at an increased risk for all three types com-pared to women w ho have borne children . Oc-currence of either breast or ovarian cancer in-creases the risk of a cancer developing at theother site. Breast cancer also increases the risk of future endometrial cancer (324). Ovarianhormonal activity may be the influencing factorin alI of these sites.

Cancer of the Cervix

There is a striking association between cancerof the uterine cervix and the nu mber of sexual

partners a w oman has h ad. The death rate forthis cancer among nu ns is much lower th an it isfor the general population (129), suggesting theinvolvement of a venereally transm itted agent.A possible candidate is a virus (Herpes simplextype II) which has been found in associationwith both cervical cancer and other cervical cellabnormalities (244). Some study results supportthis hypothesis, but th e data are considered onlysuggestive at this point, and no conclusion of causality can be draw n.

The number of deaths from cervical cancer

(ACS projects 7,200 in 1981) has decreased andcontinues to d ecline, partially du e to the m orewid espread use of cervical screening (6). Basedon the assu mp tion that th e majority of cervical

8/14/2019 8104


8/14/2019 8104



with the intensity of ultraviolet radiation. Thedisease is most prevalent in people receivinglarge amoun ts of sunlight, and it is a recognizedoccupational hazard for individuals workingoutdoors. The map s prep ared by Mason, et al.(223), which depict average age-adjusted cancermortality rates on a county-by-county basis,convincingly demonstrate higher risk of skincancer in the Southeast United States. TNCSfound the annual incidence rate for nonmela-nom a skin cancer to b e 539/ 100,000 in theDallas-Fort Worth, Tex., area as compared to174/ 100,000 in Iowa.

An individual’s risk of developing skin canceris strongly influenced by genetic makeup. Ahigher risk exists for those with phenotypiccharacteristics such as fair complexion, light eyeand hair color, and poor ability to tan. By con-trast, skin cancer is very rare in deeply pig-mented populations. In addition, several dif-

ferent types of chemicals can augment the carci-nogenic properties of ultraviolet radiation, in-cluding some used as medications and in cos-metics.

As discussed in the Air Pollution section, theuse of chlorofluorocarbons h as been restrictedin the United States because of the belief thatthey may react in the stratosphere to reduce thethickness of the ozone layer and cause an in-crease in u ltraviolet radiation reaching the su r-face of the Earth. NAS (70) estimated thatglobal release of chlorofluorocarbons at 1977rates could result in several hund red thou sand

additional cases of nonmelanoma skin cancerand several thousand melanomas in the nextcentury. (For additional information on thepossible carcinogenic impact from release of chlorofluorocarbons, see Air Pollution. )

Higginson and Muir (166) estimated that 10percent of m ale and female cancers in the Bir-mingh am and West Midland region of Englandcould be attributed to sunlight. Doll and Peto(93) attribute 90 percent of lip cancers (in con-

jun ction w ith smok ing), 50 percent or m ore of melanom as, as well as 80 percent of other skin

cancers to ultraviolet radiation. This accountsfor between 1 and 2 percent of all cancer deathsif these proportions are applied to 1978 U.S.cancer deaths.

Ionizing Radiation

About half of the average U.S. exposure toionizing rad iation comes from natu ral sources.Of the remainder, most comes from diagnosticmedical exposures, and smaller amounts comefrom med ical therapy, occup ational exposu res,and from radioactive pollutants. (Those ex-

posu res are discussed in Medical Drugs and Ra-diation, Occupation, an d Air Pollution, respec-tively. ) Cosmic rays and the minu te amou nts of radioactive isotopes that occur in our bodiesand in all natural materials are the majorsources of natur al background radiation.

The quantitative dose-response relationshipbetween cancer and low-level ionizing r adiationhas been the cause of much debate. Until 30years ago, it w as comm only assumed that ioniz-ing rad iation d id not cause cancer unless the ex-posures were high enough to cause clinically

detectable da mage to th e irradiated tissue. Thisassum ption is now kn own to be false, althoughthe dose-response relationship at low levels isstill not know n w ith certainty.

Organs and tissues differ in their susceptibili-ty to carcinogenic effects indu ced by rad iation.Leukemia w as the first form of cancer associatedwith exposure to ionizing radiation, but we nowknow that cancer may be indu ced by rad iationin many tissues of the human body and that therisk of inducing solid tumors exceeds that of leu-kem ia (268). The major cancer sites affected arethe breast in w omen and the thyroid, lung, and

digestive organs in both sexes. Solid tumorshave a longer latency p eriod (10 to m ore than 30years) than the few years before the excess risk for leukemia manifests itself. The total cancerrisk from radiation is greater for women thanmen , principally because of the contribution of breast cancer.

More information is available on the dose-response relationship between radiation andcancer than any other, but there is still muchcontroversy over the app ropr iate extrapolationmodel to use for estimating the cancer risk from

ionizing rad iation. In July 1980, NAS officiallyreleased BEIR III (268). The estimates for therisks from ionizing radiation used by BEIR 111were derived principally from human experi-

8/14/2019 8104


104 qTechnologies for Determining Cancer Risks From the Enviromnent

ences with much higher doses than most peoplereceive. The most important of these are datafrom populations exposed to atomic blasts inHiroshima and Nagasaki, patients exposed totherapeutic radiation, and various occupationalgroups such as uranium miners and radiumwatch d ial painters. Disagreement p ersists in thefield of rad iation carcinogenesis over th e ap pro-priateness of these high-exposure populationsfor estimating low-level risks.

The development and release of the BEIR 111report illustrates the diverging and changingopinions concerning the appropriate methodol-ogy for extrapolating from the m easured effectsof high doses of ionizing rad iation to th e mostprobable effects of low doses. A draft of theBEIR 111 report was released for public commentin May 1979 but was retracted when it waslearned that a majority of the committee sup-ported a mod el different from the one pr esented.

Some m embers of the comm ittee felt that forlow d oses, a linear m odel overestimated th e risk and a pu re quad ratic mod el, wh ich estimates alower risk, could be used as a lower bound.How ever, others felt that the linear mod el wasmore accurate. Because of this disagreement, aconsensus pan el was assembled and it adoptedthe linear-quadratic model: the pure quadraticmod el produ ces the lowest estimate, the linearthe highest, and the linear-quadratic estimatesan intermediate risk. Depending on which mod-el is used, estimates of mortality from all formsof cancer will differ by abou t one ord er of mag-nitude.

The committee also could not agree on w heth-er the cancer risk from radiation would have anadd itive or mu ltiplicative effect on the generalcancer ra te. Throu ghout the rep ort, excess risk is given as both an absolute (additive) and as arelative risk. The relative approach assumesthat the excess risk increases gradually and con-tinuously, proportional to the spontaneous risk,wh ich increases with age for near ly all cancers.The absolute app roach assumes a constant nu m-ber of add itional cancers throu ghou t life.

The BEIR III report (268) makes various at-temp ts at estimating cancer risk from different

types of exposure but does not quantify the risk from low-dose radiation:

. . . the degree of risk is so low that it cannot beobserved directly and there is great uncertaintyas to the dose-response function most appro-priate for extrapolating in the low-dose region.

It further states:

It is by no m eans clear wh ether dose rates of gamma or x radiation of about 100 mrads/ yr arein any way detrimental to exposed people . . .

For the purpose of this assessment, the risksderived for continuous lifetime exposure to 1rad/ yr of low-dose (gamma or X-ray) radiationis the most relevant, though it is an order of magnitude higher than average natural back-groun d rad iation. Table 21 displays both abso-lute and relative risk estimates generated byBEIR III.

The BEIR III report estimated that the averagewhole body dose of natural background radia-tion in the United States is approximately 100mrem / yr. The dose of rad iation received varieswith altitud e and geograph ical location. Using afigure of 220 million Americans, one can esti-mate that the popu lation as a w hole would beexposed to 22 million person-rem of back-ground radiation.

3Although BEIR III did not at-

temp t to estimate th e risk of cancer indu ced bylow-dose rad iation, it would not be incorrect tocompu te an up per limit of risk from the linearmodel since this is the only model whichassumes that dose and effect are proportional.Using the range of risk estimates, 158-403 ex-cess cancer death s per m illion persons p er rad ,yields estimates of 3,476 and 8,866 cancerdeaths per year attributable to backgroundrad iation. These figures r epresent betw een 0.9and 2.2 percent of all cancer deaths, and arecomparable with 3-percent estimates by Doll

31n 1972, EPA estimated that the average dose of natural back-ground radiation for a person living in the United States is 130

mrem. Based on this estimate, the Interagency Task Force on theHealth Effects of Ionizing Radiation estimated that th e U.S. pop-ulation is exposed to 20 million person-rem, but this is in error be-

cause dividing 20 million person-rem by 130 mrem produces aquotient of 154 million people. The population is nearer 22o

million.

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


Ch. 3–Factors Associated With Ca/leer q 107

The examples cited are unlikely to be the onlyways in which infection affects the risk of de-veloping cancer; but even if they were, the rangeof estimates for its contribution to the overallcancer rate would be large. Rapp (304) specu-lated that if herpesviruses are eventually shownto be involved in all cancers of the male and

female genitourinary systems, then those withactive venereal disease would have a 1 per70-100 risk of developing the d isease:

This estimate is based on the nu mber of newgenitourinary cancer cases per year in the UnitedStates (100,000 for females and 70,000 for males)in a p opu lation containing abou t 14 million ac-tive cases . . . of venereal disease due to herpes-viruses.

If one applies this reasoning to 1978 U.S. cancerdeaths, approximately 15 percent of all cancer

mortality would be associated with infection.This estimate is extremely tentative—there ismuch more certainty around the associationwith cervical cancer than other genital sites.Doll and Peto (93) suggested a figu re abou t 10percent as a very u ncertain best estimate, w ithina very wide range of acceptable estimates, of the

proportion of cancer deaths attributable to in-fection, distributed as follows:

. . . 5% perhaps attributable to the action of viruses and a token figure of 5% to allow for thepossible role of other infective agents in deter-mining the conditions under which cancer is pro-duced in vivo. The likely role of infectiousagents in the etiology of cancer of the uterinecervix provides a lower limit of at least 1%, butwe can at present m ake no useful guess at the up-per limit.

OTHER OR UNKNOWN ASSOCIATIONS

Two basic classes of “other” associations willbe considered :

1. Associations w ith cancers wh ich as yet re-main unidentified, but which, when eluci-dated , will most likely fall into categor iesthat have been d iscussed in th is chapter.

2. Associations of agents or exposures withcancers of the future. These may havebeen introdu ced recently, but have n ot yetproduced cancers, or may be introduced

in the future.

Causes of Cancers Occurring Today

The agents associated w ith and/ or responsi-ble for many of today’s cancers have not beenidentified. For example, although cancers of thebreast, colon, rectum, and stomach are associ-ated in some w ay with dietary factors, a causalrelationship has not been found for cancers atthose sites. Add itionally the incidence of somecancers, myelomatosis and n on-Hod gkins lym-phoma, for example, appear to be increasing

but no evidence associates those cancers witheven a broad category of factors.

The search for environmental factors asso-ciated with cancer will undoubtedly continue.Observed differences in rates between popula-tions, within countries, and through inter-national comparisons will result in advancingnew hyp otheses and p romoting further studies.

Some suggested associations that were notdiscussed above are likely to be further in-vestigated, but currently few data exist aboutthem . Biological factors, like imm un ologic con-

trol, may normally limit the onset of disease,and a breakdown of this mechanism which maybe brought on by an environm ental agent, couldincrease the p ropensity for cancer d evelopm ent.Similarly, psychological factors such as stressmay create an internal m ilieu suitable for tu morgrow th. There is some animal evidence to sup-port th is hypoth esis but it is limited. Stud ies of patients in mental hospitals are not supportiveof an increased risk (93). Psychological stressdoes have a recognized role in causing people tosmoke, drink, overeat, and take part in otherharmful activities which may directly or indi-

rectly increase their risk of cancer.

8/14/2019 8104


8/14/2019 8104


Ch. 3–Factors Associated With Cancer q109

Table 22.–Estimates of the Percentage of Total Cancer Associated with Various Factors (Continued)

Time periodto which

Factor Estimate estimate applies Author

Asbestos13-18°/0 U.S. cancers

30/0 (1 .4-4.4°/0) U.S. cancers

Air pollution20/0 U.S. mortality

Consumer products< 1% U.S. mortality, males and females combined

Infection5%?-10%?? U.S. cancers

150/0 U.S. mortality

Lifestyle30% male cancers, Englanda

630/o female cancers, England a

Radiation, natural and medical:

Ultraviolet10°/0 male cancers, England

a

10°/0 female cancers, Englanda

Unspecified radiation1% male cancers, Englanda

1% female cancers, Englanda

Ultraviolet and X-ray8°/0 U.S. male incidence8°/0 U.S. female incidence

Natural ionizingO-(0.9-2.2°/0) U.S. mortality30/0 U.S. mortality30/0 U.S. mortality

Medical drugs and radiation10/0 U.S. mortality, males and females combined10/0 male and female cancers combined, England

a

4°/0 U.S. incidence, males and females combined (refers onlyto exogenous hormones)

Sexual development, reproductive patterns and sexual practices7°/0 U.S. mortality, males and females combined

Other or unknown?%

15°/0 male cancers, Englanda

11% female cancers. Englanda

Near term andfuture

Now or likely innear future

Future

1977

Present andfuture1978

1968-721968-72

1968-721968-72

1968-721968-72

19761976

197819781978

19771968-72

1976

1977

Present andfuture

1968-721968-72

HEW (82)

Hogan and Heel (171)

Doll and Peto (93)

Doll and Peto (93)

Doll and Peto (93)Rapp (304)

Higginson and Muir (166)Higginson and Muir (166)



Wynder and Gori (368)Wynder and Gori (368)

OTA based on BEIR Ill (268)Doll and Peto (93)JabIon and Bailar (191)

Doll and Peto (93)Higginson and Muir (166)

Wynder and Gori (368)

Doll and Peto (93)

Doll and Peto (93)Higginson and Muir (166)Higginson and Muir (166), - - -

aBlrmlngham and West Midland Region.bEstimated from graphic presentation.

SOURCE: Office of Technology Assessment.

8/14/2019 8104


8/14/2019 8104


Contents

Page

Methods . . . . . . . . .......................113

Analysis of Molecular Structure and OtherPhysical Constants, . . . . . . . . . . . . . . . . . . .114

Short-Term Tests . . . . . . . . . . . . . . . . . . . . . . . .115The Ames Test. . . . . . . . . . . . . . . . . . . . . . . . .116Other Short-TermTests. . ................120

Use of Short-TermTestResults and PolicyStatements About the Tests . ......,.....121

Bioassays . . . . . . . . . . . . . . . . . . . . . . . . .. ....122Standard Protocols for Bioassays. . .........123Objections to the Usefulness of Bioassays ....124Expert Review of Bioassay Results . .........127Policy Consid eration s Abou t Bioassays. ... ..l28How Many Chemicals Are Carcinogens in

Animal Tests? . . . . . . . . . . . . . . . . . . . . . . .129

Selection of Chemicals forTesting in Bioassaysand Short -Term Tests . . . . . . . . . . . . . . . ..131

The Inter agency Testing Com mittee . .......131National Toxicology Program . ............133

The National Center for ToxicologicalR e s e a r c h . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 3 4

Chemical Industry Institute of Toxicology. ...l35

Tier Testing. . . . . . . . . . . . . . . . . . . ... .;. ....135

Epidemio logy . . . . . . . . . . . . . . . . . . . . . . . . . 136Descriptive Epidemiology . ...............137Observational Epidemiology. . ............137Causal Associations. . . . . . . . . . . . . . . . . . . . .138Policy Considerations About Epidemiology. ..l38Carcinogens for Which There is Human

E v i d e n c e . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 4 6Annual Report on Carcinogens . ...........140

Sources of Epidemiologically Useful Data ... ...l42Health Status Information. . ..............142Exposure Data. . . . . . . . . . . . . . . . . ... ... ..145Chemical Information Systems . ...........146Collection and Coord ination of Exposure

and Health Data. . . . . . . . . . . . . . . . . . . . . .149Government Records and Record Linkage. ...l5l

S u m m a r y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 5 2

LIST OF TABLES

Table No. Page

23. General Classification of Tests Available

24.

25.

26.

27.28.

29.

30.

31.

To Determine Properties Related toCarcinogenicity. . . . . . . . . . . . . . . . . . . . . . .114Results Obtained in Two Validation Stud ies

of the Ames Test WhenKnownCarcinogens and Noncarcinogens WereTested . . . . . . . . . . ....................118Expected Results of Examining TwoCollections of Chemicals for MutagenicityUsing a Short-Term Test That is 90-PercentSensitive and 90-Percent Specific: OneCollection of Chemicals Contains 1 PercentCarcinogens; the Oth er Contains 10Percent .119Distribution and Number of Animals in aTypical Bioassay Study of Carcinogen s. ....124Guidelines for Bioassay in SmalI Rodents ...125Some Organizations That Have Consideredand Endorsed Animal Tests as Predictors of

Human Risk From Chemical Carcinogens ...l29Advantages and Disadvantages of Case-Control and Cohort Studies .. ... ... :. ...139Chemicals and Industrial ProcessesEvaluated for Hum an Carcinogenicity bythe Intern ational Agency for Researchon Cancer. . . . . . . . . . . . . . . . . . . . . . . . . . .141EPA’s Recommend ed Base Set of Data ToBe Included in Premanufactur ing Notices ...148

LIST OF FIGURES

Figure No. Page19. Molecular Structures of Two Closely

Related Chemicals: One a Carcinogen andOne a Noncarcinogen . .................115

20. 1nterrelationships of Major TestingActivities of NTP . . . . . . . . . . . . . . . . . . . . .136

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


Ch. 4—Metkods for Detecting and ldentifying Carcinogens q115

chemical relative, 4-acetylaminofluorene (4-AAF), is not a carcinogen (see figu re 19).

Figure 19.— Molecular Structures of Two CloselyRelated Chemicals: One a Carcinogen and

One a Noncarcinogen

2-acetylamlnofluorene, a known carcinogen

4-acetylaminofluorene, a noncarcinogen

SOURCE. Off Ice of Technology Assessment

EPA recently based a regu latory decision onmolecular structural analysis. Under section 5(e)of the Toxic Substances Control Act (TSCA),EPA prohibited the entry of six new chemicalsubstances into the marketplace unless themanu facturer provided add itional informationabout toxicity. A National Cancer Institute

(NCI) bioassay had shown a related chemical tobe carcinogenic, and based on that resu lt, EPAdecided that more information w as needed be-fore manufacture could begin (88). The manu-facturer decided not to proceed with the testingand did not mar ket the chemicals.

Short-term tests are so nam ed because of therelatively short time needed to condu ct the ex-perimen ts. Some stu dies involving m icro-orga-nisms r equire less than 1 day to comp lete (87),most require a few days to a few w eeks, and thelongest, using mice, requires 8 to 9 months(172), These times may be compared to themore than 3 years required to complete a bio-assay and the months to years required to com-plete epidemiologic studies.

A number of reasons account for the growinginterest in using short-term tests to predict achemical’s carcinogenic p otential:

q shorter time period required for the tests;

q low cost ($100 to a few thousand dollars foreach test compared to $400,000 to $1 mil-lion for a bioassay);

q evidence that the m ajority of chemical car-cinogens are mutagens and that many mu-tagens are carcinogens;

q growing opinion that short-term tests

can predict which chemicals may be car-cinogens.

The third point is important because manyshort-term tests determine whether or not achemical causes mutations (mutagenicity) ratherthan if it causes cancer (carcinogenicity). Thepostulated relationship between mutagenicityand carcinogenicity stems from b iological p rop-erties common to all living organisms. Th egenetic information in both germ cells (egg andsperm) and somatic cells (nongerm or “bodycells”) is composed of deoxyribonucleic acid

(DNA), and agents that cause m utations in germcells are also expected to cause mutations insomatic cells. A germ cell mutation may prevent

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


Ch. 4—Methods for Detecting and ldentifying Carcinogens q121

The dimensions of NTP, and the significantdemand s it places on the funds and personnel of the participating agencies, should diminish by1985, as the fiscal projections suggest . . . . It isour h ope that, by then, better test systems willbegin to replace the tedious and costly animalassays now required.

Not everyone is so optimistic as to think a re-

placement for animal assays will be available in4 or 5 years. Transformation tests are p robablythe best bet for the replacement, but they re-quire more development and validation.

Transformation assays are technically moredifficult than the Am es test, but not so d ifficultas to preclude their use on a routine basis.Validation studies are being carried out on anumber of in vitro transformation systems todetermine h ow accurately they identify carcino-gens and noncarcinogens (271,272).

NTP is cond ucting add itional validation stu d-

ies of a test that u ses wh ole animals. This test,which has been in limited use for about 30years, requires about 6 months to complete. Ex-posure of a particular strain of mice to knowncarcinogens causes an increase in the frequencyof lung tumors (adenomas) and causes earlierapp earance of the tum ors. The test takes muchless time than the standard assay, and NTP(272) has found this test accurately predictsresults in bioassays. NTP tested 60 chemicals inthis system in 1980 and p lans to test another 30in 1981.

The British publication The Economist (97)singled out a transformation assay, using Chi-nese hamster ovary cells, as having promise forcarcinogen identification. Discussion of short-term tests in that publication, which seldompublishes articles about biology, reflects the in-creasing importan ce of the tests. A m ore auth or-itative source, the NCI National Cancer Ad-visory Board ’s Subcommittee on EnvironmentalCarcinogenesis (245) said: “ . . . this subcom-mittee is enthu siastic about the p ossible futureuse of in vitro [short-term] tests as part of ascreening system for potential carcinogens andbelieve that their further development and vali-

dation d eserve high priority.”

Use of Short-Term Test Results andPolicy Statements About the Tests

How best to utilize short-term tests in car-cinogen identification is hotly debated. The ma-

jority view is that th e tests are most u seful as ascreen to d etermine a chemical’s p otential car-cinogenicity. As a new chemical is developed or

as an old one comes under suspicion, an inex-pensive short-term test or battery of tests canprovide information about w hether it is or is notlikely to be a carcinogenic hazard . If the resultsof the test are negative, the chemical is consid-ered less likely to be a hazard than a chemicalthat is positive. In the case of a chemical beingcommercially d eveloped , a positive result m ightsuggest that the chemical not be produced orthat the cost of testing it in a bioassay should beconsidered in d eciding wh ether or not to p ro-duce it. A positive short-term test result on acommercially produced chemical most likely

causes more of a problem. The ma nu facturer isfaced with having to begin other tests and towarn his employees and customers of potentialhazard.

Opinions d iffer abou t the w eight to be placedon short-term test results. Peter Hutt, formerGeneral Counsel at the Food an d Dru g Ad min-istration (FDA), and now in pr ivate law p racticesays that h e adv ises his clients not to continu ethe developm ent of a prod uct wh ich is positivein a sh ort-term test. He m aintains that, “life istoo short” to invest time and effort in a chemical

that is more likely than not to be considered asuspect carcinogen. Near the ether end of thespectrum of opinion, Leon Golberg, in review-ing poor correlations between results of Amestesting and bioassays of components of hairdyes, concludes “ . . . it is very hard to acceptthe fact that the Am es test is a pred ictor of car-cinogenic potential” (143).

The OSHA docum ent “Id entification, Classi-fication and Regulation of Potential Occupa-tional Carcinogens” (279), accepts the results of short-term tests as supportive evidence for de-ciding whether a chemical will be classified as a

carcinogen or non carcinogen. TSCA test stan d-

8/14/2019 8104



ards (106) and Federal Insecticide, Fungicide,and Rodenticide Act (FIFRA) guidelines (102)accept short-term tests as measures for muta-genicity but do not consider them in makingdecisions about carcinogenicity. However, theymention that test developments are prom ising.

An important step in making decisions about

the use of short-term tests are the ongoingvalidation studies which compare predictionsmade from short-term tests to knowledge aboutcarcinogenicity from bioassays or epidemiolo-gy. These studies, although limited by the qual-ity and quan tity of data abou t carcinogenicity,are producing valuable information. In addi-tion, studies of molecular m echanisms of m uta-genicity and carcinogenicity are important indeciding about the applicability of short-termtests.

BIOASSAYS

Chemicals cannot be tested for carcinogenici-ty in humans because of ethical considerations.A substantial body of experimentally derivedknowledge and the preponderance of expertopinion support the conclusion that testing of chemicals in laboratory animals provides re-liable information about carcinogenicity.Animal tests employ whole mammal systems,and although they differ one from another, allmammals, including humans, share many bio-logical features (266):

Effects in animals, properly qualified, are ap-plicable to man. This premise underlies all of experimental biology and medicine, but becauseit is continually questioned w ith regard tohuman cancer, it is desirable to point out thatcancer in men and animals is strikingly similar.Virtually every form of human cancer has an ex-perimental counterpart, and every form of mul-ticellular organ ism is subject to cancer, includinginsects, fish, and plants. Although there are dif-ferences in susceptibility between different ani-mal species, and between individuals of the samestrain, carcinogenic chemicals will affect most

test species, and there are large bod ies of exper-imental data that indicate that exposures that are

The problem of the carcinogens that are notdetected (false negatives; lack of sensitivity) andthe noncarcinogens that are falsely detected(false positives; lack of specificity) by any onetest might be solved with additional short-termtests. The great attractiveness of a battery of short-term tests is that it m ight correctly iden-tify all carcinogens and noncarcinogens. Un-

fortunately, no such battery has yet been de-fined. The composition of the battery will de-pend on validation studies and acceptance of each comp onent test.

The growing use of short-term tests showsthat short-term tests have moved to an impor-tant position in toxicology. The speed withwhich they have been incorporated into Gov-ernment and private sector pr ograms reflects theimportance of the need to which they are ad-dressed.

carcinogenic to animals are likely to be carcino-genic to man, and vice versa.

In comp arison to short-term tests and ep ide-miology, bioassays have h ad a longer d evelop-ment p eriod and enjoy greater acceptance thanthe short-term tests; they are m ore easily m anip-ulated to p rodu ce evidence linking a p articularsubstance to cancer than epidemiology, andthey can predict human risks rather than r elyingon cases of hum an cancer to dem onstrate risk.

On the other hand, they take longer and costmu ch more than short-term tests.

The bioassay’s apparent simplicity belies thedifficulty of execut ing su ch exper iments. Brief-ly, the suspect chemical is administered to apopulation of laboratory animals. As animalsdie or are killed d uring the course of the study,they are examined for the presence of tumors.At the end of the treatment and observationperiod (generally abou t 2 years), the survivinganimals are killed and examined. A controlgroup of animals is treated exactly the same ex-

cept that they are not exposed to the suspectsubstance. The type and number of tumors and

8/14/2019 8104


Ch. 4—Methods for Detecting and identifying Carcinogens q123

other relevant pathologies present in the ex-posed animals are compared with those in thecontrol group , and statistically analyzed.

A statistical expression comm only u sed to de-scribe a positive result is “. . . it has a p valueless than 0.05" (5 percent). The p value is theprobability that the observed effect might be ex-plained by chance; in this case, the expressionmeans th at the p robability of the observed car-cinogenic effect being due to chance is less than

5 percent. A p value of 0.05 or less is comm onlyrequired to decide that a test result was statis-tically positive.

Finally a conclusion is draw n abou t wh etheror not the evid ence ind icates that the su bstancecaused cancer in the exposed animals. An excel-lent discussion of experimental design andanalysis is available from the InternationalAgency for Research on Cancer (IARC) (187).

The first successful experimental induction of cancer in animals (in 1915) showed that paintingrabbits’ ears with coal tar produced tumorswhich morphologically resembled human tu-mors associated with exposur e to the same agent(cited in 342). Most chemicals which are pres-ently known to cause cancer in humans are alsocarcinogens in animals.

Verification of the predictive power of bio-assays would require that the agent be shown tobe a human carcinogen. Currently, IARC main-

tains that convincing evidence for human car-cinogenicity is available for only 18 exposu res,including 14 chemicals. At the same time, it lists142 substances for which data are “sufficient” toconclude that they are carcinogenic in animals.It is difficult to d emonstrate hum an carcinoge-nicity. Once a substance is shown to be an ani-mal carcinogen, regulatory restrictions on andvoluntary r edu ctions in the u se of the chemicalm ay reduce human exposures, making thosedemonstrations more difficult. Reductions in ex-posure in such cases are far m ore importan t tohum an health than the foregone opportunities

to verify the p redictive p owers of bioassays.

Standard Protocols for Bioassays

An important event in bioassay design wasthe development of NCI’s Guidelines for Car-cinogenic Bioassay in Small Rodents (33 I). Theguidelines describe minimum requirements forthe design and conduct of a scientifically valid

bioassay and discuss important considerationsin un dertaking such stud ies. They are w ritten toprov ide flexibility in experimental d esign wh ilesetting certain m inimal requirements:

1.

2.

3.

4.

5.

6.

7.

Each chemical should be tested in at leasttwo species and both sexes. Rats and miceare u sually the species of choice.Each bioassay should contain at least 50animals in each experimental group.When both sexes of two species are usedand two treatment levels are administeredand a third gr oup is used as controls, a to-tal of 600 anim als is need ed (see table 26).

Exposure to the chemical should startwhen the animals are 6 weeks old oryounger and continue for the greater partof their lifespan. Mice and rats are usuallyexposed for 24 month s.One treatment group should receive themaximum tolerated dose (MTD), which isdefined as the highest dose that can begiven that would not alter the animals’normal lifespan from effects other thancancer. The other treatment group istreated w ith a fraction of MTD.The route by w hich a chemical is adminis-

tered should be the same or as close aspossible to the one by which human ex-posure occurs.Animals are closely monitored through-out the study for signs of toxic effects andother causes affecting their health.Examination of animals is conducted byor under the direction of a pathologistqualified in laboratory animal pathology.

The guidelines also specify that special proce-dures (e.g., organ function tests, body burdendeterm inations, absorption and excretion tests)may be needed for evaluating certain chemicals.

8/14/2019 8104



Table 26.—Distribution and Number of Animals ina Typical Bioassay Study of Carcinogens

Species A Species BExperimental groups Males Females Males Females

Dosage MTDa group 50 50 50 50Dosage MTD/x group 50 50 50 50Control group 50 50 50 50

— —aMaxlmum toterated dose.

SOURCE: National Cancer Institute

A National Research Council report (262)suggested a tw o-generation bioassay when thereis special concern w ith pr enatal and perinatal ef-fects of a substance. In the suggested two-gen-eration experiment, the first generation is ex-posed from 6 weeks of age through their adultlife, including periods of reproductive activityand pregnancy, and the offspring are exposedfor their lifespans. The most important two-gen-eration experiments, from the standpoint of

public policy, were the three showing saccharinto be a blad der carcinogen in second gen erationrats (282). The advantage of the two-generationexperiment is that it exposes fetuses and veryyoung animals, which may represent the mostsensitive stages of life. This procedure costsmore and requires more time to comp lete thanthe one generation experiment. Probably be-cause of those disadvantages, it has not oftenbeen used.

Table 27 compares some specifics of NCIguidelines with those proposed by EPA fortesting u nd er FIFRA (102) and TSCA (106). NCI

guidelines state that at least two-dose levels betested; EPA requires three test doses. The high-est dose, high-dose level (HDL), for TSCA isdefined as being “slightly toxic” and is to bedetermined in a 90-day-feeding stud y to precedethe oncogenicity experiment. The other twodoses are to be fractions of HDL. EPA prefersthe term HDL, which is less specifically definedthan MTD, because of the controversies whichhave erupted over defining MTD. Three-doselevels are pr oposed by EPA because a r eview of many NCI bioassays revealed that tumor in-cidence was sometimes higher at a dose of

MTD/ 2 than at MTD because the higher dosekilled some animals before tumors might h avedeveloped. The third-dose level will also pro-vide additional information about the dose-

response curve (102). The text which accom-panies the p roposed EPA guidelines for carcino-gen testing under FIFRA (102) contains addi-tional information about alternative approachesconsidered and subsequently discarded for theguidelines.

Three to five years are requ ired to comp lete abioassay. A subchronic test, to set dose levels,requiring 2 to 6 month s, is followed by an aver-age 24 months of exposure to the agent, and

sometimes an additional 3 to 6 months observa-tion period. Pathological examination of tissuesfrom the animals and evaluation of the patho-logical and other data may take an additionalyear.

The cost of a bioassay has been variously esti-mated: NCI estimates about $400,000, TSCAguidelines (106) estimated $400,000, and EPA(112) later estimated a range from $390,000 to$980,000 .

Some changes have been made in protocolssince the NCI guidelines were published, but in

contrast to the situation in spring 1977, whenOTA reviewed carcinogen testing (282), appar-ently mu ch less confusion an d contention nowexist about what constitutes an adequate bio-assay. NCI guidelines are for minimal stand-ards; certainly larger populations of animals canbe tested. For very imp ortant bioassays, largernumbers of animals or more dosage groupsmight be specified.

Objections to the Usefulnessof Bioassays

Some general aspects of test design are seldomdisputed. Examples of such provisions are therequirement of a m inimum num ber of animalsin the test grou ps an d that (generally) both sexes

8/14/2019 8104


Ch. 4—Methods for Detecting and ldentifying Carcinogen 125

Table 27.—Guidelines for Bioassay in Small Rodents

— —— .- --NCI (331)a FIFRA (102) TSCA (106)

Endpoint

Study plan:Animal speciesNumber of animals ateach doseDosages

Dosing regimenStartEndObservation period

Organs and tissues to beexamined

Personnel qual i f icat ions:

Study director

Pathologist

Animal husbandry

Cost estimate

Carcinogenicity

2, rats and mice

50 males, 50 females2, MTD

MTD/2 or MTD/4plus no-dose control

At 6 weeks of ageAt 24 months of age3-6 months after end ofdosing

All animals: external andhistopathologic examination(ea. 30 organs and tissues)

N.s.Board-qualified

N.s.

N.s.

Oncogenicity

2, rats and mice

50 males, 50 females3, MTD

MTD/2 or MTD/4MTD/4 or MTD/8plus no-dose control

In utero or at 6 weeksMice, 18-24 mos; rats, 24-30N.s.

All animals: external exami-nation; some animals: path-ologic exam of 30 organsand tissues, other animals,fewer organs and tissues

N.s.N.s.

N.s.

N.s.

Oncogenicity

2, rats and mice

50 males, 50 females3, HDL

HDL/2 or HDL/4HDL/4 or HDL/10plus no-dose control

At 6 weeks of ageAt 24-30 months of ageN.s.

All animals: external andhistopathologic examination(ea. 30 organs and tissues)b

Responsibilities detailedBoard-certified or equivalent

Board-certified vet. orequivalent

$400,000 Y 160,000

Abbreviations: MTD, Maximum tolerated dose, causes minor acute toxlclty

HDL, high dose level, causes some acute tox!clty

N s., not speclfled

a The NC I Guldellnes specify the lndl~ated m,nlmum requirements, Tfley a[lo~ for flexlblllty in experimental design so long as the minimum requirements are met,

b EPA estimates that the 40,000 microscope slldes produced In this examination WIII require more than 3/4 of a year of a pathologist’s time for analysls


be tested. On the other hand, consensus doesnot exist about some aspects of experimentaldesign, for instance: How high a dose is to beadministered? The policy of the agency that

draws up the guidelines is reflected in what itsays about th e arguable aspects of experimentaldesign . Toma tis (341) has d iscussed five debatedissues about bioassay.

1. Doses of chemicals to wh ich test anim als

are exposed are too high and are not pre-dictive for effects at levels of hu man ex-posure.

High doses of suspect carcinogens are neces-sary to increase tumor incidence to a level thatcan be d etected in the limited n um ber of animalsused for tests (180). A chemical causing tumors

at a rate of 0.5 percent in the U.S. populationwould result in over 1 million cancer cases. Butan incidence of 0.5 percent w ould very p robablygo undetected in the usual test population of 50

male plus 50 female rats or mice. Administra-tion of the chemical at a 10-times higher dosemight increase the response to a detectable levelgiven comparable sensitivity between the test

animals and hu man s. High d oses are necessaryto increase the sensitivity of the tests, but argu -ments arise over whether or n ot a dose is so highthat it is not pred ictive of wh at hap pens at lowerdoses.

The biological argument against dependingon results at high doses centers on the conten-tion that such doses may alter metabolism orcause local irritations or other toxic responseand cause cancer through a pathway that wouldnot exist at lower doses (60,139,278). A solutionsometimes offered to the problems raised by

high doses is to run bioassays with man y moreanimals tested at lower doses. The most spec-tacular attempt at this was the N ational Centerfor Toxicological Research (NCTR) ED

O1study

8/14/2019 8104



wh ich u sed 24,000 mice and w as designed to d e-tect one cancer in 100 animals. Unfortunately,neither it nor (probably) any experiment caneliminate the necessity for testing chemicals atdoses significantly higher than those experi-enced by humans. Even 24,000 is a smallnum ber compared to the num ber of people ex-posed to many chemicals.

2. Routes of exposure in animal tests do notcorrespond to routes of human exposure.

Chemicals are administered to laboratory ani-mals in either food, water, by inhalation, force-feeding (gavage), skin-painting, or injection.Few objections are raised to administration infood, water, or by inhalation when the chosenroute mimics the route of human exposure.More objections are raised to gavage, skin-painting, injection, or ingestion wh en th at is notthe normal human exposure route. However,such method s are sometimes necessary and , fur-therm ore, carcinogens appear to be d istributedthroughout the body regardless of the route of exposure. EPA (106) stipulates that “to the ex-tent p ossible, route(s) of adm inistration shou ldbe comparable to expected or known routes of human exposure.” Adherence to such sugges-tions will reduce the frequency of this objection.

3. Some anim als used for testing are so bio-logically different from hu man s that re-sults from them have no value.

Choice of animals for bioassays represents acompromise. Most current guidelines require orsuggest that chemicals be tested in two rodentspecies, generally rats and mice. The adva ntagesof these species are their small size, reducing thespace necessary for housing, short lifespans (2to 3 years), redu cing th e time need ed for a life-time stud y, and a large amou nt of informationabout the genetics, breeding, housing, andhealth of these animals. Rats and mice are cheapto buy, feed, and house, as compared withlarger animals.

Primates are sometimes used for certain toxi-cological testing. They are certainly more like

humans than rodents but their supply is limited.They are expensive, live u p to 25 years, and re-quire large areas for h ousing. Despite these dif-ficulties, NCI now maintains about 600 mon-

keys for carcinogenicity testing at a cost of about $500,000 per year (2). Dogs are thoughtto be between rodents and monkeys in their ap-parent likeness to humans, but they are morelike prima tes in costs.

Differences in metabolism, bioaccumulation,and excretion between rodents and humans

should be considered in interpreting the sig-nificance of animal results for humans. There isno question that further research in the com-parative biochemistry and physiology of manand rodents is necessary, but the comparisonswill ultimately be limited by restrictions onwh at can be determ ined by experimentation inhumans. Moreover, metabolic studies haveshown that m ost differences between hu mansand experimental animals are quantitativerather than qualitative and support the idea thatanimal results can be used to predict humanresponses.

4. Some test animals or organs of test ani-mals are exquisitely sensitive to carcino-gens, and such sensitivity invalidates useof results from such anim als.

Griesemer an d Cueto (146) have analyzed th eresu lts of testing 190 chem icals in th e NCI Bio-assay Program (see d iscussion in Expert Reviewof Bioassay Results and app . A). They identified35 chemicals that were “strongly carcinogenic”in either the rat or the mouse and noncarcino-genic in the other species. Of the 35, 18 werepositive in the mouse and negative in the rat,and 17 were positive in the rat and negative inthe mou se, wh ich indicates that neither animalwas much more often the sensitive species.How ever, 12 chemicals caused mou se liver tu -mors, no other lesion in the mouse and no le-sions in rats. Taken by them selves these resu ltssuggest that th e mou se liver is a sensitive organ.

Liver tum ors are often found in mice but areinfrequently found in U.S. citizens, althoughthey occur frequen tly in hu man pop ulations inother parts of the world. Should a chemical thatcauses mouse liver tumors be considered a haz-

ard ? An app roach to resolving the mou se liverquestion was to review how predictive such re-sults are for tumors in other animals. Tomatis,Partensky, and Montesano (343) showed posi-

8/14/2019 8104


Ch. 4—Methods for Detecting and identifying Carcinogens q127

tive correlation between a chemical’s beingoncogenic in the mouse liver and its being onco-genic either in the liver or at some oth er site inrats or hamsters. (Griesmer and Cueto (146)presented data only from rats and mice, andsome m ouse liver carcinogens in th eir comp ila-tion might be positive in animals other thanrats. )

Despite the finding of Tomatis, Partensky,and Montesano (343) that m ouse liver carcino-gens w ere often p ositive at oth er sites or in otheranim als, IARC (185,187) considers m ouse liverand lung tum ors as “limited evidence” for car-cinogenicit y . However, OSHA (272) acceptsmouse liver tumors as “indicators of carcinoge-nicity” if “scientific experience and judgment”are used in interpreting the data.

Crouch an d Wilson (72) have ana lyzed someof the NCI Bioassay Program data. They cal-culated the potency and the standard error asso-ciated w ith potency of a chemical in causing atum or at a par ticular site in either the rat or th emouse. In their analysis of 35 tested chemicals,they found that in 31 cases the potency in bothrats and mice agreed within a factor of 10. Theiranalysis and similar analysis by Ames et al.

(14), which consider the inherent error in e x-periments and the sensitivity of experiments,significantly reduces the number of chemicalsthat are p ositive in one sp ecies and n egative inanother. However, those analytical methodshave not been generally applied to bioassayresults.

5. Finding benign tumors in test animals hasno value in defining carcinogenicity.

Tumors can be divided into two classes,benign and malignant. Benign tumors do notmetastasize, the tumor cells remain in contactwith each other and do not invade other tissuesor organs. Malignant tumors can invade othertissues and metastasize, spreading to distantparts of the body and causing other tumors.Both types of tumors are found in experimentalanimals.

Hearings on pesticide regulations have beenmarked by repeated arguments about the im-portance of benign tumors. Should benigntumors found in experimental animals be takenas eviden ce that a chemical causes cancer?

The position that a benign tumor may laterbecome malignant, that th e line of dem arcationbetween benign and malignant is unclear, andthat ben ign tu mors can also be life-threateninghas prevailed in regulatory agencies. Therefore,no distinction is made in regulatory decisionsbetween benign and malignant tumors. This is

clearly r eflected in FIFRA guid elines (102) andTSCA test standards (106) in which the end-point is oncogenicity (tumor causation) ratherthan carcinogenicity (wh ich em ph asizes malig-nancy) (see table 27).

Griesemer and Cueto (146) reported no dif-ficulty in deciding between benign and malig-nant tumors in the NCI Bioassay Program andthat their conclusions about carcinogenicitywere unaffected by including or excluding be-nign tumors. IARC statements reflect difficultiesin the interpretation of benign tumors. As men-tioned above, it considers the occurrence of

some benign tumors as “limited evidence” forconclud ing th at a chemical is a carcinogen (185).It also states that “preneoplastic lesions” mayprogress to “frank malignancy.” Furthermore,IARC (186,187) considers that the occurrence of both p reneoplastic and neop lastic lesions in thesame organ strengthens conclusions about car-cinogenicit y.

Expert Review of Bioassay Results

In addition to the general objections to

bioassay procedures that have been discussed,specific objections may be raised to particulartests. For instance, animals may have been in-advertently exposed to m ore than one chemical,to pathogenic micro-organisms, to extreme tem-perature, or to temp orary d eprivation of foodor water, any one of which migh t influen ce re-sults. Another frequent item of contention iswh ether or not a p articular p athologic lesion in-dicates carcinogenicity. Critical reviews can re-du ce concern that flaws in experimental d esignor conduct mar the experiment or bring its re-sults into qu estion.

IARC periodically calls together panels of ex-perts to review the worldw ide literature aboutthe carcinogenicity of particular chemicals orexposures. The IARC Monograp h Program , be-

8/14/2019 8104


8/14/2019 8104


Ch. 4—Methods for Detecting and Identifying Carcinogens q 12 9

Table 28-Some Organizations That Have Considered and Endorsed Animal Tests as Predictors ofHuman Risk From Chemical Carcinogens

Organization

The Secretary’s (HEW) Committee on Pesticides andtheir Relationship to Environmental Health

National Research Council

Office of Technology Assessment

National Cancer Advisory Board, Subcommittee onEnvironmental Carcinogenesis

American Industrial Health Council

Food Safety Council

Occupational Safety and Health Administration

Interagency Regulatory Liaison Group

Office of Science and Technology Policy

American Federation of Labor, Congress of IndustrialOrganizations, and United Steelworkers of America

— —

—

—

—

— —

—

—

—

—

—

—

—

—

Publication

Report (325)

Evaluating the Safety of Food Chemicals (259)Safety of Saccharin and Sodium Saccharin in the HumanDiet (261)Pest Control: An Assessment of Present and AlternativeTechnologies (262)Principles for Evaluating Chemicals in the Environment(263)Food Safety Policy (269)

Cancer Testing Technology and Saccharin (282)Environmental Contaminants in Food (284)

General Criteria for Assessing the Evidence for Carcino-genicity of Chemical Substances (245)The Relation of Bioassay Data on Chemicals to theAssessment of the Risk of Carcinogens for Humans UnderConditions of Low Exposure (246)

AIHC Recommended Alternative to OSHA’s Generic Car-cinogen Proposal (8)

Proposed System for Food Safety Assessment (125)

Identification, Classification and Regulation of Potential

Occupational Carcinogens (279)Scientific Bases for Identification of Potential Carcinogensand Estimates of Risks (180)

Identification, Characterization and Control of PotentialHuman Carcinogens: A Framework for Federal Decision-making (281)

Post-Hearing Brief on OSHA’s Proposed Standard on theIdentification, Classification and Regulation of Toxic Sub-stances Posing a Potential Occupational CarcinogenicRisk (7)

SOURCE: Off Ice of Technology Assessment.

Evidence of the carcinogenic activity of anagent can be obtained from bioassays in experi-mental animals.

and,

Positive results, obtained in one species onlyare considered evidence of carcinogenicity. Posi-tive results in more limited tests (e.g., when theobservation period is considerably less than theanimal’s lifetime), but by experimentally ade-quate procedures, are acceptable as evidence of carcinogenicity. Negative results, on the otherhand , are not considered as evidence of lack of acarcinogenic effect, for operational purposes,unless minimum requirements have been m et.

As these quotes show , both AIHC and IRLG

accept bioassays as predictors of potentialhuman risk. However, they differ in the weightof evidence each requires. AIHC wants positive

test results in tw o sp ecies before m aking d eci-sions about carcinogenicity, while IRLG willconsider making a decision on the basis of posi-tive results in one species. Labor organizations(e.g., American Federation of Labor, Congressof Indu strial Organizations, and United Steel-workers of America (7)) and environmentalgroup s also consider positive results in a singleanimal as sufficient to m ake a jud gment aboutcarcinogenicit y.

How Many Chemicals Are Carcinogensin Animal Tests?

The number of chemicals that are carcino-

genic in h um ans or animals is uncertain. A fewestimates have been made, but the bases for theestimates are poor.

8/14/2019 8104



OSHA (279) states: “Most substances do notappear to be carcinogenic, ” and uses informa-tion from two lists of test results to support thestatement. OSHA cites the study of Innes et al.(176), which tested 120 industrial chemicals andpesticides . Eleven (less than ten percent) werereported to b e carcinogen ic, 20 were consid eredto require furth er study, and 89 “did not give

significant indication of tumorigenicity. ” A Na-tional Academy of Sciences review of pesticides(262) also drew attention to the low number of positive results reported in the Innes stud y. Thesmall number was judged to be especially signif-icant because of the large nu mber of biologicallyactive pesticides included in the test. Certainreservations must be attached to these conclu-sions. The maximum dose tested and the num -ber of animals used in the experiments weresmaller than now required, and consequently,the experiments were less conclusive than morerecent ones.

A more d irect criticism of relying on th e Inneset al. (176) document as a measure of how manychemicals are carcinogens is that it w as a p re-liminary report. Based on the complete report(26), Barnard (21) states that 29 percent of thetested compounds were carcinogenic.

The second sou rce cited by OSHA (279) is theseven-volume Public Health Service (PHS) listof chemicals tested for carcinogenicity (298).OSHA (279) reported that about 17 percent of the 7,000 tested chemicals were tumorigenic.However, OSHA concluded that these data“overstated the true proportion of carcinogenicsubstances in the human environment” becausesusp icious chemicals are selected for testing .

Despite such reservations, the Task Force onEnvironmental Cancer, Heart an d Lung Disease(339) said :

Of the upwards of 100,000 known chemicalsof potential toxicity, only approximately 6,000

have been tested for carcinogenicity. It is esti-mated that 10 to 16 percent of the chemicals sotested p rovide some evidence of animal carcino-genicity.

To treat the rep orted 10 to 16 percent of testedchemicals as carcinogens as a reliable number isprobably an error because many of the 7,000

tests are clearly inadequate when measuredagainst current testing guidelines. Scientists em-ployed by the General Electric Co. (141a) havealso examined the lists from the seven v olumes(298). Using a relaxed cr i ter ion for carc ino-

genicity “ . . . any listing that reported anincidence of tumors in the test animals higherthan in the control animals was scored as

‘positive, ’ “ up war ds of 80 percent of the listedchemicals were found to be positive. It isimportant to recognize that this relaxedcriterion would not be accepted in any regula-tory d ecision, and it exaggerates the nu mber of positives. Furthermore, biases toward testingchemicals that ar e likely to be carcinogenic andtoward reporting positive results push thepercentage of positives upward.

OSHA (279) had a contractor analyze a list of 2,400 susp ect carcinogens compiled by the N a-tional Institute of Occupational Safety andHealth (NIOSH). The contractor estimated that

for 570 (24 percent of the total) there were suffi-cient d ata to perm it initiation of ru lemaking toclassify them as category I or 11 carcinogensun der the prop osed OSHA p olicy.

Neither the Innes stud y nor th e PHS list pro-vides information as reliable as that which ismore recently available from NCI, IARC, andNTP. Fifty-two percent of NCI-tested and re-ported chemicals were carcinogens (146). Either“sufficient” or “limited” data existed to classifyabout 65 percent of IARC-listed chemicals ascarcinogens (186,344). NTP (273) repor ted that

252 (including the 190 reported by Griesemerand Cueto (146)) tests have been completed inthe N CI Bioassay Program. Un der conditions of the tests, 42 percent were positive, 9.5 percentwere equivocal, 36 percent w ere negative, and13 percent w ere inconclusive.

The biases toward testing likely-to-be-posi-tive chemicals cannot be ignored, an d the factthat negative test results are less likely to bepu blished an d included in any comp ilation (279)further complicates analysis of lists of testedchemicals. These factors tend to increase thepercentage of positive chemicals, and conse-quently may inflate the estimates of the percent-ages of chemicals that are carcinogens.

8/14/2019 8104


Ch. 4—Methods for Detecting and Identifying Carcinogens q131

A definitive answer to questions about how Find ing more an d m ore chemicals to be car-many chemicals are carcinogens would depend cinogenic in bioassays raises important policyon testing every chemical, and that is beyond questions and may force a decision to rank car-the capacity of the bioassay system. Tomatis cinogens for possible regulation or voluntary(341) reported that 828 chemicals were under redu ctions. It is not apparent h ow to d eal with atest worldwide in governmental programs in large number of carcinogens without ordering1975, and that 317 were r epeat t ests of chemicals them according to their riskiness.for which, in his opinion, adequ ate data alreadyexisted. He did not estimate the number of chemicals under test in private or commerciallaboratories.

SELECTION OF CHEMICALS FOR TESTING INBIOASSAYS AND SHORT-TERM TESTS

Tests develop information to aid in makingdecisions about chemicals, and the m ost impor-tant step in information development is placingthe chemical on test. Limited testing capacity

makes it necessary to pick and choose amongchemicals. Not all can be tested. The secondNTP Annu al Report (272) und erlines this pointin discussing bioassays:

It is unreasonable even to attemp t to stud y all48,000 chemicals in this w ay. Current know nworld capacity permits the initiation of perhaps300 such chemical tests each year, and the resultspublished th is year are from the tests begun 4 ormore years ago. This capacity—even if financialresources were not limiting—could be no morethan doubled in the next 5-10 years. At this rate,it would take an additional 80 years to study all

currently existing chemicals. Further, approxi-mately 500 new chemicals are introduced intocommerce each year, and this wou ld result in anadditional backlog of some 40,000 untestedchemicals.

The same report pinpoints Federal Governmenttesting capacity:

In carcinogenesis testing the NTP proposes tostart 75 new chemicals on the lifetime carcino-genicity bioassays in fiscal year 80. This is thesame as the level achieved in fiscal year 79 and isa two-fold increase over the rate of testing priorto the establishment of the program. [In fact, 50

were started in fiscal year 1980; 30 are expectedin fiscal year 1981. ]

By spring 1980, thecentered its selection

Federal Government hadactivities on two pro-

grams, the Interagency Testing Committee(ITC), and the N TP Chemical Nom ination an d

Selection Committee. ITC recommends chem-icals for testing to the EPA Ad ministrator un der

jurisdiction of section 4 of TSCA (the ITC list)and NTP selects chemicals for testing by theFederal Government. In addition, NCTR andCIIT have also published criteria for testingsubstances.

The Interagency Testing Committee

Section 4(a) of TSCA stipulates that EPAissue a rule to require testing if an individualchemical or category of chemicals “may present

an unreasonable risk of injury to health or theenvironment . . . or will be produced in sub-stantial quantities and . . . enter the envi-ronm ent in substantial quan tities or there is ormay be significant or substantial hum an expo-sure. ” The vagueness of such terminology, “un -reasonable risk, ” “substantial quantities, ” “sig-nificant or substantial human exposure, ” re-quires EPA to generate interpretations withinscientific, legal, and policy contexts.

Section 4(c) of TSCA established ITC to rec-ommend chemical substances or mixtures wh ich

should be given p riority consideration for test-ing. ITC is composed of eight members, one

8/14/2019 8104



each from EPA, OSHA, the Council on Envi-ronmental Quality (CEQ), NIOSH, NIEHS,NCI, the National Science Foundation (NSF),and the Department of Commerce.

ITC is to give p riority to those chemical sub-stances which are know n or suspected to causeor to contribu te to the causation of cancer (car-

cinogens), mutations (mutagens),- or birth de-fects (teratogens). The total number of chemicalsubstances or mixtures recomm ended for testingat any one time is not to exceed 50. TSCA spec-ifies that ITC must update, as necessary, thetesting p riority list at least every 6 mon ths. TheEPA Administrator is directed to respond to achemical’s being placed on the list within 12mon ths. The Ad ministrator mu st either initiaterulemaking to requ ire testing or pu blish reasonsfor not initiating a testing rule.

ITC’s initial selection of chemicals for con-sideration was made by combining about 20Government-compiled lists of potentially haz-ardous substances. Chemicals that did not comeunder TSCA’s purview were eliminated fromfurther consideration, and the remaining sub-stances were ranked against two measures: ex-posure and biological activity. The bases for de-termining exposure were explicitly specified byCongress in section 4(e)(i)(A) of TSCA:

q general population exposure (number of people exposed, frequency of exposure, ex-posure intensity, penetrability);

quantity released into the environment

(quantity released, persistence);qproduction volume; andq occupational exposure.

In general, chemicals w hich w ere ranked highon the exposure scale were further rankedagainst biological activity measurements. TSCAspecified the first three factors, and the otherswere included because of their significance incharacterizing biological effects (183):

q

q

q

q

q

q

q

carcinogenicity,mutagenicity,teratogenicity,

acute toxicity,other toxic effects,ecological effects, andbioaccumulation.

ITC examined the biological activity scoreand the individu al biological and exposure fac-tors, and weighed the biological scores againstthe exposu re scores to select chemicals for d e-tailed reviews. After the reviews, chemicalswere recommended to the EPA Ad ministratorfor consideration. A d etailed description of thescoring system can be found in the initial ITCreport to the Administrator of the EPA (183).

As of November 1980, ITC had filed seven re-ports to th e EPA Adm inistrator. Each report hasup dated and revised the list of chemicals eligiblefor prom ulgation of test rules und er section 4(a)of TSCA.

Chemical Categories

An important consideration in selectingchemicals for testing is how to deal w ith chem-

ical categories or classes. Section 26(c) of TSCAspecifies that any action authorized or requiredunder the act “may be taken by the Administra-tor . . . with respect to a category of chemicalsubstances or mixtures [and] shall be deemed tobe a reference to each chemical substance ormixture in such category. ” In making its recom-mendations ITC selected some categories of chemicals for testing. EPA must determinewhether to evaluate the scientific, economic,and regulatory consequences of every p resent orpoten tial member of a category of chemicals orto evaluate selected representatives from the

category. If the decision is made to test repre-sentatives, EPA has to assess whether it could“reasonably” extrapolate test results to chem-icals within the group that have not beenstudied. Some industry representatives havequestioned the validity of using categories forchemical testing, particularly the categories rec-omm ended by ITC. They express concern that itwou ld be unr ealistic to establish a general policyfor choosing w hich chemicals shou ld be testedfor every possible category. Charles Holds-worth, of the American Petroleum Institute,recomm ended that EPA select m embers of a cat-

egory once the “m etabolic actor” for the grou pis determined. Another possible approachwould be to select chemicals of interest becauseof exposure or prod uction volum e.

8/14/2019 8104


Disposition of the ITC List

As of May 1981, 3-1/ 2 years after completionof the first ITC list, EPA had issued no final rulerequiring testing of any ITC-identified chem-icals. However, EPA (110) in the summer of 1980 proposed its first health effects test rule onchemicals nominated by ITC. The p roposed rule

wou ld require testing of chloromethane and rep-resentative mem bers of the category of chlori-nated benzenes for oncogenicity and otherhealth effects. Support documents describingthe reasons for d eciding to require testing andan economic impact analysis of the tests werealso released (110,112). As of April 1, 1980,draft copies of test rules for four additionalchemicals and groups of chemicals from the 21iden tified on the first three ITC lists were avail-able from EPA.

The General Accounting Office (141) re-

ported that EPA estimates it will take 5 years toissue a final test rule for an ITC-nominatedchemical. That time is required for the agency toevaluate the published information about thechemical. EPA further estimates that an addi-tional 4 years w ill be requ ired for execution of bioassay and analysis of their results. At aminimum, then, 9 years will pass between thedecision to test an ITC-selected chemical andcompletion of testing. Certainly the long timebetween the d ecision to require testing and theproduction of results is of concern to EPA, andthe agen cy is working to red uce it (141).

There is a suggestion tha t man y ITC-selectedchemicals are curren tly und er test in other arm sof the executive branch. NTP (273) reportedthat of 20 single chemicals and 15 classes recom-mended by ITC (as of April 1980), 16 of the 20

chemicals and representative chemicals from 14of 15 classes were then un der test or scheduledfor test by NTP. NTP (273) did not make a com-parison between th e types of tests recommendedby ITC and the types of tests being carried ou tor planned by NTP. The NTP report draws at-tention to a p roblem common to an active field

such as toxicology. ITC judgments aboutwhether or not to require a test are based orwhat it knows about in-progress and completedtests. It is imp ortant tha t NTP and other testing

organizations share their latest information andplans with ITC. Accordingly, a liaison repre-sentative of NTP attends and participates in ITCmeetings and related activities.

Responsibility for Testing Un der TSCA

Once a chemical or category is selected for

testing, EPA must determine who should bearthe responsibility and burden of testing. TSCArequires EPA to indicate whether manufac-turers, processors, or both m anufacturers andprocessors bear the responsibility. EPA ispresently evaluating exposures that occur atvarious points in a chemical’s life cycle. If EPAfind s that the chem ical’s man ufacture may pre-sent a risk, only the m anu facturers m ust test. If processing m ay pose a hazard, only the proces-sors are required to test. However, if distribu-tion in commerce, use, or disposal of the chemi-cal may present a risk, both the manufacturers

and the processors are required to test. Thisdetermination has substantial economic andlegal ramifications since it will establish theuniverse of firms which must bear the cost of testing. Some of the chemicals for which testrules have been drafted are manufactured bymore than one company. EPA is urging thatfirms cooperate to sponsor single tests of thechemicals rather than have each company spon-sor its own test.

National Toxicology Program

NTP was established w ithin DHEW (now De-partment of Heal th and Human Services(DHHS)) on N ovember 15, 1978, to further the

development and validation of integrated tox-icological test methodologies. The NTP Ex-

ecutive Committee is comp osed of the head s of FDA, OSHA, CPSC, EPA, National Institutesof Health (NIH), NIOSH, NCI, and NIEHS.

NTP’s annual plan for fiscal year 1980 (272)describes methods to select chemicals for test-ing: NTP operates under the principle that in-dustry will test chemicals for health and envi-

ronmental effects as intended and mandatedby

Congress under legislative authorities. H O W-

ever, some chemicals will not be tested by theprivate sector, and NTP will select chemicals for

8/14/2019 8104



its own testing program from the followingcategories:

1.

2.

3.

4.

5.

6.

7.

8.

chemicals found in the environment thatare not closely associated with commer-cial activities;desirable substitutes for existing chem-icals, particularly therapeutic agents, that

might not be developed or tested withoutFederal involvement;chemicals that should be tested to improvescientific understanding of structure-activ-ity relationships and thereby assist in d e-fining groups of commercial chemicals

that should be tested by ind ustry;certain chemicals tested by industry, or byothers, the add itional testing of which bythe Federal Government is justified toverify the results;previously tested chemicals for whichother testing is desirable to cross-compare

testing methods;“old chemicals” with the potential for sig-nificant human exposure which are of social importance bu t wh ich generate toolittle revenu e to supp ort an adequ ate test-ing program (some of these may be“grandfathered” under FDA laws);two or more chemicals together, whencombined human exposure occurs (suchtesting probably cannot be required of in-dustry if the products of different compa-nies are involved); andin special situations, as determined by the

executive committee, marketed chemicalswhich have potential for large scaleand/ or intense hum an exposure, even if itmay be possible to require industry to per-form the testing.

NTP solicits lists of chemicals from NTP re-search (NCI, NIEHS, FDA, NIOSH) and regu-latory (FDA, OSHA, CPSC, EPA) agencies,other Federal agen cies, academ ia, ind ustry, la-bor, and the public. All of the chemicals sug-gested for study are funneled to the NTP Chem-ical Nominations Group.

The Chemical Nominations Group, com-posed of representatives from EPA, OSHA,FDA, CPSC, NIH, NCI, NIEHS, and NTP, pre-

pares a dossier describing what is known aboutthe physical properties of each chemical, its pro-du ction volum e, its use, exposures to it, and an ytoxicological information, Each chemical is

judged against the chemical selection principlesdescribed above and n ominations are forw ardedto the NTP Executive Comm ittee which makesfinal d ecisions about which chemicals to p laceon test.

A decision by NTP to test a chemical does notmean necessarily that the chemical will beplaced on a bioassay program . It may m ean thatthe chemical will be entered into less expensive,short-term tests first, and depending on theresults of those tests, subsequent d ecisions w illbe made about w hether testing should continue.

The National Center forToxicological Research

NCTR, a research arm of EPA and FDA w asestablished to develop a better un derstan din g of adverse health effects of potential l y t o x i cchemicals. NCTR’s research emphasis is on de-termination of adverse health effects resultingfrom long-term, low-level exposure to chemicaltoxicants (food additives, residues of animaldrugs, etc.); determination of the basic biologi-cal processes involving chemical toxicants inanimals in ord er to enable better extrap olationof toxicological data from laboratory animals toman; and development of improved methodolo-

gies and test protocols for evaluation of the safe-ty of chemical toxicants (good laborator y prac-tices, automated data systems, etc.). NCTRchooses substances for testing from th e follow-ing categories (271):

1.

2.

substances that have no clear industrialsponsorship and for which it is determinedthat furth er toxicological data are n eeded.Usually these are either food contami-nants, GRAS (generally recognized as safefood additives) compounds, or cosmeticingredients.

substances that can act as model com-pounds in a continuing toxicologicalmethods d evelopment program.

8/14/2019 8104


8/14/2019 8104



Figure 20.—lnterrelationships of Major Testing Activities of NTP

Phase IVPhase I Phase II Phase Ill Scientific and Public

Test program Identify toxic potential Confirm toxic effect More definitive result Health Assessment

I IGenetic toxicology Sa/nTone//a (Ames) assay (30~ EIrosophi/a spp~ Mammalian assaysS(mutagenicity) Mammalian cell culture (70) (30) (mouse heritable I I

N 7translocation)I

I

II Total evaluation

ICarclnogenesls Cell culture ~ Short-term in_ Lifetime rodent~ of all data

tran:;formatlon wvo (mouse bloassays I

1

lung aaenoma) (75)~1 i I

Toxicology Rodent toxicologyscreen (75)

a ( ) ~earlY capaci ty , fiscal year 19E0 fi9ures.

SOURCE: Off Ice of Technology Assessment, adapted from NTP (272)

-+

I

I

I

sence of arrow head s on th e lines is intentional;according to David Rail, NTP Director (303),there has not yet been enou gh experience withthe scheme to be certain which phase I testsshould come first or whether all chemicalsshould go through all phase I tests.

A critical feature of tier testing is the ability tomake decisions about whether or not to con-tinue testing from phase I to II, or from 11 to III.Guidelines are necessary for making the deci-sion that a chemical is sufficiently without risk

and that no further testing is necessary. Rail(303) said that developm ent of d ecisionmakingguidelines is a priority item for NTP in 1980 and1981. NTP intends to analyze the testing his-

tories of chemicals that have gone through allthree ph ases (albeit not necessarily un der N TPaegis) to determine w hich test results w ere mostpredictive of the ultimate decision about thechemical. NTP will take advantage of the factthat the most expensive and time consumingtesting, ph ase III, has been comp leted on som echemicals which have not otherw ise been tested.Such chemicals will be entered into phase I andII testing to provide additional informationabout wh ich tests are m ost pr edictive. Finally,the NTP decision to continue testing a few

chemicals that a re negative in p hase I tests willprovide additional information.

EPIDEMIOLOGY

Lilienfeld (210) defined epidemiology as the exposur e to carcinogenic agents or betw een as-study of the distribution of disease in human pects of lifestyle and increased cancer risk.pop ulations and of the factors that influence dis-ease distribution. Epidemiologic techniques are The earliest association of a factor withuseful for identifying causative agents an d con-

cancer was made by Ramazzini in 1713. Heditions that pr edispose for cancer. Stud ies can found that nuns had a higher rate of breastdeterm ine associations in pop ulations between cancer than other w omen an d, in his Treatie on

8/14/2019 8104


the Diseases of Tradesman of 1700, he at-tributed the increase to celibacy (322). Thatassociation has been sharpened to include theobservation that wom en w ho d eliver a child atan early age are less likely to develop breastcancer. In addition to identifying cancer risks,epidemiology may play a positive role by poin-ting out less hazardous d iets or agents wh ich areprotective against cancer.

For the purpose of this discussion, epide-miologic stud ies are divided into three generaltypes: 1) experimental, 2) descriptive, and 3)observational. While several basic strategies ex-ist, there are no rigid stu dy d esigns w ithin anyof these categories. Flexibility is importantsince, unlike laboratory experimen ts, epidemi-ology examines groups of unpredictable peopleliving in dynamic environments. The impor-tance of flexibility is underlined by IRLG Guide-lines, which state that epidemiologic study

design m ust “be d escribed and justified in rela-tion to th e stated objective of the stu dy” (181).

Experimental Epidemiology

The ideal procedu re for investigating cause-and-effect hypotheses is through experimentalepidemiology. This type of study requires thedeliberate application or w ithholding of a factorand observing the appearance or lack of appear-ance of any effect. Given th e severity of cancer,ethical considerations p reclud e the ad ministra-tion of suspected carcinogens to people, though

it is possible to test agents thought to aid inprevention (292).

Experimental epidemiology studies are dif-ficult to condu ct because of the n eed to securethe cooperation of a large group of people will-ing to permit an experimenter to intervene intheir lives. The investigator must have reason tobelieve that the proposed intervention, whethera d eliberate app lication or w ithholding, w ill bebeneficial, but at the same time, he must besomewhat skeptical of the effects. Once suffi-cient eviden ce leads to the conclusion th at theintervention is or is not beneficial, the experi-

ment must be terminated.

Descriptive Epidemiology

Descriptive epidem iology stu dies examine thedistribution and extent of disease in popu lationsaccording to basic characteristics—e.g., age,sex, race, etc. The primary purpose of con-du cting d escriptive epidem iologic stud ies is toprov ide clues to the etiology of a disease which

may then be investigated more thoroughlythrough more deta iled studies. Descriptivestudies have focused on international com-parisons and comp arisons amon g smaller geo-graphical regions, such as U.S. counties (29).

The identification of high bladder cancer ratesin New Jersey males and excess mortality ratesfrom cancer of the mouth and throat, esopha-gus, colon, rectum, larynx, and bladder in theindu strialized N ortheast have suggested that oc-cupational factors might be incriminated andhave prompted additional investigations. Blot

et al. (29) describe NCI’s stepwise approach tosearch for etiological clues. Examination of age-specific rates of disease occurrence or mortalityacross time (see ch. 2) is another example of de-scriptive epidemiology.

Observational Epidemiology

Observational epidemiology depends on dataderived from observations of individuals or rel-atively sm all groups of people. These stud ies areanalyzed using generally accepted statisticalmethods to determine if an association exists be-tween a factor and a disease and, if so, thestrength of the association. Often the hypothesisto be investigated arises from the results of a de-scriptive study. NCI has embarked on severalobservational studies based on findings fromtheir county-correlation studies. For example,high rates of lung cancer were found in theTidew ater Virginia area, and a large stud y wasinitiated which found elevated risk for lungcancer in shipbuilders and smokers.

Cohort Studies

Two types of observational epidemiology

stud ies, cohort and case-control stu dies, differ

8/14/2019 8104



in the selection of the population groups forstudy.

A cohort study starts with a group of people,a cohort, considered free of the disease understudy, and whose disposition regarding the risk factor under consideration is known. Usuallythe risk factor is an exposure to a suspect car-

cinogen or a p ersonal attribute or behavior. Thegroup is then stud ied over time and the healthstatus of the individu al members observed. Thistype of study is sometimes referred to as“prospective” because it looks forwar d from ex-posure to development of the disease character-istic (210,225). Cohort studies can be eitherconcurrent or nonconcurrent in design. Concur-rent cohort studies depend on events which willoccur in the future, w hile nonconcurrent cohortstud ies rely on past data or p ast events,

Case-Control Studies

In a case-control study, individu als with thedisease under study (cases) are compared to in-dividuals without the disease (controls) withrespect to risk factors which are judged rele-vant. Some authors label this study design“retrospective” because the presence or absenceof the pred isposing risk factor is d etermined fora time in the past (210,225). However, in somecases the p resence or absence of the factor an dthe d isease are ascertained simultaneously.

The choice of appropriate controls is rarelywithou t pr oblems. Often, for practical reasons,

controls are chosen from h ospital records. H ow-ever, they m ay not be repr esentative of the pop-ulation, and they therefore may introduce “se-lection bias,” as discussed by MacMahon andPugh (218).

In case-control and cohort stu dies, the group sselected should be comparable in all characteris-tics except the factor under investigation. Incase-control studies, the group s shou ld r esembleeach other except for the presence of the disease,wh ile in cohort studies, the stud y and comp ari-son groups should be similar except for expo-sure to the suspect factor. Since this rarely is

possible in practice, comparability betweengroups can be improved by either m atching in-dividual cases and controls (in case-control

studies) or by standard statistical adjustmentprocedures (in either case-control or cohortstudies). Demographic variables, e.g., age, sex,race, socioeconomic status, are most commonlyused for ad justment or matching.

There are advantages and disadvantages withboth the case-control and cohort studies (see

table 29). Case-control studies tend to be less ex-pensive to cond uct, require relatively fewer in-dividu als, and many have been especially usefulin studying cancer. The great advantage of cohort stud ies is that they allow observation of all outcomes, not only those originally antic-ipated. Bias is somewh at redu ced in cohort stud-ies since classification into an exposure categorycannot be influenced by prior knowledge thatthe disease exists. In a concurrent cohort study,it is often n ecessary to wait many y ears for themanifestation of enough disease cases to con-du ct an analysis. The cost and time of the stud y

can be reduced if conducted nonconcurrently.Cohort studies tend to require many more subjects than case-control studies and assignment of individu als to the correct cohort for an alysis isdifficult.

Causal Associations

A pragmatic view of causality is necessary,particularly when stud ying complex, mu ltifac-torial diseases such as cancer. Analysis of theassociation between exposure and disease in anepidemiologic study dep ends on tests of statisti-

cal significance. However, finding a positivestatistically significant association is not suffi-cient to conclude a causal relationship. Arti-factual and indirect associations must be con-sidered. As MacMahon and Pugh (218) state,"

. . . only a minority of statistical associationsare causal within the sense of the definition,which requires that change in one party to theassociation alters the other. ”

Policy Considerations AboutEpidemiology

While short-term tests and bioassays are u sedto evaluate a chemical’s carcinogenic potentialin the laboratory, the effect on humans is direct-

8/14/2019 8104


Ch. 4—Methods for Detecting and identifying Carcinogens q 13 9

Table 29.—Advantages and Disadvantages of Case-Control and Cohort Studies

Type of study Advantages Disadvantages

Case-control Relatively inexpensive Complete information about past exposures oftenunavailable

Smaller number of subjects Biassed recall

Relatively quick results Problems of selecting control group and matchingvariables

Suitable for rare diseases Yields only relative risk

Cohort Lack of bias in ascertainment of risk factor status Possible bias in ascertainment of disease

Yields incidence rates as well as relative risk Large numbers of subjects required

Can yield associations with other diseases as by- Long follow-up periodproduct

Problem of attrition

Changes over time in criteria and methods

Very costly

Difficulties in assigning people to correct cohort


ly assessed by epidemiologic techniques. Well-

conducted and properly evaluated epidemiologystudies which show a positive association areaccepted as the most convincing evidence abouthuman risks.

Negative epidemiologic results show that ex-posure of a certain number of people to a sub-stance at a sp ecified level did n ot cause cancer.From such resu lts, it is poss ible to calculate thathuman risk is no higher than what the studycould have detected. For instance, a study of 1,000 people which showed no excess cancerwould be “more negative” than one of 100 peo-

ple exposed at the same level. Neither studywou ld show that a risk exists, and neither showsthat no risk exists, but th e larger study sh ows alower probability of risk.

The OSHA Generic Cancer Policy (279) pro-posed that OSHA, after ascertaining the ade-quacy of the study design, would interpretnegative epidemiologic studies as setting an up -per limit for hu man risk. AIHC (8) wan ts nega-tive human evidence to be considered alongwith an imal data in m aking decisions about car-cinogenicity. The OSHA position, and that of Federal Governm ent regulatory agencies in gen-

eral (306), is to use epidemiology to estimatelimits of risk, but n ot to weigh n egative hum anevidence against other positive evidence in

deciding whether or not a substance is a car-

cinogen.

The Regulatory Council (306) considers prop-erly designed and conducted epidemiologicstud ies, which show a significant statistical re-lationship between human exposure to a sub-stance and increased cancer risk, to provide“good evidence” that a substance is carcino-genic. The Council mentioned some difficultiesin epidemiology, e.g., long latency periods,multiplicity of exposures, and cautioned thatoften “even large increases (which could involvethou sand s of peop le) . . . cannot be d etected. ”

For these reasons they cite two caveats in usingepidemiological studies:

The failure of an epidemiological study to de-tect an association between the occurrence of cancer and exposure to a specific substanceshould not be taken to indicate necessarily thatthe substance is not carcinogenic.

Because it is unacceptable to allow exposureto potential carcinogens to continue until hu mancancer actually occurs, regulatory agenciesshould not wait for epidemiological evidencebefore taking action to limit hu man exposure tochemicals considered to be carcinogenic.

OSTP (281) states that “a positive finding in awell-conducted epidemiologic study can beviewed as strong evid ence that a chemical poses

8/14/2019 8104


140 Technologies for Determining Career Risks From the Environment

a carcinogenic risk to h um ans. ” Alternatively,“a negative finding is not nearly so meaning-ful . . . “ and OSTP emphasizes the importanceof examining the sensitivity of a negative study,and suggests that the upper limit of risk thatmight have gone undetected in the study becalculated and p resented.

Carcinogens for Which There IsHuman Evidence

Through various means, epidemiologic stud-ies have identified several hu man carcinogens.The first use of epidemiologic principles torelate environmental contaminants to humancancer is credited to Pott in 1775 (322). Pott, aphysician in London, suggested that scrotal can-cers, which he observed in men who hadworked as chimney sweeps when boys, werecaused by exposure to soot. Pott is an examp leof an astute physician recognizing an unusual

cluster of cancer cases. A more r ecent example isvinyl chloride which was identified as a carcino-gen after three cases of a rare liver tumor(hepatic angiosarcoma) were diagnosed inworkers in a manufacturing plant (71). In thecase of vinyl chloride, evidence for its carcino-genicity in laboratory animals w as available inadv ance of the hu man evidence. In both cases,control action followed the demonstration of occupational risk. The Dan ish Chimn ey SweepsGuild instructed its members about protectiveclothing and to p ractice preventive hygiene soonafter Pott’s report was published, and OSHA

regulated v inyl chloride.

IARC bases its evaluation of carcinogenic risk to humans on consideration of both epidemi-ological and experimental animal evidence.IARC considered hu man evidence bearing on 60chemicals and industrial processes and classified18 as human carcinogens (see table 30). Many of the hu man data considered by IARC are fromstudies concerning workplace and medical ex-posure. This does not necessarily reflect thedistribution of carcinogens b ut m ore likely thehigher exposure an d relative ease of performingepidemiologic studies on patients and occupa-tional groups. For instance, the availability of medical records facilitates locating people ex-

posed to a d rug and provides information abouttime of exposure and dose level.

IARC also classified 18 additional compoundsas probably carcinogenic for humans but therewas insufficient evidence to establish causalassociations. These 18 were furth er su bdivid edaccording to the degree of evidence, high or

low, as displayed in table 30. Insufficient evi-dence was available to decide about the carci-nogenicity of 18 chemicals listed in table 30.Finally, because of time limitations, IARC wasunable to evaluate six compounds for whichhu man d ata exist.

Annual Report on Carcinogens

In an effort to provide information on car-cinogens which would be useful to regulatoryagencies, Congress passed an amendment to theCommunity Mental Health Act (Public Law 95-

622). It requires the Secretary of DHH S to pu b-lish an annual report containing a list of sub-stances which are known to be carcinogens ormay reasonably be anticipated to be carcino-gens and to which a significant number of per-sons residing in the Un ited States are exposed .The task was assigned to NTP, and th e first re-port (82) includes th e 26 exposures which IARChad determined in 1978 to be human carcino-gens (344). Candidates for the 1981 and 1982 listalso will be drawn from IARC beginning withchemicals and processes judged “probably carci-nogenic for humans. ” The initial report did not,

as required by the statute, list “ . . . all sub-stances which either are known . . . or . . .may reasonably be anticipated to be carcino-gens . . . . “

One limitation enu merated in the first reportwas that, “Science and society have n ot arrivedat a final consensus on the definition of carcino-gen either in human populations or in experi-men tal animals. ” The report, by includingchemicals and industrial processes already clas-sified by IARC, has sidestepp ed d ealing w ith thequestion of what is a carcinogen. A definition

for “carcinogen” remains elusive unless it isgiven in the context of a particular methodol-ogy.

8/14/2019 8104


Ch. 4—Methods for Detecting and Identifying Carcinogens q 14 2

Table 30.—Chemicals and Industrial Processes Evaluated for Human Carcinogenicity by the InternationalAgency for Research on Cancer (IARC)

Chemicals and processes judged carcinogenic for humans

4-aminobiphenyl. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . c .. ..0. “ 0. “ . . P “ 0. “ . ~ . .Arsenic and certain arsenic compounds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . “ . . .Asbestos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 0..... . . “ 0.0 “ “ ..0. “ 0. “ ..0.Manufacture of auramine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.0 “ . . . “ . “ 0 . . . .Benzene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...........”.”....” ““””.””.””.”.Benzidine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ““.......””.”.”..”..”” “.N,N-bis (2-chloroethyl)-2-naphthylamine (chlornaphazine) . . . . . . . . . . . . . . . . . . . . . . . . . . .Bis(chloromethyl)ether and technical grade chloromethyl methyl ether . . . . . . . . . . . . . . . . .Chromium and certain chromium compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......Diethylstilboestrol (DES). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .......”..”..”.”” .“.Underground hematite mining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .........”””.””.. ““Manufacture of isopropyl alcohol by the strong acid process . . . . . . . . . . . . . . . . . . . . . . . . .Melphalan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .....,.-.”.”..”..””” “...”..”.*.oMustard gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .“..”””.o”c”.”.””.”.o2-naphthylamine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .........”.”...””. “...”...’”<Nickel refining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ““”..””....”.. -.Soots, tars and mineral oils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .....””...”..”.””Vinyl chloride . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...”....”..”””..”’””. “.

Chemicals and processes judged probably carcinogenic for humansGroup A: Chemicals and processes with “higher degrees of evidence.”

Aflatoxins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .......”...”””. .“’””..”....””... “..Cadmium and certain cadmium compounds . . . . . . . . . . . . . . . . ...........”...” ...”...Chlorambucil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .“......”...”””- ..””.Cyclophosphamide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..”.”.......”..”Nickel and certain nickel compounds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ~ . . . . .Tris(1-aziridinyl)phosphine sulphide (thiotepa) . . . . . . . . . . . . . . ..;.... . . . . . . . . . . . ..”

Group B: Chemicals and processes with “Iower degrees of evidence.”

Acrylonitrile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....’.”.”...”..”.. “..”.”....””””Amitrole (aminotriazole) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...”..”....Auramine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .........”,..’’... ....”.....Beryllium and certain beryllium compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Carbon tetrachloride . . . . . . . . . . . . . . . . . . . . . . ................o....o .“..””..””....Dimethylcarbamoyl chloride . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .........”.””.” ....Dimethylsulphate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .......”......”.”””.”. .Ethylene oxide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..”.”.....””.”.......”Iron dextran . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....”,........... ““.”””.”.”..”Oxymetholone . . . . . . . . . . . . . . . . . . . . . . ........o.o.....”’”o.”. “.”..””..”...”..- ..Phenacetin . . . . . . . . . . . . . . . . . . . . . . ..........,.,”....”.. ....”..”..””””.”.””””. “Polychlorinated biphenyls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....”........... ...”.

Degree of evidence

ExperimentalHumans animals

Sufficient SufficientSufficient Inadequate

Sufficient SufficientSufficient NonapplicableSufficient InadequateSufficient SufficientSufficient LimitedSufficient SufficientSufficient SufficientSufficient SufficientSufficient NonapplicableSufficient Not applicableSufficient SufficientSufficient LimitedSufficient SufficientSufficient NonapplicableSufficient SufficientSufficient Sufficient

Degree of evidence


Limited SufficientLimited SufficientLimited SufficientLimited SufficientLimited SufficientLimited Sufficient

Degree of evidence


Limited SufficientInadequate Suff ici en tLimited LimitedLimited SufficientInadequate Suff ici en tInadequate Suf fic ientInadequate SufficientLimited InadequateInadequate SufficientLimited No dataLimited LimitedInadequate Sufficient

Chemicals and processes that could not be classified as to their carcinogenicity for humans

Degree of evidence

Experimental

Humans animalsChloramphenicol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ““”..”””. Inadequate No data

Chlordane/heptachlor . . . . . . . . . . . . . . . . . . . . . . . . . . . ........c....o.”0”..o”.. .“.”””. Inadequate Limited

Chloroprene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....”’”””..”..””.”””.. Inadequate Inadequate

8/14/2019 8104


—.


Table 30.—Chemicals and Industrial Processes Evaluated for Human Carcinogenicity by the InternationalAgency for Research on Cancer (IARC) (Continued)

Chemicals and processes that could not be classified as to their carcinogenicityDegree of evidence

for humans—continued ExperimentalHumans animals

Dichlorodiphenyl trichloroethane (DDT). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Dieldrin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Epichlorohydrin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... , . . . . . . . . . . . . . . . . . . . .

Hematite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Hexachlorocyclohexane (technical grade HCH/lindane). . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Isoniazid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Isopropyl oils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Lead and certain Iead compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Phenobarbitone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .N-phenyl-2-naphthylamine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Phenytoin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Reserpine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Styrene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Trichloroethylene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Tris(aziridinyl)-para-benzoquinone (triaziquone) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

InadequateInadequateInadequate

InadequateInadequateInadequateInadequateInadequate

LimitedInadequateLimitedInadequateInadequateInadequateInadequate

LimitedLimitedLimited

NegativeLimitedLimitedInadequateSufficient (forsome solublesalts)LimitedInadequateLimitedInadequateLimitedLimitedLimited

Chemicals and processes for which human data are available, but which were not considered by the IARC Working Group

Ortho-and para-dichlorobenzene

DichlorobenzidinePhenylbutazone2,3,7,8-tetrachlorodiberzo-para-dioxin (TCDD, the ’’dioxin” of Agent Orange)Ortho-and para-toluidineVinylidene chloride

SOURCE: Office of Technology Ajsessmen~ adapted from IARC (185)

SOURCES OF EPIDEMIOLOGICALLY USEFUL DATA

Three major types of information are u sefulin assessing the carcinogenic risk of a substance:l) health status of exposed and unexposed pop-ulations; 2) exposure data; and 3) physical,

chemical, and biological properties of the sub-stance. Information related to each of these cate-gories can come from a variety of sources andcan be used in different w ays. Testing the sub-stance generates information about its potentialhazard , but information about its distribution inthe environment and any impacts on humanhealth are n ecessary to describe its hu man risk

Health Status Information

DHH S is pr imarily responsible for adm inis-tering health d ata collection, storage, and anal-

ysis projects. An overview of existing DHHSprograms and other departments’ health datacollection activities can be found in Selected Topics in Federal Health Statistics (283).

National Center for Health Statistics

The National Center for Health Statistics(NCHS) located within the DHHS Office of

Health, Research, Statistics, and Technology,was established to collect and d isseminate d ataon the health of Americans. Since 1960, it hasplayed a major role in the development of na-tional health statistics policy an d program s. TheNCHS Division of Vital Statistics collects infer-mation on natality, mortality, marriage, anddivorce from th e individual states and regions.

(See ch.2 for a discussion of cancer mortalityand incidence statistics.) In addition to vitalstatistics, NCHS conducts several general-pur-pose surveys that provide statistics about thehealth status of the U.S. popu lation. Add itional

information can be obtained from Data Systemsof the National Center for Health Statistics

(253).

8/14/2019 8104


Ch. 4—Methods for Detecting and Identifying Carcinogens q 143

Health Interview Su rvey (HIS) .–HIS is theprincipal source of information on the health of the civilian noninstitutionalized population of the United States. Initiated in 1957, interviewsare cond ucted each week in a pr obability sam-ple of households to provide data on a range of health measures, including the incidence of ill-ness and accidental injuries, the prevalence of diseases and impairments, the extent of disabili-ty, and the use of health care services. Eachyear, approximately 40,000 households contain-ing about 120,000 persons are sampled.

HIS collects information only about condi-tions which respondents are willing to report.The basic questionnaires are similar from yearto year but supplemental questions may beadd ed. In 1978 and 1979 questions on smokingwere added, but these were discontinued in1980.

The NCHS Study of Costs of Environment-Related Health Effects, mandated by Public Law95-623, may u se HIS as a data sour ce (179).

Health and Nutrition Examination Survey(HANES).–HANES, initiated in 1970, is a

modification and expansion of the earlier HealthExamin ation Su rvey (HES). These surveys col-lect and use data from interviews an d p hysicalexaminations to estimate the prevalence of chronic diseases, establish physiological stand-ards for various tests, determine the nutritionalstatus of the population, and assess exposurelevels to certain environmental substances. The

sampling techniques employed provide rep-resentative national data. Two surveys, HANESI (1971-75) and HANES II (1976-79) have beenconducted. Both surveys examined approx-imately 20,000 persons.

HANES is the most extensive national assess-ment of health and nutritional status of theAmerican people. The nutritional component of HANES includes: information on dietary in-take; data from hematologic and biochemicaltests; body m easuremen ts; and chemical exam-ination for various signs of high risks of nutri-tional deficiency. Preliminary findings from theHANES II pesticide monitoring program have

found an ap paren t rise in tissue levels of DDTand PCBs. The implications of the observedlevels are un certain.

HANES surveys might become valuablesources of information for cancer epidemiologyif sufficient resources were available. Because of its representative nature, aggregate data from

the survey can be used to represent “normal” orbackground levels. For example, white cellcount levels determined in HANES I were usedfor comparative purposes in an epidemiologicstudy of laboratory workers exposed to sus-pected toxic chemicals. HAN ES II contains cer-tain information about dietary intake of sub-stances wh ich h ave been associated w ith a lowerrisk of cancer, vitamins A and C, and sub-stances such as fats which are associated withhigher risks.

HANES might be linked with other health

data systems, such as the National Death Index(see below) to facilitate assessment of whetherparticular exposure levels or certain nutritionalstatuses were associated w ith cancer mortality.NCHS, with its HANES capabilities, has beenasked to participate in studies near Love Canal,and to evaluate the health status of certain high-risk industry groups. It was unable to do sobecause of limited resources.

The NCHS overall monitoring survey budgetfor fiscal year 1981 is $28 million. This is a $3million increase over 1980 and includes $1.1million for a special HANES study wh ich w illfocus on Hispanics in selected areas of theUnited States. The study is designed to d escribethe health an d nu tritional status of the Mexican-American, Puerto Rican-American an d Cu ban-American populations. Studies of specificgroups are necessary to acquire data in suf-ficient d etail to describe subgroup s of the pop -ulation which differ from the “average.” Gen-eral national surveys such as HANES I and IIproduce data about the “average” citizen bysampling groups in proportion to their repre-sentation in the total pop ulation, and th is oftenresults in too small a sample size to be useful foridentifiable smaller grou ps.

8/14/2019 8104



As examples of data useful for cancer studies,the HAN ES Hispanic study w ill determine:

q name, date of birth, and social securitynumber recorded in machine readable formfor subsequent use w ith the National DeathIndex;

qhistory of toxic substance exposure;q

nutritional status including dietary inter-views, serum vitamin A levels, prevalenceof vitamin C d eficiencies; and

the quantity and frequency of alcohol con-sumption.

Hospital Discharge Survey .—This surveywas established in 1964 to p rovide r epresenta-tive statistics for the U.S. population dischargedfrom short-term hospitals. The survey collectsinformation on the characteristics of patients,the lengths of stay, diagnosis and su rgical oper-ations, and p atterns of use of care. Comp letionof each medical abstract form is estimated to

take app roximately 5 minu tes. Only short-stayhospitals with six or more beds and with anaverage length of stay of less than 30 days areincluded in th e samp le (177).

Vital Statistics Followback Surveys. —NCHS-conducted m ortality “followback” surveys haveprovided information on possible relations be-tween environmental and lifestyle factors anddeath from cancer. Information is sought aboutdecedents through inquiries addressed to thoseproviding information for the death certificate,such as the m edical certifiers, fun eral directors,

and family members. These surveys are an effi-cient mean s to augment th e routinely reportedinformation contained in the v ital records.

The efficiency of the followback approach foreliciting ad ditional information about d eaths isrelated to the relative rareness of death as anevent in the U.S. pop ulation. About 1 percent of the pop ulation d ies annually; and cancer-relateddeaths are reported for about 20 percent of this1 percent, or a total of 0.2 percent of the totalpopulation. The followback approach permitssampling directly from the file of those deathcertificates of interest in order to supplement

existing information.

Mortality followback surveys were con-du cted in the Un ited States annu ally from 1961

through 1968. They have since been discon-tinued due to inadequate resources, includingpersonnel.

National Death Index (NDI).—Deaths in theUnited States are registered by the States orother death registration areas (e.g., District of Columbia). The records are tran smitted on m i-

crofilms or on m agnetic tape to NCHS for com-pilation. Historically, because there was no inte-gration of records for the country as a wh ole, nomechanism h ad existed at th e national level todetermine if a person had died. In 1981, afterseveral years of planning and preparation, theNDI w ill be placed into op eration to serve thatpurpose. The NDI will code deaths that oc-curred in 1979 and each year thereafter. Al-though there has been discussion of codingdeath s that occurred before 1979, no p lans arenow in place to do so.

NDI, administered by NCHS, is designed toprovide medical and health researchers withprobable fact of death, the death certificatenumber, and the location of the death certifi-cate, when sup plied with a minimu m set of iden-tifiers (generally the person’s name and socialsecurity number or date of birth). The re-searcher then m ay contact the registration areawh ere the p ossible match has occurred to obtainthe death certificate or the required infor-mation.

NDI w ill be of imm ediate use in ongoing long-term stud ies which includ e mor tality. Beebe (23)

described the NDI as the most important recentadvance in making vital statistics accessible toresearchers. NCI’s Surveillance, Epidem iology,and End Results (SEER) program plans to usethe NDI to d etermine deaths of all persons in theSEER registries. This should reduce the numberof people lost to followu p by SEER and provid ebetter information about survival. Currently,deaths of peop le who m oved or cease to partici-pate ar e not alw ays recorded by SEER.

National Cancer Institute

NCI, 1 of 11 research organ ization s of NIH ,receives more than one-fifth of all Federal healthresearch funds (283). NCI operates the SEERprogram, which provides cancer incidence data

8/14/2019 8104


on approximately 10 percent of the population.Additional information on the SEER programcan be found in chapter 2.

Centers for Disease Control

The overall mission of the Centers for DiseaseControl (CDC) (56) is “to p revent u nnecessary

illness and death and to enhance the health of the Am erican p eople. ” CDC serves as a focusfor DHHS efforts in the areas of disease preven-tion and control, environmental health, healthpromotion, and health education.

NIOSH, located within CDC, assists OSHAin establishing w orkplace health stand ard s. Be-tween 1972 and 1974, NIOSH conducted theNational Occup ational Ha zard Survey (NOH S)to provide estimates of the proportion of em-ployees exposed to potential health hazards invarious ind ustries. NOH S estimates of exposure

are often used in assessing risks from occupa-tional carcinogens. NIOSH periodically con-ducts studies to identify the health effects of par-ticular indu strial processes and to determine thehealth experience of selected employee pop-ulations. The National Surveillance Network,which is operated by NIOSH, collects data fromState safety and health inspection p rograms.

Social Security Administration

The Social Security Administration (SSA)collects information on economic and demo-

graph ic data in ad ministering th e social securitysystem. SSA makes available an annual l-per-cent continuous work-history sample (CWHS)to outside users which provides informationabout employment, migration, and earningstatus. Six different types of files, all of whichcontain sex, race, and age d ata, are available tooutside users. For purposes of confidentiality,the employee and employer identification areincluded in scrambled form. The usefulness of the CWH S for ep idemiologic stud ies is limitedby p rivacy constraints of access and other char-acteristics, e.g., only wages UP to the taxable

maximu m are reported.SSA has recently initiated efforts to amass a

10-percent samp le of the w ork force because of the Office of Management and Budget’s (OMB)

need for better estimates of intercensal pop ula-tion. This file would constitute the most detailedinformation on the structure of the labor forceso that employment distributions by sex, race,age, wages, and wage chan ges, work force par-ticipation, industry, and regional migration pat-terns could be analyzed systematically.

The Disability Insurance Fund, managed bySSA, contains information regarding benefitcomputation and actions related to employeeentitlemen t. SSA rou tinely prepa res reports r e-garding specific disease entities and has pub-lished characteristics of workers disabled bycancer (283).

Exposure Data

The principal data deficiencies for assessingcancer risks are inadequate information aboutexposures and lifestyle characteristics. Since

cancer h as a long latent period , relating cancerin today’s population to particular exposuresmight require information from 20 or moreyears ago. Even when information was col-lected, records may have been destroyed beforethey became useful in cancer epidemiologystudies.

As the lead agency for regulating chemicals,EPA adm inisters n um erous exposure-monitor-ing p rograms. Several studies hav e been criticalof EPA’s monitoring data collection efforts, andas a result, EPA established a Depu ty AssistantAdministrators Committee to review and makerecommendations regarding agency monitoringand information management activities (113).Three of the major conclusions found by thecommittee are:

1.

2.

3.

A considerable quantity of collected ambi-ent environmental information has notbeen analyzed or presented to top man-agement.

The most serious problem foun d was thelack of consistent, integrated informationon toxic and hazard ous pollutants.

There is little coord ination betw een EPAoffices focusing on the same area. In addi-tion, there is a lack of comp arability an dsharing of these data.

8/14/2019 8104



Exposure data in the workplace are limitedeven though a number of cancer causing agentshave been identified in the occupational envi-ronment. NIOSH’s National Occupational Haz-ard Survey collects exposure d ata in a sam ple of indu stries, and O SHA requires monitoring for 6of the 20 substances regulated because of car-cinogenicity.

Death certificates generally include questionson the u sual occup ation and industry or bu si-ness of the decedent, However, those questionsare not always answered and there is uncertain-ty abou t the accur acy of the information that isprovided. NCHS, along with NIOSH, is cur-rently assessing the feasibility of using and im-prov ing occupational descriptors on death cer-tificates as a surrogate for exposure infor-mation. Approximately a dozen States nowcode the usu al occupation of the d ecedent, andabout half of these also code the reported usual

industry or business of the decedent. One State,Wisconsin, now publishes such tabulated infor-mation in their annual public health report.Other States, with support from NIOSH, haveexecuted special studies on occupational mor-tality based on data reported on the standarddeath certificate (45,236).

In England, information reported on thedeath certificate has been used as a basis for oc-cupational mortality analyses every 10 yearssince 1851 (with the exception of the war year,1941). In the United States, a study of this typewas condu cted in 1950. It involved coding over300,000 death certificates for the occupationand industry of males aged 20 to 64. The in-formation was used in conjunction with the de-cennial census information for that year to p ro-du ce measures of relative mortality risk associ-ated with occupation and industry of decedents(148,149,150,151,196).

The National Human Adipose Tissue Moni-toring Program and HIS and HA NES, which aredescribed above, are the principal mechanismsfor mon itoring levels of toxics in the bod y. Thisinformation is used not only to determine nor-

mal baseline levels but also to id entify p opu la-tions which may beat high risk.

This assessment did not concentrate on moni-toring methods and programs, but the generalimpression is that data collection efforts are in-complete and th at man y generate data of limitedusefulness. For instance, measurement tech-niques are not always specified for collected ex-posure data and ignorance of the sensitivity of the instrument used makes it difficult to com-

pare measurements from different times andsites. Furthermore, a nondetectable measure-ment does not necessarily indicate that a sub-stance is not present, and may mean only thatthe instru ment wa s not su fficiently sensitive tomeasure it. Such negative readings are often notreported, and w hen they are, they may be mis-leading. The efforts of organizations such as theEPA Comm ittee (113) mentioned above to im -prove collection and analysis of monitoringdata m ight be encouraged.

Chemical Information Systems

The lack of toxicological information aboutmany substances and concern over perceivedtoxic substance problems prom pted Congress toenact more than tw o dozen statutes dealing withtoxics. Many of the statutes delegate informa-tion-gathering functions to Federal regulatoryand research agencies.

Toxic Substances Control Act —TSCA

In 1976, Congress p assed TSCA to stren gthenthe ability of the Federal Government to ac-cumulate information on potentially hazardous

chemicals and better enable the Federal Govern-ment to protect the public from toxic sub-stances. TSCA required the establishment of several new programs at EPA and a completelynew infrastructure had to be organized. Thisnecessitated recruiting an Assistant Admin-istrator and placed a large burden on a smallstaff. Subsequently, the programs have beenslow to get off the ground (141). Recruitmenthas continually lagged beh ind au thorized staff ceilings which may be inadequate to meet ex-pected program needs. EPA estimated that ap-proximately 1,500 peop le were needed in fiscal

year 1979: 382 permanent positions were au-thor ized, 313 were actually filled.

8/14/2019 8104


Ch. 4—Methods for Detecting and ldentifying Carcinogens q 14 7

New Chemicals. —One of TSCA’s primaryobjectives—as stated in the op ening p aragrap hof the Committee on Interstate and ForeignComm erce Report (165), is to p rovide “ for theevaluation of the hazard-causing potential of new chemicals before commercial productionbegins.” TSCA requires man ufacturers and im-porters to notify EPA at least 90 days prior to

the manufacture or import of a new chemicalsubstance by subm itting a p remanufacture n o-tice (PMN). Along with notification, manufac-turers are to provide specific information aboutthe new chemical, includ ing any test data w hichrelate to the effects of the su bstance on h um anhealth or the environment. “New” is synony-mous with not being listed on the inventory of existing chemicals and subject to TSCA’s au-thor ity. In May 1979, EPA (105) issued a stat e-ment of interim policy covering su bmission an dreview of PMNs. Final PMN rules have yet to beissued. In the 2-year period, April 1979 to

March 12, 1981, EPA received 488 PMNs, andabou t 800 are anticipated in fiscal year 1982.

TSCA w as not th e first Federal law to r equirea review of new chemicals entering the market.FIFRA an d the Food, Dru g, and Cosmetics Act,have registration/ certification p rovisions forpesticides and pharmaceuticals respectively.Unlike those laws, TSCA requires neither li-censing nor registration, but only notification of intent to manu facture. As mentioned , the PMNmu st contain any test da ta available to the sub-mitter, but EPA cannot require testing of a new

chemical substance just because it is “new. ” Theact places the burd en of proof on EPA to dem-onstrate that the information available to EPA:

. . . is insufficient to permit a reasoned evalua-tion of the health and environmental effects of achemical substan ce . . . and . . . [that] in theabsence of sufficient information . . . [the sub-stance] may p resent an u nreasonable risk or w illbe produ ced in substantial quantities, and theremay be significant or sub stantial exposure.

If such a finding is made, EPA may issue anord er un der section 5(e) to prohibit or limit themanufacture, processing, distribution, use, or

disposal of the chemical. EPA has p roposed twosuch orders and each time the comp any w ith-drew its notice and decided not to manufacture.

On tw o other occasions notices were w ithdraw nwh en the companies learned ord ers to requiremore information were in preparation.

Lack of regulatory action on a new chemicalby EPA does not imply that the substance is“safe” or has been “approved. ” However, it doesgrant the manufacturer the right to produce and

use the chemical as desired, subject to any otherregulations that m ay be ap plicable. Und er sec-tion 5(a)(2) EPA can issue a “sign ificant new use

ru le” (SNUR) for a chem ical wh en th ere is con-cern that a sp ecific use of the substan ce, otherthan those proposed in the PMN, may p ose anunreasonable risk. Issuance of an SNUR re-quires that persons must notify EPA 90 daysprior to manufacture or processing of a sub-stance for a use subject to a SNUR. ThroughMarch 1981, one chemical specific SNUR hadbeen proposed and EPA was considering SNURson m ore than 40 chem icals. Once a substance is

in production, EPA can require testing u nderTSCA section 4 and / or subm ission of hum anhealth and environmental monitoring d ata un-der a TSCA section 8 rule (see Existing Chem-

icals).

EPA recently pu blished a p olicy statement de-scribing a recommended list of premanufacturetests for new chemical substances (114). Thetests are identical to those u nd er considerationby the Organization for Econom ic Cooperationand Development (OECD). OECD considersthat its base set of tests would generate theminimum amount of information normally suf-ficient to perform an initial hazard assessmentof a chemical (see table 31). EPA is r ecommen d-ing that flexibility be used by m anu facturers totailor this “base set” of tests for the particularchemical substance and intended uses. Use of this base set is voluntary du e to EPA’s lack of statutory authority to require testing of newchemical substances. The estimated costs of thisbase set, should all tests be employed, rangefrom $53,000 to $67,850.

EPA has reported that no toxicity data weresubmitted in 60 percent of the first 199 notices

received (141). In fact, 25 percent of th e no ticescontained n o data on p hysical or chemical prop-erties. No chronic test data have yet to be

8/14/2019 8104



Table 31.—EPA’s Recommended Base Set of DataTo Be Included in Premanufacturing Notices

Type of data Estimated cost

Physical/chemical dataData about 11 characteristics . . . . . . . . $ 3,800

Acute toxicity dataAcute oral toxicity. . . . . . . . . . . . . . . . . . 2,000Acute dermal toxicity . . . . . . . . . . . . . . . 2,800

Acute inhalation toxicity . . . . . . . . . . . . - 3,300Skin irritation. . . . . . . . . . . . . . . . . . . . . . 700Skin sensitization . . . . . . . . . . . . . . . . . . 3,200-6,700Eye irritation (for chemicals showingno skin irritation). . . . . ., . . . . . . . . . . . 450

Repeated dose toxicity data14-28 day-repeated dose test(s) usingprobable route(s) of human exposure. . 10,200-12,800

Mutagenicity dataGene (point) mutation data . . . . . . . . . . 1,350Chromosomal aberration data . . . . . . . 18,000

Ecotoxicity dataData about killing of three lowerorganisms . , . . . . . . . . . . . . . . . . . . . . . . 4,100

Degradation/accumulation data . . . . . . . . 3,100-11,850

SOURCE Office of Technology Assessment, adapted from EPA ( I 14)

presented to EPA in a PMN . In cases where tox-icity data are given they are usu ally limited. Thethen-Assistant Administrator for Toxic Sub-stances has indicated (192) that the lack of testdata:

has placed an extraordinary burd en on. . .EPA’s limited resources . . . . Furthermore, we[EPA] believe that ou r objective w ill not b eachieved un til indu stry assumes m ore of theburden of generating ad equate risk informationand assessing the risk of its products.

Unless add itional information is received withthe not ices, EPA’s reviews will be “based u pon afund amental lack of information and data. Thisin turn m eans that our information will be high-ly u ncertain” (192). In order to evaluate a chem-ical’s carcinogenic potential, EPA staff have hadto rely on structure activity relationships andmu tagenicity data w hen available.

Existing Chemicals, –Sections 4 and 8 of TSCA relate directly to the issue of acquiringadequ ate information for assessing th e carcino-genic risk of existing chemicals.

Section 4 grants EPA the au thority to requ ireindu stry testing of a potentially harm ful chem-ical if available information is insufficient for a

reasoned evaluation of risk and if the substance:1) may present an unr easonable risk or 2) resultin substantial or significant exposure. TSCA es-tablished ITC to recomm end chemicals for test-ing under section 4; ITC and EPA responses to itare described above. To require industry test-ing, EPA must demonstrate that available in-formation is insufficient to conclude that thesubstan ce “pr esents an un reasonable risk, ” yetsupp orts the find ing that the chemical “may pr e-sent an unreasonable risk. ” If available informa-tion were sufficient to show that the chemicalpresents an unreasonable risk, EPA could reg-ulate the substance under section 6 of TSCA.One difficulty EPA faces in using this testing au-thority is how to define “may present an unrea-sonable risk. ”

Section 8 of TSCA required EPA to compileby November 1977 an inventory of chemicalsubstances manufactured in the United States.

An initial inventory was pu blished about 2-1/ 2years late in Jun e 1979 and up dated in July 1980.Information originally requested for the in-ventory was limited and EPA has proposedrules requiring additional information for cer-tain substances. In February 1980, rules re-quiring exposure information on approximately2,300 substances were proposed (108). Addi-tional information-gatherin

grules are scheduled

for proposal in 1981.

Section 8(c) of TSCA also requires manu-facturers, p rocessors, and distributors to notifyEPA of information wh ich reasonably sup ports

the conclusion that a substance “presents a sub-stantial risk” of injury to h ealth or the en viron-ment. In addition, Section 8(c) requires themaintenan ce of records ind icating “significantad verse reactions” alleged to have been causedby a chemical substance. Allegation by employ-ees are to be retained by industry for 30 yearsand all other allegations for 5 years. These rec-ords are to be submitted to EPA up on request,and EPA is investigating mean s of establishingan au tomatic reporting system wh enever a cer-tain nu mber of allegations are received in a 12-month period for the same substance, process,

or discharge (109). Final ru les to imp lement thesignificant adverse reaction reporting require-ment are expected in 1981.

8/14/2019 8104


Ch. 4—Methods for Detecting and ldentifying Carcinogens q 149

Chemical Substances InformationNetwork (CSIN)

In February 1978, under mandate of TSCA,EPA and CEQ established the Toxic SubstancesStrategy Committee (TSSC) to facilitate inter-agency coordination of chemical informationcollection, dissemination, and classification.

The Departments of Health and Human Serv-ices, Agriculture, Commerce, Defense, Energy,Interior, State, and Tran sportation, and OSHA,CPSC, and NSF participate in TSSC. IRLG an dthe DHHS Committee to Coordinate Environ-mental and Related Programs are also members.TSSC has formed a number of committees forspecial tasks, and the data committee recom-mended the development of a broad-based net-work of data systems—CSIN. CSIN wasadopted by TSSC, and if sufficient resourcesand personnel are committed, it should go along way tow ard the goal of providing conveni-ent access to information about chemicals. Itsmaster file will contain all information collectedunder TSCA; a subfile, stripped of confidentialdata about trade secrets, will be available forpublic use. CSIN will identify about 1 millionchemicals and for each one provid e selected r e-search and test data, references in the tox-icologic and biomedical literature, and in-formation about regulations that pertain to thechemical (345). The system is still in the d evel-opmental stages, although some aspects wereexpected to be operational in 1980 and the entiresystem is to be comp leted w ithin a decade.

CSIN w ill not be a single new system; ratherit will incorporate several systems already inuse. To facilitate locating information about asingle substance in m ore than one system, EPAis developing an unambiguous identificationnumber system, which was another recommen-dation of TSSC’s data committee. In this way,information about structure, chemical andphysical properties, production volume, uses,application, distribution, and toxicity, nowstored in different systems, can be linkedtogether. Such systems do not n ecessarily con-

tain information about human health effects,but they can be used in combination with healthinformation systems.

Collection and Coordination ofExposure and Health Data

Congress has responded to concern about thecollection and availability of data for assessingenvironmental health risks. It has mandatedcommissions and studies directed at improving

data collection, storage, and dissemination.The Health Services Research, Health Statis-

tics, and Medical Libraries Act of 1974, PublicLaw 93-353, mandated the U.S. National Com-mittee on Vital and Health Statistics to assistand adv ise PHS with statistical problems bear-ing on h ealth and the d elivery of health serviceswhich are of national interest. A committee rec-ommendation was important to the establish-ment of the N DI.

The 95th Congress passed tw o acts which in-cluded sections pertaining to data collection for

assessing and redu cing cancer r isks. The CleanAir Act Amendments of 1977 (Public Law95-95) established the Task Force on Environ-mental Cancer and Heart and Lung Disease, tofocus efforts by EPA and various branches of DHH S on issues relating to these d iseases. Thetask force is composed of representatives fromEPA, NCI, NHLBI, NIOSH, NIEHS, NCHS,CDC, and FDA (340). The task force was di-rected to recommend , among other things (340):

. . . a comprehensive research program to deter-mine and quantify the relationships between en-vironmen tal pollution and h um an cancer . . .

[and] . . . recommend research and such othermeasures as may be appropriate to prevent orreduce the incidence of environmentally relatedcancer . . . .

Initial efforts were focused on defining prob-lems, categorizing relevant research programs,and exchanging information among memberagencies and other appropriate groups. Activ-ities related to p revention and redu ction of envi-ronmental risks are planned to begin during1981. ,

The task force established several project

groups to address specific areas of interest. Of particular relevance to this assessment is theProject Group on Exposure and Metabolic

8/14/2019 8104


150 . Technologies for Detemining Cancer Risks From the Environment

Mechanisms, established in May 1979. Thegroup is examining th e interrelationship of ex-posure to a toxicant and body u ptake, metabo-lism, and affected tar get organs. It is hoped thatthis information will increase the ability of re-searchers to p red ict a chemical’s potential toxici-ty and establish which symp toms maybe associ-ated w ith trace chemical levels in th e body.

The Project Group on Standardization of Measurements an d Tests is primarily concernedwith obtaining reliable, comparable data on en-vironmental and disease measurements. Thegroup has focused on “potential contributions tothe state of the art of environmental and oc-cupational monitoring and testing that wouldcomplement, rather than duplicate or overlap,the efforts of individuals or agencies” (340).

They identified two problem areas common totask force agen cies:

1. There is a need for better resource alloca-tion to optimize the qu ality of data sinceagencies and laboratories charged withmonitoring activities typically have lim-ited resources.

2. Researchers currently have limited meansof assessing the r elationship and validityof published environmental monitoringdata.

The Project Group on Standardization of Meas-urem ents and Tests expects to develop recom-mend ations in these areas.

The Health Services Research, Health Sta-tistics, and Health Care Technology Act of 1978(Public Law 95-623) mandated the Secretary of DHH S to initiate several efforts relating to th eimpact of the environment on health. He is todevelop a plan for the collection and coord ina-tion of statistical and epidemiological data onthe effects of the environment on health andprepare guidelines for the collection, compila-tion, analysis, publication, and distribution of this information. The law also authorized theSecretary of DHH S to “consult w ith and take in-to consideration any recommendations of theTask Force” in d eveloping the p lan.

NCHS (252) recently pu blished Environmen-

tal Health: A Plan for Collecting and Coordi-

nating Statistical and Epidemiologic Data,which reviews information currently availablefrom Federal data collection systems. A series of recomm endations are m ade to Con gress for cor-recting gap s and deficiencies in en vironmentalhealth data systems. These recommend ationsinclud e the need for priority setting in new datacollection efforts, interagency coordination inthe environmental data collection process,assurance of the quality of data, and linkingdata on the environment and health.

NCHS (252) identified 64 ongoin g systems in18 agencies that gather environmental health-related d ata. At least two-thirds of these collecteither information on cancer incidence/ mortali-ty or data on cancer risk factors. Most of thedata collection systems identified are designedto:

q collect health-related d a t a ,

q

measure environmental pollutants and in-dividual exposures,q test specific interr elationships, orq link d ata on the environment and health.

Linking systems that collect different kinds of data appear to have the greatest utility for as-sessing associations of environment and health.For examp le, the Up grade system, wh ich is con-cerned with water quality and health is a jointeffort of CEQ, EPA, NIOSH, and NCHS, andintegrates data from the Bureau of the Census,mortality data from NCHS, and water qualitydata from EPA and the U.S. Geological Survey.

OTA encountered an example of the often-voiced complaint that data are not in a formeasily accessible and useful to researchers. Na-tional mortality data were necessary to carr y

out the an alyses reported in chap ter 2, but th osedata w ere available in comp uter read able formonly for the years since 1968. OTA had to com-pu terize the da ta back to 1933 to carry ou t theanalyses. The OTA computer tapes of cancermortality data will be made available to re-searchers. These data include deaths by age,race, and sex for selected cancers, for each of the

years 1933 through 1978. All of the data onthese tapes ar e consistent w ith those availableon pap er from N CHS.

8/14/2019 8104


8/14/2019 8104


privacy and confidentiality. Restrictions havebeen tightened considerably in r ecent years bythe Privacy Act of 1974 and the 1976 amend-ments to the Internal Revenu e Code. Althou ghepidemiologists have not been implicated, w ell-publicized breaches of confidentiality and pri-vacy have engendered suspicion among legisla-tors and the public, about possible misuse of easily accessible files. MacMahon (217) hassummarized the epidemiologist’s point of view:

To determine cancer risks among persons ex-posed to particular environmental factors, weneed to be able to link information relating tothe same individual at different times in his lifeand to determine whether an individual exposedin the past is now dead or alive and in w hat stateof health . . . . Maximum confidentiality m eansminimum epidemiologic information and min-imal effectiveness in identifying new cancerhazards. In my opinion, we are well beyond thepoint at which concern for confidentialityseriously impairs the extraction of v a l u a b l e

knowledge, even from routinely collected in-formation. Working as an epidemiologist, onecomes to recognize the readiness with whichmost people, patients or nonpatients, will sup-ply even sen sitive inform ation if they believe thecause is reasonable. Someh ow , the issue of con-fidentiality becomes more difficult when it isinstitutionalized or p oliticized. We mu st attemp tto convince the public’s representatives that areasonable balance can b e achieved.

Other linkages, and th e means to accomp lishthem, have been suggested. One current pro-

posal involves drawing a sample of several mil-lion individuals from those people who filledout the 1980 long census form for futu re match-

SUMMARY

Interest in cancer prevention an d acceptanceof the idea that some substances cause cancerhave spurred developments in methods to deter-mine which substances are associated withcancer.

Epidemiology has played an importan t role inidentifying both carcinogenic substances and ex-posures which are associated with cancer al-though the causative agent may not be know n.

ing to the N DI. This w ould facilitate matchingcause of death with personal characteristicsreported in the census. Another proposal hasbeen made to add cause of death information tothe SSA’s l-percent CWHS. However, since col-lection of cause of death data is not part of themission of SSA, it would presumably requiremoney from outside SSA for imp lementation.

The cited desirabilityof making records m ore

available for research purposes runs into con-flicts with society’s intention of protecting theindividu al’s p rivacy. The linking of individu alrecords between agencies would allow a personwith access to the system to obtain most or allGovernment-held information about any indi-vidual. The potential for abuse is apparent. Atthe same time, linkage, in the hands of a bio-medical researcher, might quickly provide in-formation about behaviors, exposures, an d

health that could be obtained only with great

difficulty in any other system.

The Workgroup on Records and Privacy of the Interagency Task Force on the Health Effectsof Ionizing Radiation (182a) has ad dressed theproblems arising from the conflict betw een ac-cess to records for research and the right toprivacy. It suggested changes in the PrivacyAct, the Tax Reform Act, and commented onpending bills (in 1979) that would have per-mitted access to records u nd er tightly controlledconditions. The w orkgroup’s report is an excel-lent starting point for discussion of this com-

plicated subject and provides possible directionsfor, Federal efforts.

When available, epidem iologic data abou t can-cer risks are the most convincing. At the sametime, identifying a carcinogen on the basis of human disease and death means that other test-ing methods failed to identify the agent before

hum an illness resulted from it.The most important laboratory method for

determining carcinogencit y is the long-term bio-assay in laboratory an imals, generally rats and

8/14/2019 8104


mice. The 1960’s saw marked increased interest

in the test, and N CI has played a m ajor role indesigning appropriate test methods. NCI devel-oped a large-scale bioassay p rogram for testing,evaluating, and documenting the carcinoge-nicity of environmental chemicals. It has alsosupported the IARC program to review infor-

mation on environmental carcinogens and pub-lish findings in its authoritative monographseries.

While the bioassay has been improved and iswidely accepted as an appropriate method toidentify carcinogens, it is expensive (each onecosts $400,000 to $1 million) and time-con-

suming (each takes from a minimum of morethan 2 years to a mor e realistic 5 years). Becauseof those costs and limited long-term bioassaycapacity, other tests are being d eveloped.

The short-term tests measure biological ef-

fects other than carcinogenicity (often mutagen-icity) in bacteria, yeast, cultured mammaliancells, or in intact lower organisms. Major inter-national efforts have been completed an d othersare ongoing to validate th e ability of these teststo identify correctly carcinogens and noncar-cinogens. Some tests perform well in the vali-dation experiments, and it is expected that theywill play an increasingly imp ortant role in iden-tifying carcinogens.

Deciding to test a chemical in a short-termtest or a long-term bioassay involves studying

its structure. Some classes of chemicals are morelikely to be carcinogenic than others, some arepresent in high concentrations in the environ-ment, and some are viewed with suspicion forone reason or anoth er. Based on such informa-

tion a chemical may be selected for testing. NTPis developing a tier-testing scheme w hich firstassays a chemical in a number of short-termtests. The chemicals that appear most risky as aresult of short-term tests will be accorded priori-ty for further testing in long-term bioassays.This NTP program promises to improve both

the use of information from short-term tests andthe usefulness of the bioassay program.

The establishment of NTP, which app ears tobe moving toward an ordered, careful devel-opment of methods for testing and interpreta-tion of results, and IRLG’s appearing as a singlevoice for Federal regulatory procedures anddecisions promise further improvements. Addi-tionally, IARC’s distinguishing between non-carcinogens and carcinogens and orderingamong carcinogens on the basis of test results isan imp ortant step in increasing the usefulness of

the results.

Not included in testing systems, but discussedhere are already useful and potentially moreuseful data collection systems. Several of thesesystems were m and ated by Congress to collectinformation about exposures and health status.Currently, inadequate resources and coordina-tion may be hampering the performance of thesystems. Another m ajor sou rce of information,the administrative data systems, were not de-veloped as health information resources. How -ever, both SSA and IRS collect some informa-

tion which might be of great value for epidemi-ologic studies. Perhaps the m ost pressing needin adapting the administrative data systems forthese uses is more consu ltation between epide-miologists and the d ata systems experts.

8/14/2019 8104


5Extrapolation From

Study-Generated Data toEstimates of Human Risk

8/14/2019 8104


Contents

Page

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ............157

Numeric Extrapolation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..........157

The Question of Thresholds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... ...158Nu meri c Extrapolation To Project Risk at D oses Below Th ose Tested . ......160

Quantitative Effects of Selecting a Model . . . . . . . . . . . . . . . . . . . . . ...........162

Virtually Safe Doses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .............163

What Quantitative Projections Can Be Made From Negative Results . ..........164

Other Extrapolation Models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ........164

Extrapolation From Short-Term Tests to Human Risks. . . . ..................165

Carcinogen ic Activity In dicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .......166

Potency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .........................166

The ED 01 Experiment and Extrapolation Models . . . . . . . . . . . . . ..............167

Biologic Extrapolation From Animal Tests . . . . . . . . . . . . . . .................169

Comparing M easured Hu man Cancer Incidence and Mortality to Estimates

Made Using Extrapolation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... ... .170Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ................172

LIST OF TABLES

Table No, Page

32. Expected Incidence of Tumor s Calculated by Three Mod els When a Doseof l.O Caused Tumors in 50 Percent of the Tested Animals. . . . . . . . . . .. ....163

33. Relative H um an Risk Depend ing on H ow Dose Rate is Scaled FromExperimental Animals to Humans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....170

34. Comparison of Tumor Rates in Laboratory Tes tAnimals and HumansFollowing Lifetime Exposures to Comparable Amounts of Each of Six Agents .171

LIST OF FIGURES

Figure No. Page

21.22.23.24.25.26.

A Stylized Dose-Response Curve and Some Extrapolated Curves . .........158Proportion of Mice With Liver Tumors v. Dose . . . . . . . . . . . . . ....,......167Proportion of Mice With B1adder Tumors v. Dose . . . . . . . . . . . . .,........167Proportion of Mice With Bladder Tumors v. Dose . . . . . . . . . . . ...........168Proportion of Mice With Liver Tumors v. Dose . . . . . . . . . . . .............168Prop ortion of M ice With Bladd er Tu mors v. D ose . . . . . . . . . . . . . . . . . . . . .. 168

8/14/2019 8104


5

Extrapolation From

Study-Generated Data to

Estimates of Human Risk

INTRODUCTION

Information abou t carcinogenicity is obtainedfrom exposing animals to measured doses of suspect substances in the laboratory or bystud ying associations between exposures to sus-pect carcinogens and developm ent of cancer inhumans. In practice, both the animals and cer-tain groups of humans, particularly those ex-posed in the w orkplace, are exposed to d oses far

larger than those encountered by most citizens.A nu mber of “num eric extrapolation” m ethodshave been developed to estimate the effect of ex-posu re to low d oses based on observed effects athigh doses. When information about carcino-genicity is obtained from anim als, “biologic ex-trapolation” techniques are employed to projectfrom animal results to estimates of human risk.(In this report the word “hazard” is applied to asubstance or exposure that harbors a “risk” topeop le wh o come in contact with it—i.e., a car-cinogenic chemical is itself a hazard . Risk is theprobab ility of cancer developing as a result of a

particular exposure to the hazard. )

NUMERIC EXTRAPOLATION

This discussion describes information thatcan be obtained from extrapolation, and com-ments on various extrapolation methods. It isneither rigorous nor inclusive, and the inter-ested (and mathematically sophisticated) readeris referred to Heel et al. (170), Crum p et al. (75)and the Food Safety Council (FSC) (125) forsuch treatments.

Toxicity testing produces data relating tumorincidence (I) to dosage (D) as shown in figure21. Generally, a smooth curve drawn between

In most cases, no adequate human data areavailable for estimating risks, and it is necessaryto make both numeric and biologic extrapola-tion from measur ed resp onses in an imals to esti-mate human risk. Less uncertainty attends mak-ing numeric extrapolations from observation of human responses at high-exposure levels thanextrapolations from animal data. However,

such extrapolations are often complicated bypoor exposure data.

Extrapolation, like testing, is employed tomake regulatory decisions, and as in the case of testing, a nu mber of agencies have m ade state-ments about the methods they will employ.These policy statements are necessary to ex-plain, in the absence of agreed-on un iversal pro-cedures, how the agencies intend to use bioassayand epidemiologic data to estimate risk throughthe use of extrapolation m ethods.

the experimental points, P 1-P 5, (solid line infigure 21) is sigmoidal, or S-shaped. It can beseen that the incidence of tumor formation de-creases with decreasing dosage. The crux of theextrapolation problem is w hat sort of line bestapproximates the response in the region forwhich data are not available. Or, what kind of line should be d rawn from point P1 to lower, un-

measured, response levels.

Grap hic representations such as figure 21 donot fully show the d ifficulties in estimating in-

157

8/14/2019 8104


8/14/2019 8104


Ch. 5–Extrapolation From Study-Generated Data to Estimates of Human Risk q 159

Institutes of Health, private testing laboratories,and industry. The conclusion of the article, ap-parently shared by the interviewed scientistsand administrators, is that resolution of thethreshold qu estion is not now available:

. . . it is extremely unlikely that it would bepossible to distinguish between a linear dose-response curve and a highly nonlinear one(threshold), even in a large-scale experiment in-volving several thousand animals per dose level.

. . . statistical analysis of standard animalcarcinogenicity experiments, Schneiderman[then-Associate Director of the N ational CancerInstitute (NCI)] concludes, does not now, andprobably never w ill, resolve the threshold qu es-tion, There are, he says, simply too many “bio-logically reasonable” mathematical models,both implying and denying the existence of thresholds, that will fit the observed results.

. . . there is so little data and so many inter-pretations, Gehring [Dow Chemical Co. ] says,

arguing about thresholds is an exercise in fu-tility.

The view that thresholds cannot be demon-strated is accepted in publications of the N a-tional Research Council (NRC) (262), Interagen-

cy Regulatory Liaison Group (IRLG) (180), FSC(125), and, for tumors resulting from somaticmutations, by the American Industrial HealthCouncil (AIHC) (8). Despite the apparent agree-ment that it is impossible to demonstrate athreshold, man y individu als object to the ideathat thresholds should not be considered inmaking decisions about carcinogens.

In particular, some tum ors are thought to re-sult from irritation or mechanical injury, andthreshold models are postulated for those. Anoften mentioned example is Clayson’s (60) find-ing that bladd er stones are found in conjun ctionwith some bladder tumors. Stone formationmay be d epend ent on intake of large quan titiesof a chemical; if exposure to the chemical is suf-ficiently low so that no stones are formed and if stone formation is necessary for carcinoge-nicity, then a threshold should exist. This andother examples of “epigenetic” tumors can be

considered separately from tumors that orig-inate from somatic mutations. Just as thresholdsfor acute toxic responses differ, thresholds for

stone formation most likely differ am ong ind i-vidual animals. Therefore, it may be impossibleto determ ine exactly wha t d ose is necessary toproduce a stone and therefore pave the way to atum or. Nevertheless, if the threshold for stoneformation in animals is found to be 100 or 1,000

times greater than human exposure levels, theobservation takes on great significance. If stone

formation is a necessary precursor to tumor ap-pearan ce, and no stone formation occurs at hu -man exposure levels, it becomes difficult tomaintain the position that human exposure lev-els presen t a carcinogen ic risk. The d ifficulties inthis argum ent are the possibility that some hu -mans m ay produ ce stones at very much lowerdoses than animals and that carcinogenicitymight not proceed through stone formation inhumans. Such caveats seem reasonable to somepeople; unreasonable to others. Experimenta-tion and m ore sophisticated models may even-tually settle such questions. However, for the

present, positions on these issues reflect orga-nizational policy and ind ividual jud gment.

Gehring, Watanabe, and Young (139) de-scribe differences between metabolism of chem-icals administered at low and high d oses. Meas-urem ent of these differences is called p harm a-cokinetics. In gen eral, biochem ical mechanismsto detoxify chemicals and to repair damage toDNA are seen as having a better chance of de-toxifying and repairing damage at low dosesthan at high doses. High doses can swamp de-fense mechanisms resulting in toxic effects;

lower doses are seen as presenting little or norisk.

Metabolic differences measured by pharma-cokinetics might have im portan t effects on theshape of the dose-response curve. If a metabolizeof an ingested substance is the actual carcino-gen, and the production of the metabolize in-creases out of proportion to dose above a cer-tain level, then the d ose-response curve m ightbend upward at that point. Pharmacokineticdata ar e not always collected, and the standa rdbioassay wh ich prod uces data at only two dose

levels does not provide sufficient information tosee if differences in metabolism might be impor-tant in th e slope of the dose-response curve.

8/14/2019 8104



Biochemistry of a particular compou nd maybe affected by other compounds present in theorganism. Cancer is a relatively common dis-ease, and if avoiding cancer involves biochem-ical detoxification processes, it may be thatsmall doses of many substances can swamp theprocess as well as large doses of a single sub-stance. If this is correct, the ad dition of the car-

cinogens, even at low d oses, might be enough toovercome the detoxification mechanism.

Cornfield (67) described quantitative extrap-olation m odels to take into account differencesin metabolism at high and low doses. His paperwas criticized because it postulated a threshold(see 36,73,229,276,.319), and his hope that thepaper would lead to discussion of the merit of his models app arently was not realized (68). Thedifficulty with models which propose detoxi-fication activities for producing thresholds isthat th e detoxification has to be instantaneous

and complete. Otherwise molecules mightescape detoxification and initiate a carcinogenicevent.

Donald Kenned y, a few m onths after leavingthe post of Commissioner of the Food and DrugAdministration (FDA), expressed his opinionthat thresholds may exist for some chemicals:

Dr. Kennedy said that the [Delaney] clause inthe FDA’s aut horizing legislation codifies the hy-pothesis that th ere is no th reshold concentrationbelow which a chemical does not cause cancer.And although this hypothesis “probably holdsmost of the time, ” Dr. Kennedy said th at he w as“as certain as I can be of an y scientific pred ictionthat some day, very soon, some compoun d w illbe demonstrated to have a threshold level forcancer causation” . . . (274).

It is difficult, if not impossible, to marshalmore evidence on one side of the threshold ques-tion than on the other. The ascendancy of themore conservative view, that thresholds cannotbe identified for human populations, can betaken as a p olicy decision mad e in the interest of protecting the public health. Such a general pol-icy can not exclude the possibility that a thresh-

old may someday be demonstrated.

Numeric Extrapolation To Project Riskat Doses Below Those Tested

The shape of the line in figure 21 dep ends onthe num ber of tum ors observed at p oints P1-P5.No matter what method is used to extend theline below P1, that extension r epresents an esti-mate. Any number of smooth curves can be

draw n from point P1 to the origin; for conveni-ence, the p ossible lines w ill be d ivided into threefamilies: supralinear, linear, and infralinear. Adetailed d iscussion of these mod els as they relateto radiation and cancer is available in a paperby Sinclair (329).

Sup ralinear Extrapolation

A supralinear extrapolation is presented onfigure 21. It says that some doses less than IDare relatively more effective in inducing tumorsthan doses equal to ID. Conceptually the con-

tention that lower d oses are more carcinogenicis easy to add ress. Further tests at lower d oseswould resolve the question, but additional testsare costly and time consuming.

Supralinear models are considered for tworeasons. In several NCI bioassays, the tumoryield was lower at the high dose than at the lowdose (102). In other words, the lower dose wasmore efficient at producing tumors. The ex-planation is that the higher dose was so toxicthat it killed animals before they developedtumors. The other reason for considering supra-linear responses is that some stud ies of

radiation-induced cancer have been interpretedas prod ucing supralinear dose-response curves(e.g., 241), but those interpretations are hotlydisputed.

Such respon ses might result from th e presenceof a subpopulation of more sensitive individ-uals. On figure 21, the sup ralinear response be-tween the origin and P1 represents the tumorsindu ced in the p roposed sensitive fraction of thepop ulation; the solid line dr awn from the originto P5 represents th e sensitivity of the remainingmem bers of the p opu lation. The difference be-

tween the two lines between the origin and P 1

8/14/2019 8104



represents the contribution of the sensitive sub-

population to the total response at doses below

ID. It can be seen that the sensitive subpopu-lation accoun ts for the majority of tum ors thatoccur below P1. Clearly, if this model describes

risks, reducing doses to 1/ 2D or l/ 4 D would not

significantly reduce tumor incidence, and a

supralinear dose-response curve would forcelowering doses to very small fractions of D to

significantly reduce tumor incidence.

Nonartifactual , supralinear dose responses

have rarely been observed in bioassays butneither would they be expected. Laboratory

animals are highly inbred and each animalshould be more nearly equally sensitive than aremembers of human populations. Supralinearresponse models have been advan ced bu t do notnow receive the acceptance accorded to theother two general models.

Linear Extrapolation

A linear model is show n by the straight linethat extends from P1 to the origin on figure 21. If the true dose-response curve is represented bythe solid curved line from P 1 to the origin, thenthe linear model is “conservative” and over-estimates the num ber of tum ors at all doses be-tween P1 and the origin.

The paper by Crump et al. (75) is an often-cited and important argument for linear dose re-sponses at low doses. The paper points out that

25 percent of the U.S. pop ulation w ill developcancer as a result of existing carcinogenic in-fluences (see ch.3). Crump et al. (75) p r o p o s e

that any new carcinogenic substance interacts

addit ively with exposures and behaviors al-ready p resent in the environment. Their mathe-matical theories predict that regardless of theshap e of the dose-response curve at high expo-sures, at low d oses cancer incidence shou ld beproportional (linear) with exposure to the sub-stance under study.

Gaylor and Kodell (137) argue that no risk

estimate can be very reliable for doses belowthat associated w ith the lowest data p oint (Pl infigure 21) because there is no information avail-able below point P1. They propose the use of “linear interpolation” and the 95-percent upper

confidence level to estimate the maximum risk posed by a su bstance.

Error is associated with any experimental de-termination, and standard methods can be usedto calculate “confidence limits” for each es-timate. Usually “95-percent confidence limits”are calculated for carcinogenicity exper iments;they are plotted as vertical bars extending fromthe data points, as shown on the figure. The 95-

percent confidence limit says th at given th e ex-perimentally determined incidence and the sizeof the experiment, we can be 95-percent certainthat the actual incidence represented by thepoint estimate lies inside the error limits.

In the method of Gaylor and KodelI (137) a

line is drawn from the upper limit of the errorbar on the lowest data point to the origin. In -spection of figure 21 shows that this methodprojects a larger risk than does linear inter-

polation from the point P 1 to the origin. This isnot an estimate of risk; it is an estimate of theup per bound of risk.

Objections to including upper confidencelevels in extrapolation are frequently voiced.The practice of including them is seen as in-trodu cing a “safety factor. ” Indu stry spokesmen(9a) and others contend that the best risk es-timate should be made and the safety factorsadded after the estimate is made.

As a p ractical matter , there is often n o alter-native to the linear model. The dose-response

curve in figure 21 is an outrageous overstate-ment of the data that are generally available.

Bioassays carried out according to NCI’s cancer

testing guidelines (331) produce only two datapoints. The Environmental Protection Agency’s(EPA) (102) analysis of many such tests showedthat tumor incidence was sometimes higher andsometimes only measu rable at the lower of thetwo doses because other toxic effects killedanimals at the higher dose. Left with only theresponse at the lower dose, there is little choiceavailable bu t to estimate respon ses at still lowerdoses on the basis of simple proportionality.

Such calculations prod uce a straight line fromthe experimental point to the origin.

IRLG (180) did not discuss how to extrapolate

when more than one data point is available; it

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104



WHAT QUANTITATIVE PROJECTIONS CAN BE MADEFROM NEGATIVE RESULTS (ZEROIN A TEST POPULATION)?

No tumors occurring in a test population of 100 animals or no excess tumors among 100 ani-

mals as compared with the number of tumors in100 control animals does not show that zerocancer risk is associated with the chemical. In-stead stand ard statistical calculations based onzero excess tumors in 100 animals show th at w ecan be 95-percent confiden t that the actual in-ciden ce of tum ors is no more than 4.5 percent.This estimate of the incidence of tumors thatmight have gone undetected is called the upperconfidence limit. The percentage can be reducedby testing more animals, for instance, findingzero excess tumors in 1,000 animals would meanthat we can be 95-percent confident that the ac-

tual incidence is no m ore than 0.45 percent.Proceeding with the illustration of zero excess

tum ors in 100 animals, assume that the dose ad-ministered to the animals was 1,000 times higherthan th at to which hum ans are exposed. Linear

The supra linear, linear, and infralinear mod -els are all dichotomous. They compare th e num -ber of tumors or tumor-bearing animals in the

exposed population to the number in the con-trols. In both populations, the analysis dependson the presence or absence of tumors. Othermod els can be u sed to make inferences about thetimes (or ages) at which animals develop tu morsin response to exposures.

Two of these models have been used exten-sively to describe animal and hum an “time-to-tumor” data. The lognormal model described inChand and Hoe] (57), predicts that the averagetime-to-tumor is longer at low d oses. An impor -tant outcome of this model is that at sufficientlylow d oses, the time necessary for tu mor d evel-

opment may exceed the expected lifespan. Sucha long latent period would produce a “practicalthreshold.” The Weibull model (also described

EXCESS TUMORS

extrapolation (proportionality)posure of the U.S. popu lation

pred icts that ex-to that chem ical

at 1/ 1,000 the level fed to) the animals will resultin fewer than 10,000 cases of cancer assumingequal sensitivity between m an an d animals. Theestimated risk would be reduced if the exper-iment on which it is based is more sensitive. Forinstance, finding zero excess tumors in 1,000

animals (instead of 100) would reduce the esti-mated risk to fewer than 1,000 cases.

The statistics of the above exercise are notquestioned, but they are seldom applied.Although a test cannot show that a substancepresents n o risk, mu ch less concern is attachedto substances that cause no excess of tumors.

The risk associated with substances that arenegative in bioassays is qualitatively low er, andthe consideration of qu antitative risk estimatesfrom negative experiments is of minor im-portance.

OTHER EXTRAPOLATION MODELS

in 57) predicts that the average time-to-tumor isnearly independent of dose. This predictionmeans that an increase in dose simply causes

more cancers; it d oes not shift the age d istribu-tion at wh ich they occur. The assum ptions andpredictions of these two models are quite dif-ferent. Unfortuna tely both are apt to give ad e-quate fits to any available data set, making itdifficult to reject one in favor of the other. IRLG(181) concluded that th ese mod els have not r e-ceived the attention that has been focused on d i-chotomous d ose- incidence relation mod els, andit recommends more research be directed to-ward exploring them. The mathematics forthese mod els is sophisticated, and the interestedreader is referred to Chand and Heel (57).

At this time, discussion of other models ismore an academic than a policy exercise. Op-posing camps ar e for or against quantitative ex-

8/14/2019 8104


Ch. 5—Extrapolation From Study-Generated Data to Estimates of Human Risk q165

trapolation, and among those favoring it, the cl). There is no agreement about an other singleargum ent is between those for and against using model being offered as an alternative at thislinear extrapolation (includ ing EPA’s new mod - time.

EXTRAPOLATION FROM SHORT-TERM TESTS TO HUMAN RISKS

McCann et al. (227) and McCann and Ames(226) showed that about 90 percent of the car-

cinogen s tested in the Ames test were m utagen ic(see ch. 4). Meselson and Russell (235) devel-oped a model to compare the mutagenic andcarcinogenic potency of tested chem icals. Four -teen chemicals were analyzed because therewere sufficient m utagenic and carcinogenic dataavailable to construct d ose-response curves foreach. The correlation between animal carcino-genicity and bacterial mutagenicity was ex-cellent for 10 of the 14 compounds. The otherfour comp ound s (all nitroso-comp ound s) were

more p otent as carcinogens than as mu tagens.

A spirited exchange of views resulted fromthe suggestion that quantitative relationshipsexist between m utagenicity in the Ames test andcarcinogenicity in animal tests. Ashby andStyles (18) challenged the idea that such rela-tionships were common, and Ames and Hooper(12) respond ed th at they w ere.

The International Program for the Evaluationof Short-Term Tests for Carcinogenicity (188)

distributed 42 chemicals to each of 12 lab-

oratories for testing in the Ames system: To

eliminate bias, none of the laboratories knewthe identity of the chemicals. There was ex-cellent agreement among test results obtained indifferent laboratories and about 80 percent of

the carcinogens were scored as mutagens and

about 80 percent of the noncarcinogens were

scored as nonmutagens. These numbers com-pare well with those in the literature that de-scribe results from experiments in which the in-vestigators knew before the mutagenicity testwas ru n that th e chemicals were or were not car-

cinogenic. Although the 80-percent is lowerthan the 90-percent accuracy sometimes re-ported, the program included a number of chemicals that are known to present difficultiesfor the Ames test. In a qualitative sense, the testperformed very well, but m utagenic potency didnot correlate with carcinogenic potency. Inother words, the results from this program donot sup port the idea that th ere is a quan titativerelationship between Ames test results and car-cinogenicity in animals. Similar results showinggood qualitative and poor quantitative agree-ments betw een mu tagenicity and carcinogenic-

ity were reported by Bartsch et al. (22).

Meselson and Russell (235) reported that suf-ficient qu antitative data w ere available for twohuman carcinogens, aflatoxin B and cigarettesmoke, to allow comparison of mutagenic po-tency to carcinogenic potency in hu man s. Cor-relation betw een carcinogenicity in hu man s andmutagenicity was good for the two compounds,and the Meselson and Russell paper raised thepossibility that human cancer risks might bepredicted from mutagenicity data. Acceptanceof such a pr ocedu re is far away and will depend

on much more data being available to supportthe prop osed quan titative relationships.

Clearly, controversies now exist about thevalue of extrapolations made from short-termtests. It seems r easonable that, as u se of short-term tests increases, such projections are goingto be mad e, but some initial optimism about thevalue of quan titative extrapolation from short-term tests to carcinogenicity is apparentlyfading.

8/14/2019 8104



CARCINOGENIC ACTIVITY INDICATORS

The NRC Pesticide Com mittee (267) recom-mends calculating potency expressions, “Car-cinogenic Activity Indicators” (CAI), for testedchemicals. For each point of a dose responsecurve (fig. 21), the number of chemical mole-cules ingested divided by the animal bodyweight can be related to the excess percentage of tumor-bearing animals in the exposed popula-tion.

excess percentage of subjects in w hich tum ors tumor

CAI = are observed = incidence

lifetime dose dose(molecules/ kg of

body w eight)

CAIs do not have to be based on total tumors,for ins tance, s i te specif ic tumors may becoun ted, the analysis may b e limited to on e sex,

only malignant tumors may be counted, orother alterations can be made as wanted. Inpractice, the number of molecules of differentsubstances, S1, S2, and S3 required to inducethe same percentage of excess tumors can becompared to d etermine which is the most potentcarcinogen. CAIs will probably be different foreach point on a dose--response curve because the

POTENCY

Ames et al. (14) have analyzed more than

l,500 bioassays carr ied ou t on som e 600 chem i-cals. For each experiment, they have calculateda p otency index, TD50, which is calculated as thetotal dose of substance necessary to producetumors in 50 percent of the animals. They ex-pect to compare p otency:

1.

2.3.

4.

5.

among m ultiple tests run on the same sub-stance in the same strain and species;between male and female animals;between d ifferent strains of the same an i-mal;between different sites in different ani-

mals; andbetween rod ent tests and 26 tests that havebeen carried out: in monkeys.

points seldom fall in a straight line, but com-parisons can be mad e at comp arable doses.

Using information abou t exposure levels forhuman populations, the number of moleculesthat compare to human exposures can be calcu-

lated. Linear extrapolation is then to be used toestimate the animal response at hu man exposur elevel. This method is especially app ropriate forcomparing chemicals with similar uses, and if applied, would assure that the m ost and leastrisky ones based on animal data a re identified.

The procedure does not make predictions of human cancer risks from animal data andavoids th e p roblems associated w ith biologic ex-trapolation. The NRC committee (267) urgesthat only epid emiologic data be used to estimatehu man risk; it restricts the use of animal data to

making comparisons of carcinogenicity in ani-mals. (Such an approach is especially attractivewh en d eciding abou t regulating p esticides. Sub-stitutes are often available and CAIs offer amethod to decide on th e less or least risky one.This suggestion correspond s to th e second possi-ble use of extrapolation discussed in Quan-titative Effects of Selecting a Model above. )

The results of this massive project is expected to

provide much information about biologic ex-trapolation from species to species and aboutpotencies.

Ames et al. (14) mention that preliminaryresults show that th ere is usually less than a ten-fold variation in potencies among rodents. Thislevel of agreement was also found by Crouchand Wilson (72) for most of 70 chemicals testedin both rats and mice by NCI. Crouch and Wil-son included tu mors wh ich w ere not present at“statistically significant” levels in their an alyses.Therefore, some chemicals judged positive in

only one species by NCI (146) have potenciesthat agree w ithin a factor of 10 between r ats andmice although one is statistically significant and

8/14/2019 8104



the other is not. Results showing a substance is fers from species to species. This information

mu ch more potent in one species than in an other will also be imp ortant in efforts to imp rove ex-

will suggest that metabolism of the chem ical dif - trapolation methods.

THE

Thesearch,known

EDO1 EXPERIMENT AND EXTRAPOLATION MODELS

National Center for Toxicological Re-a joint FDA/ EPA laboratory, tested thebladder and liver carcinogen 2-ace-

tylaminofluorene (2-AAF) in approximately24,000 female mice. This experiment was de-signed to study dose responses down to a 1-percent tumor incidence, i.e., the effective dosefor a l-percent (0.01) response (ED01).

Figures 22 and 23 show the results of post-

mortem examination for liver and bladder tu-

mors in animals exposed to 2-AAF for between21.5 an d 28.5 months. The curve that describes

the liver tumors (fig. 22) shows no thresholdandThe(fig.dose

increases almost proportionally to dose.curve that describes the bladder tumors23) gives an impression of a “threshold”

below 60 ppm of 2-AAF. Gaylor (136)

Figure 22.—Proportion (P) of Mice WithLiver Tumors v. Dose (21.5 to 28.5 months)

0.80

0.60

P

0.40

0.20 .

.-.

4 0 80 120 160

2-AAF (ppm)

SOURCE Llttlefleld, et al (213)

Figure 23.— Proportion (P) of Mice With BladderTumors v. Dose (21.5 to 28.5 months)

0.80

0.60

P0.40

D

0.20 -

I

40 80 120 160

2-AAF (ppm)

SOURCE Llttlefleld, et al (213).

ascribes the apparent threshold to a lack of

resolution of the graph for low tu mor rates. Hereplotted these same data on an enlarged scale(figure 24) to show that the num ber of bladdertum ors increases with dose even at lower d oses.This diagram supports the idea that there wasno threshold. However, there is a dramaticchange in slope of the line between 60 and 75ppm, and the efficiency of 2-AAF in causingtum ors increases greatly at doses above 60 ppm(fig. 24).

Carlborg (50) has also analyzed the data fromED01. He does not argue that th e bladder cancer

data show a threshold, but he does contend thatneither the bladd er nor the liver data fit a one-hit (linear) model. In addition to trying to fit aone-hit model to the data, he also tried the

8/14/2019 8104


—


Figure 24.— Proportion (P) of Mice With Bladder

.008

.006

P.004

I

.002

0

Tumors v. Dose (21.5 to 28.5 months)

20 40 60 70 75 a

2-AAF (ppm)

NOTE: The vertical scale has been expanded 100 times compared to figure 23— line fitted to data points between () and 60 ppm

-—- slope of line between 60 and 75 ppm (derived from figure 23)

SOURCE: Gaylor (136) and Office cf Technology Assessment.

Weibull mod el, a “generalization of the one-hitmodel. ” Figu res 25 and 26 repr esent his ana lysisof the two sets of data, and the “best-fittin g one-hit and Weibull models” are plotted in bothfigures. It can be seen that the Weibull modelprovides a better fit, especially with the bladd ertumor data. Carlborg (50) also states that the

one-hit model fits well with only the threelowest dose points from the liver tum or da ta.

The Weibull mod el produ ces a different slopefor the line to be draw n from 30 to O p pm . Carl-

borg calculates that a 0.000045 ppm dose of 2-

AAF is necessary to produce a one in a million

risk of liver cancer using the one-hit model.

Using the Weibull model, the calculated dose for

a one in a million risk is 100 times higher,

0.0045 ppm. In other words, use of the Weibullmodel wou ld allow exposures 100 times higherthan the linear model if there were agreement

that a one in a million risk were acceptable.

Despite the large number of data points fromED 01, Gaylor (136) argues that a “best” model

Figure 25.—Proportion (P) of Mice WithLiver Tumors v. Dose (24 months)

.5 0 - w

Welbull

.40 -

P.30 -.20 -

.10 -

.

0 20 40 60 80 100 120 140 160

2-AAF (ppm)

SOURCE: Carlborg (50)

Figure 26.— Proportion (P) of Mice WithBladder Tumors v. Dose (24 months)

1 .0 ‘ 4

.80 -

.60 -d

P Wdbuli /

.4 0 -“ .,

/

.2 0“

, ,

-+

1 I

o 20 40 60 80 100 120 140 160

2-AAF (ppm)

SOURCE: Carlborg (50)

cannot be chosen on the basis of fit with the dataand that the Weibull model cannot be singledout above all others. Fur therm ore, he says thatseveral models fit the observed data about

equally well, and that they pred ict very d ifferentlevels of risk at d oses between O and 30 pp m 2-

AAF (see discussion of different models in35,49,51,123,170).

8/14/2019 8104


Ch. 5–Extrapolation From Study-Generated Data to Estimates of Human Risk q169

Obviously, animal data can be best usedquantitatively to estimate potential risks withinthe same an imal species und er test cond itions.These estimates cannot be used directly to es-

timate human risks. Other “biologic” assump-tions are necessary to extrap olate from animaldata to estimates of hu man r isk.

BIOLOGIC EXTRAPOLATION FROM ANIMAL TESTS

Bioassay guidelines call for testing in twospecies. Qualitative judgments are based onwhether a tested substance causes tumors inboth, only one, or neither sp ecies. Sometimes, asubstance is reported positive (causes tu mors) inonly one species and negative in another. Thefrequency with which this problem is encoun-tered can be estimated from Griesemer andCueto’s (146) discussion of the results of the NCI

bioassay program (see app. A). Of the 98 chem-icals for which “very strong” or “sufficient”evidence of carcinogenicity was found in either

rats, mice, or both, 54 were positive in only onespecies. An analysis of test results from 250

substances (300) found 38 percent were carcino-genic in neither rats nor mice, 15 percent w erecarcinogenic in either the rat or the mou se butnot both, and 44 percent were carcinogenic inboth. Crouch and Wilson (72) and Ames et al.(14) tend to dismiss these discordan t results be-tween rats and mice because tests analyzed bytheir methods, which consider experimentalerror and test sensitivity, are in agreement m uchmore often.

In cases in which there are apparent speciesdifferences in sensitivity, the positive result isgenerally accepted as m ore impor tant. The Of-fice of Science and Technology Policy (281),

IRLG (180), the Occupational Safety and Health

Administration (279), and American Federationof Labor/ Congress of Indu strial Organizations

(7) al l present arguments for fol lowing this

course. AIHC (8) and Purchase (300) present

arguments against deciding that the more sensi-

tive animal is predictive for human response.

The disagreement continues, but in Federal pr o-grams, extrapolation is based on results fromthe more sensitive species.

Accepting that experimental animals provideappropriate data for extrapolation to estimates

of human risk, a decision has to be mad e abouthow to adjust the dose measured in the bioassayto the dose experienced by humans. A mouse orrat, of course, is much smaller than a human,and the d ose necessary to cause a carcinogenicresponse is less than that required in humans.Three “scaling factors” are in general use tomake allowance for the different sizes and ratesof metabolism between experimental animalsand humans. The three are listed below in orderfrom least conservative, that is the one thatpred icts the lowest hum an risk, to the most con-

servat ive. The four th scaling factor is less oftenused, but it is included because it was used forthe estimates rep orted in table 34. For rats andmice, the m ost comm only used laboratory ani-mals, the relationship between scaling factorsand estimated risk in m an for the same d oses areshown in table 33.

1.

2.

3.

Exposur es may be ad justed on the basis of relative bod y w eights, milligram of agent/ kilogram of body weight/ day (mg/ kg/ day), for animals and humans. This meth-od is most generally used by toxicologists.

In cases where the experimental dose ismeasured as parts p er million in food, air,or water and h uman exposure is throughingestion, the dose of the chemical is ex-pressed as p arts per million. This methodis generally used by FDA and in somecases by EPA.

Exposure may be ad justed on the basis of the relative surface areas of the test animaland hum ans: milligram of agent/ surfacearea/ day, for animals and hum ans. It isgenerally expressed as milligram of sub-

stance/ square meter of surface area/ dayor mg/ m

2 / day. EPA uses this scaling fac-

tor.

BII-4 31 9 - 81 - 12

8/14/2019 8104



Table 33.—Relative Human Risk Depending on How Dose Rate is Scaled FromExperimental Animals to Humans

Risk projected for humans when an identical dose isscaled by different factors

Milligram Parts per million Milligram Milligram

Kilogram KilogramExperimental animal body weight/day In diet m2 body area body weight/lifetime

Mouse 1 6 14 40Rat 1 3 6 35

SOURCE: Off Ice of Technology Assessment.

4. Exposu res may be ad justed on the basis of relative body weight over lifetime, m illi-grams of agent / kilogram of body w eight/ lifetime (mg/ kg/ lifetime).

As can be seen from table 33, the choice of scaling factor can make a difference of up tofortyfold in estimating human risks. The mg/ kg/ day scaling factor wa s arbitrarily set equal to1.0. Use of the mg/ m

2 / day factor (for instance)

projects that humans would have 14 times the

risk of a mouse for equivalent doses measured inmg/ kg/ day. The information given in table 33allows a comp arison to be mad e among th e scal-ing factors. However, it is imp ortant to rem em-ber that great uncertainties surround biologicextrapolation because of possible differences be-tween laboratory animals and man, and nogreat assurance is attached to any number intable 33.

COMPARING MEASURED HUMAN CANCER INCIDENCE ANDMORTALITY TO ESTIMATES MADE USING EXTRAPOLATION

The more troubling and more fundamentalproblem with biologic extrapolation concernsquestions about how closely the test animalresembles humans. This problem is partially re-lated to d ifferences in the g reater genetic com-plexity of human populations. Populations of test animals are highly inbred and are almost ge-

netically identical. Populations of humans areoutbred and includ e greatly differing genotypes.There is no way to deal with the problem of hu man s that m ay d iffer in sensitivities becausethere is seldom, if ever, a way to associate sen-sitivities with individuals. The other problemconcerns differences in metabolism between testanimals and hum ans. Few are w ell und erstood,and m any may be u nidentified.

Laying these problems aside, a few effortshave been made to compare human cancer inci-den ce or mor tality to the levels of incidence or

mortality extrapolated from animal studies. Thenumber of such attempts is limited by the fewcases for wh ich d ata are av ailable both from ex-perimental animals and from humans. The Con-

sultative Panel on Health H azard s of ChemicalPesticides of the National Research Council’sStudy of Pest Control (262), identified sixchemicals for which such comparisons could bemad e. Comparisons were mad e on the basis of lifetime d osage (expressed as m illigram chemi-cal/ kilogram body w eight/ lifetime) in animals

and in hu man s. Table 34 shows those find ings.It can be seen that for three chemicals the in-cidence of human tumors was essentially thatpredicted from animal studies, and for the otherthree, extrapolation from animal data over-estimated human risk as measured by epide-miology. Crouch and Wilson (72) have madesimilar comparisons for 13 chemicals (includingthe 6 in table 34). They reported good agree-ment between predicted and observed humantumors rates, using a linear nonthreshold ex-trapolation model. Crouch and Wilson (72) alsocalculated “scaling factors” from their experi-

ments and conclud ed that hu mans are twice assensitive as mice and between one-third andthree times as sensitive as rats to the same doseexpressed as mg/ kg/ day. These values differ

8/14/2019 8104



Table 34.—Comparison of Tumor Rates in Laboratory Test Animals and HumansFollowing Lifetime Exposures to Comparable Amounts of Each of Six Agents

(comparison based on mg agent/kg body weight /lifetime)

Animal tumor Human tumor Relative tumor

Chemical Test animal site(s) site ra tea

Benzidine Mouse Liver Bladder ca. 1Rat Bladder

Cigarette smoking Mouse Lung Lung ca. 1

Hamster LarynxN, N-bis(2-chIoro-ethyl) -2-naphtyl-amine Mouse Lung Bladder ca. 1

Aflatoxin B, Mouse Liver Liver ca. 10Rat Liver

Diethylstilbestrol(DES) Mouse Mammary Daughters’ re- ca. 50

productive tractMouse Cervix and

vagina

Vinyl chloride Mouse Lung Liver ca. 500Mouse MammaryRat KidneyRat Liver

ar~lativ~ ~U~Or ~at~ = ~mor Incidence predicted from~ost Sensltlve animal sPecles — ————

tumor incidence observed In humans

SOURCE Adapted from National Research Council (262)

from unity by a factor of 3 or less which isassumed for doses scaled on the basis of mg/ kg/ day (see table 33).

A decision about whether the reasonablygood agreement between extrapolated valuesand observations (72,262) are of significance,and whether or not these findings mean extrap-olation is accurate enough for making quanti-

tative decisions about hu man risks dep ends onthe observer. The NRC Committee (262) con-cluded that:

Although there are major uncerta int ies in ex-

t rapola t ing the resul ts of animal tes ts to man,

this is usually the only available method . . .Despite the uncertainties, enough is known toindicate what dependencies on dose and timemay operate and to provide rough p redictions of indu ced cancer rates in human popu lations.

Regulating Pesticides,a report prepared bythe National Academy of Science’s Committeeon Prototype Explicit Analyses for Pesticides(267), says that seven previous National Re-

search Council reports have recommend ed ex-trapolating from animal data to projectedhu man risk, The Committee (267) took a differ-ent position and recommended that only epi-dem iologic data be u sed to estimate hu man r isk:

OPP [Office of Pesticide Programs, EPA]should abandon its attempts to produce numer-ical estimates of the effects of the use of pes-

ticides on human mortality and morbidity ex-cept when reliable human epidemiological dataar e available. In the usual case, in which majorreliance has to be placed on the results of bio-assays, those results should be used to constructindicators of the relative pathological activity of the pesticide under review in comparison withother pesticides and compoun ds.

Documented differences between the m etab-olism of a chemical in test animals and in hu-mans would be very useful in any attempt atbiologic extrapolation. Poor und erstanding of comparative biochemistry hamp ers research inthe basic biology of cancer. As research con-

8/14/2019 8104



tinues, knowledge of metabolic differences be- limited by strong constraints on studying me-

tween animals and humans may provide clearer tabolism of carcinogens in humans.direction, but su ch information w ill always be

SUMMARY

Animal testsdata that relate

or epidemiologic studies yieldcancer (or tumor) incidence to

exposure levels (dosage) of the substance understud y. The accuracy of the relation betw een ex-posur e and incidence is always limited. Practicalrestraints on the num ber of animals that can betested means that the data are always sub ject tosignificant experimental error; it also means thatonly relatively high in cidence almost alwaysgreater than 10 percent, can be m easured in theexperiments. Epidemiologic studies may be lim-ited by small num bers of people available forstudy, or by unknown or uncertain exposure

levels. In all cases, deficiencies in experimen taldesign and execution may furth er limit the ac-curacy of relating incidence to dose.

Quantitative extrapolation begins with theexperimentally d etermined relationship betweenincidence and exposure and may u se one of sev-eral method s to derive an estimate of incidenceat exposure levels likely to be encountered in theenvironment. When animal d ata are used for ex-trapolation one of four scaling factors can beused to extrapolate from animal results to ex-pected human response. The scaling factorsvary some fortyfold in the risk they project for

humans, and agreement has not been reachedabout w hich one is most approp riate.

There is also no agreement about whichmath ematical mod els best extrap olate from th eexposure levels measured in stud ies to those en-countered in the environment. Linear models,wh ich assum e that incidence is prop ortional to

exposure at low-exposure levels, are used byFederal agencies. Some other organizationsfavor nonlinear mod els in w hich estimates of in-cidence d ecrease faster than dose d ecreases. Aspecial feature of some models is the incorpora-tion of a threshold, a low, but n onzero exposurelevel at which the estimated incidence is zero.Nonthreshold models, which are used by Fed-eral agencies, associate some positive estimateof incidence with all doses above zero.

Suggestions are frequently made that carefulinspection of available d ata and testing var iousextrapolation models against them will allowselection of the best m odel. Unfortunately dataare not sufficient to make such judgments.Another method to decide which model is ap-propriate is to make projections from animaldata an d comp are those to observed incidence inhu mans. The cases where hum an d ata are avail-able to make comparisons are few, but the con-ceptu ally simp le, linear, nonthreshold mod el isreported to estimate human incidence reason-ably well.

The increasing importance of short-term testshas led to efforts to extrapolate from them to

estimates of carcinogenic risks in humans oranimals. Qualitatively short-term tests performwell in pred icting wh ether a substance w ill becarcinogenic or noncarcinogenic in an animaltest. Quantitative agreement between muta-genic potency in short-term tests and carcino-genic potency in animal tests for carcinogenicityis not nearly so good.

8/14/2019 8104


6 q

Approaches toRegulating Carcinogens

8/14/2019 8104


8/14/2019 8104


6

Approaches to Regulating Carcinogen;

PUBLIC PERCEPTION ABOUT CANCER AND

THE LEGISLATIVE RESPONSEThe importance of cancer in U.S. policies

about disease is illustrated by the attentionfocused on cancer research. The first institute of the U.S. Public Health Service to be devoted toa single disease was the National Cancer Insti-tute (NCI), established in 1937. Initially a free-standing institute, it was incorporated into theNational Institutes of Health (NIH) w hich wasorganized in the 1940’s, Thirty-four years after

NCI’s establishment, a nearly successful effort

was mounted in Congress, in 1971, to separate

NCI from NIH and to establish a National Can-cer Authority, that would have set cancer fur-ther ap art from other biomed ical research activ-ities. While the National Cancer Act of 1971

was unsu ccessful in establishing a new authori-ty, it elevated NCI to bureau status, a higherorganizational level than all other institutes atNIH until the National Heart, Lung, and BloodInstitute was also made a bureau. The 1971

legislation established a three-person cancerpanel, appointed by and responsible to thePresident; no other d isease has been singled outin such a way. The attention bestowed on can-cer research reflects the importance of the dis-ease to the p ublic. It is the nu mber two killer inthe United States, the num ber one d isease killeramong people younger than 55, and it is themost d readed disease (307).

In the 1960’s and 1970’s, public and congres-sional interest in cancer p revention was sp urredby associations being draw n between environ-mental exposures and cancer. Congressionaltestimony mentioned associations between theenvironment and cancer, and several laws wereenacted to pr ovide Federal agencies with regu la-tory mechanisms to reduce exposures to car-

cinogens.

Public fear and dr ead of cancer is not likely todecrease, and despite the current antiregulatory

mood , Americans still favor health and environ-mental regulations. A survey of 2,000 people,

commissioned by Union Carbide in 1979, foundcontinued public support of Government efforts“to protect individual health and safety and theenvironment.” Seventy percent of those sur-veyed favored stronger measures to protectworkers from cancer; 65 percent favored strong-er measures to protect consumers from cancer(352).

Concern about cancer is likely to provide im-

petus for continued efforts to reduce its in-cidence and to imp rove its treatment. Efforts toimprove treatment are seen as highly desirableand excite little controversy, but efforts toreduce cancer incidence by regulatory interven-tion generate great passion about whether theexpected benefits from the regulations justifytheir costs. As is pointed out in the earlierchapters of this report, some uncertainty isassociated with estimates made of the cancerrisk posed by p articular exposures. Part of thecontroversy about regulatory intervention,whether it is worth it or not, flows from those

uncertainties, but controversy also stems fromthe fact that the regulations bring tw o societalgoals into conflict. The majority of people w antprotection from carcinogenic risks, and at thesame time want to redu ce regulatory costs andburd ens. Choosing between these tw o goals orreaching compromises between them will re-main an important point of contention in pol-icies about the control of cancer.

Several Federal agencies administer regula-tory programs for the control of carcinogenicand other health risks to humans from chemicalsubstances. These programs differ in their objec-tives and regulatory authority. Some were de-signed by Congr ess to deal with several differenttypes of risks, including carcinogenic risks, in

175

8/14/2019 8104



the workplace or in consumer products. Others cerning the existence and magnitude of risk.were designed to protect human s and the envi- This chap ter first d iscusses statutory m and atesronment through control of toxic substances in related to carcinogen regulation, then moves toair, water, and food. an examination of risk assessment issues. A

Regulatory decisionmaking for control of concluding section focuses on the process of

cancer risks to humans is guided by specificmaking regulatory decisions for controlling

legal mandates and administrative procedurescarcinogens.

and depends on technical determinations con-

STATUTORY MANDATES

Regulations of carcinogenic substances aredesigned to reduce health risks. The laws thatrequire such regulations differ in wh ether theydirect regulators to consider only h ealth risks orto consider both health risks and other factors.The other factors to be considered m ay includ ethe costs of reducing the exposure an d the costsof foregone benefits from reduced availability of

the substance.

Table 35 lists 10 laws under which some ac-tion has been or may be taken to reduce ex-posure to carcinogens. Of the applicable stat-utes, the Federal Food, Drug, and Cosmetic Act(FDCA), the Clean Water Act (CWA) of 1977,

and the Toxic Substances Control Act (TSCA)specifically mention carcinogens. The remainingstatutes provide for regulating all toxics, andcarcinogens are included in the more generalterm. The list includes the laws most often dis-cussed in relation to carcinogens bu t not all laws

under which carcinogens might be regulated.For instance, laws govern ing transp ort of haz-ardou s substances might be used to r egulate car-cinogens, and the U.S. Department of Agricul-ture regu lates carcinogens in poultry and meat.

The earliest laws reflecting congressional con-cern about toxics centered on th e food sup ply.Those laws, enacted around the turn of the cen-tury, established the Food and Drug Adminis-tration (FDA). In line with the importance socie-ty attaches to a safe food supply, the first law toapp ly directly to carcinogens was aimed at car-cinogenic food additives. The Delaney clause,

incorporated into the Food Additives Amend-ment of 1958, forbids the incorporation into

food of any ad ditive shown to ind uce cancer inhum ans or other animals.

The late 1960’s and the 1970’s saw the identifi-

cation of carcinogens in various parts of the en-vironment, and Congress provided legislativeauthor ity to regulatory agencies to reduce suchexposures. The number and diversity of laws

produ ces a “balkanized” Federal regulatory ef-fort. Whether or not carcinogen regulationwould be better accomplished under fewer,broader laws is a question worthy of considera-tion, but it is beyond the scope of the presentassessment.

Bases for the Laws

Although many of the laws deal with othertoxics in addition to carcinogens, the discussionhere will focus on carcinogens to the exclusionof other health and environmental risks. The ex-

istence of the laws clearly states that Congresshas seen cancer risks as deserving Governmentattention. At the same time, despite the fact thatsome of the laws are attacked as prop osing anunobtainable risk-free society, Congress hasrecognized that cancer-risk management cansometimes involve balancing and comparing of risks against other societal goals. The 10 laws intable 35 can be d ivided into “ risk-based laws”(or zero-risk laws) w hich allow no balancing of health risks against other factors, “balancinglaws” w hich require balancing of risks againstbenefits of the substance, and “technology-

based laws” wh ich d irect regulatory agen cies toimpose specified levels of control (306).

8/14/2019 8104


Risk-Based Laws

For this discussion, “risk-based” refers to leg-

islation that provides for regulations to reducerisks to zero without considering other factors.

The primary example is the Delaney clausewhich specifies that carcinogenic food additivesare to be eliminated from th e food sup ply. Sec-

tion 112 of the Clean Air Act (CAA) and section307(a)(4) of CWA call for the reduction of ex-posu res to levels which allow an “amp le marginof safety. ” Because Federal agencies do n ot ac-cept threshold levels below w hich carcinogenspose no risk, strict interpretation of “amplemargin of safety” would also require reductions

to zero exposures. The Resource Conservationand Recovery Act (RCRA) is also risk-based,but no regulation about carcinogens has yetbeen issued under it.

When it enacted the Delaney clause, Congresswas aware that over 1,000 substances w ere pres-ent as ad ditives in th e U.S. food su pp ly. Manyof those substances had been poorly tested (if atall) for acute toxicity and hard ly any had beentested for chronic toxicity. Congress recognizedthat some of the additives might pose healthproblems, and the Delaney clause reflects itsconclusion that no benefit could militate againstbanning a carcinogenic add itive.

The most recent application of the Delaneyclause was the proposed removal of saccharinfrom the “generally recognized as safe” (GRAS)list of food additives. This action, based on the

find ing that saccharin causes bladd er cancer inrats, would then have resulted in the ban of sac-charin from use as a food additive (282). Fewpeople qu estion th at saccharin is a rat carcino-gen, and the Delaney clause is clear: Saccharinshould be removed from the market on the basisof the animal study. Peter Hutt, former GeneralCounsel of FDA, points out that a zero-risk ap -proach, such as the Delaney clause, may work reasonably well so long as there are su bstitutes.When no substitute is available, controversyflares, It flamed in the case of saccharin.

Because FDA had banned cyclamates on th ebasis of animal tests in 1969, the banning of sac-

charin would have meant that there would be nononnutritive sweetener on the market. No one

could have the p leasure of sweetness without thecost of calories. Citizens who objected to theban, aided by postage-paid postcards insertedinto cartons of saccharin-sweetened soft drinks,deluged Congress with m ail. Congress delayedthe imp osition of the saccharin ban an d calledfor more studies. In 1980, it continued the

delay.

At the request of Congress, both OTA (282)

and the National Research Council (NRC) (269)reviewed the scientific data and conclusionsabout saccharin and agreed w ith the FDA deci-sion that saccharin is a carcinogen. The con-gressional decision to forestall action on sac-charin can be viewed as the Con gress reopeningdiscussion about its earlier decision that benefitsof carcinogenic food additives could not beweighed against their risks. It is aware of theevidence of risks (from animal studies) andbenefits (from pu blic outcry), and by d elaying

the ban , it is giving weight to the benefits.Whether the saccharin moratorium presages

an eventu al voiding of the Delaney clause is anopen qu estion. Congress may retain the Delan-ey clause but, from time to time, exercise itsprerogative to overru le agency d ecisions w hen itdecides that benefits, whether measured or not,outweigh risks.

Comments are sometimes heard that the De-laney clause was appropriate for the state of knowledge in 1957 when it was enacted, but thattimes have changed. It is said that at that time

there was general agreement that few su bstanceswere carcinogenic and that those few could beeliminated from commerce with little difficulty.However, more and more substances, includingmany useful and some apparently essentialones, have been identified as carcinogens. Somepeop le make a connection betw een these discov-eries and the apparent turning away from r isk-based laws such as the Delaney clause.

Despite the fact that some of the laws writtenin the 1970’s were technology-based or balanc-ing (see below ), section 112 of CAA, 1970, andRCRA, 1976, are risk-based. They direct that

risks to health be reduced or eliminated withoutspecifically calling for consideration of otherfactors, An important consideration in these

8/14/2019 8104


8/14/2019 8104


C/I. 6—Approaches to Regulating Carcinogens q179

Table 35.—Public Laws Providing for the Regulation of Exposures to Carcinogens (Continued)

Definition of toxics or

Legislation hazards used for regula- Agents regulated as carcinogens Basis of the

(Agency) tion of carcinogens Degree of protection (or proposed for regulation) Iegislation Remarks

Resource Conserva-tion and RecoveryAct (EPA)

Safe Drinking WaterAct (EPA)

Toxic SubstancesControl Act (EPA)

Sec. 4 (to requiretesting)

Sec. 6 (to regulate)

Sec. 7 (to com-mence civil actionagainst Imminenthazards)

Federal HazardousSubstances Act

(CPSC)

Consumer Product

Safety Act (CPSC)

One which “may cause, orsignificantly contribute toan Increase in mortality oran increase in serious irre-versible, or Incapacitatingreversible, illness; or, posea. hazard to human

health or the environ.ment. .” sec. 1004(5) (A)(B)

“contaminant(s)which. . may have anadverse effect on thehealth of persons.” sec.1401(1) (B)

substances which “maypresent an unreasonable

risk of injury to health orthe environment. ” sec. 4(a)(1) (A) (i)

substances which “pre-

sent(s) or will present an

unreasonable risk of injuryto health or the environ-ment,” sec. 6(a)

“imminently hazardouschemical substance ormixture means a.substance or mixturewhich presents an immi-nent and unreasonable riskof serious or widespreadinjury to health or the en-vironment. ”

“any substance (other thana radioactive substance)which has the capacity toproduce personal injury ori ll ness “15 USC sec.

“products which present

unreasonable risks of in-jury. in commerce,” and“ ‘risk of Injury’ means arisk of death, personal in-jury or serious or frequentinjury. ” 15 USC sec. 2051

“imminently hazardousconsumer product’ meansconsumer product whichpresents Imminent andunreasonable risk of

death, serious illness orsevere personal in jury.” 15USC sec. 2061

“that necessary to protecthuman health and the envi-ronment. .” sec. 3002-04

“to the extentfeasible. (taking costs in-

to consideration). .” sec.1412(a) (2)

Not specified

74 substances proposed for Iisting ashazardous wastes

Trihalomethanes, chemicals formed byreactions between chlorine used asdisinfectant and organic chemicals.

Two pesticides and 2 metals classifiedas carcinogens by CAG, but regulatedbecause of other toxicities.

Risk. The Admini-strator can ordermonitoring andset standards forsites.

Six chemicals used to make plastics Balancing: "unrea-pliable. sonable risk”

Based on degree of protec-tion in sec. 6

Balancing

“to protect adequately PCBs regulated as directed by the law.

against such risk using the

least burdensome require-ment” sec. 6(a)

“establish such reasonablevariations or additional

label requirements.necessary for the protectionof public health and

safety “15 USC sec.

“standard shall be Five substances: asbestos, benzene,

reasonably necessary to benzidine (and benzidine-based dyesprevent or reduce an and pigments), vinyl chloride, “tris”unreasonable risk of in jury.”15 USC sec. 2056

Balancing. “unrea-sonable risk. ”

Risk “Highly toxic”defined ascapacity tocause death,

thus toxicitymay be limitedto acute toxicity.

Balancing: “unrea- Standards are to

sonable” be expressed,wherever feasi-ble, as perfor-mance require.ments.


8/14/2019 8104



two laws is that they deal with pol lutants .

Pollutants benefit no one, and their reduction or

elimination improves the environment and pub-lic health.

A problem arises because eliminating the pol-

lutants costs money. Furtherm ore, in th e case of carcinogens, the evidence that a substance poses

a risk is not always accepted by everyone. As aresult, the cost of redu cing exposur e to a pollut-ant is often offered as an argument against theprojected health benefits expected from regula-tion, and suggestions are mad e that the costs beconsidered against the benefits before regula-tion, even in cases wh ere the law d oes not callfor such considerations.

Balancing Laws

The “balancing laws, ” such as the ConsumerProduct Safety Act (CPSA), the Federal Insec-ticide, Fungicide, and Rodenticide Act (FIFRA),

and TSCA, pu t a qualifying word , such as “un -reasonable” in front of the word “risk. ” Thisconstruction implies that some risks are to betolerated, and, in practice, means risks from asubstance are to be weighed against other fac-tors in the process of deciding wh ether and howto regulate. TSCA requires that “the benefits

for various uses, . . . the economic conse-. . .quences of the rule, . . . the effect on thenational economy, small business, technologi-cal innovation, the environment, and publichealth” [TSCA sec. 6(c)(D)] be considered in d e-ciding whether a substance does or does not

pose an u nreasonable risk.

Balancing is equated with some kind of com-parison of benefits and costs, but none of thelaws explicitly requires formal benefit-costanalysis. For instance, the Committee on Inter-state and Foreign Commerce Report (65) onTSCA says that, “a formal benefit-cost analysisund er which a monetary value is assigned to therisks” is not required. And a court decisionabout an action taken und er CPSA d eclared thatthe Consumer Product Safety Commission“does not have to conduct an ‘elaborate cost-

benefit analysis’ to conclude that ‘unreasonablerisk’ exists” (145).

All of the laws provide for the regulation of carcinogens which threaten human health. Inthe case of the balancing laws Congress requiresthat other considerations be balanced againstthe health risk. In practice health risk signals anagency that it should consider regulation; thestringency of that regulation is at least partiallydetermined by b alancing.

Technology-Based Laws

CWA and CAA are, in general, technologybased. For instance, CAA directs the Ad minis-trator of th e Environm ental Protection, Agency

(EPA) to reduce particulate emissions to somepercentage of existing levels. The regulationsmay be “technology-forcing” because new tech-niques may be required to achieve the reduc-tion. In oth er cases, the laws specify th at p ollu-tion control is to be achieved by using “bestpractical technology” (BPT) or “best available

technology” (BAT). Such regulations do notforce new technology, but br ing all control ef-forts up to standards established by existingcontrol technologies.

An important consideration of the technol-ogy-based law s is that EPA has n ot yet been re-quired to prod uce studies to show that the im-position of new standards will imp rove pu blichealth. Imp osition of the stand ard s redu ces ex-posu res, and in the case of carcinogens, given anonth reshold ap pr oach to carcinogenic risks, itfollows that redu cing exposures should imp rovepu blic health.

The Occupational Safety and Health Act(OSH Act) requires:

The Secretary, in promulgating standards. . . shall set the standard which most adequ ate-ly assures, to th e extent feasible, on the b asis of the best available evidence, that no employeewill suffer material impairment of health orfunctional capacity . . . . In addition to the at-tainment of the highest degree of health andsafety protection for the emp loyee, other consid-erations sha ll be . . . the feasibility of the stand -ards . . . (OSHA; sec. 6(b)(5))

In the sen se that feasible has a techn ologicalmeaning, the OSH Act can be considered as a

8/14/2019 8104


Ch. 6—Approaches to Regulating Carcinogens q181

technology-based law. However, the SupremeCourt m ay issue a d ecision in a case involvingan OSH Act standard for exposure to cottondu st, that w ill determine wheth er or not benefitsand costs have to be calculated to justify thestand ard . That case may not be heard since theOccupational Safety and Health Administration

(OSHA) has w ithdrawn its proposed standard.Freeman (131) has argued that the “technol-

ogy-based” laws require balancing, pointingout that BPT implies balancing. How else can“practical” be d efined? Likewise, deciding w hatis BAT involves ba lancing costs of the technol-ogy against the expected gains.

Doniger (95) cites a sign ificant d ifference be-tween technology-based and balancing laws. Hesuggests that once a hazard is identified un der atechnology-based law , the next step is to d eter-mine the best means to control it and then de-

cide if there are any compelling reasons to back off from the best m eans. Under a balancing law,he says, once a hazard is identified, the next stepis to quantify the risks it presents in order tobalance those against costs of control.

The tripartite division of the laws—risk, bal-ancing, technology—while useful , does n o tneatly describe all the laws when subjected to

closer inspection. Complex laws contain sec-tions that have different bases, and carcinogen

regulations are generally developed under risk-based or balancing sections of the those laws.

An Example of Balancing

FDA can balance costs and benefits in regulat-ing carcinogens in food except when the car-cinogen is a food ad ditive. An example of thatbalancing is the FDA (124) regulation of poly-

chlorinated biphenyls (PCBs) in fish. FDA con-sidered three possible levels for PCBs in fishfrom the Great Lakes (see table 36). Fish thatcontain PCBs up to the FDA-established toler-ance level can be sold ; those having m ore PCBscannot be sold.

Few (perhaps no) people dispute that PCBsare a human health hazard. The acceptance of that fact is amp ly dem onstrated by TSCA (sec.6(e)) directing that PCBs be regulated. The in-

Table 36.—An example of Balancing Cancer Risk v.Revenue Loss. The FDA’s Setting a Tolerance

for PCBs in Fish

EstimatedProposed tolerance Projected cancer loss of

ppm cases/vear revenue

5 46.8 $ 0 .6 million2 34.3 5.7 million

1 21.0 $16.0 million

formation in table 36 may illustrate that once arisk is accepted as real, i.e., worthy of regula-tory attention, the stringency of the regu lationis set by economic or oth er factors. It is reason -able to assume that more cancer would resultfrom PCBs at 5 ppm than at 1 ppm, but giventhe uncertainties of quantitative risk assess-ment, it is difficult or impossible to accept thatthe projected number of cases is accurate.Neverth eless, FDA d ecided that “th e balance be-

tween public health protection and loss of foodis properly struck by a 2-ppm tolerance. ”

The Precautionary Nature of theDefinitions of Toxic Risks in the Laws

Reduction of exposures to carcinogens are in-tended to prevent cancer. Given the long latentperiod between exposure and overt disease

symptoms, prevention must depend on the iden-tification of carcinogens in test systems. Thealternative, waiting for hu man evidence of car-cinogenicity exposes some por tion of the pop -ulation to a carcinogen . Even if the substan ce isthen withdrawn completely, the legacy of theexposure wou ld be a continuing n um ber of casesas some of the exposed peop le develop cancer.Reflecting these concerns, the definitions of tox-ic substances und er w hich carcinogens are regu -lated do not require evidence of human disease.In accepting evidence from other sources, eachof the laws is precautionary.

The Delaney clause is most direct; it acceptseviden ce of animal carcinogenicity as su fficientto ban a food add itive. Section 112 of CAA callsfor regulation of pollutants that “may cause . . .[an] increase in m ortality;” RCRA and th e SafeDrinking Water Act also use a “may” construc-

8/14/2019 8104


tion in d efining toxics. TSCA d irects EPA to re-quire testing of new chemicals wh ich “m ay p re-sent an unreasonable risk” (sec. 5), but has amore stringent, but still precautionary phrase,“presents or will present an unreasonable risk”in section 6 which authorizes regulating a toxicsubstance already pr esent in commerce.

Regulatory Definitions of Carcinogens

The Delaney clause contains an operationaldefinition for carcinogens:

no additive shall be deemed safe if it isfound, after tests wh ich are ap prop riate for theevaluation of food ad ditives, to ind uce cancer inman or animal.

Although the other laws do not define whatprop erties of a substance make it a carcinogen, anumber of Federal documents (e. g., 180,279)specify what technical results agencies will con-

sider in d eciding abou t carcinogenicity.

In most cases, animal d ata are th e only basisfor decisions about carcinogenicity. Seeminglyendless argum ents can be m ounted about testresults: that the tested animal may not be a sur-rogate for hu man s, that the dose was too high,that lesions in animals may not p arallel hu mandisease states. Given the current state of knowl-edge, those argument cannot be answered toeveryone’s satisfaction. In the absence of agree-ment among all concerned p arties, agency state-ments about m ethods to be used in making deci-

sions about carcinogenicity represent the Fed-eral Government’s position. It can only be ex-pected that the methods will remain disputeduntil basic science provides more informationabout carcinogenic mechanisms and h um an re-sponse to carcinogens.

Degree of Protection

A balance between health and other consider-ations is struck by defining the d egree of protec-tion in each law. The Delaney clause, in whichthe balance is on the side of health, requiresbanning and the maximum d egree of protection.

The other definitions of d egree of protectionin table 35 are not so clear, and at least oneposed a difficult task for a regulatory agency.Section 112 of CAA is the basis for EPA’s (111)prop osed airborn e carcinogen p olicy. It d irects:

The Administrator shall establish any suchstandard at the level which in his jud gment p ro-

vides an am ple margin of safety to protect thepublic health from such hazardous pollutant

[CAA sec. 112(b)(l)(B)].

The proposed airborne carcinogen policy(111) states that EPA has, “as a m atter of p ru-dent health policy, taken the p osition that in theabsence of identifiable effect thresholds, car-cinogens p ose some risk of cancer at any expo-sure above zero. ” The position that no thr esholdcan be assumed for carcinogenicity makes it im-possible to achieve an “amp le margin of safety”unless zero emissions are imposed. However,EPA also decided that zero emissions for somesubstances would impose a too heavy economicburden, and that such controls were not whatCongress had in mind . EPA solved this problemby proposing that BAT controls will be im-posed, and if they leave an unreasonable residu-al risk, further controls will be considered (111):

Final standards for source categories present-ing significant risks to public health would, as aminimum, requ ire such sources to use best avail-able technology to reduce emissions. If, how-ever, the risk remaining after the application of best available technology is determined to be

unr easonable, further control is required.

A striking contrast in the use of the words“am ple margin of safety” is seen in comparingCAA (sec. 112) and CWA (sec. 307). Under

CWA, the first level of regulation is to be BAT

which is also the language chosen for the pro-posed airborne carcinogen policy, after it hadbeen concluded that an “am ple margin of safe-ty,” the language of section 112, was unat-tainable. If after BAT has been applied, residualrisk remains, “effluent standards” may be w rit-ten to reduce effluents to achieve “an amplemar gin of safety” (see table 37).

8/14/2019 8104


Ch. 6—ApproacheS to Regulating Carcinogens q 183

Table 37.—Ample Margin of Safety as Used in Two Laws

Clean Air Act Clean Water Act

Level of protection Sec. 112(b) (1) (B)“ample Sec. 307(a) (2) in accordancemargin of safety to protect with sections 301(b) (2) (A) andthe public health. . .” 304(b) (2) sets “effluent limita-

tions” according to BATa.

Regulatory language No threshold level isassumed for carcinogens

and an ample margin of safety is unattainabletherefore BATa is imposed.

What happens if BATa Stricter measures can be em- Sec. 307(a) (4) an “effluent stan-is judged inadequate? ployed to control residual un- dard” can be promulgated to

reasonable risk. provide “an ample margin of safely. ”

aBe5t available technology

AGENCY ADMINISTRATIVE PROCEDURES FORCARCINOGEN REGULATION

The Administrative Procedures Actof 1946

The executive branch of Government adm in-isters hu nd reds of d ifferent laws. Agency p ro-cedures for carcinogen regulation as well asother subject areas are substantially dictated bythe Ad ministrative Procedu res Act of 1946, spe-

cific formulas mandated by Congress withincertain enabling statutes, and, in certain cases,by Executive ord ers. To varying degrees theseformulas aim to safeguard individual rights anddue process, while balancing potentially con-flicting national goals and policies.

The Administrative Procedures Act (APA)was passed in 1946 afte r more than 10 years of painstaking study and drafting. Since then,there has been no major reform of regulatoryprocedures. According to Senator McCarran(228), who supported and explained the billbefore the Senate, its purpose is to:

. . . improve the administration of justice byprescribing fair administrative procedure . . . .[it] is a bill of rights for the hundreds of thousands of Americans whose affairs are con-

trolled or regulated in some w ay or another byagencies of the Federal Government. It is de-signed to provid e guaranties of due process inadm inistrative procedures.

APA is generally applicable to all regulatoryagencies and sets forth requ ired p rocedures foragencies to follow when they engage in rule-making (e. g., rules which set stand ards) and ad-

jud ication (e.g., licensing). The procedu res in-volved may range from those which allow in-formal, mostly written, decisions w ithout p riorhearings, to those which require formal ad-

judicatory hearings complete with the right tocross-examination. While APA contains no spe-cific guidance on informal decisionmaking re-quirements, d ifferent levels of procedu ral d etailhave developed d epend ing on the kind s of issuesinvolved. As has been noted :

In terms of ordering the procedural values,one might organize (categories of administrativedecisions) on a scale of maximum to minimumprocedures. At the top of the scale, the hearingprocedures em ployed m ay come close to the fulladjudicative model, since the issues at stakeresemble those decided in the civil or criminalprocess. Toward the bottom of the scale (e.g.,planning and policy making) there may be few,if any, procedural requirements . . . . Even inthese categories, however, certain procedural in-gredients appear; in effect, notice and reasonsrequirements approximate a minimum proce-

du ral m odel (355).

Most of the agencies discussed in this studyuse “informal rulemaking” to make regulatory

8/14/2019 8104


decisions. Therefore, the m inimal pr ocedu ral re-quirements ap ply of providing p ublic notice of proposed and final rules, and an opportunityfor affected interests to comment. However,specific procedur al requirements are im posed bysome enabling statutes. TSCA, for instance,provides an opportunity for an informal hearingwith cross-examination as a p art of the ru lemak-

ing p rocess. It goes beyond the m ore simplifiedAPA informal rulemaking, but does not go asfar as to require APA formal rulemaking pro-cedures.

Judicial Review

APA and some enabling statutes provide for jud icial review of the process by w hich agencydecisions are made. In the case of informalrulemaking, the tendency is to require thatagency performance not be, “arbitrary, capri-cious, an abu se of discretion, or otherwise notin accordance with law” (5 U.S. C. sec. 706).

Judicial review then examines the agency record

to see whether the agency has provided notice,

responded to comments, and given reasons for

its actions to the public.

A more stringent judicial review is applied to

formal rulemaking. In those cases, the courtmay “set aside agen cy action, findings, and con-clusions found to be . . . unsupported by sub-stantial evidence” (5 U.S. C. sec. 706). The courtmust decide whether or not the agency recordcontains substantial evidence to support the

proposed action, but there is the presumptionthat th e court w ill defer to agency expertise ontechnical matters contained in the record.

There is a tendency for judicial review of in-formal rules under the “arbitrary and capri-cious” stand ard to be increasingly stringent tothe d egree that it is converging on the “substan-tial evidence” standard prescribed by APA for

judicial review of adjudicatory and formal rule-making decisions (20). The reasonable conclu-sion is that both standa rd s are coming to meanthe same thing in terms of agency accountability

when subjected to judicial review. As recentlynoted by Judge McGowan (231):

If you’re raising the q uestion of the d ifferencebetween arbitrary and capricious review andsubstantial evidence in the record, I think the jud iciary is finally having to accept the fact that,because Congress has u sed them so loosely andso interchangeably, we have to assure ourselvesthat there is very little difference, if any, be-tween them.

The practical effect of this merging of judicialreview standards is to require closer judicialscrutiny of agency records and evidence in allreviewable actions. Some argue that agencyflexibility and innovation, even to the extentallowed by APA, to meet the particular situa-tion of each case has been almost eliminated bythe threat of more critical judicial review. Thereis some concern that these developments havecaused confusion in the courts an d agencies, aswell as worked considerable mischief on proce-dural regularity and the unifying function thatAPA was originally designed to perform for all

branches of government and the p ublic (355).

As the su bject matter of regu lation becomesincreasingly comp lex, agencies are required towork more and more with incomplete or ap-proximate data. Scientific uncertainty, com-poun ded by the need to balance and value oftenconflicting goals, has raised questions about thecontinued workability of the present regulatoryframework.

Regulatory Reform

The following remarks reflect some of thefrustrations with the present system which arefeeding the calls for reform. In a p aper presentedbefore the National Conference on Federal Reg-ulation, September 1979, Richard Neustadt,then-Assistant Director of the White House Do-mestic Policy Staff, observed (275):

[The 1946 Administrative Procedures Act]predated most of the health and safety pro-grams, concentrated on issues of p rocedural fair-ness, and took little account of the problems of economic impact and inconsistent policies.

Cutler (76), in expressing the views of the

American Bar Association Commission on Lawand Economy, has charged:

8/14/2019 8104


[W]e have a regulatory system in which . . .the buck stops nowhere. We pursue each of ournum erous and conflicting and competing goalswith single-minded devotion regardless of the ef-fect of one upon another.

Numerous regulatory reform proposals have

been generated during the past few years, creat-

ing a wid espread attack on the existing Federalregulatory system. The lack of effective balanc-ing mechanisms in light of pr esent economic andother n ational goals and p olicies has been raisedwith increasing frequency as a major failing.Furthermore, as observed by Costle (69), theconcern for reform extends to cancer regulation:

There is no qu estion abou t the need for bettermanagem ent of the regu latory process. For ex-amp le, I will be annou ncing in the next severaldays a national p olicy on the regulation of car-cinogens [306]. There are 21 statutes on thebooks that au thorize the regu lation of carcino-

gens. Thus, the necessity of having a consistentnationa l policy is self-eviden t.

But the issue of regulating carcinogens alsoillustrates a larger p oint, which is that we havenot had a national road map of the cumulativeeffect of regulation, nor of where that regulationis taking us in terms of conflicting nationalgoals . . . [We] have lacked a systematic way of tracking, effectively and intelligently, the m ulti-tude of regulatory activities that are ongoingevery day.

Many of the recent regulatory reform pro-

posa l s have ca l l ed fo r subs t an t ive change

through deregula t ion or severely restrictedregulation.

Procedural Reforms

One category of reform p roposals focuses onprocedural improvements within the existingregulatory framework to make regulation moreefficient, effective, and responsive to publicneeds. Through various means these efforts aimto improve agency administrative procedures,agency analytic and management capability,and pu blic inpu t for better regu latory decision-

making. The executive branch, in the presentand in previous administrations, has acted toimprove the procedure through Executive or-

ders. Legislative reforms are p roposed in Con-gress.

Executive Order No. 12044 (Under PresidentCarter) and No. 12291 (Under PresidentReagan)

In a major step toward regulatory reform at

the executive level, President Carter issued Ex-ecutive Order No. 12044 in March 1978. It di-

rected each executive branch agency to publish a

semiannual agenda of significant regulations

u n d e r r e v i e w o r d e v e l o p m e n t , to p rov ide

greater opportunity for public part icipation,

and to prepare regulatory analyses on all pro-

posed regulations that may have major econom-ic consequences (an annual economic effect of $100 million or more). To assist individualagencies in meeting the goals of this order,President Carter established the RegulatoryAnalysis Review Group, to prepare reports onparticularly imp ortant prop osed rules, and theRegulatory Council, to prepare a biannual regu-latory calendar and deal with areas of overlap-ping and conflicting regu lations.

This order had a significant and controversialimpact on agencies charged by Congress withregulating risks to health, safety, and the en-vironment. Office of Management and Budget(OMB) guidance encouraged the use of cost-benefit and other economic analyses to resolvehealth, safety, and environmental problems.However, there was no uniform policy or guid-

ance for d ealing w ith the m ethod ological limita-tions of cost-benefit analysis—use of discountrates, how to value health and environmentalbenefits, how to allocate costs and benefits todifferent societal groups, etc. (20). In spite of theorder’s alleged fau lts and deficiencies, the Ad -ministrator an d oth ers at EPA, one of the agen-cies most experienced with this order, defendedit as an encouraging beginning to resolving reg-ulatory conflicts and shaping agency decisionsin a manner that reflects overall policy objec-tives (69).

President Reagan issued Executive Order No.12291 in February 1981. It preserved many fea-

tures of the earlier order and provides for an in -

80-481 9 - 81 - 13

8/14/2019 8104


creased role of OMB and cost-benefit analysesin deciding about regu lations.

A conflict that a rose und er the old order is ex-pected to continue, Some segments of societywant to kn ow the contents of a proposed d raftregulation both when it leaves the agency and

when it returns from OMB. Presently no publicrecord is required of such drafts, and, in fact,such materials cannot be disclosed under theFreedom of Information Act. Protection of suchdocum ents is seen as necessary for the smoothfunctioning of the agencies and to allow regu-lators and d ecisionm akers to explore ideas andpositions before going to the p ublic with them.Resolution of the conflict between the public’sright to information and the Government’sdesire for confidentiality m ay be reached in thecourts.

LegislationIn 1979, the administration transmitted to

Congress a bill that would have strengthenedthe reforms enacted by Executive Order No.12044, mad e them permanent, and app lied themto all regulatory agencies, including the inde-pendent regulatory commissions (S. 755). It alsowould have overhauled key parts of the Admin-istrative Procedur es Act. Other bills of the 96thCongress proposed p rocedu ral changes, some of which paralleled portions of the administra-tion’s bill (e.g., S. 262, S. 755, S. 1291, S. 2147,

and H.R. 3263).

Several proposals require an agency regula-tory impact analysis before issuance of majorrules. Many require that the agencies set dead-

lines for rulemaking; and a schedule for review

of significant existing rules. There are new pro-

visions for agency rulemaking and adjudication,

appointment of administrative law judges, and

greater involvement of the Administrative Con-ference. Measures to increase public input in-

clude establishment of a Government-wide pro-

gram of assistance to public interveners.

The Regulatory Flexibility Act of 1980 (PublicLaw 96-354) was enacted to lessen the impact of Federal regulation on small businesses and smallGovernment units. This law requires, wherethere is a likelihood that agency rules may have

a “significant economic imp act on a su bstantialnumber of small entities, ” that the agencyprepare annual agendas for such rules and aflexibility analysis of each proposed rule. Thisanalysis is for the pu rpose of explaining th e ra-tionale for agency action, considering flexibleregulatory proposals, and examining alterna-

tives which migh t minim ize economic impact.While the scope of this law is somewhat limited,determined movement toward proceduralreform is clear.

Structural Reforms ThroughShifts in Oversight

Another category of reform proposals offers amore far-reaching approach by shifting existingregulatory authority to Congress, the courts, orthe executive branch. The proposals call forstructural reform in the sense that regulatory

authority and, ultimately, political power areredistributed. They challenge the fundamentalrole of adm inistrative agencies as they now ex-ist, by imposing new outside oversight andreview controls. Examples of these proposals in-clude the legislative veto), the presidential veto,and the Bumpers amendment.

The Legislative Veto

A major study on Federal regulation by theSenate Committee on Governmental Affairsduring the 95th Congress recommended that

Congress substan tially change its agency over-sight processes to improve evaluation, coordi-nation, and systematic review of agency pro-grams. H owever, w hen it came to the legislativeveto, it concluded that although this approach“may be appropriate in limited situations, theCongress sh ould reject use of the legislative vetofor regulatory agency rules . . . [and] shouldalso refrain from routinely adding a legisla-tive veto provision to regulatory agency stat-utes” (63).

Nevertheless, a number of reform proposals

would enact various forms of the legislativeveto. Depending on the bill, the approachwould subject some or all agency proposed andexisting rules to congressional scrutiny, witheither Hou se having authority to veto the ru le

8/14/2019 8104


within a specified time (e. g., H.R. 1033, 96th

Cong. ; H.R. 460, H.R. 495, H.R. 532, H.R.1858, and S. 1463, 95th Cong. ).

Serious questions have been raised about theefficiency of such an all-encompassing ap-proach, par ticularly in complex areas of regula-tion, such as those dealing with carcinogens.The legislative veto option could conceivablyresult in technical decisionmaking being trans-ferred from expert agencies to generalists inCongress. There may be some constitutionalproblems as w ell, concerning the separation of pow ers, congressional d elegation of au thority,the role of the President, and bicameralismwh ere only a one-Hou se veto is required.

Increased Presidential Authority:The Presidential Veto

Another structural reform proposal with far--reaching consequences would give the Presidentincreased authority over regulatory decisions.Again, there are a var iety of forms to this pro-posal. One generating ongoing debate was de-veloped by the American Bar Association (4). Itwould authorize the President to direct an agen-cy to take up or reconsider an d m odify certaincritical kind s of regulation , require cost-benefitregulatory analysis, and subject the President’sactions to limited congressional review. Whilean agency would continue its norm al rulemak-ing procedu re, the President w ould h ave, in ef-fect, final rulemaking authority over criticalareas, since he could direct the agency to recon-sider, modify, or reverse its decision.

As with the legislative veto prop osals, a fun-damental impetus for this kind of approach isthe p erceived loss of Governm ent accountabilityresulting from overly broad d elegations of auth -ority to the agencies. Because of the many oftenconflicting national goals, to which the Presi-

dent m ust be responsive, and the narrower re-sponsibility of single-mission agencies, greaterPresidential authority is seen as n eeded for aneffective balancing p rocess an d more responsi-

ble and accountable Government.

Questions similar to those raised w ith the leg-islative veto are again raised here. Constitu-

tional and political issues related to overlybroad delegation of legislative authority; Presi-dential efficacy, particularly when overseeingtechnical agencies; further delay and bureau-cratic overload; and maintaining proceduralsafeguard s are all concerns w hich m ust be exam-ined with this approach to regulatory reform.

Additionally, the shift in decisionmaking fromthe agen cies, which hav e technical expertise, toindividuals in the Office of the President whomay lack technical expertise, can be viewed asinappropriate.

A Greater Role for the Cou rts:

The Bumpers Amendment

Some structural reform proposals would shiftpower to the courts. A major example, whichcontinues to spark interest is the Bumpersamendment, adopted by the Senate as a flooramendment in September 1979 (S. 111, 96th

Cong., 1st sess., 1979). The amendment woulddo two things. First, it would remove the pre-sumption of validity that accompanies a regula-tion when it is challenged in court. Second, itwould require an agency to support the validityof a rule by preponderance of the evidence, ahigher standar d than either “arbitrary or capri-cious” or “su bstantial evidence. ”

The amendment is seen as an attempt to curbproblems associated with overregulation, andthe courts would p lay a greater role and havegreater influence. The role of the courts w ould

extend to both overregulation and underregula-tion, since the courts also wou ld be the forumfor challenges that an agency is not regulatingvigorously enough.

Levin (206) says that application of theamend ment raises many am biguities and d iffi-culties of interpretation. While agencies couldstill be expected to condu ct normal ru lemakingactivities, it app ears that the p roposal wou ld re-quire the cour ts to give little or no weight to anagency’s decision. If so, the very rea son for ad -ministrative agencies to retain expertise may be

un dercut. The likely outcome wou ld be more lit-igation and challenge of regulations. Some ob-servers have expressed d oubts about the amend -ment’s ability to facilitate more effective and ef-

8/14/2019 8104


8/14/2019 8104


Figure 27.—Assessing Carcinogenic Risks andConsidering Other Factors in Making a Decision

about Risk Management

[U) H=ard identlficatiofl “ I

laboratory studiesepidemiology

I-Qif no hazardidentified

stop

(2) Hazard evaluation

extrapolation to generate estimates of poten-cies and of unmeasurable carcinogenicrisks at low exposure levels

T

(3) Risk evaluation (or assessment) I

estimation of number of people exposed atvarious levels combined with potencies toproduce estimates of risks to indhridualsandpopulations

IQif risk issmal Ienough

stop

(4) Risk management I

Evaluation of risk (step 3)plusevaluation of costs, cost-effectiveness,benefits, risk vs risks of substitutes,economic feasibility, technical feasibility, etc.

c)Strategy

to beimplemented

SOURCE: D!esler (81)

cussed above, require only this step to make aregulatory decision. The second step, risk evaluation, produces qualitative and quantita-tive expressions of the hazard associated w iththe substance. The quantitative expressionsallow a rough ordering of hazards.

The third step, risk assessment, which canproduce a quantitative estimate of the potentialeffect of the substance on humans, is limited byall the uncertainties involved in hazard iden-tification and evaluation. Additionally, in thisstep, other uncertainties are introduced alongwith th e estimates of the num ber of people ex-posed and the levels of their exposure. Someobservers object to the third step because of these many u ncertainties. How ever, uncertain-ties do not make it impossible to attempt toquantify risk assessments; they do affect theform the assessment takes and the form in w hichthe result is presented.

The third step is necessary before quan titativecomparisons can be made between risks andother factors. Such comparisons are shown instep 4.

The critical fun ction of step 3 is quantitativerisk assessment. The generally good agreementthat epidemiology and laboratory tests canidentify carcinogens does not extend to agree-ment that currently available data and risk assessment techniques can accurately predict thelevel of human risk, and the usefulness of quan -

titative risk assessment is argued.

Some observers view quantitative risk assess-ment efforts, especially when applied to chronichealth problems, including cancer, as prema-ture. They see the techniques as u ncertain, theresults they produ ce as potentially misleading,and express concern that the u sers of the resultsdo not appreciate the reservations attached tothe numbers generated in quantitative risk assessments.

The advocates of quantitative risk assessment

say that the methods are necessary to decidewh ich of the many identified carcinogens pre-sent intolerable risks and which do not. To

8/14/2019 8104


counter the argum ent that the techniqu es are im-perfect, they cite the great amount of interest inrisk assessment an d say that th e methods are be-ing improved.

Quantitative risk assessment is likely to takeon increasing importance in Federal decisionmaking about carcinogens. For instance, the re-

quirement for benefit-cost analysis under Execu-tive Order No. 12291 will result in greater de-mands for quantitative risk assessment. Addi-tionally, the Supreme Court decision aboutworkplace exposure to benzene (337) apparentl y

requires that OSHA make some estimate of risk to sup port r egulations. The fact that benzene isa carcinogen was not disputed in the case, butthe court returned the p roposed standard to theagency for reconsideration because OSHA hadnot found ,

. . . that th e toxic substance [benzene at th e lev-el currently encountered in the workplace] . . .

poses a significant health risk in the workplaceand that a new lower standard is . . . “reason-ably necessary or app ropriate to provide safe orhealthful employment’’(20).

The requirement to show a “significant h ealthrisk” will surely result in greater pressure onOSHA to estimate the nu mber of w orkers likelyto suffer ill-effects from exposures. The pres-sures for use of risk assessment will probably in-crease, and opponents will continue to voiceconcern about its un certainties.

Limits to Quantitative Risk Assessment

Rowe (316) draws attention to limits that hesees surrounding the accuracy of quantitativerisk assessment. In p articular he cites problemswith choosing an extrapolation model, thewidely variant estimates produced by the use of different models (see ch. 5), and the problemsinherent in animal tests because of the relativelysmall num bers of animals that can be tested (seech. 4). Rowe suggests that results of testing par-ticular substances in short-term tests and bio-assays be established as benchmarks. Then, asother substances are tested, they can be clas-

sified as less hazard ous, equally hazard ous, ormore hazardous than the benchmark sub-stances. At the same time, data about use and

exposure can be obtained, so that the number of

people at r isk, along with potency, can be con-sidered in deciding what next to do.

Importantly, in Rowe’s scheme, a next step isto consider whether or not additional testingmight change th e conclusion reached on p resent-ly available data. To facilitate making that deci-sion, Rowe developed a statistic to allow con-

sideration of the probability that a test will pro-vide information that might alter a previouslymade decision. Application of the statistic esti-mates the limits of knowledge that are attain-able. Considering the limits of tests and extrap-olation models will allow decisionmakers toselect ad ditional testing only w hen the tests willyield more definitive information. When thatcondition does not prevail, the decisionmakercan choose to do nothing or to consider variousmethod s to restrict exposures.

“Trigger” Levels for Regulation

Individuals in th eir private lives accept n on-zero risks. In some instances larger risks are ac-cepted, presu mably, because the benefits asso-ciated with the risk-taking are also large. Inother situations, smaller risks are avoided,again p resumably, because the benefits are seenas not w orth even th e small risk. Such observa-tions lead to th e suggestion that individu als bal-ance risks and benefits in deciding w hich volun-tarily assumed risks are “acceptable,”

Exposures to carcinogens in air, water, theworkplace, and other environments are dif-

ferent from voluntary risks. In few cases is itpossible for an individu al to know th e identityand magnitude of the risk. In many cases, theindividual has no choice but to bear the un-know n risk because contact cannot be avoid ed.

It is not necessary, in all cases, to quantify arisk in order to decide it is not acceptable. Anexample of such an unacceptable, unreasonablerisk is the use of thalidomide by pregnantwom en. The association of that dr ug w ith chil-dren suffering multiple anatomic abnormalitiesled to a suspension of its use. A more recent ex-

ample was the association of a particular brandof tamp on w ith toxic shock syndrome. Before aquantitative evaluation of the lifetime risk of toxic shock syndrome was published, the manu-

8/14/2019 8104


facturer recalled the product. In these two cases,substitute prod ucts were read ily available, andthe cost of avoiding the risk was largely re-stricted to the lost sales of the manufacturers.

Regulatory agencies identify and quantifyrisks, and the ability to qu antify risks has p ro-duced suggestions for managing risks based on

their size. A small chance of risk can be set as a“floor,” and substances presenting a risk lessthan the floor m ight be considered acceptable.Risks above the floor level wou ld be d ivided in-to two grou ps. Intermediate risks wou ld be con-sidered for reduction, and benefits and riskswould be balanced in making a regulatory deci-sion. Above a still higher level of risk, no bal-ancing would be necessary, and any su bstancepresenting a risk of that magnitude or higherwou ld have to be regulated (3).

Products of a quanti ta t ive r isk assessment

(see fig. 27, step 3) can be estimates of the

lifetime risk of cancer to: 1) members of the

general population; and 2) to members of highly

exposed populations. These risks are expressedin scientific notation as 1 0

-4, 10

-5, 10

-6(or as

fractions, 1 chance in 10,000, 1 in 100,000, and1 in 1,000,000, respectively). Carcinogen s d iffera millionfold in potency as measured in labora-tory animals, and m ore potent on es are associ-ated w ith higher risks (10

-3) and less potent ones

with lower risks (10-7).

Albert (3) and others have suggested that sub-stances associated with individual lifetime risks

of 10-5

might be considered as presenting risksso low that they require no action to reducethem further. Such low, negligible risks wouldrepresent a “floor” on risks. An idea of themagnitud e of this risk is furnished by recallingthat a 10

-5lifetime risk means that one person of

100,000 exposed at that level is expected to

develop cancer from that exposure. Currently,

about 20,000 of every 100,000 Americans die of

cancer. Exposure of 100,000 people to a sub-

stance that increases risk by 10-5

could increase

the number of cancer deaths to 20,001. Whether

or not this number seems reasonable, it may be

important to consider that a lifetime risk of 10-5

for the U.S. population (220 million people) isequal to 2,200 cancer cases in th ose peop le’s life-

spans. The annual number of cancer deathsfrom exposing the U.S. population to a 10

-5risk

would be about 30, assuming a lifespan of 70

years.

A fundamental object ion to the idea that

some fraction of the population might be al-

lowed to die because of exposure to a risk that is

viewed as acceptable is expressed in an argu-ment called “the murder of the statistical per-son.” If the identity of the 30 people who died

annually as a result of a 10-5

risk were known,

there is little doubt that much greater effort

would be expanded to reduce the risk. Objec-

tions to the idea of society deciding that some

risks are negligible are raised by those who con-

sider that statistical people as well as identified

people deserve protection.

It is apparently more difficult to suggest alevel of risk so high that it demands regula-tory action than it is to suggest a level so lowthat it requires none. Nevertheless, toward thehigh end of the risk scale, agreement m ight bereached that an exposure that produced alifetime risk of cancer of 10-

2(1 in 100) is so high

that the Government should regulate it as ahealth risk with little or no regard for other con-siderations. The magnitud e of this risk may becompared to the tenfold higher chance (10

-1

lifetime risk) of cancer that is volunta rily borneby lifelong smokers (362).

Another consideration is that the low-level

risks, those of 10-5

or less, add up. If there are

ten 10-5 risks and the risks are additive, theirtotal risk is 10

-4. (Of course, if synergism exists

among any of the 10, the combined risk from

both might greatly exceed 10-4). What is to be

done about the next identified 10-5

risk? Shouldit be regulated, or should small risks, even if

there are 100 or 1,000 or more of them go unreg-

ulated?

If agreement could be reached on such limits,risks above a certain level (10-

3to 10

-2) might be

declared unreasonable no matter what, andrisks below a certain level (10

-5) might be de-

clared reasonable or acceptable or negligible. Inbetween, the risks that range from 10-5

up to10

-3or 10-

2wou ld requ ire balancing of the risks

8/14/2019 8104



and benefits to decide whether or not toregulate.

The inaccuracy of the risk estimates alarmsmany people. Crouch and Wilson (72) find“good correlations” between cancer rates m eas-ured in animals and those detected in hu mansdue to exposure to the same chemicals. The

num ber of chemicals for w hich data are avail-able is small (fewer than 20) and the “goodcorrelation” is good within a factor of 10 or 100.In other w ords, an estimated risk of 10-

5might

be as high as 10-3or as low as 10-

7.

Deisler (80) has approached the problem of the existence of many small risks by suggestingthat a ceiling be p laced on total risk in “exposuresituations. ” He cites industrial exposure as onesituation and discusses a maximum allowableamou nt of cancer that m ight be set as tolerablefor workplace exposures. As was pointed outearlier, estimates for occupational contributionsto cancer vary from less than 5 percent to abou t40 percent, and Deisler (80) suggests:

. . . the first interim goal should be . . . toassure that industrially related cancer becomesless than a truly small fraction of today’s totalcancer incidence in the United States . . .

As an example of a possible interim goal, hesuggests that w orkplace exposur es be set so thatthey account for 1.5 percent of the total cancerburd en. The 1.5 percent is only an example, andDeisler suggests that the selection of a goalmight be negotiated in a public forum, possibly

throu gh a congressional commission.

When a ceiling is established, the risks withinthe exposure situation wou ld be inventoried. If total risk exceeds the ceiling, the risk would bereduced by addressing one or more individualexposures. This appears to be a cost-effectivesystem because the easiest-to-control, leastcostly-to-control exposur es w ould be attackedfirst. If the inventory of risks totaled less thanthe ceiling, the ceiling could be reduced. The in-dividu al exposures in the inventory ar e expectedto interact add itively, but allowances could bemade for synergisms if they occur. The ceilingalso provides a method to d eal with the identi-fication of previously unrecognized carcinogensin an exposure situation. Such a carcinogenwou ld be, as a first step, controlled to the p ointthat the ceiling is not exceeded.

Despite the dem onstrated existence of unac-ceptable, unreasonable (and generally unq uanti-tated) risks, it has been impossible to assign

“trigger” risk levels. Deciding on the “trigger,”10

-3or 10

-2on the high side, 10

-5or 10

-6on the

low side would be d ifficult, and problems withthe accuracy of the estimates and equity in thatthose who most directly bear risks may notmost d irectly benefit are p roblems to be solved.Currently, quan titative risk assessment d oes notprovid e a tidy fix for the d ilemm a of deciding if and when to regulate. At the same time the factthat “acceptable” and “reasonable” risks arediscussed implies that not all risks are equal andthat society has the task to d ecide w hich are tobe regulated and w hich are not.

LOCATIONS OF FEDERAL CARCINOGENICRISK ASSESSMENT ACTIVITIES

Carcinogenic risk assessment involves agroup of people considering the available evi-dence, drawing conclusions from the data, and,using method s they accept, deciding w hether ornot th e substance is carcinogenic and, in som ecases, estimating carcinogenic potency or

human risk. Groups which have preparedstatements about methods to be used in evaluat-ing data differ in their interests and respon-

sibilities, and their statemen ts reflect the d iffer-ences. In addition, there is much discussionabout who should consider the data, apply themethod s, and make the decisions.

Scientists in each regulator y agency now eval-

uate data about carcinogenicity of substancesthat may be regulated by their own agency.They may consider advice from expert commit-

8/14/2019 8104


8/14/2019 8104


review. Most expert panels are now advisory,and the SAB is a reviewing p anel. Some of theproposals call for the technical panel to makethe d ecision for the agencies, others to r eviewcontested decisions.

OSTP Proposal for Decisionmaking

OSTP (281) divided Federal decisionmakingabout carcinogens into two phases. Phase I isthe identification and quantitative characteriza-tion of risks and encomp asses steps 1 and 2 andpart of 3 in figure 27. Phase II is making theregulatory d ecision to control the iden tified an dcharacterized risk. OSTP prop oses that p hase Iactivities for all agencies be brought togetherand located w ithin N TP.

The current responsibilities of NTP do not in-clude phase I judgments for regulatory deci-sions. However, OSTP argu es that just as N TPhas a centralized coordination role in toxicolog-ical research so should it have a central role ininterpretation of data.

Gilbert Omenn, an au thor of the OSTP pap er(281), insists that decisionmaking about car-cinogenicity should remain the responsibility of Federal officials. He m akes the p oint that lawsdealing with carcinogens are d esigned to protectpu blic health an d that Federal officials are en-trusted with resp onsibility for adm inistering th elaws.

Having a panel of Federal officials decideabout carcinogenicity is a major difference be-tween the OSTP proposal and the proposalsfrom AIHC and Representative Wampler. Thelatter two proposals centralize decisionmaking,but they would delegate carcinogen decision-making authority to new non-governmentalorganizations which would include non-Federalexperts.

The American Industrial Health CouncilScience Panel Proposal

The AIHC (10) proposal for a science paneldr aws a d istinction betw een scientific (“PhaseI“) and regulatory (“Phase II”) decisionm akingas d oes the O STP prop osal. The science pan elwould be composed of “the best scientists

available” and located centrally within Govern-ment or elsewhere. The panel would review ex-isting data and not conduct or control research.Briefly, when a regulatory agency reached thepoint of considering regulatory action against achemical, it would bring its data and conclu-sions about carcinogenicity to the panel. The

pan el wou ld solicit add itional data from indu s-try, public interest groups, and other Govern-ment agen cies and r eview all such da ta. It wou ldassess the evidence and commu nicate its conclu-sions to the agency about whether or not thesubstance was a carcinogen. If sufficient ex-posure information and dose-response datawere available, the panel would also evaluatethe human hazard and risk posed by the car-cinogen.

In the AIHC p roposal, the panel would h aveto reach a decision within a certain time limit

wh ich w ould assure that it worked to a sched-ule. The panel would consider only scientificquestions and its find ings would not be binding.An agency could reject the panel’s conclusionsby explaining its reasons for rejection.

AIHC recomm ends that Congress establish ascience panel consisting of 15 members whowould assemble from time to time to considerdata. It could appoint ad hoc members to work-groups to consider particular cases, and itwould be provided with full-time professionalstaff to manage its workload.

Representative Wampler’s NationalScience Council Proposal

Representative Wampler’s bill to establish a

National Science Coun cil (NSC) was introducedon February 13, 1980 (H. R. 6521) and reintro-duced as H .R. 638 on Janu ary 5, 1981. It wou ldestablish a 15-member council to review d eci-sions concerning chemical toxicity. A companyor individual who objected to an agency’sassessment of toxicity during the agency’s ad -

judication could request NSC review of the

agency decision. Subjects of meetings of NSCwou ld be annou nced in the Federal Register andopen to the pu blic except w hen tr ade secrets orconfidential information were discussed. Thebill offers amendments to CPSA, FDCA, the

8/14/2019 8104


Federal Meat Insp ection Act, the Pou ltry Prod-ucts Inspection Act, the OSH Act, and TSCA.The amendm ents would make the d ecisions of NSC final with respect to scientific fact underthose laws.

Mr. Wamp ler’s prop osal stipu lates that NSCmembers wou ld be appointed by the President

to full-time Federal posts for 2 years and haveappropriate staff support. This full-time servicediffers from AIHC’s proposal. The AIHC Sci-ence Panel would convene “periodically toassess materials, ” but its members would main-tain their usual employment while serving.

Judge Markey’s Proposal forLegislative Branch Reviewof Risk Decisions

Each proposal so far discussed—conveningof Federal committees, non-Federal advisory

groups, or creation of scientific boards within oroutside of the Government—seeks to improvedecisionm aking before regulations are written.A strikingly different idea has been ad vanced byJudge Howard Markey of the U.S. Court of Customs and Patent Appeals.

Markey (222) pr oposes no chan ge in the w ayagencies m ake d ecisions abou t carcinogenicityand regulations. However, he does propose thatOTA, as an agen cy of Congress, review agencydecisions about risks. The r eview would be in -itiated if a regulated ind ustry or a pu blic interest

group objected in court to a regulation on thebasis that the agency had made a mistake inscience.

Markey goes on to sa y that if OTA or someother agency designated by Congress cannotreach an agreement about the correctness of theagency decision about risk, “OTA w ould turn toCongress, where the final decision on acceptablerisk could be m ade by the people through theirrepresentatives. ”

Summary Comments About TechnicalPanels and Study of Their Feasibility

The num ber of pr oposals for risk determ ina-tion panels almost guarantees that the p anelswill remain an issue in Federal policy about car-

cinogens. Establishment of su ch a pan el wou ldrepresent a significant change in the processused by the Federal Governm ent to make d eci-sions about health risks.

Proponents of panels claim they would im-prove the efficiency of the regulatory process. Apanel w ould make technical decisions for all theagencies, which is seen as assuring consistentscientific findings. Secondly, a time limit im-posed on panel deliberations would ensure thatit complete its work quickly. Finally, a regula-tory agency could initiate the panel review of data about a suspect substance, and thereforethe review could take p lace wh en it best fits theagency schedule.

Public interest, labor, environmental or gani-zations, and th e Federal regulatory agencies op-pose these suggestions. They see regulatoryagencies as the ap prop riate and lawful locationsfor making decisions about risk. In general theysee a science panel as an other layer of bu reauc-racy that m ight hind er regulatory activities, andworry that a single panel might be more sen-sitive to pressure from interested parties. Fur-thermore, they see the division between “sci-ence” and “policy” in decisions about cancer asillusionary. They argue that such a panel mighthave the p ower to d elay decisions by imp osing ahigher standard of proof that a substance was acarcinogen than is required by law. This wouldstymie preventive “precautionary” governmen-tal action wh ich they v iew as necessary to pro-

tect lives and health when certainty cannot beachieved.

Congress has app ropriated $500,000 to FDAto place a contract to investigate the feasibilityof a centralized science panel, and a report is ex-pected in 1982. The results of that stud y shou ldanswer questions about how the panel mightfunction and how, or if, scientific and technicaldecisions can be made separately from policy orregulatory d ecisions. The FDA-sponsored stud ymay also reveal w hether difficulties associatedwith chemical regulations have h inged on scien-

tific or on regulatory controversies. If the stu dyshows that scientific errors have seldom resultedfrom processes used by the Government or thatthe errors have been of little importance,

8/14/2019 8104


changes in the process of making scientific deci- ly, if many examples of incorrect scientific deci-

sions wou ld seem to h ave little merit. Converse- sions are foun d, changes might be app ropriate.

MAKING REGULATORY DECISIONS

Possible Decisionmaking Frameworks

Each environmental health law seeks to re-duce risks to the public health. Lave (204) dis-cusses “frameworks” under which regulatorydecisions can be m ade, and his categorizationsare the basis for this section. He arrays theframew orks from that requ iring the least infor-mation and analysis––the no-risk framework—to that requiring the most information andanalysis —the benefit-cost framework. Some arefamiliar, describing the way in w hich decisionsare made today; some suggest possible futuredirections.

1. The Delaney clause is m ost frequently men -tioned as an example of the no-risk frame-work. It is a statement that Congress, con-cerned about the safety of food additives,has d one the balancing of risks and benefitsand has decided n o benefits of food add itivescan outweigh a demonstrated cancer risk.FDA, the agency that adm inisters Delaney,mu st show a risk before it bans an add itive;it does not have to identify and measure ben-efits. Even if other organizations identifyand measure benefits, FDA cannot balance

benefits against risk.2. The risk-risk framework requires a compari-

son of the risks associated with continueduse of a substance to any risks generated byits use being controlled or discontinued.Generated risks would include those inher-ent in substitutes that might be employed inplace of the substance. The best and mostdirect example of risk-risk analysis is the useof dangerous d ru gs to treat life threateningdiseases. The risk from use of the d rug canbe comp ared w ith the risk of death if it is notused. A regu latory risk-risk determination is

required under section 202 of CAA. As anexample, a control device that redu ced par-ticulate emissions might, at th e same tim e,produ ce chem ical emissions. The risks from

3.

4.

allowing continued particulate emissions

would be compared to the risks from thechemical emissions.

Direct risk-risk analysis could also havebeen app lied to nitrites add ed to food if theFederal Government had sustained its casethat n itrites were carcinogenic and deservingof regulation. Eliminating or red ucing use of nitrites as a preservative would have in-creased the risk of botulism (severe foodpoisoning). Its continued use m ight be asso-ciated with a cancer risk. Under the risk-risk framework, a d ecision to ban , to redu ce use,or to allow curent use patterns might have

been made by comparing expected deathsfrom botulism and from cancer under thedifferent levels of regulation.

General balancing of risks against benefitsdiffers from the first two frameworks andpara llels all others in that it allows consid-eration of effects other than health. Lavesays that the framework is purposefullyloose and vague; it requires enumeration of all effects, but weighing and balancing them,as well as deciding w hich to quantify, is leftto the regulator. He includ es determination

of “feasibility” u nd er the OSH Act and reg-ulations of air and water pollution which re-quire BPT and BAT under this framework.In reality there is no best available technol-ogy. It can always be made better, (e.g., byplacing tw o control devices in series) but atsome point costs are judged to be prohibi-tive. The hallmark of this framew ork is bal-ancing of risks and benefits withou t qu antifi-cation and ana lysis. Because of d eficienciesin available and attainable information ageneral balancing app roach is probably usedin most d ecisionmaking.

Cost effectiveness compares the costs of dif-ferent ways to reach the same goal. (The Im-

plications of Cost-Effectiveness Analysis on Medical Technology (285) discusses this

8/14/2019 8104


method and its application to medical prac-tice. ) In general, the framework assum es afixed budget and produces an allocation of resources (money) among different pro-grams which share a common objective.Programs selected for fund ing are those thatgo farthest toward reaching the goal, in th is

case, reduced cancer incidence and mortali-ty, at the least cost.

The expenditure of public funds is re-quired to d evelop, prom ulgate, and enforceregulations. Private fund s are expend ed toargue against regulations and to buy andmaintain control devices. Cost-effectivenesstechniques can be used to compare the totalcost of a regulation, i.e., agency costs, in-

dustry costs, and costs passed onto con-sumers to the benefits of the regulation.Alternatively, cost-effectiveness analysis cancompare benefits to either p ublic or privatecosts separately.

Regulations that are expected to preventthe most cancers at the lowest cost would beidentified in a cost-effectiveness framework.The difficulties of evaluating costs and bene-fits make it u nlikely that the technique candistinguish between regulations of nearequal benefits and costs, but distinctionsshould be possible between the most andleast cost-effective.

Informal discussions with officials at EPAand OSHA suggest that Government costs

are about the same for every regulation be-cause each regulation is likely to be chal-

lenged in court, and each one requires aboutthe same amount of staff time, whether itsimpact is large or small. Althou gh th is opi-nion w as common ly expressed, no attemptwas made to verify it. If it is true, an agencycan select its goals by concentrating on thosethat will produce the biggest health benefit.

The regulatory budget strategy appliescost effectiveness to non-Government costs.Under it each agency might be granted anamount of pr ivate sector costs that its

regulations could generate. Working withinthat amount, the agency would then propose

regulations that it intends to promulgate and

submit them for executive approval. This

5.

approach considers only costs to be borneby the private sector and could set an upperlimit on those costs. OMB (280) repor tedthat difficulties in estimating such costsmake this app roach infeasible.

Benefit-cost or cost-benefit analysis is simi-lar to general balancing of risks against ben-efits but is mor e formal and quan titative. Inthis framework, all costs and benefits are ex-pressed in dollars, and this analysis requiresplacing a monetary value on human life.After all items u nd er consideration ar e con-verted to dollar terms, the benefits and costsare compared and if the benefits of the reg-ulation exceed the costs, the decision isweighted toward making the regulation.Placing a d ollar value on hu man life is one of the m ost controversial aspects of this meth -od, and it is simply repugnant to many peo-

ple.Lave claims the benefits of this framework are that it is the most flexible, requires themost information and analysis, and drivesqualification where possible. Formal benefit-cost analysis is not now required by any car-cinogen regulating law, but it is required byExecutive Order N o. 12291 where legislationdoes not forbid it, and it is frequen tly men-tioned in plans for regulatory reform. Ahearing of the Subcommittee on Oversightand Investigations (334) of the then HouseCommittee on Interstate and Foreign Com-

merce provides a juxtaposition of opp osingviews of the applicability of the technique toregulatory decisionmaking.

Feasibility and Limits ofBenefit-Cost Analysis

Traditional benefit-cost analysis is an eco-nomic tool that requires listing of all benefitsand costs and assigning a dollar value to each.Carcinogen regulations involve the ultimateconsiderations —life and death—and opinionsdiffer about whether or not a monetary value

can be placed on life. In what ever way thatcontroversy is settled, certain comm ents can bemade about the usefulness of benefit-costanalysis.

8/14/2019 8104



In the case of a carcinogen regulation, ex-pected benefits include the health gains fromreducing exposure to the agent, and uncertaintyis attached to the circulation of these gains.Presenting the expected health gain as a numberdoes not add certainty to the estimate, but itadds to the estimate’s importance. A frequentobservation is that caveats and reservations at-tached to the numbers in the analysis are lost.The number of prematu re deaths to be averted(“lives saved”) as well as the number of dollarsto be spen t to achieve the benefit take on lives of their ow n and are un encumbered by statisticalreservations about accuracy. In this way, withno m ore information behind them, the num bersbecome more certain in the p ublic’s and the d eci-sionmakers’ minds. Too much reliance on im-precise numbers becomes a special problem inbenefit-cost analysis which reduces the benefitsand costs to d ollar figures. This problem is com-monly acknowledged but a practical solution isnot readily apparent.

Many criticisms are directed at benefit-costanalysis. Lave (204) claims that a well-doneanalysis will always favor the status quo;change costs money. Change can be producedby going either from regulation to no regulationor from no regulation to regulation. Lave goeson to say that it would be surprising if the pres-ent state is tru ly the p innacle of social evolution.If such a tilt toward the status qu o exists whenall measurements and calculations are accurate-ly done, it is easy to imagine how a bias on the

part of the analyst could affect the analysis. Notall economists agree with Lave’s statement thatbenefit-cost analysis will always favor the statusquo, and Freeman (132) cites the NAS 1974study, Air Quality and Automobile EmissionsControl, as an example of cost-benefit analysiswhich favored tighter controls.

Two other criticisms are directed at benefit-cost analysis. Equity considerations are not apart of benefit-cost analysis. For instance, thehealth benefits from a regu lation accrue to thoseindividu als whose risks are d ecreased; the cost

of reducing the risks is borne by those who payfor the control devices or procedure. However,the situation before the control is imp lementedhas the exposed people bearing a health risk

wh ich spar es anyone in society from having topay to reduce the exposure. The technique of cost-benefit analysis is silent about whethereither case is just.

Another difficulty encountered with ben efit-cost analysis is its inability to consider in-tergenerational effects. As a specific example,many substances are both carcinogens andmu tagens. Such a substance may cause cancer inpeop le exposed to it and m utations in their germcells. The mutations are expressed in the nextgeneration. Quantifying genetic damage is atleast as difficult as expressing the value of a lifein dollars.

A quite different problem in dealing with thefuture is econom ic. The discount rate chosen toproject mon etary costs and benefits is very im -portant to these analyses, but th ere is no agree-ment about the appropriate rate, especially in

inflationary times.

A resp onse to objections about valuing livesin dollar term s is seen in th e suggestion of theConservation Foundation (78) that differenttypes of analyses can be carried out under thebenefit-cost rubric. It suggests use of the term“single-value analysis” to describe benefit-costanalyses which express all items in dollar termsand “two-value analysis” for analysis that com-pare “lives saved” to “dollars cost. ”

The Conservation Foundation found “two-

value” analysis appropriate for carcinogenswhen the major concern is death. It is less easilyapplied when additional considerations, such asdam age to an ecosystem, are involved. “Multi-value analysis” is suggested as a m ethod to con-sider three or m ore irredu cible elements. It en-ables tradeoffs to be made, e.g., human healthrisk v. costs of reducing exposure, and eco-system risk v. costs, but the analysis becomesmore d ifficult. The claimed adv antage of bene-fit-cost analysis (that it forces a detailing of wh at is being considered) is equally app licableto the one-, two-, and multi-value methods. The

two- and multi-value methods involve balanc-ing of health risk against other factors, and fitwithin Lave’s third framework of general bal-ancing of risks and benefits.

8/14/2019 8104


While two-value or multivalue benefit costanalysis may be attractive because of its rigorand its not placing a dollar value on life, it isnot, strictly speaking, benefit-cost analysis. Abenefit-cost analysis drives toward a singlenumber, the quotient obtained when benefitsare divided by costs. If the quotient exceeds 1.0,the benefits are greater than the costs, and thepr oject shou ld, on an economic basis, proceed.If it is less than 1.0 it should not. Two-value and

mult ivalue benefi t -cost analyses produce no

such qu otient. Instead they comp are the benefitsin on e term (i. e., lives) v. costs (in d ollars). Thecomparison is useful in a cost-effectiveness ap-proach, where various approaches to a commongoal can be ranked, but it does not produce anumber that indicates “yes” or “no. ”

A workingPrinciples of Chemicals in

that:

panel of the NAS Committee onDecision Making for Regulatingthe Environment (263) concluded

The systematic application of the tools of decision analysis and benefit-cost analysis canprovide the d ecision m aker with a useful frame-work an d langu age for d escribing and discussingtrade-offs, noncommensurability, and uncer-tainty. This framework should help to clarifythe existence of alternatives, decision points,gaps in information, and value jud gments con-cerning tr ade-offs.

Decision analysis, as described in the NAS re-port, is a careful detailing of regulatory options,expected outcomes and uncertainties of risks,

benefits, and costs. Its “main contribution . . .is to organize information for the decisionmakerto assist him in his un avoidable balancing task. ”

The NRC committee (263) endor sed th e use of benefit-cost analysis and concluded that thetechnique is useful in making decisions, but thatit should not be the only consideration in thedecision. Furthermore it did not recommendcontinued research to improve the techniquesbecause “highly formalized methods of benefit-cost analysis can seldom be used for makingdecisions about regulating chemicals in theenvironment. ”

If market prices and shadow prices are fullyutilized to va lue econom ic efficiency effects, the

initial list of noncommensurate effects of thedecision will have been reduced to:

fully commen surate economic efficiency ben-efits and costs measu red in d ollars; and

qnoncommensurate effects , descr ibed andquantified in other units, of which the mostsignificant are likely to be hazards to healthand life, damages to the environment and

ecosystems, and the distribution of benefits,costs, and hazards among individuals andgroups.

The NRC definition of benefit-cost analysis doesnot require that all costs and benefits be ex-pressed in a common unit (dollars), and it toowou ld fit in Lave’s third framew ork.

This assessment, which deals with the scienceand policy of making decisions about carcino-gens, has emph asized un certainties in d etermin-in g risk. The Con servation Found ation (78) and

NRC (263) also draw attention to u ncertainties

in projecting costs of regulations. Both sides of the ben efit-cost analysis ar e difficult, subject tohuman error, and encumbered by uncertainties.Nevertheless, this method, whether one-, ortwo- or m ulti-value has its ad vocates, and it isincreasingly mentioned as at least a tool to beused by decisionmakers.

An Alternative to Benefit-Cost Analysisfor Making Regulatory Decisions

Sometimes the world seems divided betweeneconomists and lawyers. Economists favor deci-

sion frameworks th at rely on qu antitation an d,in benefit-cost approaches, on converting allvalues into dollars. Lawyers are more com-fortable with qualitative concepts as developedon a case-by-case basis un der common law. Inmaking decisions about risks under commonlaw, courts rely on concepts of reasonablenessof behavior, of duties and responsibilities tolearn and to inform, of assum ptions of risk thatare made by different parties, and of con-tributory behavior. Application of these con-cepts is constrained by common law p recedents,procedu ral rules, and rules about ad missibility

of competent evidence. Quantitative considera-tions of costs and benefits play a minor role(20).

8/14/2019 8104


Regulatory law d iffers from th e common law.Its history is shorter, and it is constrained andgoverned by many specific and varied enablingstatutes, the Adm inistrative Procedu res Act, theRegulatory Flexibility Act, and Executive Order12291. However, more importantly, the

regulatory agencies rely heavily on semiquan-titative and quantitative risk assessment techni-

ques, on experts, and, increasingly on a benefit-cost framework for decisionmaking. Agencyemphas i s on quan t i t a t ion , wh ich “ in -vites . . . playing with the numbers” to reachcertain analytical outcomes is producing a “lack of credibility or acceptance in the p ublic and theregulated ind ustries” according to Baram (20).

Offered against these criticisms o f benefit-costanalysis are “the rational app roach to decision-making th at it allegedly fosters” and that “it em-phasizes economic considerations, it retards ex-cessively zealous regulations, it ensures that in

general only incremental controls will be pro-mulgated. ”

This has now become th e centra l controv ersyfor the regulatory agencies: How to implementtheir ma nd ates to control carcinogenic risk in afair, objective and accountable manner by usinga “rational” framework (e.g., cost-benefit)which emphasizes economic cost factors, whenthe health and environmental benefits at stakeare generally considered as being unm easurablein economic terms (20).

Baram offers a d ecision framework tha t con-siders costs, but which also relies on qualitative

considerations for regulatory agency considera-tion. His framework has six steps. The first twoare common to any such decisionmakingscheme—hazard identification and risk meas-urement—but beyond those steps, he considersrisk management options that he says are nowoverlooked.

1. Hazard identification, —This step involvesthe technologies discussed in th is report andis initiated w hen a test shows a su bstance is acarcinogen. The developm ent of th e initialfinding of hazard can be accomplished byGovernment or non-Government testing,because of a governm ental rule requiring atest by private industry, or by presentationof a petition for rulemaking to an agency.

2.

3.

4.

Risk measurement. --This step, too, in-volves some technologies discussed here.Baram emphasizes that agencies can im-prove th eir performance at this step.

Since agencies can use evidence in theirrulemaking which wou ld not be consideredcompetent for admission in a trial in court,there are few legal protocols governing the

quality of the evidence . . . [used] . . . inrulemaking. Clearly, there is a need for theagencies to establish such admissibility of evidence protocols and to bind themselvesto the p rotocols if the competency and pro-bative quality of the evidence used is to beimproved and the confidence of the pu blicand regulates in agency findings is to be in-creased. Costs will be a major factor in es-tablishing such protocols . . . . Thus, it isimperative that agencies jointly add ress howto man age their resources for risk measure-men t collectively in th e most efficient ma n-ner, and to establish collectively the costs to

be imposed on industry for risk measure-ment w hen the m easurement task is man-dated for indu stry by law (20).

Risk management options selection. — Thisstep is absent from current regulatory pro-grams. It would identify and roughly assessthe efficacy of regulatory and nonregulatoryapproaches to risk management. Regulatoryoptions include both a) setting new stand-ards and b) enforcing in-place standards.Nonregulatory approaches include a) re-course to comm on law, b) voluntary ind us-trial standard setting, c) restrictions on fu-

ture Federal procurement from sources of risk, and d) education or public disclosureprograms.

Economic and technical feasibilit yanal-

yses.—These analyses are required by someenabling statutes, Executive Order No.12291, and the Regulatory Flexibility Act.They would focus on the m anagement op-tions identified in step 3.

The second, third, and fourth stages of measuring risk, assessing options, andcosting options should ideally be kept

separate and indep endent to ensure that risk measurement and option identification re-sults are objectively arrived at on the basisof the best data and analytic methods for

8/14/2019 8104


Ch. 6—Approaches to Regulating Carcinogens q 201

risk estimation. This accords with the find-ings of many critics of the regu latory man -agement of risk who have found a break-dow n in risk measurem ent objectivity wh enit is influenced by cost considerations. Thisimportant reform need not be Congres-sionally mand ated, but can be accomp lishedby the responsible exercise of agency discre-tion un der existing statutes. It permits eachagency then to conduct its fourth stage of economic and technical analysis in a struc-tured fashion—as cost-effectiveness analysesof each option separately and in certaincombinations (20).

5. Ordering of risk management initia-tives. —

Having identified and measur ed risks andselected the most efficacious and cost-effective management options which aretechnically feasible, each agency separately

(and in conjun ction if several find th e samecarcinogen falling within their regulatory jurisdiction) should then . . . [order] . . .risks it will choose to manag e on th e basis of the carcinogenic risk reduction benefits to beachieved.For instance, EPA might decide that the ben-efits of managing the carcinogenic risks of asbestos alone far outweigh the benefits of managing the risks of many chemical sub-stances, and thereby wou ld be in a positionto allot rationally a proportionate amountof its resources to asbestos (20).

6. Deployment of risk management options—

regulatory and nonregulatory—for selected priority chemicals. —

Finally, an agency is at the point w here itcan proceed to engage in rulemaking andother regulatory efforts, and foster the useof available alternatives to regulation, forreducing the risks from exposure to the se-lected priority chemicals. The use of alter-natives in conjunction with regulation canbe particularly appropriate in those caseswh ere the num ber of exposed person s is rel-atively small (as d etermined in step 2); thealternatives are promising in terms of theirefficacy for reducing the risk (as determined

in step 3); regulatory approaches do not ap-pear to be cost-effective or technically feasi-ble (as determ ined in step 4); or the agencyhas determined that an optional allocationof its resources on the basis of its selectedpriority chemicals militates against pro-

mu lgating an d en forcing regulations (as de-termined in step 5). Thus, although the ra-tional approach to risk management pre-clud es regu lating such chem icals, the agencyassumes the continuing responsibility of fostering alternative approaches to managesuch risks in keeping with societal valueswhich support the protection of individualsfrom the dan gerous a cts of others. Followingthese actions, the agency has the respon-sibility to m onitor results and take necessarycorrective step s—e. g., in the case when theuse of alternatives has failed, the agencymay renew its efforts to foster the use of alternatives or decide to regulate (20).

Agencies can adopt either this specific risk management approach or some other approach,without congressional intervention or mandate.Baram proposes that agencies publish two rulesto govern their use of the framework in ord er to

guar d against ad hoc or case-by-case subjectivedeterminations. The first would describe theprocedures it will follow; the second, any agen-cy assump tions abou t subjective elements of itsanalyses.

This scheme shares draw backs with any thatinvolves ordering of regulatory goals. The deci-sion that a risk is number one may be chal-lenged, and concern about the accuracy withwhich the actual number one is identified mayprolong the ordering process. Also, the publica-tion of a number one risk, say, asbestos in step 5

above, might be accompanied with a list of “also-rans.” Whether such risks would go un-checked because it is clear that agency resourceswou ld never reach them or w hether voluntaryrisk reductions wou ld flow from the pu blicationis a pond erable question.

Baram’s step 3 may be the most critical dif-ference between what is now done and whatmight be done. Considering alternatives, dis-cussing and comp aring them, would better en-sure the pu blic and the regulated indu stry thatthe most efficacious approach was being se-

lected. It also incorporates the cost-effectivenessframework to choose between alternatives. Thescheme enjoys a powerful attractiveness in thatit considers nonregulatory management of riskswhile reserving the regulatory “club” if it isnecessary.

8/14/2019 8104


8/14/2019 8104


Ch. 6—Approaches to Regulating Carcinogens q 203

officials evaluating the evidence and analysis,sorting through their own personal values, per-haps testing the p olitical wind s, and th en comingto a conclusion. It is definitely not a p rocess thatcan be subjected to tidy r ules or guid elines.

The Administrator, who is appointed by thePresident, makes an d pu blishes the d ecision. He

is responsible for the decision mad e, and if it isshown to be incorrect, he can be removed fromhis position.

Not “Unreasonable Risks” andWhat To Do About Them

It seems likely that testing of chemicals willreveal a substantial number that present“limited” but not “sufficient” evidence for carci-nogenicity. The wor ds “limited” and “sufficient”are borrowed from the International Agency forResearch on Cancer’s (IARC) classification (see

ch. 4 and app. A). In the case of “limited,” theavailable evidence supports the idea that the

substance is a carcinogen, but it is less than “suf-ficient” to force a conclusion.

A regulator might consider a substance forwh ich on ly “limited” evidence exists as deserv-

ing of regulation. He m ight come to this d eci-sion because man y peop le or very sensitive peo-ple, such as young children, were exposed orbecause a substitute w as readily available.

Whatever h is reasons for considering regula-tion, economic assessment of most chemicals incommerce will probably show that significant

costs would attend any regulatory scheme. Inother w ords, often th ere may be a small risk andlarge costs. With those assessment results, it islikely that further analysis will not result in ananswer that says “regulate” under the formal,

rigorous balancing mechanisms discussed here.

A regulator who sees his first responsibility tobe the protection of public health might be u n-comfortable with that decision, and the pro-ducer of the substance, although not wanting tobe regulated, might be uncomfortable also withcontinuing exposure at cur rent levels. Methods

to deal with these situations through discussionand incentives might reduce the antagonism be-tween regulators and the private sector and pro-mote p ublic health (19, 20).

RESPONSIBILITY FOR MAKING DECISIONS

Quantification of risks and benefits is neverlikely to be so precise that estimates of these

variables can be p lugged into a formu la to pro-duce a number that dictates a decision. Instead,after all the an alyses, a decision about wh ether arisk is acceptable or unacceptable and reason-able or unreasonable will be made by a fewindividuals. Those judgments reflect societalvalues and wou ld, ideally, be made by citizensas a w hole. In ou r form of Government, electedrepresentatives are responsible for expressingsocietal values, and this responsibility is some-times delegated by elected representatives to ex-ecutive or jud icial bran ch officials.

Executive Branch DecisionsThe size and complexities of Government

have resulted in elected representatives del-

egating authority to make judgments aboutacceptable risk to executive branch agencies. A

number of commentors, including Markey(222), point out that many executive branch of-ficials are civil servants with almost lifetimetenure. While high-level appointees, such as theEPA Administrator, are at some risk if theymake p oor d ecisions, tenured civil servants ar enot. Citizens w ho feel aggrieved as a result of executive branch decisions have little oppor-tun ity for red ress (222).

Field (120) cites a number of legal s c h o l a r s

who contend that “vague mandates” such asTSCA’s directions to reduce or eliminate “un-reasonable risks” give too much discretion toagencies:

If regulatory decisions are to be broadly ac-ceptable, the governing statutes mu st do more

8/14/2019 8104



than p rovid e for decisions about wh at is safe orwhat is an un reasonable risk. They must also domore than m erely list factors to be considered.The legislative process must make the basicvalue judgments and tell the agencies how tomak e the n ecessary tr ad e-offs. . . .

Insofar as statutes do not effectively dictateagency actions, individu al autonomy is vulner-

able to the imposition of sanctions at the u nruledwill of executive officials, major questions of social and economic policy are determined byofficials wh o are not form ally accoun table to theelectorate, and both the checking and validatingfunctions of the traditional model are impaired(120).

Baram (19) draws attention to the frequent

absence of confessional direction about what

factors agencies are to consider in reducing risk.Absence of that direction, he says, causes theagencies to face extensive litigation.

Judiciary Branch Decisions

Markey (222) expresses concern about the judiciary becoming too involved in makingdecisions about acceptable risk. He sees suchdecisions as best made in the political arena byofficials responsible to the electorate. Judges, hepoints out, often enjoy lifetime tenur e and aremore r emoved from the electoral process thanare executive b ranch officials.

Vague definitions such as “un reasonable risk”that occur in the balancing laws are seen as in-viting legal challenge and judicial involvement

in risk decisions. The courts have generallygiven great weight to “agency procedu ral safe-guards, substantiality of evidence, and con-sistency” (120), but two recent developmentsmay alter such preference. As discussed above,the Bump ers amendment would erase the jud i-cial preference shown to agency expertise, andthe courts w ould entertain challenges to agencyexpertise. Judicial involvement is also increasedbecause of ind ustry challenges to agency ru lesand consumer, environmental, and labor orga-nization challenges to agencies for not regu-lating. Whatever the source of challenge, judi-

cial reviews have an impact on wh at level of risk will be acceptable. Should a court modify anagency-decided risk level, it is reasonable to

assume that the agency, in m aking its next deci-sion, will consider the cour ts’ decree.

Directions From Congress

Congressional attention to details about whatto balance and how to balance are seen as solu-tions to some of the problems of the regu latory

agencies. Interestingly, regulatory agency at-torneys interviewed by Field (120) generallyfavored balancing laws. They see their agencieswell able to do an adequate job of balancingrisks and benefits and evidently do not shar e theconcerns about vague man dates.

Attorneys for four environmental interestgroups were also interviewed (120). Three wereopposed to balancing laws and favored that theCongress impose clearer mandates for regula-tory action. Vague, balancing laws were seen asfavoring industry because of its greater re-

sources for influencing decisions in the agenciesand in the courts. The fourth en vironmental at-torney suggested that lack of money hamperedpublic interest group representation and thatFederal funding for preparation of their caseswould ease their difficulties.

Clearly, different laws impose different stand-ards, and the balancing laws are not specificabout w hat is to be balanced. Litigation resultsfrom these characteristics of the laws. Con-gressional intervention to d efine stand ard s andto detail wh at to balance should red uce litiga-tion from those sources.

However, there is no guarantee that clearerdirections from Congress will eliminate liti-gation—e.g., FDA’s banning of the artificialsweetener, cyclamate, begun in 1969, was con-tested until 1980 (and may be reopened). Of all

the carcinogen laws, the Delaney clause, which

was the basis of the cyclamate ban, provides the

clearest definition of a carcinogen, and it is thesimplest in the sense that it allows no balancing.Neverth eless, administrative law hearings con-testing th e qu ality of the evidence about carci-nogenicity were held off-and-on for a decade.

The FDA d ecision w as up held..Congress cannot engage in the day-to-day

business of agencies; it does not have the time. It

8/14/2019 8104


C/I. 6—Approaches to Regulating Carcinogens 20 5

must delegate authority to the agencies. If it agencies that no preventive regulatory action is

shares the opinions of some observers that it possible. On the other hand, Congress cannot

does not Provide su fficient direction to th e agen- intervene too often in regulatory m atters with-.cies about acceptable risk and balancing, it out hobbling

might provide the direction or it might more issues.

often exercise its right to intervene in regulatoryRegulations

activities. some and too

its capacity to deal with other

are seen as too burd ensome byweak by others, but the regula-

Some risks are inherent in either remedy. tions are required by law. The laws are congres-Overly strict directions, which provide many sional expressions that public policy requires apoints for jud icial review m ight so encumber th e certain level of protection for public health.

REGULATED CARCINOGENS

Table 38 lists 102 substances and categories of

substances regulated under the laws discussed

above. In every case, some evidence existed toindicate that the substance is a carcinogen. Inman y cases, evidence about th e substance was

generated in the NCI bioassay program (146)and/ or evaluated by IARC (185,186). The left-

hand columns of the table describe what conclu-sions were drawn about the human and animalsubstances by NCI and IARC.

IARC has classified 18 substances or proc-

esses as human carcinogens and another 18 asprobable human carcinogens. The data in table38 show that 20 of those chemicals are regu-

lated. Each of those 20 is identified by a “C , ”carcinogen or “PC” probable carcinogen, in the“H,” hum an evidence, colum n un der IARC.

About one-third of the chemicals tested byNCI or reviewed by IARC present “sufficient”evidence to conclude that th ey are carcinogensin animals and th ey can therefore be assum ed topresent a carcinogenic risk for humans (see,e.g., 185). Chemicals from those classes are in-

dicated with an “S” on the table in the “A,”animal evidence, column; the “I” classificationmeans the available evidence is limited butpresents a strong w arning of carcinogenicity.

About half of the chemicals reviewed byIARC and/ or tested by N CI presented neithersufficient nor limited evidence of carcinogenici-

ty. Only one chemical for which there is only in-adequate (I) evidence of carcinogenicity appearsin table 38. It is not possible to decide from the

data in the table that risky chemicals are being

regulated at the proper pace, but the data do

lead to the conclusion that nonrisky chemicals

(as judged by IARC and NCI) are not often reg-

ulated. The conclusion then suggests that reg-ulations are not so haph azard ly draw n as to reg-ulate large num bers of chemicals that p resent noor very little risk.

The absence of an entry under “NCI” or“IARC” does not necessarily mean that there ispoor or limited evidence about carcinogenicity.For instance, although the first substance, 2-ace-tylaminofluoren e, does not occur on either theNCI or the IARC list, it is an accepted animalcarcinogen (see ch. 5). Furthermore, otherchemicals have been reviewed since the IARC(186) publication, but the results of the reviewsare not yet available.

Some complexities of regulating carcinogensare demonstrated by the table. Some substancespresent a risk in locations covered by differentlaws, and separate regulations are necessary foreach exposure. Under CAA, EPA has pr oposedregulation for, or regu lated, 6 carcinogens andis considering an additional 24. Section 311 of CWA deals with oil and h azard ous spills, and isnot focused on regulating carcinogens, but “haz-ardous discharge reporting levels” have beenpromulgated for the listed chemicals, and carci-nogenicity was considered in setting thoselevels. The 49 substances for which regulation is

required un der section 307 of CWA were includ -ed in the law in 1977. Stand ard s have been setfor trihalomethanes, including chloroform,under the Safe Drinking Water Act. A few

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


Appendix Am —A Comparison of the National CancerInstitute’s and the International Agency for Research

on Cancer’s Evaluation of Bioassay Results

A working group assembled by th e InternationalAgency for Research on Can cer (IARC) reevaluated

data about 354 chemicals pr eviously evaluated anddescribed in the first 20 IARC monographs (185,186).

The working grou p d eveloped the following criteriafor grading evidence about carcinogenicity. The firstthree categories are essentially the same as those usedin earlier IARC compilations (344), and two newcategories were added:

Sufficient evidence of carcinogenicity indicates that

there is an increased incidence of malignant tumours:

(a) in m ultiple species or strains; (b) in mu ltiple experi-

ments (preferably with different routes of administra-tion or using different dose levels); (c) to an unusual

degree with regard to incidence, site or type of tum our, or age at onset. Add itional evidence may beprovided by data concerning dose-response effects, as

well as information on mutagenicity or chemicalstructure. Limited evidence of carcinogenicity means that the

data suggest a carcinogenic effect but are limitedbecause: (a) the studies involve a single species, strain,or experiment, or (b) the experiments are restricted byinadequate dosage levels, inadequate duration of ex-posur e to the agent, inad equate period of follow-up,poor survival, too few animals, or inadequate report-ing, or (c) the neoplasms produced often occur sponta-neously or are difficult to classify as malignant byhistological criteria alone (e.g., lung and liver tu-mours in mice).

Inadequate evidence indicates that because of m ajor

qualitative or quantitative limitations, the studies can-

not be interpreted as showing either the presence or

absence of a carcinogenic effect. Negative evidence means that within the limits of

the tests u sed, the chem ical is not carcinogenic. Thenumber of negative studies is small, since in general,studies that show no effect are less likely to be pub-lished than those suggesting carcinogenicity.

No data indicates that data were not available tothe working group (185).IARC (185) made the following statement about

the valu e of bioassay results:. . . in the absence of adequate d ata in hu mans it isreasonable, for practical purposes, to regard chemi-cals for wh ich there is sufficient evidence of carcino-genicity (i. e., a causal association) in animals as if they presented a carcinogenic risk for humans. T h e

use of the expressions “for practical purposes” and “asif they presented a carcinogenic risk” indicates that atthe present time a correlation between carcinogenicityin animals and p ossible human risk cannot be madeon a scientific basis, but rather only pragmatically,with the intent of helping regulatory agencies in mak-

ing decisions related to the primary prevention of cancer. (Emphasis in original. )

The largest single testing effort is the NationalCancer Institute’s (NCI)’s Carcinogenesis TestingProgram (now a part of the National ToxicologyProgram (NTP)). Griesemer and Cueto (146) ana-lyzed the resu lts of NCI’s testing 190 chem icals andplaced each tested chemical into one of nine clas-sifications (see table A-l). Ninety-eight of the 190were judged to be carcinogenic (classifications 1throu gh 5); 28 were equ ivocal (classification 6); and64 were noncarcinogenic (classifications 7 through9). The NCI results had earlier been reviewed by apanel of governm ental and nongovernm ental expertsand Griesemer and Cueto drew on that review intheir comp ilation.

A significant difference between the data analyzed

by Griesemer an d Cueto (hereafter referred to as the“NCI list”) and by IARC is that all the NCI experi-ments were carried ou t according to a stand ard p ro-tocol (331). The IARC evaluations consider experi-ments carried out u nder a variety of protocols includ-ing those done by NCI.

Thirty-three chemicals appearing on the IARC listof 354 also appear in the NCI list. In table 39, suchchemicals are listed in the class to which they wereassigned by Griesemer and Cueto. For instance, thechemical chloroform appears in the NCI class 1, andit is also present on the IARC list. Following eachlisted chemical is an S, L, or 1, The letters indicateIARC’s classifications of sufficient, limited , or inad e-

quate evidence for carcinogenicity. In some cases,additional data that became available between thecompilation of the 1978 (185) and 1979 (344) IARClists changed a chemical from I or L to S; thesechanges are reflected in the tab le.

Comparisons between the NCI and IARC lists arenot direct because different criteria were used, butthe following scheme m ay be u seful.

NC] Evidence IARC classification classification

St ron gly p os it iv e Su fficien t evid en ce (S) Classes 1, 2Positive Limited evidence (L) Classes 3, 4, 5

Not Positive Inadequate evidence (I) Classes 6, 7, 8, 9

Agreement is good about the strongly positivechemicals, Seven of the 12 NCI class 1 chemicals

were found to have “sufficient evidence” for car-cinogenicity by IARC. Of the remaining five, someare likely to be reevaluated by IARC. For instance,reserpine was classified in the “inadequate evidence”grou p by IARC before the results of the NCI bioassay

211

8/14/2019 8104


8/14/2019 8104


Appendix A—A Comparison of NCI’s and IARC’s Evaluation of Bioassay Resultsq 213

Table A-1 .—National Cancer Institute (NCI) Analysis of the Results of Testing 190Chemicals n and the International Agency for Research on Cancer (IARC) Analysis

of Chemicalsb c

That Appear on the NCI List (Continued)

Evidence for carcinogenicity, IARC1979c 1978b

Equivocal evidence in 1 or 2 species.28 chemicals

Chemicals among the 28 that appear on theIARC listsPhenacetin (tested by itself by I ARC; tested in combi-nation with aspirin and caffeine, APC, by NCI) . . . . . . L —

Styrene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . L —

NCI Group C: Three categories showing no evidence of carcinogenicity.

No evidence in limited experiments.51 chemicals

Chemicals among the 51 that appear on theIARC listsDieldrin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Endrin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Ethionamide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Lindane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Methoxychlor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

No evidence in 1 species.10 chemicals

1 chemical among the 10 appears on the IARC listsAnthranilic acid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

No evidence in 2 species.3 chemicals

None of the 3 chemicals appears on the IARC lists.

L — — — —

—

aGrlesemer #md Cueto, 1979.

%omatis, et al., 1978; Tomatis, 1979.clARC, 1979; IARC, 1980.ds Indicates “sufficient evidence of carcinogen icity. ”

L Indicates “limited evidence of carclnogenicity.”I indicates “Inadequate evidence of carcinogen icity.”

‘See text

of that dru g were available, and ortho-toluidine hy-drochloride is being reevaluated by IARC. Both NCIand IARC found that p henoxybenzamine hydrochlo-ride is carcinogenic. Therefore th e lists may only d if-fer significantly over two chemicals, and those dif-ferences, too, may disappear as more data becomeavailable.

Agreement is not as good about chemicals that areless positive. NCI’s classes 3, 4, and 5 roughly corres-

pond to IARC’s L group, but 5 of the 11 NCI class 3

compounds common to both lists were put in the “inadequate evidence” category by IARC. Additionally,NCI found no evidence for carcinogenicity for any

chemical in classes 6 through 9, but IARC found“limited evid ence” for carcinogenicity for five of theeight chemicals that it rev iewed in those classes.

As was men tioned, the data considered by IARCand NCI are not independent of each other; IARCconsiders NCI results in addition to all others. Thedifferences in classification may result from the d if-ferent criteria used by the two organizations, as wellas from IARC consider ing other resu lts.

Comparison of the 1978 and 1979 IARC listingsshows that the evidence for the carcinogenicity of some chemicals became m ore positive. The p rogres-sion to m ore-significant-evidence classes ma y be im-portant to the contention that repeated testing in-creases the chances that a chemical will be deter-mined to be a carcinogen.

Neither the IARC nor the NCI list include reviewsof all chemicals of interest. For instan ce, neither sac-charin nor sodium nitrite appears on either list, but

8/14/2019 8104


.


both compou nds have been mu ch discussed by regu-lators and potentially regulated indu stries.

The last sentences of the Griesemer and Cueto(146) paper draw attention to the importance at-

tached to positive results.

Those comp ound s for which evidence for carcino-genicity w as not foun d. . . cannot necessarily be con-sidered as noncarcinogens since the tests were con-ducted under a limited set of circumstances. It is possi-ble that evid ence for carcinogenicity m ight be foun dif, for example, a different strain of animal or a dif-ferent route of exposure were used.

The quote illustrates the imp ossibilityof proving a

negative, and in a more immediate sense, it alsoshow s that ru les are not established to allow classify-ing a chem ical as safe. A workable app roach to allowmaking a decision that a chemical should be regardedas safe rather than as a p otential risk may be neces-sary to separate important problems from minor andnonexistent ones. In qfact, Griesmer and Cueto’s

group C wh ich includes four grad es of negative evi-dence shows that conclusions can be drawn thatchemicals were found not to be risky under condi-tions of the tests.

8/14/2019 8104


Appendix B.— “Unreasonable Risk”

To learn m ore about th e concept of “un reasonablerisk, ” the assessment m ade inqu iries of a group of ex-perts and cosponsored a workshop about the subject.The inquiry was directed to 36 individ uals, selected

by the Assessment Advisory Panel and OTA Staff,and the workshop was held in March 1980 at theNew York Academy of Sciences (NYAS).

Responses to Letter Inquiries AboutUnreasonable Risk

Informed individuals were asked to comment onunreasonable risk. Each individual received the letterand attachments shown in figure B-1. The 22 w horespon ded are acknowled ged in table B-1.

Responses ranged from telephone calls or notes, toreprints of papers and speeches, to discursive letters.No attribution of a particular opinion to a specific in-

dividual is mad e in the d escription of responses thatfollows. The inquiry letter w as couched in r eferenceto the Toxic Substances Control Act (TSCA), but itdid not specify that discussions of “unreasonablerisk” were to be restricted to a balancing ap proach tocontrolling carcinogens. Both zero-risk and balanc-ing appr oaches were mentioned in the responses, but

no response mentioned a technology-based approach

to regulation.

Zero-Risk Approaches to Unreasonable Risk

A number of respondents treated cancer, especiallyworkplace-related cancer, as an unreasonable risk,regardless of the number of people affected. Pro-

ponents of a purely health-based reading of un-reasonable risk contend that workers should sufferno imp airment of health as a result of workp lace ex-posures.

These responses did not reflect a naive positionthat all workplace exposures can be eliminated im-mediately. Instead, it represented a starting positionfor regulatory efforts. A respondent said that, “cur-rently wh en a carcinogen is iden tified, the first step isto estimate how much of it can be tolerated. ” Hewould favor instead that the first step be to set as agoal the elimination of exposure. If that is impossi-ble, each “essential” exposure could be considered inturn to construct a pattern of allowed exposures

above zero. Every effort should then be made toreduce the allowed exposure.Another respondent likened the present w orkplace

situation to Thomas More’s Utopian caste system, inwhich laborers may be sacrificed for the greatersocietal good:

Not accepting the caste system, only necessary riskswould be taken and a standard would be promulgatedacceptable to labor, but not to management . . .[Workers] want standard s set so there will be no in-creased risk of getting cancer as a result of their expo-

sure to chemicals in the working environment.Recognizing that zero exposure levels will not be pos-sible in all situations, limiting exposure to low levelsis seen as an interm ediate goal to be accomplished byimposing every available control measure. There isno balancing in this app roach and the respond ent ob-

jected to “th e entire p rocess of weighing d ollar coststo employers against the value of a human life. ”

Another respondent advocated applying all avail-able controls for workplace exposures to known car-cinogens and specifically concludes that the time forquantitative risk assessment has not yet arrived. Hecited the continued presence of such obvious and in-disputable hazard s as asbestos, benzidine dyes, and

aluminum -redu ction pot-room carcinogens as ex-amples of risks wh ich need immediate attention. Hestated that even when control techniques are avail-able, they are not required or applied to the fullestextent.

Interestingly, an alternative health-based approachto unr easonable risk involves a very different meth-odology and leads to different actions. Quantitativerisk assessment would be used to estimate hum an risk for each identified hazardous substance. All esti-mates wou ld be expressed as the risk that a personmight d evelop cancer dur ing his lifetime as a result of exposure to the substance at the level now en-countered. The risks would be expressed as 1/ 10,

1/ 100, 1/ 1,000, etc. Some v alue for th e risk factor,would be designated as a critical value. Risks lessthan that value wou ld be tolerated; risks greater thanthat value wou ld be declared u nreasonable and can-didates for regulation. Costs of regulation w ould notintrude into the d ecision to regulate.

Balancing App roaches to Unreasonable Risk

The majority of respondents indicated a preferencefor a form of balancing some combination of risks,costs and benefits. The spectrum with in this ideologi-cal grouping how ever, ranges from su ggesting verysubjective, case-by-case determinations, to the use of

formal quantitative calculations that could be ap-plied to all cases. Individuals who place little faith inquantitative risk assessment and economic analysisof costs and benefits cluster around the “subjectiveand qualitative” position. The “objective and quan-titative” position is occup ied by p eople who are more

215

8/14/2019 8104


8/14/2019 8104


Appendix B. –“Unreasonable Risk” . 217

Table B-1.

The OTA thanks the following individuals who respondedto a letter inquiry about unreasonable risk.

John T. Barr, Air Products andChemicals, Inc.

Jan Beyea, Audubon SocietyEula Bingham, Occupational Safety and

Health Administration

Ralph Engel, Chemical SpecialtiesManufacturers Associa-tion

P. J. Gehring, Dow Chemical Co.Harold P. Green, Fried, Frank, Harris,

Shriver, and KampelmanFred Hoerger, Dow Chemical Co,Peter Hutt, Covington and BurlingKenneth L. Johnson, Clement Associates, Inc.Lorin E. Kerr, United Mine Workers of

AmericaArnold Kuzmack, Environmental Protection

AgencyLinda B. Kiser, Consumer Products Safety

CommissionLester Lave, The Brookings InstitutionWilliam J. McCarville, Monsanto Co.Richard A. Merrill, University of VirginiaFranklin E. Mirer, United Auto WorkersF. W. Mooney, The Proctor & Gamble Co.Parry M. Norling, E. 1. du Pent de Nemours

& Co., Inc.Glenn Paulson, Audubon SocietyDavid P. Rail, National Institute of

Environmental HealthSciences

Sheldon W. Samuels American Federation ofLabor/Congress ofIndustrial Organizations

confident of the precision of quantitative risk assess-ment and economic analysis.

A representative of the Consum er Produ ct Safety

Commission (CPSC) described CPSC interpretationof un reasonable risk:The legislative history of the Consumer Product

Safety Act (CPSA) indicates that unreasonable risk of injury is to be determined by balancing the prob-ability that the risk will result in harm and the gravityof the harm against a rule’s effect on the product’sutility, cost, and availability to the consu mer. Thus,in addition to an assessment of the risk that a rule willeliminate or reduce, an important component of un-reasonable risk of injury is the economic imp act of aplann ed regu latory initiative. The legislative historyexplains that an unreasonable risk is one which can beprevented or reduced with little or no economic im-pact; or one wh ere the ru le’s effect on a pr odu ct’s utili-ty, cost or availability is outweighed by the need toprotect consumers from the hazard associated withthe product.

One respondent made the point that our knowl-

edge of carcinogenicity forces a balancing approach.

As a policy matter, there is no recognition of a“thr eshold d ose:” the safe haven of a “no effect” leveldoes n ot exist for carcinogens.

The subjective qualitative approach is exemplifiedby a lawyer wh o wrote: “A substance constitutes an

‘unreasonable risk’ only where the probabilityand/ or the severity of harm are deemed to outweighits utility.” He explained that all p ertinent elements,

“probability, severity, harm, utility, ” in that state-ment are highly subjective, and that their definitionsare shaped by p articipating individuals, who may in-clude: “legislators, regulators, business persons,voters, writers of letters to the editor, interestedcitizens, etc. ” This type of balancing em ph asizes sub-

jective judgments in qualitative determinations andlimits to purely objective determinations.

Another lawyer describes the p rocess of unreason-able risk determinations as:

. . . paramount to a cost-benefit or risk-benefit de-cision. In essence, it relies as much on procedure as onsubstance for correct decisions. The concept is that, if all potentially relevant information is taken into ac-

count and all interested parties have an opportunity tocontribute to the proceedings, the ultimate decisionwill be rational and as sou nd a s it possibly can be inan area where there is no certainty. While that isperhaps not acceptable to a mathematician, 1 see littlepossibility of improving upon it as long as the conceptremains in statutory language.

Another respond ent sees the language of TSCA asallowing a broad range of options for regulatory ap-proaches. He suggests that TSCA rulings might fol-low the pattern of the Food and Drug Administra-tion’s develop men t of criteria for dr ug effectivenesstests, beginning as loose, intuitive, ad h oc decisions,later becoming codified into lengthy regulations.Once that state is reached, it is difficult to depart

from the formula.A variation on this theme came from an indu strial

executive, who feels that d eterminations shou ld bequalitative, and decided on a case-by-case basis.How ever, he calls for explicit criteria to be add ressedin balancing, listing as exam ples:

q

q

q

q

q

q

q

level of hazard inh erent in exposure to the chem-ical;types of exposure w hich create the hazard;popu lations likely to be exposed;extent of the exposed p opulations;whether the exposure would be voluntary or in-voluntary;availability of substitutes;

the w orth to society of continued availability of the substance;cost inherent in various levels of control orelimination;

8/14/2019 8104



nature of the data suggesting the existence of ahazard;ability to extrapolate from those data to predictthe hazard in other situations; andrelative imp ortance of regulating the p articularchemical in view of the risks presented by otheras yet unregulated chemicals.

A chemical trade association representative ad-

dressed the issue broadly:Many criteria or ap proaches for one substance will

not be ap prop riate for another. We believe that thereis no formula for such an assessment, but th at eachcase mu st be decided on its ow n m erits. This is not tosay that scientific methodology should not be u sed,but it does say that the foundation and assum ptionsthat lead to the d etermination of “unreasonable risk”should be clearly identified, along with their limita-tions, by the regulatory agency on a case-by-casebasis and not throu gh a generic app roach.A pu blic interest advocate deplored th e high de-

gree of quantification of risks, costs, and benefitsemanating from the Environmental Protection Agen-cy (EPA), wh ich are seen as based o n meth ods inad e-

quate to the task. The large areas of uncertainty sur-roun ding these estimates are rarely specified, espe-cially for economic evaluations. Risk estimates, hefeels, are good only for the grossest distinctions be-tween the most and least potent carcinogens. Bal-ancing, therefore, should rely heavily on qualitativedecisions, in w hich d oubts m ust be resolved, by thelangu age of the law, in favor of greater pr otection.

A middle ground position is taken by many re-spondents, who would like to see all of the availablequantitative information enter into the final reg-ulatory d ecision, but modified by such factors thatare less readily and more controversially qu antified.Within this grou p, there are th ose who feel that even-

tually all factors will be quantified, and that moreenergy should be directed at methodologies forreaching that goal. Others feel that certain measures,especially the value of a human life, cannot be con-verted to monetary terms.

Some industry respondents lean toward quanti-tative ap proaches to d eciding un reasonable risk. Oneexpressed his belief that risks from hazard ous chem-icals are overstated, and that regulation probablywill not materially benefit health. This respondentplaces confiden ce in risk assessmen t:

The question of “risk” can be assessed with somecertainty. “Unreasonable” is by definition a matter of perception, and would include consideration of thebenefits of a substance and the cost to regulate it.

As a part of risk analysis, respondents from acrossthe spectrum advocate risk comparisons of d ifferenttypes. An industry rep resentative thinks it would be

useful to compare the risk from exposure to a carci-nogen ic chem ical to other r isks we accept willingly indaily life.

A highly quantitative cost-effectiveness approachis promoted by one economist. Measures thatachieve the greatest risk reduction, most economical-ly, would be the first to be required. This method re-quires risk assessments for known carcinogens and

identification of carcinogens with exposures that canbe redu ced. This respondent suggested a 2-stage ap-proach, accepting red uctions to an intermediate levelwh ich maximize risk red uction for each d ollar sp ent,How ever, final limits wou ld be set requ iring an ex-plicit valuation of life. He acknow ledges tha t an im-plicit value for life is foun d abhorrent by som e, but,he says: “it is inherent in all toxic subst ance regu la-tion. ” This suggested scheme w ould not allow for aregulation that was less cost effective than an alter-native, nor any final regulation which cost more perhum an life saved th an the value placed on a life.

Acceptable Risk and Consum er Choice

Unreasonable risk d eterminations w ere not alwaysdiscussed in the context of a deliberative bodymaking a decision. Individuals, it is pointed out,make risk-benefit decisions continuously, and muchmight be learned from those behaviors in which in-dividuals decide what risks are acceptable. Some re-spond ents prefer to approach u nreasonable risk froma perspective of “acceptable risk. ”

An indu stry representative expressed the view :. , . this more positive direction of attack is useful inconsidering prospectively the extent to which regula-tion should be carried, as opposed to a retrospectivereview to see if regulatory actions were adequate.

Amplifying, he explains:

In the long run, acceptable risk is the perceived risk to which informed persons do not object. That state-ment implies a higher level of knowledge than thegeneral population now has, and acknowledges theemotional issues present in any cancer controversy.

This respondent joins those who call for improve-ments in present methods of risk and benefit es-timation, but add s:

It seems probable that them is no general answer toacceptable risk, and no numerical value that is correctfor all situations.

The issue of whether a risk is taken voluntar ily orinvoluntarily is a major determinant of levels of haz-ard that are considered acceptable. It is sup posed, in

general, that higher risk is acceptable for a purelyvoluntary exposure than for an involuntary on e, anddifferent approaches to regulation may be appro-priate for different degrees of voluntarism.

8/14/2019 8104


Appendix B.–“Utlreas(~?lab/c Risk” q219

A mem ber of the legal commun ity, concerned par-ticularly with consum er prod ucts, advocates a “con-sumer choice” approach:

. , . under which the Government would assess therisks but would not have authority to ban on the basisof this risk assessment, would be responsible forassuring that information about risk is provided toconsumers through product labeling and other educa-tional techniques, and would leave the ultimate risk-

benefit judgment to consumers. The uniqueness of thisapproach is that the consumer, rather than the gov-ernment, would make the “regulatory” decision of whether human exposure would occur. Some con-sumers would conclude that their benefits outweighthe risks and would use the product; others wouldconclude the opposite and w ould not u se it.

Concerning the benefits derived from consumergoods, he explains:

Most benefits are psychological or in any eventnonobjective and nonquantifiable in nature. In thearea of food . . . virtually all people eat not for healthor nutrition reasons but rather for pleasure. If one eatsenough of virtually anything that one likes, one will,after all receive sufficient nutrition. Thus, choice of

food—i.e., consumer benefit—is wholly subjectiveand personal in nature. With the exception of a veryfew essential nutrients, in our country no food can besaid to have any “benefits” in the sense that they are inany way essential or irreplaceable. And if you define“benefits” to include simple consumer desire forpleasure, the term soon becomes meaningless.

1 believe this same analysis holds true not just in thearea of food, but indeed for all consumer products.And because TSCA deals with chemicals broadly, itultimately impacts upon all consumer goods.

Op inions Expressed in the Responses Parallelthe D iversity of Laws

From banning of carcinogens based only on con-sideration of risk, to balancing risks, costs, andbenefits, to an antiregulatory position of letting theconsumer decide, the responses parallel develop-ments and discussions in regulatory law. This rangeof attitudes is not surprising because many re-spondents have testified before the legislature or beenparty to lawsuits about carcinogen regulation. Onthe other hand , the spectrum of opinion confound sconclusions that environmental laws can be fit intoan evolutionary pattern and that some approaches,for instance, zero-risk laws, no longer haveadherents.

The evolutionary analysis suggests that more so-phisticated laws, such as TSCA, with its balancing,

have n ow displaced risk-only app roaches. Overall,responses received by OTA indicate that risk-onlyregulations are favored by some for the w orkplaceand that “evolution” has not displaced all interest inrisk-based regulations. In fact, the responses show

that the breadth of carcinogen regulatory ap-proaches, including the old and the new laws, reflectsthe spectrum of current thinking. The responses d idnot produ ce a “new” app roach to determining unrea-sonable risk. Quantitative risk assessment is an in-tegral part of many of the responses, and its frequentmention reflects the perceived importance of thistechnique.

The New York Academyof Sciences Meeting

The title of the NYAS meeting, “Workshop onManag emen t of Assessed Risk for Carcinogens, ” waschosen to show that the focus of the meeting w as notrisk assessment. The meeting lasted 2-1/ 2 day s. Therewas some overlap betw een participants in th e meet-ing and responden ts to the OTA letter, and becauseof that and th e inherent problems of captur ing a longmeeting in a few pages, no attempt will be made hereto sum marize the m eeting. This short description w illhighlight some p oints made there that d id not arise in

any oth er context in this assessment.William Ruckelshaus of the Weyerhaeuser Co., a

former EPA Administrator, was asked why industrydoes not remove risks in advance of regulation. Hisanswer had two parts. First, he said, “industryleaders often do not know the extent of the health orenvironmental problems. ” Second, a “company thatspends m oney on a pr oblem in ad vance of regulationmay fall behind its competitors who m ake the sameprod uct but do not spend for risk reduction. ” He sug-gested that regulations are sometimes welcomed be-cause all competitors ar e affected.

Peter Preuss described CPSC’s approach to reg-ulation of carcinogens. Each of CPSC’s actions has

originated from a petition asking for agency action.All but one of the chemicals had been regulated byanother agency. His analysis of CPSC actions ledhim to conclude that CPSC has employed no over-arching method to settle unreasonable risk qu estions;each d ecision w as largely indep endent of the others.

A number of speakers made the point that manychemicals are carcinogens wh en tested and that thereare perhaps hundreds of candidates for regulation.Lively discussions broke out between those whowan t to ord er carcinogens on the basis of their poten-cy and those who hold that extrapolation is tooimprecise to estimate p otency.

Surp risingly enough, there w as general agreement

about the value of cost effectiveness as a method toplot regu latory strategy. Cost effectiveness involvesordering the risks, estimating the cost of reducing oreliminating the risk, and d eciding w hich expend itureof resources will accomplish the largest red uction in

8/14/2019 8104


8/14/2019 8104


Appendix C.—Abbreviations and Glossary of Terms

Abbreviations

2-AAFACS

AIHCB(a)PBATBPTCAACAGCAICDCCEQCIITCPSACPSC

CSIN

CWADESDHEW

DHHSEPAFDAFDCAFIFRA

FILSFSC

gGAOGRASHANES

HCFA

HDLHISIARC

ICD

IOMIRLGITCkg

LASSmmg

2-acetylaminofluorene

Amer ican Cancer Society

American Industrial Health Councilbenzo(a)pyrenebest available technologybest pr actical technologyClean Air ActCarcinogen Assessment Grou p (EPA)Carcinogenicity Activity In dicatorCenters for Disease Contr ol (PHS)Council on Environmental QualityChemical Indu stry Institute of ToxicologyConsumer Product Safety ActConsumer Product Safety CommissionChemical Substances Information Network

(EPA)

Clean Water ActdiethylstilbestrolDepartment of Health, Education, and

WelfareDepartment of Health and H uman ServicesEnvironm ental Protection AgencyFood and Drug Ad ministration (PHS)Food, Drug, and Cosmetic ActFederal Insecticide, Fu ngicide, and

Rodenticide ActFederal Inform ation Locator SystemFood Safety Coun cilgramGeneral Accoun ting O fficegenerally recognized as safe

Health and Nu trition Examination Survey(NCHS)

Health Care Financing Administration(DHHS)

high d ose levelHealth Interview Survey (NCH S)Internationa l Agency for Research on

Cancer (WHO)Internation al Classification of Diseases

(WHO)Institute of Med icine (NAS)Interagency Regulatory Liaison Grou pInteragency Testing Com mitteekilogram (1,000 g)

Linked Ad ministrative Statistical Samp lemetermilligram (one-thousand th of a gram;

10 -3

g)

MTDNAS

NCABNCHS

NCINCTR

NDIngNIEHS

NIHNIOSH

NOHS

NORS

NRCNRDCNSFNTPOECD

OHRST

maximum tolerated doseNationa l Academy of Sciences

Nationa l Cancer Advisory BoardNation al Center for H ealth Statistics(DHHS)

National Cancer Institute (NIH)Nation al Center for Toxicological Research

(EPA/ FDA)Nationa l Death Index (NCHS)nanogram (one-billionth of a gram; 10

-9g)

Nationa l Institute of Environm ental HealthSciences

Nationa l Institutes of H ealth (PHS)Nationa l Institute of Occupational Safety

and Health (CDC)National Occup ational Hazar d Survey

(NIOSH)Nation al Organics Reconnaissance Survey(EPA)

National Research Coun cil (NAS)Natural Resources Defense CouncilNationa l Science Found ationNational Toxicology Program (DHHS)Organization for Economic Cooperation

and DevelopmentOffice of Health Research, Statistics, and

TechnologyOSH Act Occupational Safety and Health ActOSHA

OSTP

PAHPCBsPH SPMNRCRARPARSABSDWASEER

SNURSSATSCATSSC

ug

WH O

Occupational Safety and HealthAdministration (Department of Labor)

Office of Science and Technology Policy

polycyclic aromatic hydrocarbonspolychlorinated biphenylsPublic Health Service (DHHS)preman ufacturing noticeResource Conservation and Recovery Actrebuttable presum ption against r egistrationScientific Advisory Board (EPA)Safe Drinking Water ActSurveillance, Epid emiology, and End

Results program (NCI)significant new use ru leSocial Security Ad ministration (DHH S)Toxic Substances Con trol ActToxic Substances Strategy Com mittee

microgram (one-millionth of a gram;10-’ g)

World Health Organization (UnitedNations)

221

8/14/2019 8104



Glossary of Terms

Benign tumor: A tumor confined to the territory inwhich it arises, not invad ing surroun ding tissueor m etastasizing to distant organs. Benign tu morscan usu ally be excised by local surgery .

Carcinoma: Cancers of the epithelia, including theexternal epithelia (mainly skin an d linings of the

gastrointestinal tract, lungs, and cervix) and theinternal epithelia that line various glands (e.g.,breast, pancreas, thyroid).

Bioassay: In general, a test in living organisms. Asused in this report, a test for carcinogenicity inlaboratory animals, generally rats and mice,which includes near-lifelong exposure to the agentunder test. Used interchangeably with “animaltest .“

Carcinogen: A substance that cau ses cancer.Epigenetic: As used in reference to cancer, an effect

on cancer causation that does not directly involvean interaction w ith DNA.

Epitheliums: The covering of internal and external

surfaces of the body, including the lining of ves-sels and oth er sma ll cavities.Incidence: The number of new cases of a disease,

usually expressed as a rate:

Number of new cases of a d isease occurring in a popu lationduring a specified period of time

Num ber of persons exposed to risk of developing the disease

during that p eriod of time

The inciden ce rate is a direct estimate of the p rob-ability, or risk, of developing a disease during aspecified p eriod of time.

Initiator: An external stimulus or agent that pr odu cesa cell that is “latently premalignant. ” An initi-ation event, or more generally, an early event,

may be a mutational change in the cell’s geneticmaterial, but the change is unexpressed, and itcauses no detectable change in the cell’s growthpattern. The change is considered to be irrevers-ible.

Leukemia: Cancers of the blood-forming organs,characterized by abnorm al proliferation and d e-velopment of leukocytes (white blood cells) andtheir precursors in the blood and bone m arrow.

Lymphoma: Cancers of cells of the immu ne system,i.e., the various types of lymphocytes. Hodgkin’sdisease is included among the lymphomas.

Malignant tumor: A tumor that has invaded neigh-boring tissue and/ or und ergone metastasis to dis-tant bod y sites, at wh ich point the tu mor is calleda cancer and is beyond the reach of local surgery .

Melanoma: A tumor m ade up of melanin-pigmentedcells. As used in this report, “malignant mela-noma, ”

Mesothelioma: A tum or d eveloping from a cell onthe surface of the peritoneum (the membrane lin-ing the abd ominal cavity), pericardium (the mem-

brane enclosing the heart), or pleura (the mem-bran e lining each half of the thorax).Metastasis: The spread of a malignancy to distant

body sites by cancer cells transported in blood orlymph circulation.

Morbidity: The cond ition of being d iseased.Mortality rate: The death rate, often made explicit

for a particular characteristic, e.g., age, sex, orspecific cause of death. A m ortality rate containsthree essential elements: 1) the number of peoplein a population group exposed to the risk of death; 2) a time factor; 3) the number of deaths

occurring in the exposed population during a cer-tain time period. For example, the annual U.S.cancer morta lity rate is:

Num ber of deaths from cancer in the United States during 1 year

Number of people in the population at midyear

Mutagen: A chemical or phy sical agent tha t interactswith DNA to cause a permanent, transmissiblechange in the gen etic material of a cell.

Myelomatosis: A m alignant neoplasm of p lasma cellsusually arising in the bone marrow. Also calledmu ltiple myeloma.

Neoplasm: A new grow th of tissue in which growthis uncontrolled and progressive. A tum or.

Nonmelanoma: Skin cancer of tw o typ es: basal celland squamou s cell carcinomas. Though these tu-mors may invade surrou nd ing tissue, and there-fore are technically cancers, they seldom metas-tasize and are usu ally successfully treated withlocal surgery. Because they are relatively easilytreated, often outside hospitals, cause relativelyfew deaths, and are not often enumerated, theyare usu ally excluded from cancer statistics.

Prevalence: The number of existing cases of a dis-ease, usually expressed as a rate:

Number of cases of a disease present in the population at(or during) a specified time (period)

Number of persons in the populaton at

(or during) the specified time

Promoter: An influence or agent causing an initiatedcell to produce a tumor. Promotion events, ormore generally, late events, can occur only in

8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104


8/14/2019 8104
