?My" yuan \l Else manila Egme?etas the Politieai [enemy or Eisk Social and power relationships between workers, management, and inspectorates are as important as science in any realistic analysis of industrial risk. A shift in this balance of power is a pre-requisite of any policy for the control of risks This Series has shown us just how complicated a ?risk equation" is. Certainly, quantitative scienti?c studies of hazards are essential. But their results cannot be treated like weights that are simply balanced against estimates of ?values?. Science, power and ethics are intimately related in every assessment of risks. The more we know about risks, the more we become aware of the ways their assessment is conditioned by goals and. values now dominant in society. Through studying risks, we can come to question and to improve these. And a realistic analysis of the relations involved in the "political economy of risks" is necessary for any real improvement in their control. Such a study requires a new sort of scien- ti?c work, in which the committed amateur can match his skills against the expert. Some risks, particularly those occurring in ordinary experience such as car accidents, can be described quite well by statistical methods. By relating statistical indi- cators of personal and political reactions to risks, as Lord Ashby has done (New Scientist, vol 74, 598), we get some idea of the practical limits of "acceptability" of various risks. But the more we know about these limits, the more complex and even bizarre they seem. For example, pres- sure groups concerned with particular industrial or medical risks will argue endlessly about hypothetical scenarios, while the twin drugs of speed on the road and alcohol are allowed (and, by of?cial inaction, encouraged} to regularly claim thousands of victims each year. . Any effective policy of risk control must come to terms with such paradoxical features of risks. I believe that they can be expressed in terms of three ?unthinkables?. First, as a matter of personal hazards with a low probability are unreal; we have no intuitions to help us balance costs and payods when the odds against losing are Secondly, and related to this, is the severe dif?culty of studying hazards scienti?cally. Elaborate calculations of the probabilities of complex harmful events are all too likely to be mere pseudoprecision, for accidents and disasters do not follow the statistical model of successively picking balls out of separate urns. Finally, there is the moral dilemma of anyone who imposes a risk on another, through their activity as designer, inSpector, supervisor or worker. A zero probability of harm is just impossible; but if harm occurs, there will always be the question. ?though I did my best. was it as good as it should have been?? Because the ?unthinkable" dilemmas cannot be sup- pressed in the day-to?day practice of risk control, that practice can never be reduced to the ?puzzlesolving within a paradigm" that characterises the "normal science? Thomas Kuhn describes. Although there is a great need for more competence, any attempt to create a closed, esoteric body of technical expertise would be deceitful, and (in the present climate of public involvement in science) ineffective. Some might indeed wonder whether there is any chance of achieving rational dialogues on particular contested risks and {more dif?cult) well-planned programmes in broad areas of risk. This problem emerges 'Dr Jerry Raver: drafted the Council for Science and Society's report, The Acceptability of Risks more forcefully than perhaps any other from the protracted Windscale inquiry. The risk triangle We can make a start on managing risks as a social phenomenon by recognising that there are three sides involved in every hazard: those who create it; those who experience it; and those who regulate it. Sometimes all the ?sides" come together in the same person (say, a mountain- climber). But for most technological risks the sides are largely separate. Although all share in creating, experienc- ing and regulating risks, broadly speaking, in industry it is managers, workers, and health and safety inspectors whose basic responsibility and commitment lie on the three different sides of the triangle. Since risks are so dif?cult to study objectively or even to imagine, it is. only natural that the way each ?side" sees a hazard depends strongly on the values and expecta- tions of its role, and that this perception will be very dilierent from that of another side. Hence a manager need not be callous or inhumane to allow a hazard to persist even when warned about it; he just doesn?t necessarily see it the same way as others. Nor need the worker take it too seriously, especially if he or she has real ?danger money? to compensate for an unreal slig chance of future harm. or, as is often the case, does not know the extent of the risk. This argument does not, however, excuse crimes: no doubt a murderer or rapist sees the crime in a different light to its victim or to a policeman! Another consequence of the risk triangle is that the style of the control of risks will mimic, and be in?uenced by. wider social relations of power. A hazardous environment (at work or at home) is a part of social powerlessness. Hence all debates on risks have an inevitable and inescap able element of politics in them. Although bargaining on risks certainly involves the paradoxes of ?the three un? thinkables", a risk policy that excludes such bargaining is all too likely to serve the interests of only one side of the risk triangle: that with the most power to shape perceptions and values. Recognising this basic structure makes it easier to isolate the in?uences that make the control of risks much more than a technical exercise. For example, the government agencies that regulate risks work in a bureaucratic context. where ?success" is assessed by many criteria besides the satisfaction of their publicly-stated function?most notably the classic civil service aims of departmental well-beinu and a peaceful life. Such constraints, when added to the dif?culties of imagination and analysis mentioned above. can lead to a total fragmentation of perception and, responsibility among agencies charged with regulation. Further, mechanistic solutions, such as ?more flow of information" can be counter-productive. Faced with more memos, people throw away all the ?bumph?. The existence. of such in?uences, analysed by Barry Turner, indicates that the modelling of disasters as a series of independent random events is grossly over-simpli?ed. We can appreciate more of the distortions of the func- tions of ?science" in risk control by putting aside the ?pub lic knowledge? model of the social uses of science. and thinking instead .of ?corporate know-how". For manage- merit, and also regulators, desire (or are even legally re- quired) to keep to themselves much of the knowledge about risks that gives them power. And the data that are collected, indeed the very categories in which they are cast, reflect a conception of the problem that is in?uenced by the perceptions and values of the ?side" that has the power to collect and process the data All sides of the risk triangle are involved in its dilemmas of "designing for death"; but the of?cial regulators are particularly vulnerable to corruption of their work. For, in reality, they depend on the goodwill of managers to get improvements carried out, despite the great formal powers they may hold. This forces them to accept unsatisfactory conditions in the short, medium, and even longer terms. Yet, through all this, they must present themselves to the public as effective and sole guardians of safety. When the workers actually running the risks react cynically to the inspectorate?s claims, the regulatory agencies are driven still closer to management, in order to preserve their public image and self-esteem. The end result of this process is that factory inspectors have often appeared to workers as little more than adjuncts ??even adjutantswof management, trapped by their de- pendence on the philosophy of persuasion. Lest this seems too harsh, we need only refer to the recently-revealed evidence about the ineffectiveness of the Factory Inspectorate in controlling asbestos hazards at the Hebden Bridge and east London factories. O?icial muddle and subsequent cover-ups are as much an essential part of the hazard as the physical causes and social conditions which led to the event. We can now begin to understand why the poor and powerless are worse afflicted by accidents and disasters. both in the frequency of their occurrence and in their efiects. We also see how the ?Superstar Technology" phenomenon _(to quote Professor John Ziman?s report for the is more a matter of social organisation than of scale of technical Operation. A classic example of this was provided by an ?establishment" Speaker at a US National Academy of Sciences forum on genetic engineering last March. Dr D. Callaghan confessed that hitherto the burden of proof of a hazard from this kind of research had been on those who warned about risks; the effective principle being to carry on with experiments unless convinced of the dangers. Since the US machinery set up to regulate genetic engineer- ing has been monopolised by researchers in the field, excluding even the ?dangerous pathogens? experts as well as workers? representatives and lay persons, we have had there the elements of a textbook case of a future science- based catastrophe. If risks are so deeply embedded in the human condition, no simple administrative device can solve the problem. A change in the balance of power is a pro-condition of improvement: such a shift depends not only on laws but on changes in consciousness and conscience. We should think not of a once-for-all solution, but in terms of cyclic process, of advances on different fronts according to effectiveness. And we should remember that there will inevitably be a reaction to every action that threatens existing power relations. The Latin motto Quis custodiet ipsos custodes? (who guards the guardians?) ShOl-id be- displayed on the portals and letterheads of every inspec- torate! In this context, plant-level bargaining about hazards, perhaps spurred by the creation of union safety representa- tives next year, is one step in increasing consciousness. Similarly, proposals in the CSS report The of Risk (to be published later this month) for community risk advisory centres could, if implemented, provide a framework for bargaining. The study of risks is a clear example of the sort of science that is now emerging as salient in the management of our high?technology society. Here, as in the ?elds of resources and environmental problems. the established scientists have lost their professional monopoly of legitimate expertise. Much of the accepted research and argument has been achieved by students, journalists. amateurs and pressure- groups. In this sort of science the political commitments are open rather than smuggled in; there is no fear of over- democratising the issues. And the distinction between ?facts" and ?politics" is achieved practically by detailed debates, rather than ideologically by a fundamentalist faith. .. Hu- nu v-u pun-V. v. Javelin Night Viewing Devrces bring pho ographs out of the dark. No infrared to taint studies. More and more. physical and social scientists. technical photog- raphers and others are turning to Javelin Night Viewing Devices (NVDs) for photographing and see- ing in the dark. For those performing experiments. the elimination of infrared light subtracts one more variable in their research data. Javelin NVDs are presently being used for emission or "smokestack" research; studies of the nocturnal habits of mammals. reptiles and insects; and sleep patterns of humans. A major TV network exposed drug use of American soldiers in Germany. Another network verified Highway Patrol complaints of night- time driver abuses Whatever you?re studying or photographing?don't be kept in the dark. Let a Javelin NVD Open your eyes. A range of models is available to fit on any camera?stiltmovie or TV For details. contact: JAVELIN ELECTRONICS Subsrdiary of Walter Kiddo 8. Company Inc KIIE 6357 Arizona Circle Los Angeles. CA 90045 Phone (213) 641-4490 Telex 698204 Circle No. 359 on Flaaders' Servlce Card lism experiments are essential in eluci- dating what causes tumors in the animal model. The implications to public health differ depending on whether one is deal- ing with a potent direct-acting carcino- gen. or an opportunistic carcinogen ca- pable of doing harm only to mammals of severely impaired resistance. or an agent providing the opportunity for ubiquitous carcinogens to become effective. The ?bioassay? approach imposes a dogmat- ic and narrow interpretation of tumor in- cidences and discourages broader stud- ies needed to advance our knowledge of what contributes to tumor formation. Only full consideration of physiological effects on a case by case basis can lead to credible risk assessment. An encourag- ing note is that Food and Drug Admin- istration's advisory committees have provided ?exible responses to "bio- assay" data. pointing the way to more balanced risk assessments M. SHIACH vos Sit/ivy and Drug Metrihm?is-ut. Biker Research. P?zer. Inc. Grown. 06340 References and Notes 1. I. Berenhlum. J. ("urn-er lrt?Xl?. 60. 723 ?97th. 3. V. Riley. Stir-rin- 189. 465 ?975). 3. C. Peraino. R. J. M. Fry. E. Sta?eltt. J. Natl. Cancer Inst. I349 1973}. 4. "Maximum tolerated doses" are gener- ally insisted upon for inclusion in carcinogenic- ity bioassays. and their effects are considered relevant to risk assessment. This appears to be an inappropriate application to biology of the mass law as known in chemistry. The can be expressed as tantmals) ?_tc_h_emicali liumorl In chemistry the mass law in this form holds on- ly as long as ?rst-order kinetics prevail. As it is well known that in biology the ranges for linear kinetics are limited. and because the develop- ment of tumors in response to an animal?s ex- posure to a chemical involves many enzymatic and other processes. application of the mass law to tumor formation seems unsound. unless [in- earity ofthe relevant reactions has been demon- strated. 5. D. S. Salsburg. J. Tm. Environ. Health 3. (ill ll977i. "Using the standard formulation of tests of hypothesis. it is shown that there is a 20-500:- chance of having a false positive. . . These are irreproduciblc artifacts. I would like to de- fine a reproducible artifact as one which. though real. is not relevant to the question of risk from low-level exposure. A good example appears to be NTA lnitrilotriacelic acid). a chemical useful as a detergent additive which caused bladder tu- mors in rats when fed in concentrations above 0.7 percent in the diet. At these dosage levels NTA concentrations are so high in the urine as to cause the formation of calculi: R. Anderson. "Discontinuities of dose response curves in tox? icological testing." paper presented at the Soap and Detergent Association 52nd Annual Con- vention. Boca Raton. Fla. 25 to 28 January IW9. 6. Rigorous proof for the causal relationship be- tween carcinogenic and mutagenic activities of chemicals is still missing. However. attractive mechanistic theories and the generally good cor- relation such activities suggest that short-term mutagenicity tests are useful for the assessment las opposed to identi?cation] of car- cinogenic risk. in that probable mechanisms of action may be de?ned. Since many mutagenic and tissue-transformation tests are designed to be exquisitely sensitive. positive results in such experiments are expected to be obtained with those mutagens andlor carcinogens also. which should be designated as "opportunistic." that is. capable of doing harm to cells and organ only under the most unusual circumstances. 7. The work ofR. W. Hart and R. B. Setlow [Proc Natl. Acad. Sci. 71. 2169 ?975? suggests a greater resistance of humans to carcin ns af- fecting DNA because of a greater capability for DNA repair by the human cell as opposed to that of the rodent cell. 8. See the summary minutes of the Food and Drug Administration's Toxicology Advisory Com- mittee meeting on the role of prolactin in mammary carcinogenesis (12 to 13 May 1977) and the minutes of the FDA Endocrinology and Metabolic Drug Advisory Committee meeting on cio?bratevlikc drugs (15 to [6 February 1979) (available from the Supervisor. Public Records and Documents, Food and Drug Ad- ministration. Rockville. Md). The FDA Drug Advisory Committee on Pulmonary-Al- lergy Drugs proposed (3 to 4 May 1977) the use of a class of drugs. some of which had caused tumors in rodents. and did not distinguish be- tWeen ?tumorigenic? and ?nontumorigenic? compounds. Mercury in Sperm Whale Meat Japanese whaling interests have long resisted international whale conserva- tion initiatives. A major argument used by the Japanese to support their plunder of great whale stocks has been that the meat is needed for human consumption (even though whale meat supplies less than 1 percent of yearly Japanese protein consumption) (I). However. data re- leased by Masashi Taguchi of the Uni- versity of Tokyo's College of Fisheries. at the June 1979 meeting of the Inter- national Whaling Commission in London. indicates that sperm whale meat offered for sale in Japanese food stores contains unsafe levels of organic mercury. The whale meat contained mercury levels of 2.3 parts per million. which is six times the level the Japanese govern- ment considers acceptable (0.4 part per millionl. . . . Because of modern industrial activity. the world?s oceans are polluted with mercury. This is not so important a factor in the contamination of ?sh with short life-spans. However. sperm whales live for 60 years and concentrate mer- cury in their ?esh. When humans con- sume contaminated whale meat. the lipid soluble is concentrated in the cells of the nervous system and very slowly eliminated from the body. even when all intake is stopped. It is hoped that the Japanese will act on Taguchi?s data. Their action could have the double bene?t of protecting public health (2) and preventing threat- ened whale species from diminishing further. JOHN F. Cornell University Medical Center. New York l002l References and Notes I. The Whale Manual [Friends of the Earth. San Francisco. l978]. p. 47. 2. Brit. .Hr?d. J. 1.599(1978l. SCIENCE. VOL. 206 . oses to cleanse our environment of ti] potential pathogens: yet it'is recognized Lhat some must be scrupulously avoided. "Positive? results in a carcinpgenicity Dioassay may arise for a variety of rea- sons. The chemical tested may be a po- tent carcinogen ttrue "pathogen"). lt rnay be an ?opportunistic" carcinogen Jetectable only because the defense of :he host was weakened. for example. by senility or stress induced by a drug over- dose: or the test substance may have erided the opportunity for an environ- mental or endogenous carcinogen to ex- press itself. Any treatment which signi?- cantly in?uences the physiology of the rodent can change the pattern of back- ground lesions in the senile animal. in- sofar as tumors are a part of the disease pattern observed in old rodents. a dif- ference in tumor distribution between medicated and control animals does not necessarily have a mechanistic basis that calls for zero exposure of humans. The possibility that hormonal imbal- ance can in?uence tumor formation is well recognized (1). Stress alone can af- fect the tumorigenic response in mice (2). and it has been reported that the tu- mor incidence was higher in mice housed one per cage than in those housed ?ve per cage t3). Some chemicals. particular- ly at high doses (4). may support tumor formation in rodents through stress-re- lated mechanisms. The term "stress" may stand for several phenomena includ- ing the pharmacological activity of a chemical or the depletion of groups needed for deactivating metabo- lites. Statistical artifacts are another source of difficulty that can lead to re- sults incorrectly perceived as pusitive (5). The results from carcinogenicity stud- ies in our laboratories with many chem- icals clearly indicate that no one mode of interpretation can serve adequately in every case. We have encountered chem- icals that caused carcinoma at a time when untreated animals were still in ro- bust health. Only relatively few animals were needed to detect this effect. When tests for mutagcnicity (6) are positive. it is reasonable to conclude that such sub- stances are "true" carcinogens (patho- gens) in rats. Whether or not these chem- icals would have the same effect in hu- mans. or even in another rodent species. can. of course. only be a matter of specu- lation at this time. It is. however. pru- dent pot'i'cgv to assume that carcinogenic potential which coincides with mutagen- ic activity might also be expressed in hu- mans. although di?'erences in metabo- lism and host susceptibility make the es- timation of potency in the human situa? l4 DECEMBER I979 tion tenuous at best (7). In any case. prudence dictates that precautions should be taken to minimize human ex? posure to such chemicals. We have also studied chemicals that appeared to in?uence the tumor pattern normally observed in old animals. As- sessing the risk of such physiologically active agents requires a very different approach from that for genotoxic chem- icals. One example. a potentially valu- able drug. was labeled "carcinogenic" and barred from further development. al- though in a certain segment of the rele- vant patient population it would have been the treatment of choice. This com- pound was inactive in a battery of muta- genicity tests. Numerous observations established that it affected the endocrine balance of rodents. In a mouse experi- ment lasting 20.5 months. malignant uterine tumors were found only in con- trol animals. and mammary carcinoma only in medicated females. Numerically. an untreated mouse had a threefold greater probability of dying with uterine tumor than a medicated mouse did of dying with a mammary carcinoma. The mammary carcinomas were seized upon to label the drug carcinogenic. although the incidence of all tumors. when results from both sexes were pooled. was 46 percent in control mice and 36 percent in the highest dose level group. The appro? priate conclusion should have been that the drug at the high doses administered altered the pattern of tumor distribution by upsetting the hormonal homeostasis. Such data can only be meaningfully as- sessed in a manner very different from that appropriate for mutagenic carcino- gens. Any application of one-hit or simi- lar models is clearly inappropriate. while considerations familiar to pharmacolo- gists of dose-response and inlerspecies differences most probably yield a realisv tic and credible estimate of the human hazard. Other data support the thesis that a complete "biological" evaluation of carcinogenicity "bioassay" results is needed. For example. we know ofcom- pounds that appear to suppress tumors. A hypoglycemic agent showed such ef- fects in mice and rats. Another agent. known to affect prostaglandin concentra- tions. caused diminished occurrences of spontaneous pulmonary tumors in mice. without concomitant increase in other tumors. Such ?ndings clearly indicate that. while statistical treatments are valuable tools. they cannot be used in isolation from other facts in deciding whether or not a chemical should be con- sidered a hazard. Pharmacological stud- ies. tests for mutagenicity. and metabo- Prepared by electrofocusing f_o; electrofocusing to? agarose 8?95? Usrng 4 .. 5 twat-I Ampholine?l? carrier ampholytes are prepared by electrofocusing a range of polyamino-polycarboxylic acids into nine narrow, speci?c pH fractions. Is there any better way to prepare materials used in a biochemical technique than by the very technique itself? We know of none. Are you also aware that Am: pholine carrier ampholytes have the sharpest and lowest MW range of any ampholytes on the market? And that only ampholytes have been shown to be easily sepa- rated from proteins with no ar- tifactual binding? For the highest resolution, for the highest reliabil- ity, you can put your trust in Ampholine ampholytes. Contact LKB today for full in? formation on Ampholine solutions. Ask, too, about IEF workshops, seminars and a free subscription to Acta Ampholinae, a bibliography of over 2000 papers on IEF using Ampholine carrier ampholytes. New: agarose for electrofocusing! LKB LKB Instruments Inc. 12221 Parklawn Drive Rockville, MD 20852 301: 881-2510 Circle No. 292 on Readers Servace Card 1 BIA-303 print. I've tried to corroborate his tigures but can't. Not because no one vt-ill give me the data. but because no one appears to have them. Yet they are. thanks to Ja- cobson. now part of the public record? to be quoted and requoted. I suspect it will take more than one or even a series of editorials in Si fence to change the public image ol~ our food sup? ply?-a potpourri of carcinogens. Mi 1 A. Bi snot Department Medicine and Em'irririmi'mrii Ht'ru?th. Hospital rif- Philadelphia. Philadelphia. References l. Philadelphia Birth-rm. 2 t'tctohei' 1979. 0. It). Biotechnology and Profit There is one aspect nhich I thought was omitted from the otherwise com- plete factual account by Nicholas Wade tNews and t'omment. t) Noe. p. will of the lounding. funding. and management of research of the smaller new biotech- nological companies. Much of vvltat these companies are doing is based on fundamental research. mostly the use ot? restriction enzymes in recoml?linant DNA work. research funded by public moneys. sortie of it i am sure in direct grants to some of the biologists who are non so involved with these companies. This is how it has been vi ith pharmaceu- tical companies: there is no bar against this. but it seems to me that there is an ethical principle being violated. 'l?hat principle has to do vtith the reason ?by public money is being spent on biological research: namely. that the fruits of this research will be available to the public vvho has supported it. Ul'course it will be available. but in the process. there vvil] be profits. great and small. for the com panies involved and. I gather. for some of the individual scientists involved. (if course the public will eventually bene?t if. for example. a large supply ot? insulin is available: but at but price?.? Now that these companies are set Lip and are going concerns. may I suggest to those scientists who either manage the companies. sit on their boards. or advise them. that they see to it that the prolit margins to the investors are small: and that if large protits accrue. that these be placed in research funds to be plowed back into basic research. preferably to support toting scientists a ho have not had the opportunity to dip into the public trough for private gain. 14 DECEMBER 1979 0N GENETIC AND CYTUGENETIC TOXICULUGY BFIDDKHAVEN NATIONAL LABORATORY Upton, Long Island New York February 25-29,1980 A one-week lecture program will be presented covering the principal meth- ods in current use in genetic toxicology testing with special emphasis on in vivo and in vitro cytogenetic methods. A second week (March 3-7) of intensive laboratory work will be available to a small number of applicants who attend the lecture series. The Workshop stall include both Brookhaven National Laboratory per- sonnel and distingurshed lecturers trom other institutions. A tee of $350 will be charged for the lecture program. and will include housing. There will be an additional fee for those accepted for the second week's laboratory train- ing. For applications and further information, write to Dr. A. Bender, Medical Department Brookhaven National Laboratory Associated Universities. . Upton, NY 11973 COMBATING THE KILLER The SCIENCE Report on Heart Research JEAN MARX and GINA KOLATA - a. direct, unbiased report with information for 8.11 investigators in the ?eld, makers of public policy, scientists and the general public. $17.00 casebound 87.50 paperbound 10% discount to members Send name. address and mmittance to AMERICAN ASSOCIATION FOR THE ADVANCEMENT OF SCIENCE Department 8% 1515 Massachusetts Avenue, NW Washington, DC. 20005 l3? In the time it takes you to read this ad you could have loaded 20 samples onto your electrofocusing gel That?s how easy it is with Multiphor" unit. And duration of the runs is also short: the precisely engineered all-glass cooling stage means that you can apply higher power for faster runs? higher ?eld for sharper resolution. With the Multiphor unit and LKB's power supply you can do up to 48 samples in less than two hours! Besides being the system of choice for analytical and prepara- tive electrofocusing, the Multiphor unit is excellent for elec- trophoresis as well. Simply add the required kit and you?re ready to work with SDS?polyacrylamide gels, agarose gels even immunoelectrophoretic methods. For safety the Multiphor unit is also unique. There is no metal in the cooling stage to invite short circuits, the electrode design makes it almost impossible to come into contact with high volt- age, and the power supply has a safety interlock so you can con- nect it to your own equipment without additional risk. If you think that a system which offers so much in speed, repro- ducibility, versatility and safety has to be costly, think again. The Multiphor system is one of the least expensive ?at bed instru- ments available. Send for details today. (And be sure to ask for pertinent LKB Application Notes, a free subscription to Acta and information about forthcoming electrofocusing seminars and workshops.) LKB Instruments Inc. 12221Parklawn Drive Rockville. MD 20852 30]: 881-2510 Circle No. 293 on Readers' Service Card I258 Finally. a debatable question must be raised based on the premise that there must be some avocations in a capitalistic society that are not tied in to the profit motive. and that scientific research. based on the socialistic principle of funding for the public good. must be one of them. PHILIP SIEKiavn?x lerfli'lh'r New York 100.?! Cancer Risk Assessment The controversy over the appropriate method for assessing cancer risks. as de? scribed in Srimw tNeos and Comment. 25 May. p. 8 I i. cannot be resolved sat- isfactorily without considering the di- verse causes that can lead to an apparent positive result in a hioassay for carcino- genicity. Indeed. the very term ?bio-as- say" is unfortunate. in that it implies that chemicals which are ?active" share one particular characteristic that can be quanti?ed and mechanically extrapolat- ed to yield an estimate of risk. Such in- terpretations. though possibly appealing to the decision~maketu lack credibility. If experiments were perceived as "investi? gations" of biological activities. rather than as ?bioassays." studies would be designed differently and would yield in- formation morc meaningful to risk as. sessment. The resulting estimates. based on recognition of the ditfering kinds of effects chemicals can exert on the whole animal. would appear more be- lievable than the rigid mathematical in- terpretations currently proposed. A helpful perspective on the issue is afforded by considering certain analogies to infectious disease. Bacteriologists dis- tinguish between pathogenic organisms and those incapable of causing illness. They also recognize ?opportunistic? pathogens. those capable of infecting a host only if the defense of the host has been weakened. Virulence depends upon several factors. including the host spe? cies. and bacteriologists will pause be- fore concluding that the risk to humans is the same as that observed for another mammal. They certainly do not View all pathogens as representing an equal health hazard and know that the progres- sion of an infection depends upon more factors than just the size of the in? oculum {exposure}. Death associated with infectious disease. jitst as death from cancer. frequently occurs under conditions {impaired defenses) which in- dicate that such infection should be viewed as a rather than the cause of an organism's failing. Nobody SCIENCE. you 206 a general norms of scienti?c activity (2). They were, ?rst, the requirement that the quality of a scienti?c work should be judged on the basis of its scienti?c merits or signi?cance alone; this he called ?uni- versalism." Second. the requirement that scienti?c works be judged provision- ally and only after the relevant evidencebrought together. is at hand: this he called the "principle of or- ganized scepticism." Third. the pre- scription that whatever the personal mo- tives of scientists. the advancement of scienti?c knowledge must be the primary concern in the evaluation of scienti?c achievements; this he designated the ?principle of disinterestedness." And fourth. the requirement that an individ- ual scientist should share the knowledge acquired through his research with the scienti?c community. which has a right to that knowledge. This principle he called ?communism." but "commu- nalism? might be more satisfactory. the part of a scientist. Such choices per- tain not only to the selection of problems and hypotheses in relation to which the facts have meaning. but more particular? ly to the selection of means of presenting data according to their proper impor- tance. The criteria applicable to such choices depend on sensitive discernment and a strict conscience (5). From these obligations of objectivity and honesty ?ows the obligation to con- quer one's vanity and to acknowledge priority of'discovery by other scientists when the evidence calls for it. Tolerance. The norm of tolerance is based partly on the recognition that re- spect for the creative potentialities of other scientists is closely related to re- spect for their good faith. They must be seen as engaged in a common enterprise. The scientist should not discard new ideas out of hand. while at the same time not waste time on obvious nonsense. He should be suf?ciently receptive to new Summary. Scientist?s norms (principally honesty. objectivity. tolerance. doubt of certitude. and unselfish engagement) are in danger of serious distortion unless broad- ened to apply to the relations between scientists and nonscientists. Also needing supplementation is an ethic of development appropriate to a fast-changing society and advanced as an approach to the more effective and humane regulation of cultural and technological development. Because of their genetic relationships the code of the scientist and the ethic of development are probably complementary and together may overcome the shortcomings of each taken separately. Taken together. furthermore. they indicate the possibility of a humane world order based on the cooperation of a community of scientists and its public. Twenty-?ve years later I attempted to reformulate the norms of science with explicit reference to the conduct of indi- vidual scientists. In successive publica- tions 1 analyzed and discussed the future of the code of the scientist. in collabora- tion ?rst with Harriet Zuckerman t3) and later with Michael Meyer (4). in this for- mulation the norms of scientists are mainly honesty. objectivity. tolerance. doubt of certitude. and unsel?sh engage- ment. Intellectualr integrity and objectivity. The ?rst obligation of scientists is in- tellectual integrity. 1 take this to mean not merely that scientists must be unre- mittingly honest in their investigation of the natural world: they must as well avoid the undisciplined introduction of subjective elements into their per- ceptions. They must prevent their de- sires and aversions from penetrating their observations of the phenomena that they study and their analyses of these observations. Of course. the observation and analysis of facts in a form certi?able by the appropriate rules entail choices on 700 ideas to see whether they are consistent with established knowledge or furnish links to new and valuable concepts. Doubt of certitude. The next principle is doubt of certitude: that is. questioning, what is asserted authoritatively. i agree with what Michael Polanyi said about the importance of respect for the authority of science (6). This does not con?ict with my belief that an attitude of readiness to question what is accepted as certain by established authorities in science is one of the primmn mavens in the generation of new knowledge. Recognition of error. The cruder forms of error can quite easily be avoid? ed by scientists: it is the more subtle forms of error that are more dif?cult to discern. Yet the recognition. acknowl- edgment. and admission of error favor progress in understanding. Urisel?sit engagement. The ?fth norm is unsel?sh engagement on behalf of the growth of scienti?c knowledge. The sci- entist's purpose should be to extend our knowledge and understanding ofthe uni- verse. and not to secure personal gain or to promote the supremacy of a part philosophy or ideology: Communal spirit. Finally. it is in . cumbent on scientists to appreciate and respect their dependence on the commu- nity of scientists. S?ientists must recog- nize that their own work is part of the larger scienti?c enterprise and that they themselves are linked to their colleagues through submission to its tr?itions and participation in its ethos. as well as through their common effo to increase and improve the body of certi?ed knowl- edge. The other rules ofthe code help to pro- tect the fabric of the scienti?c commu? nity and in doing so they reinforce them- selves. Threats and Stresses Bearing on Science and Its Code 1 believe that the operating code I have de?ned has been instrumental in making possible the advance of scienti?c knowl- edge and the trend toward scienti?c uni- versalism. But I also believe that these principles should be extended beyond the domain of science. Before I can make this argument, must comment on what [judge to be the main threats to the scienti?c enterprise. and on possible remedies. It is convenient to discuss the stresses on science by reference to the pressures to which they give rise for modifying the code of the scientist. These pressures can be divided into four main categories: failures to observe the code. (ii) the undermining of the concept of objectivi- ty. the trend toward over-speciali? zation. and (iv) the increasing concern of society about science. Nonobsen'unce of the norms of the code. Failures of observance of the norms by scientists reveal themselves usually under the form of intolerance. abuse of authority. or nonrecognition of priority. In recent years the principles of integrity and of disinterestedness and sel?ess engagement in scienti?c activi- ties have been occasionally infringed by referees and other readers of scienti?c reports. They have abused their privi- leges by selectively disseminating the contents prior to formal publication or by using the knowledge gained from the work for their own or their immediate colleagues' advantage. Some of these in? fringements are known to scientists from their own experience. Scientists? attachment to the code is under greater stress than it used to be. partly because there is. on the whole. SCIENCE. VOL. 198 mi in social use some 1: of misinformation" (26. p. 19). thus "subvert the possibility of in- formed consent? (26. p. 21). ?Prior gen- eral consent" or "presumptive'?onsent" {26. p. 211 have been proposed to deal with this ethical problem. Recombinant DNA research makes it at least theoretically possible to combine the genetic characteristics of plant and mammal. to produce a "plammal" or a ?mam." We need to ?nd a balance be- tween possibly inadvertently producing the means to cause catastrophe to man? kind.,and potentially high bene?cial de- . velop?ents. The genetic splicing of re- combinant DNA technology has already been Used to transfer the rat gene for in- sulin production to bacteria (27). This development has the potentially high bene?cial consequence of making pos- sible massive commercial production of human {instead of other species) insulin for diabetics. It also has. in the eyes of some. the possibility of catastrophe should insulin-producing bacteria get out of the laboratory into the body of a hu- man. to multiply and throw the person into insulin shock. One argument is that knowledge is power. and if we do not acquire the knowledge. other countries will. Re- member that in World War 11 the other side was also working on an A-bomb. If we vauire the knowledge. we can also vauire the means to control the knowl- edge. If we do not. the controls may be in other hands. These. too. are ethical Considerations. As Jonas notes 128). generally there is something experimental because tenta- tive about every individual treatment. beginning with the diagnosis itself. He would be a poor doctor who would not learn from every case for the bene?t of future patients. and a poor member of the profession who would not make any new insights gained from his treatments available to the profession at large. In summary. we recognize that acquir- ing new information while retaining old ethics demands adherence to the funda- mental rule that a person should not be subjected to avoidable risk of death or physical harm unless be freely and in- telligently consents. The problem is to balance rights against bene?ts with re? spect for human dignity in the quest for the cure of human diseases. References and Notes 1. S. J. Reiser. A. .1. W. J. Curran. Eds. Ethic! in (MIT Press. Cambridge. Mass. 1977). . W. A. Silverman. Sci. Am. 236 (No. 6). 100 (1977). 3. B. Barber rt ttl.. Research on Human Subjects [Russell Sage Foundation. New York. I973). 4. G. J. Annas. The Rights of Hospital Patients. American Civil Liberties Union Handbook tAvon. New York. 19751. 5. V. Herbert. in Pmceedingr. Western Hemi- sphere Nutrition Congress W. P. L. White and N. Sclvey. Eds. {Publishing Sciences. Acton. 19751. pp. 1'14?91. h. E. E. Conn. in Naturally Occurring in Foods (National Academy of Sciences. Wash- ington. D.C.. 1973). pp. 299?308: J. P. Levvts. West. J. Med. 127. 55 t19771: Fed. Regist. 42. 39768 119771. 7. H. K. Beecher. J. Am. Med. Assoc. 159. 1602 11955): Research and the Individual: Hurrmn Studies tl-ittlc. Brown. Boston. 1970). 8. J. W. Tukcy.5cit'nce 193.67911977). 9. J. P. Gilbert. B. Mcl?eck. F. Mosteller. ibid.. p. 684 10. H. C. Black. Black's Law Dictionary. Revised tSt. Paul. Minnesota. ed. 4. 19681. In "United States of America v. Articles of Food and Drug Consisting of. . . apricot kernels . . . amygdalin . . Civil No. District Court. Eastern District of Wisconsin. 29 July 1977). Judge Reynolds closed a lacttile factory after Carcinogenic Risk Assessment Man is exposed to a variety of natural and substances that are known to be harmful to experimental animals in high doses and consequently are under suspicion of being harmful to humans in low ones. Exposure to many of these substances. particularly those involving involuntary exposure through food. wa? ter. air. or the workplace is subject to 18 NOVEMBER 1977 Jerome Corn?eld regulation by governmental agencies. In some instances the bene?ts conferred by a suspected substance can be achieved by other safe substances in equally satis- factory ways. in which case the most ap- propriate regulatory action is an outright ban. no regard being given to the strength of the suspicion. But in many cases important bene?ts are lost if the holding as 3 Finding of Fact that. "Anecdotal and testimonial evidence as to cures or e??ects of treatments on cancer victims as described by lay persons. or persons possessing either an MD. or but who are not quali?ed by scienti?c training and experience as experts in the ?eld of cancer therapy. is not probative or substantial evidence of the safety and ef?cacy of cancer treatments." Judge Reynolds held as 3 Con- clusion of Law. "The testimony of lay witnesses as to the existence of cancer and the safety and e?icacy of an alleged cancer treatment based on their personal experience with the treatment is entitled to no weight and is therefore in- admissible as irrelevant and non-probative evi~ dence." As precedents for this Conclusion of Law. Judge Reynolds cited United States v. Hoxsey Cancer Clinic. 198 Fed. Rep. 2nd ser. 273 (Sth Cir. CL. 1952); United States v. Wier. 281 Fed. Rep. 2nd ser. 8511 Cir. CL. 1960}. and Federal Rules of Evidence M11. 402. 4113. and 701. 1. J. Corn?eld. Science 198. 693 (19771. 2. J. A. Robinson. Columbia La? Rev. 76. 48 (1976). 13. V. Mlk? and R. A. Good. Science 198. 677 {1977). 14. L. A. Altman. N. Engl. J. Med. 286. 346(1972). 15. V. Herbert. Traits. Assoc. Am. Physicians 75. 307119621. 16. Am. J. (?iin. Nutr. 28. 55511975). 17. L. Thomas. Science 198. 675 119771. 18. S. Barrett and G. Knight. Eds. The Health Rob- bers: to Protect Your .llunpy and Your Life (Stickley. Philadelphia. 19761. J. W. Miner. J. Forensic Sci. 9. I 11964]: Califomia Phillips. 75 Calif. Rep. 721) (I969). certior'ati denied. .96 U.S. 11121 (1970). 19. S. 1217. A bill to regulate activities involving re- combinant deoxyribonuclcic acid. 19 May 1977 tintroduced by Senator Edward Kennedy. 0? Mass.): Newsletter. Fed. Am. Soc. Exp. Biol.. 101No. 8). 311977]. 20. B. D. Davis. E. Charga?'. S. Krimsky. Chem. Eng. News. 30 Ma 1977. pp. 26?42. The latest cvndence is that ears regarding recombinant DNA research may be greatly exaggerated Abelson.St-irnce 197. 721 11977): W. Gaylin. N. Engl. J. Med. 297. 665 (1977)]. 2 I. Zinder. hearings before the Subcommittee on Health. U.S. Senate. on 3. 1217119771. 22. T. M. Powledge. Hastings Cent. Rep. 7 (No. 2). 1811977): D. Callahan. ibid.. p. 20: K. Dis- mukes. ihid.. p. 25. 23. C. Cohen. N. Engl. J. Med. 296. 1203 [1977). 24. R. Goldstein. ibid.. p. 1226. 25. G. Robinson and A. Memv. Ann. Thoma Surg. 22. 209 [1977). 26. S. Milgrim. Hastings Cent. Rep. 7 (No. 2). I9 (1977). 27. A. Ullrich at 111.. Science I96 I3l3 11977): N. Wade. lht?d. 197. 1342(1977). 28. H. Jonas. in Ethics in Medicine. S. J. Reiser. A. J. W. J. Curran. Eds. (MIT Press. Cam- bridge. Mass.. 1977). agent is banned. and the magnitude of the risk must then be balanced against the bene?t conferred. The risk may be of such magnitude that banning is appropri- ate even in the face of the bene?tsthe levels to which hu- mans are exposed that a ban is not con? sidered appropriate. Risk assessment is therefore an essential component of reg- ulatory decisions. it is also a particularly appropriate topic for consideration be- came of the mixture of statistical. scien- ti?c. and public policy considerations that it presents. The problem of risk as- sessment is the same formally. no matter what the route of exposure. but since much of the exposure is by way of food. 1 will con?ne my discussion to that topic. The author is professor of statistics at George Washington University. Washington. DC. 20052. 693 um SanKQw?r.) 3 V3 vniaa. '5 u. -. nun-A Ammu- Carcinogenic Risk Assessment: A Guide -.- to the Literature 1' 1; DANIEL AND CHARLES Baown?r Introduction health hazards of environmental chemicals, considerable ef- fort is being devoted to the identification and regulation of carcinogenic chemicals. Although the primary concern of this i research is human health, information on the carcinogenic poten- i tial of chemical substances is necessarily derived mainly from bio- 4 assays conducted with animal models. The carcinogenicity of a 33 A RESULT of the increasing awareness of the potential Mathematical Statistics, Carleton University. Daniel Krewski is head of the chemical statistics section, food statistics and operational planning division, of the health protection branch of Health and Welfare Canada. Mathematical Statistics. George iVashington University. 'arles Brown is a mathematical statistician at the biometry branch of the National Cancer Institute. The authors are indebted to Dr. Roger McCullough for bringing to their attention several references included in the regulatory considerations section, to Dr. David Salsburg for providing references in the sections on carcino- genicity screening and quantitative" risk assessment, and to Mr. John Kora! for his help in organizing the material. This article will be published simultaneously in Biometrics, Volume 37, Number 2, June 1981. 84 aide potential :lerable ef- regulation am of this nic poten- from bio~ deity of a Krewski is rial planning Canada. :ity. Charles the National ring to their ions section, on carcino- Iohn Kocar Volume 37, No. I Carcinogenic Risk Assessment 85 substance is established when its administration to test animals in an adequately designed and conducted experiment results in a greater incidence or decreased latent period of one or more types of neoplasia than control animals maintained under identical con- ditions but not exposed to the compound under study. A guide to the statistical literature on carcinogenic risk assess? ment using animal models is presented in this article. Included are sections on the general principles of carcinogen bioassays, statisti- cal analysis of screening bio-assays, quantitative risk assessment, and regulatOry considerations. Some references have been included in more than one section, where appropriate. The practical aspects of conducting an adequate and valid ear- cinogen bioassay are discussed in the references provided in sec- tion 1. Statistical procedures for the analysis of screening bioassays designed to detect carcinogenic compounds may be found in the references in section 2. Although simple binomial comparisons of the tumor incidence rates observed in the control and test groups may be apprOpriate for conventional bioassay designs (section 2.1), other procedures are required for two-generation studies where the litter rather than the individual animal may be the appropriate experimental unit for purposes of statistical analysis (section 2.2). Time-adjusted analy- sis may be used whenever it is desirable to consider the time at which lesions were observed (section 2.3). Such an analysis may be appropriate, for example, when different survival rates among the treatment groups are encountered. Information on the time- to?tumOr development may be available in the case of grossly vis- ible, palpable, or rapidly lethal tumors and may be assessed using techniques for the analysis of censored survival data. (In a car- cinogen bioassay, survival time would correspond to the tumor- free time on test with animals sacrificed or dying without a tumor being censored.) Serial sacrifice experiments may be used to obtain information on time-to-tumor development in cases where this is not directly observable. In a carcinogen bioassay, a number of organs and tissues are examined for neOplastic changes {Breach sex and species of the experimental animals used. Because of the large number of statisti- cal tests that may be performed using the data from a single study, measures of the overall error rates for a screening bioassay should consider the multiplicity of comparisons made (section 2.4). For 86 Toxrc SUBSTANCES JOURNAL Vol. 3 a large-scale screening program in which the carcinogenic poten- tial of several different chemicals is being assessed, sequential ex- perimental designs that minimize the average numbers of experi- mental animals required to achieve prespecified false positive and false negative rates have been proposed (section 2.5). Although the screening bioassay might be an appropriate instru- ment for the assessment of carcinogenic potential 011 a qualitative basis, a dose-response study covering a range of dose levels is re- quired to provide a quantitative estimate of the carcinogenic risk to man (section 3). This risk-assessment process involves the use of mathematical models of dose-response to extrapolate from the high doses necessarily used in animal studies to lower doses cor- responding to expected human exposure levels (section 3.1). An important aSpect of these models is whether they provide for a threshold dose below which the population risk is zero (section 3.2). In View of the Uncertainties invoIVed in extrapolating beyond the experimental dose range as well as from animal to man, relevant biological information should be considered in the risk-assessment process (section 3.3). When prior information concerning the shape of the dose?response curve is available, procedures for constructing experimental designs leading to improved estimates of risk at low doses recently have been proposed (section 3.4). Section 4 contains a variety of references involving the applica- ?tion of the methodology discussed above in a regulatory setting. Included are references on risk/benefit analysis, which may be required in cases where a safe alternative for a hazardous com? pound is unavailable. In preparing this guide, an attempt has been made to ensure that the bibliography in sections 3 and 4 are as comprehensive as possible. The references provided in the recent review papers by Cart, Chu and Tarone (1979), Haseman and Kupper (1979), Barton and Turnbull (1979), and Lagakos (1979), as well as the list of references prepared by Wahrendorf (1979), have precluded the need for as extensive a bibliography in sections 2.1, 2.2, and 2.3. 4- Carcinogen Bioassay Food and Drug Advisory Committee on Protocols for Safety Evaluation (1971). Panel on carcinogenesis report on cancer testing in the safety evaluation of food additives and pesticides. Toxicology and Applied Pharmacology 20, 419438. ..-Q-. Vol. 3 in poten- ential es- Jf experi- ;itive and rte instru- [unlitative eels is re- genic risk 5 the use from the doses cor? 3.1). An ride for a action 3.2). eyond the 1, relevant issessment the shape instructing isk at low 1e applica- Iry setting. 11 may be dons com- to ensure :hensive as papers by 79), Barton the list of 2d the need 2.3. Lation (1971)- evaluation of oncology 20. wam-MM?- - In No. 1 Carcinogenic Risk Assessment 87 Food Safety Council (1980). Proposed System for Food Safety Assessment (Chapter 10: Chronic Toxicity Testing). Food Safety Council, 17:25 Street, ??ashington, D.C. 20006, 121-136. Fox, Thibert, P., Arnold, D.L., Krewsl-ri, DR. and Crice, MC. (1979). Toxicology studies II. The laboratory animal. Food and Cosmetics Toxicol- ogy 17, 661-675. Crice, H.C., Munro, I.C., Krewski, 13.11. and Blumenthal, II. (1981). In utero exposure in chronic toxicity/carcinogenicity studies. Food and Cosmetics Toxicology. In press. Health and Welfare Canada (1975). The Testing of Chemicals for Carcino- genicity, Mutagenicity and Teratogenicity. Health and Welfare Canada, Ottawa. National Cancer Advisory Board (1977). General criteria for assessing the evidence for carcinogenicity of chemical substances. Journal of the National Cancer Institute 58, 461-465. Munro, LC. (1977). Considerations in chrome toxicity testing: The chemical, the dose, the design. Journal of Environmental Pathology and Toxicology 1, 183-197. Munro, LC. and Willes, RF. (1978). Reproductive toxicity and the problems of in utero exposure. In Chemical Toxicology of Food. C.L. Calli, R. Paoletti and C. Vettorazzi Elsevier/ North Holland, New York, 133-145. Page, NR. (1977). Chronic toxicity and carcinogenicity guidelines. Journal of Environmental Pathology and Toxicology 1, 161-182. Sontag, (1977). Aspects in carcinogen bioassay. In Origins of Human Cancer (Book C: Human Risk Assessment). H.H. Hiatt, Watson and LA. Winston (eds), Cold Spring Harbour Laboratory, Cold Spring Har? bour, 1327-1338. Sontag, J.M., Page, NP. and Saffiotti, U. (1976). Guidelines for Carcinogen Bioassay in Small Rodents. DHEW Publication No. (NIH) 76-801. Depart- ment of Health, Education and Welfare, Washington, D.C. Wei], CS. (1972). Guidelines for experiments to predict the degree of safety of a material for man. Toxicology and Applied Pharmacology 21, 194?199. World Health Organization (1978). Principles and Methods for Evaluating the Toxicity of Chemicals. Part 1. Environmental Health Criteria 6, World Health Organization, Geneva. 2. Carcinogenicity Screening 2.1 Tumor Incidence Data Bishop, Y.M.M., Feinberg, S.E. and Holland, P.W. (1975). Discrete Multi? variate Analysis: Theory and Practice. SIIT Press, Cambridge, Massachusetts. Casagrande, J.T., Pike, M.C. and Smith, RC. (1978). The power of the ?exact" test for comparing two binomial distributions. Applied Statistics 27, 176-801. . nan-.5111 .: .3. . - I..vToxn: SUBSTANCES JOURNAL Vol. 3 Cox, DE. (1970). The Analysis of Binary Data. Chapman and Hall, London. Drinkwater, NR. and (1981). Statistical methods for the analysis of tumor multiplicity data. Cancer Research 41, 113-119. Feinberg, S.E. (1977). he Analysis of Cross?Classified Categorical Data. MIT Press, Cambridge, Massachusetts. Fleiss, (1973). Statistical Methods for Rates and Proportions. John Wiley Sons, New York. Cart, (1971). The comparison of proportions: A review of significance tests, confidence intervals and adjustments for stratification. Rcoicw of the International Statistical Institute 39, 148-169. Cart, (1979). Statistical analyses of the relative risks. Environmental Health Perspectives 32, 157-167. Cart, Chu, KC. and Tarone, RE. (1979). Statistical issues in interpreta- tion of chronic bioassay tests for carcinogenicity. Journal of the National Cancer Institute 62, 957-974. Haber, M. (1980). A comparison of some continuity corrections for the chi- squared test on 2x2 tables. Journal of the American Statistical Association 75, 510-515; Haseman, and Hoel, DO. (1979). Statistical design toxicity assays: Role of genetic structure of the test animal population. Journal of Toxicology and Environmental Health 5, 89-101. . Hutchison, T.P. (1979). The validity of the chi-squared test when expected frequencies are small: A list of research references. Communications in Sta- tistics, Series A 8, 327-335. Mantel, N. (1977). Tests and limits for the emnrnon odds ratio of several 2 2 contingency tables: Methods in analogy with the Mantel-Ilaenszel Procedure. Journal of Statistical Planning and inference 1, 179-189. McDonald, L.L., Davis, B.M. and Millikan, CA. (1977). A nonrandomized unconditional test for comparing two proportions in 2 2 contingency tables. Technometrics 19, 145-157. National Cancer Institute (1977). Bioassay of chlordane for possible carcino- genicity. DHEW Publication No. (NIH) 77-808. Department of Health, Education and Welfare, \Vashington, D13.1 Fate, 3., Pike, Day, N.E., Cray, R.C., Lee, P.N., Parish, 5., Peto, Richards, S. and \Vahrendorf, (1980). Guidelines for simple, sensitive significance tests for carcinogenic effects in long?term animal experiments. In Long-Term and Short-Term Screening Assays for Carcinogens: A Critical Appraisal IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, Annex to Supplement 2, IARC, Lyon, 311-425. Salsburg, D. (1980). Are carcinogenicity tests useful? In Controversies in Therapeutics. Lasagna W.B. Saunders, Philadelphia, 151?162. Santner, TI. and Snell, ELK. (1980). Small sample confidence intervals for p1 p: and pI/pg in 2 2 contingency tables. Journal of the American Statistical Association 75, 386-394. -v --..-. ?blm?l' a Vol. 3 tall, London. the analysis :1 Data. MIT John Wiley i significance review of the znoironmental in interpreta- the National for the chi- al Association assays: Role oncology and 'hen expected ations in Sta- tio of several mtel-Haenszel ?9-189. onrandomized 2. contingency ssible carcino- nt of Health, 5., Peto, aple, sensitive 1 experiments. ens: A Critical :inogenic Risk .yon, 311-425. mtrooersies in is, 151-162. .e intervals for the American .n?Carcinogenic Risk Assessment 89 Tarone, RE. and Cart, (1980). On the robustness of combined tests for trends in proportions. Journal of the Anierican Statistical Association 75, 110-116. Thomas, DC. (1975). Exact and methods for the combination of 2 2 tables. Computers and Biomedical Research 8, 423-446. Thomas, D.C., Breslow, N. and Cart, J. (1977). Trend and homogeneity analy- ses of proportions and life table data. Computers and Biomedical Research 10, 373-381. 2.2 Two Generation Bioassays Claden, B. (1979). The use of the jackknife to estimate proportions from toxicological data in the presence of litter effects. Journal of the American Statistical Associatiou 74, 278?283. Halperin, LL, Ware, 1.11. and Wu, M. (1980). Conditional distribution-free tests for the two?sample problem in the presence of right censoring. Journal of the American Statistical Association 75, 638-645. Hascman, and Kupper, LL. (1979). Analysis of dichotomous data from certain toxicological experiments. Biometrics 35, 281-293. Soms, A.P. (1977). An algorithm for the discrete Fisher's permutation test. Journal of the American Statistical Association 72, 662-664. Tarone, RE. (1979). Testing the goodness of fit of the binomial distribution. Biometrika 66, 585-590. Wei, (1980). A generalized Cehan and Gilbert test for paired observa- tions that are subject to arbitrary right censorship. Journal of the American Statistical Association 75, 634-637. Williams, DA. (1975). The analysis of binary responses from toxicological ex- periments involving reproduction and teratogenicity. Biometrics 31, 949-952. 2.3 Time-to-Tu mor Data Altshuler, B. (1970). Theory for the measurement of competing risks in ani- mal experiments. Mathematical Bioscicnces 6, 1-11. Barton, RR. and Tumbull, B.W. (1979). A survey of covariance models for censored life data with an application of recidivism analysis. Communica- tions in Statistics, Series A 8, 723-750. Bratcher, TL. (1977). Bayesian analysis of a dose-response experiment with serial sacrifices. Journal of Environmental Pathology and Toxicology 1, 287-292. Breslow, N. (1979). Statistical methods for censored survival data. Environ- mental Health Perspectives 32, 181-192. Breslow, N., Day, N., Tomatis, L. and Turusov, V. (1974). Associations be- tween tumor types in a large-scale carcinogenesis study of mice. Jour- nal of the National Cancer institute 52, 233-239. .1. ?hpf ?A?u'sgiuam'dguy' .1. l?rau .. 1-5. if . 90 Toxrc SUBSTANCES JOURNAL Vol. 3 Faulkner, LE. and McHugh, RB. (1972). Bias in observable cancer age and life-time of mice subject to spentancous mammary carcinomas. Biometrics 28, 489-498. Fleming, TR. and Harrington, DP. (1981). A class of hypothesis tests for one and two sample censored survival data. Communications in Statistics. In press. Fleming. T.B., O'rallon, 1n, O'Brien, RC. and Harrington, DP. (1930). Modified Kolmogorov-Smirnov test procedures with application to arbitrar- ily right censored data. Biometrics. In press. Cail, M.H., TJ. and Brown, CC. (1980). An analysis of comparative carcinogenesis experiments based on multiple times to tumor. Biometrics. In press. Cart, (1975). Letter to the editor (with reply R. Peta). British Journal of Cancer 31, 696-699. Cart, Chu, KC. and Tarone, RE. (1979). Statistical issues in interpre- tation of chronic bioassay tests for carcinogenicity. Journal of the National Cancer Institute 62, 957-974. Caylor, D.W. and Hoel, DC. (1980). Statistical analysis of carcinogenesis data from chronic animal studies. In Carcinogens in Industry and the En- vironment; J. Sontag Marcel Dckker, New York. In press. Guess, H.A. and Hoel, DC. (1977). The effect of dose on cancer latency period. Journal of Environmental Pathology and Toxicology 1, 279-286. Heel, DC. (1972). A representation of mortality 'data by competing risks. Biometrics 28, 475-488. Hoel, DC. and Walhurg, HE. (1972). Statistical analysis of survival experi- ments. Journal of the National Cancer Institute 49, 361-372. Kalbfleish, and Prentice, KL. (1980). The Statistical Analysis of Failure Time Data. John Wiley 6: Sons, New York. Kodell, BL. and Nelson, (1980). An illness-death model for the study of the carcinogenic process using Sunrival/sacrifice data. Biometrics 36, 267-277. Kodell, R.L., Shaw, and Johnson, AM. (1980). Nonparametric joint estimators for disease resistance and survival functions in survival/sacrifice experiments. Submitted to Biometrics. Lagalcos, SW. (1979). General right censonn' and its impact on the analysis of survival data. Biometrics 35, 139-156. Lagakos, S.W. and Mosteller. F. (1980). A case study of statistics in the regulatory process: The red 40 experiments. Journal of the National Cancer Institute. In press. Mantel, N. (1980). Assessing?laboratory evidence for neoplastic activity. Bio- metrics 36, 381-400. Mantel, N., Bohidar, NR. and Ciminera, (1977). analy- ses of litter-matched time-to-response data, with modifications for recovery of interlitter information. Cancer Research 37, 3363-3868. - ?irv 0- . ?snow. Vol. 3 meet age and .s. Biometrics nesis tests for in Statistics. DR (1980). in to arbitrar- rf comparative Biometrics. *ish Journal of as in interpre- the National carcinogenesis and the En- 1353. cancer latency I, 279-286. Impeting risks. rurvival experi- [ya-is of Failure for the study Biometrics 36. arametric joint Lrvival/ sacrifice on the analysis satis?es in the iational Cancer to activity. Bio- Haenszel analy- ms for recovery .mu? .4 Mada-l- - Mm_. \Ml?mu .1t. "v.1 - - N0. 1 Carcinogenic Risk Assessment 91 Mantel, N. and Ciminera, (1979). Use of logrank scores in the analysis of litter-matched data on time to tumor appearance. Cancer Research 39, 4308-4315. 'Mitchell, T.J. and Turnbull, (1979). Log-linear models in the analysis of disease prevalence data from survival sacrifice experiments. Biometrics 35, 221?234. Patel, KM. and Hoel, DC. (1973). A generalized k-sample test against ordered alternatives when observations are subject to arbitrary right censorship. Communication in Statistics, Series A 2, 373-380. Peto, R. (1974). Guidelines on the analysis of tumor rates and death rates in experimental animals. British Journal of Cancer 29, 101-105. Peto, R. and Lee, P. (1973). Wcibull distributions for continuous-carcinogene- sis experiments. Biometrics 29, 457?470. Peto, 11., Lee, P. and Paige, W. (1972). Statistical analysis of the bioassay of continuous carcinogens. British Journal of Cancer 26, 258-261. Peto, 151., Pike, M.C., Day, N.E. Cray, B.C., Lee, P.N., Parish, 5., Pete, Richards, 5. and \Vahrendorf, J. (1980). Guidelines for simple, sensitive significance tests for carcinogenic effects in long-term animal experiments. In Long-Term and Short-Term Screening Assays for Carcinogcns: A Critical Appraisal. IABC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, Annex to Supplement 2, IARC, Lyon, 311-425. Pike, MC. and Boo, .I.C. (1963). An actuarial method of analysis of an ex- periment in two-stage carcinogenesis. British Journal of Cancer 17, 605-610. Sachs, R. (1959). Life table technique in the analysis of response?time data from laboratory experiments on animals. Toxicology and Applied Pharma- cology 1, 203-227. Salsburg, D.S. (1979). The effects of life-time feeding studies on the pattern of senile lesions in mice and rats. Drug and Chemical Toxicology 3, 1-33. Staffa. LA. and Mehlman, MA. (eds.) (1979). innovations in Cancer Risk Assessment (EDOI Study). Pathotox Publishers, Park Forest South, Illinois. Tsiatis, A. (1978). A large sample study of the estimate for the integrated hazard function in Cox's regression model for survival data. Technical Re- port No. 526, Department of Statistics, University of \Visconsin, Madison, Wisconsin. Tumbull, B.W. and Mitchell, TJ. (1979). Exploratory analysis of disease prevalence data from survival/ sacrifice experiments. Biometrics 34, 555-570. Wahrendorf, J. (1979). Statistical analysis of survival data: Short course and workshop lift of referenms. July 2-5, Deutsches Krebsforschungszeutrum, Heidelberg. 2.4 Multiple Comparisons Fears, T., Tarone, R. and Chu, K. (1977). False-positive and false-negative rates for carcinogenicity screens. Cancer Research 37, 1941?1945. - . aut?Toxrc SUBSTANCES Vol. 3 Cart, Chu, K.C. and Tarone, RE. (1979). Statistical issues in the inter- pretation of chronic bioassay tests for carcinogenicity. Journal of the Na- tional Cancer Institute 62, 957974. Salshurg, D. (1977). The use of statistics when examining lifetime studies in rodents to detect carcinogenicity. Journal of Toxicology and Environmental Health 3, 611-628. Responses (by Fears, T. and Tarone, and Haseman, I.) 629?636. 2.5 Sequential Procedures Elashoff, R, Preston, D. and Fears, TH. (1979). Comparison and evaluation of some experimental designs for use in carcinogen screening. Journal of the National Cancer institute 62, 1209-1219. Fears, T.R. and Douglas, (1977). Suggested procedures for reducing the pathology workload in a carcinogen bioassay program, Part I. Journal of Enviroumental Pathology and Toxicology I, 125-137. Fears, TR. and Douglas, (1978). Suggested procedures for reducing the pathology workload in a carcinogen bioassay program, Part II: Incorporat- ing blind pathology techniques and analysis for animals with tumors. Journal of Environmental Pathology and Toxicology 1, 211-222. Preston, D. and Elaslioff, R. (1977). An investigation of the properties of some two sample two-stage designs. In Applications of Statistics. RR. Krish- naiah North Holland, New York, 407-431. 3. Quantitative Risk Assessment 3.1 Mathematical Models Albert, R. and Altshuler, B. (1973). Considerations relating to the formula- tion of limits for unavoidable population exposures to environmental car- cinogens. In Radionuclide Carcinogenesis. J. Ballou, R. Busch, D. Mahlum and C. Sanders (eds), AEC Symposium Series, Cont.-720505, NIIS, Spring- field, Virginia, 233-253. Albert, R. and Altshuler, B. (1976). Assessment of environmental carcinogen risks in terms of life shortening. Environmental Health Perspectives 13, 91-94. Arley, N. (1961). Theoretical analysis of carcinogenesis. In Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability (Vol. 4). J. Neyman University of California Press, Berkeley and Los Angeles, California, 1-18. Annitage, P. and Doll, R. (1961). Stochastic models for carcinogenesis. In Proceedings of the Fourth Berkeley Symposimn on Mathematical Statistics and Probability (Vol. 4). J. Neyman University of California Press, Berkeley and Los Angeles, California, 19-38. Brown, C. (1978). Statistical aspects of extrapolation of dichotomous dose response data. Journal of the National Cancer Institute 60, 101-108. vv- or .roqoI-w-ou-a?v .. .. .. Cp?lw? 1- .- Vol. 3 in the inter- of the Na- 18 studies in ncironmental ad Haseman, 1d evaluation g. Journal of reducing the 1. Journal of reducing the L: Incorporat- mors. Journal properties of RR. Krish- the formula- onmental car- I, D. Mahlum N115, Spring- tal carcinogen sspectioes 13, proceedings of mi Probability tacky and Los cinogenesis. In stiml Statistics alifomia Press, hotomous dose 101-108. -- um- . . ?Mu-uur-ek . \hciLyana . .n - .M Inu- . sud.? No. 1 Carcinogenic Risk Assessment 93 Bruce, RD. (1980). Low dose extrapolation and risk assessment. Chemical Times and Trends. In press. Carlborg, (1979). Cancer, mathematical models and allntoxin. Food and Cosmetics Toxicology 17, 159-166. ?Chand, N. and Hoel, D. (197-1). A comparison of models for determining safe levels of environmental agents. In Reliability and Biometry. F. Proschan and (eds), SIAM, Philadelphia, 681-700. Chou, T.C. (1980). Comparison of dose effect relationships of carcinogens following low-dose chronic exposure and high-dose single injection: An analysis by the median effect principle. Carcinogenesis .I, 203-213. Copenhaver, T.W. and Miclke, P.W. (1977). Quantit analysis: A quanta] assay refinement. Biometrics 33, 175-186. Cornfield, J. (1977). Carcinogenic risk assessment. Science 198, 693-699. Cornfield, Cariborg, 17.1.11. and Van Ryzin, J. (1978). Setting tolerances on the basis of mathematical treatment of dose-response data extrapolated to low doses. In Proceedings of the First International Congress on Toxicology. G.L. Plaa and W.A.M. Duncan (eds), Academic Press, New York, 143-164. Cornfield, Rai, K. and Van Ryzin, J. (1980). Procedures for assessing at low levels of exposure. Archives of Toxicology, Supplement 3 (Quantitative Aspects of Risk Assessment in Chemical Carcinogens), 295-303. Crump, KS. (1979). Dose response problems in carcinogenesis. Biometrics 35, 157-167. Crump, KS, Guess, H. and Deal, K. (1977). Confidence intervals and test of hypotheses comerning dose response relations inferred from animal car- cinogenicity data. Biometrics 33, 437-451. Crump, K.S., Hoel, D., Langley, C. and Peto, R. (1976). Fundamental car- cinpgenic processes and their implications for low dose risk assessment. Can- cer Research 36, 2973-2979. Dafler, P.Z., Crump, KS. and Masterrnan, MD. (1980). theory for analyzing dose-response survival data with application to the low-dose extrapolation problem. Mathematical Biosciences. In press. Food Safety Council (1980). Proposed System for Food Safety Assessment (Chapter 11: Quantitative Risk Assessment). Food Safety Council, 1725 Street, Washington, D.C. 20006, 137-160. - Gaylor, D.W. and Kodell, KL. (1980). Linear interpolation algorithm for low dose risk assessment of toxic substances. Journal of Environmental Pathology and Toxicology. In press. Gaylor, D.W. and Shapiro, RE. (1979). Extrapolation and risk estimation for carcinogenesis. In Advances in Modern Toxicology (Vol. I), New Con- cepts in Safety Evaluation (Part 2). MA. Mehhnan, RE. Shapiro and H. Blumenthal (eds), John Wiley and Sons, New York, 65-87. Guess, H.A. and Crump, KS (1976). Low dose extrapolation of data from animal carcinogenicity experiment?analysis of a new statistical technique. Mathematical 32, 15-36. aura nib-"diAAiLLnL .1 94 Toxrc SUBSTANCES Vol. 3 Guess, H.A. and Crump, K.S. (1977). Can we use animal data to estimate ?safe" doses for chemical carcinogens? In Health: Quantitative ?-Iethods. AS. Whittemorc SIAM, Philadelphia, 13-30. Guess, ILA. and Crump, KS. (1977). Maximum likelihood estimation of dose- response functions subject to absolutely monotonic constraints. Annals of Statistics 101-111. Guess, HA. and Crump, KS. (1978). Best?estimate low-dose extrapolation of carcinogenicity data. Environmental Health Perspectives 22, 149-152. Guess, II.A., Crump, KS. and Peto, R. (1977). Uncertainty estimatES for low-dose-rate extrapolations of animal carcinogenicity data. Cancer Research 37, 3475-3483. Guess, H.A. and Hoe], D.G. (1977). The effect of dose on cancer latency period. Journal of Environmental Pathology and Toxicology 1, 279-286. Hartley, HO. and Sielken, BL. (1977). Estimation of "safe doses" in car- cinogenic experiments. Biometrics 33, 1-30. Hartley, HO. and Sielken, BL. (1977). Estimation of ?safe doses? in car- cinogenic experiments. Journal of Environmental Pathology and Toxicology 1, 241-266. Discussion-(by N. Mantel), 267-269, and response (by HG. Hartley and BL. Sielken), 270-278. Hartley, HO. and Sielken, BL. (1978). Development of statistical method- ology for risk estimation. Technical Report prepared for National Center for Toxicological Research, Institute of Statistics, Texas A 6: University, College Station, Texas. Hartley, H.O., Tolley, HD. and Sielken, ILL. (1981). The product form of the hazard rate model in carcinogenic testing. In Current Tapies in Proba- bility and Statistics. M. Csorgo, D. Dawson, J.N.K. Rae and E. Saleh (eds), North Holland, New York. In press. Haseman, Hoel, DC. and Jennrich, RI. (1981). Some practical prob- lems arising from use of the gamma multi-hit model for risk estimation. Journal of Toxicology and Environmental Health. In press. Heel, DC. (1976). Statistical extrapolation methods for estimating risks from animal data. Annals of the New York Academy of Sciences 271, 418?420. Heel, DC. (1979). Animal experimentation and its relevance to man. En- vironmental Health Perspectives 3:2, 25-30. Heel, DC. (1980). Incorporation of background response in dose-response models. Federation Proceedings 39, 67-69. Hoel, D.G., Gaylor, D., R, Saffiotti, U. and Schneiderman, M. (1975). Estimation of risks of irreversible delayed toxicity. Journal of Toxi~ cology and Environmental Health 1, 133-151. Krewski, D. and Van Ryzin, I. (1981). Dose response models for quantal response toxicity data. In Current Topics in Probability and Statistics. M. Csorgo, D. Dawson, J.N.K. Rae and E. Saleh (eds), North-Holland, New York. In press. .. . .9. .. vawn'w- um r: . l} . Vol. 3 a to estimate Quantitative 16011 of dose- ts. Annals of extrapolation 2, 149-152. estimates for war Research aneer latency 279-286. .oses? in car- loses? in car- Toxicology 1513 (by H.O. tical method? tional Center ii University, duct form of lies in Proba- Saleh (eds), ractical prob- estimation. 1g risks from EU, 418-420. to man. En- dose-response :iderman, M. imal of Tort- i for quanta! Statistics. M. iolland, New A .. .?lnD: . I'd ad?s.? MA. -. I V?hl. u. No. Carcinogenic Risk Assessment 95 Mantel, N. (1977). Aspects of the Hartley?Sielken approach for setting "safe doses" of carcinogens. In Origins of Human Cancer (Book C: Human Risk Assessment). ILH. Hiatt, \V'atson and ?"inston (eds), Cold Spring Harbour Laboratory, Cold Spring Harbour, 1397-1401. Mantel, N. (1978). Letter to the editor and reply by K.S. Crump. Cancer Research 38, 1835-1838. Mantel, N. (1980). Limited usefulness of mathematical models for assessing the carcinogenic risk of minute doses. Archives of Toxicology, Supplement 3 (Quantitative Aspects of Risk Assessment in Chemical Cureinogcns), 305-310. Mantel, N., Bohidar, N., Brown, C., Ciminera, I. and Tukey, J. (1975). An improved Mantel-Bryan procedure for ?safety" testing of carcinogens. Can- cer Research 35, 865-872.. Mantel, N. and Bryan, (1961). "Safety" testing of carcinogenic agents. Journal of the National Cancer Institute 27, 455-470. Pete, R. (1977). Epidemiology, multistage models and short-term muta- genicity tests. In Origins of Human Cancer (Book C: Human Risk Assess- ment). H.H. Hiatt, JD. Watson and LA. Winston (eds), Cold Spring Har- bour Laboratory, Cold Spring Harbour, 1403-1428. Pete, R. (1978). Carcinogenic effects of chronic exposure to very low levels of toxic substances. Environmental Health Perspectives 22, 155-159. Prentice, R. (1976). A generalization of the probit and legit methods for dose response curves. Biometrics 32, 761-768. Kai, K. and Van Ryzip, I. (1980). A generalized multi-hit dose-response model for low-dose extrapolation. Biometrics. In press. Raf, K. and Van Ryzin, J. (1979). Risk assessment of toxic environmental sub- stances using a generalized multi-hit dose response model. In Energy and Health. N. Breslow and A. Whittemore (eds), SIAM, Philadelphia, 99-117. Salsburg, D. (1977). Open query: Theoretical problems in the modified Mantel-Bryan procedure. Biometrics 33, 419-421. Responses (by RS. Crump and N. Mantel). Biometrics 33, 752-757. Schneiderman, 31. and Mantel, N. (1973). The Delaney clause and a scheme for rewarding good experimentation. Preventative Medicine 2, 165-170. Schneiderman, M.A., Mantel, N. and Brown, C.C. (1975). From mouse to man?or how to get from the laboratory to Park Avenue and 59th Street. Annals of the New York Academy of Sciences 246, 237-246. Turner, M. (1975). Same classes of hit theory models. Mathematical Biosci- ences 23, 219-235. Van Ryzin, J. and Bai, K. (1979). The use of quanta! responSe data to make predictions. In The Scientific Basis of Toxicity Assessment. H. Witschi Elsevier/ North Holland, New York, 273-290. Van Ryzin, I. (1980). Quantitative risk assessment. Journal of Occupational Medicine 22, 321-326. Cumh'Mu.- 96 Toxrc SUBSTANCES JOURNAL Vol. 3 (1979). The problem of estimating sale dose levels in chemical carcinogenesis. Journal of Cancer Research and Clinical Oncology 95, 101- 107. ?Walsh, and Richmond, C.B. (eds) (1981). Health Risk Analysis. Franklin Institute Press, Philadelphia. In press. Watson, (3.5. (1977). Age incidence curves for cancer. Proceedings of the National Academy of Sciences 74, 13-41-1342. Wetherill, C.B. (1963). Sequential estimation of quanta] response curves. Journal of the Royal Statistical Society 25, 1-48. Whittemorc, AS. (1980). Mathematical models of cancer and their use in risk assessment. Journal of Environmental Pathology and Toxicology 3, 353-369.. Whittemore, A. and Keller, (1978). Quantitative theories of carcino- genesis. SIAM Review 20, 1-30. 3.2 Existence of Thresholds Brown, C. (1976). Mathematical aspects of dose-restionse studies in carcino- genesis?The concept of thresholds. Oncology 33, 62-65. Brown, C., Fears, T.R., Cail, 31.11., Schneiderman, 31., Tarone, R.E., and Mantel, N. (1978). Letter to the editor and reply by J. Cornfield. Science 202, 1105?1108. Claus, (3., Krisko, I. and Bolander, K. (1974). Chemical carcinogens in the environment and in the human diet: Can a threshold be established? Food and Cosmetics Toxicology 12, 737-746. Cornfield, J. (1977). Carcinogenic risk assessment. Science 198, 693-699. Cornfield, 1., Carlborg, F.W. and Van Ryzin, J. (1978). Setting tolerances on the basis of mathematical treatment of dose-response data extrapolated to low doses. In Proceedings of the First International Congress on Text- cology. C.L. Plan and Due-an (eds), Academic Press, New York, 143-164. Crunip, K.S., Iloel, D.C., Langley, C. and hto, R. (1976). Fundamental carcinogenic processes and their implicatitm for low dose risk assessment. Cancer Research 36, 2973?2979. Falk, H.L. (1978). Biologic evidence for the existence of thresholds in chemi- cal carcinogenesis. Enciromnental Health Perspectives 22, 167-170. Frecse, E. (1973). Threshold in toxic, tcratogenie, mutagenic and carcinogenic effects. Environmental Health Perspectives, Eweritnental Issue 6, 171-178. Cehring, P. (1977). The risk equations-The threshold controversy. Sci- entist 75, 426-428. Hoel, D.C., Gaylor, D., R., Sal??ti, U. and Schneiderman, M. (1975). Estimation of risks of irreversible :blayed toxicity. Journal of Toxi- cology and Environmental Health 1, 133-15}. Hooper, K.N., Harris, ?1.11. and Ames, 13.24. (1979). Chemical carcinogens. Science 203, 602-603. .1 Vol. 3 in chemical an 95, 101- sis. Franklin iings of the mse curves. their use in oxicology 3, of carcino- 5 in carcino- 3, 11.13., and ield. Science ogens in the ?shed? Food 693-699. mg tolerances extrapolated Fess on Toxi- New York, Fundamental assessment. ilds in chemi- 67-170. It carcinogenic 6, 171-178. my. New Sei- ieiderman, M. arm! of Taxi- 1 carcinogensCareinogenic Risk Assessment 97 Jones, H. (1978). Dorm-effect relationshipsin carcinogenesis and the matter of threshold of carcinogenesis. Health Perspectives 2'2, 171-172. Kotin, P. (1976). Dose-response relationship and threshold coacepts. Annals of the New York Academy of Sciences 271, 22-28. Lowy, FL, Albrecht, R., Pelissier, M. and Manchon, P. (1977). of the "no effect levels" of two pesticides, lindane and zineb, on the micro- somal enzyme activities of rat liver. Toxicology and Applied Pharmacology 42, 329-338. Lu, F.C. (1976). Threshold doses in chemical carcinogenesis. Oncology 33, 50. Mantel, N. (1963). The concept of threshold in carcinogenesis. Clinical Phar- macology and Therapeutics 4, 104-109. Mantel, N., Heston, W. and Curian, J. (1961). Thresholds in linear dose- response models for carcinogenesis. Journal of the National Cancer Institute 27, 203-215. Mantel, N. and Schneidennan, M. (1975). Estimating ?safe" levels, a haz- ardous undertaking. Cancer Research 35, 1379-1386. National Cancer Institute (1975). Proceedings of the nineteenth meeting of the interagency collaborative group on environmental carcinogenesis: Thresh- old evaluation. Division of Cancer Cause and Prevention, National Cancer Institute, Bethesda, Maryland. Preussmann, B. (1976). Chemical carcinogens in the human environment: Problems and quantitative aspects. Oncology 33, 51-57. Rail, DP. (1978). Thresholds? Environmental Health Perspectives 22, 163-165. Schneiderman, MA. (1974). Safe dose? Problems of the statistician in the world of trans-science. Journal of the Washington Academy of Sciences 64, 68-78. Sehneidennan, MA. and Brown, CC. (1978). Estimating cancer risks to a population. Environmental Health Perspectives 22, 115-124. Schneider-man, M.A., Decoull?, P. and Brown, C.C. (1979). Thresholds for environmental cancer: Biologic and statistical considerations. Annals of the New York Academy of Sciences 329, 92-130. Truhaut, R. (1979). An overview of the problem of thresholds for chemical carcinogens. In Careinogenic Risks/Strategies for Intervention. W. Davis and C. Rosenfeld (eds), IARC Scientific Publications No. 25, Lyon, France. Truhaut, R. (1980). The problem of thresholds for chemical carcinogens?Its importance in industrial hygiene, especially in the field of permissible limits for occupational exposure. American Industrial Hygiene Association Journal 41, 685-692. 3.3 Biological? Considerations American Society for Pharmacology and Experimental Therapeutics (1980). Extrapolatiofi of laboratory toxicity data to man: Factors influencing the dose-toxic response relationship. Federation Proceedings 39, 53-82. - - 93 Toxic JOURNAL Vol. 3 Brown, (1977). The shape of the close-response curve for radiation car- cinogenesis. Radiation Research 71, 34-50. Brown, 1M. (1976). Linearity vs. non-linearity of dose response for radiation carcinogenesis. Health Physics 31, 231-245. Cairns, J. (1975). The cancer problem. Scientific American 233, 64-78. Campbell, TC. (1980). Chemical carcinogens and human risk assessment. Federation Proceedings 39, 2467-2484. Cook, P., Doll, R. and Fellingham, S. (1969). A mathematical model for the age distribution of cancer in man. International Journal of Cancer 4, 93-112. Dedrick, BL. (1973). Animal scale-up. Journal of Pharmacokinetics and Bio- pharmaceutics 1, 435-461. Dixon, BL. (1976). Problems extrapolating toxicity data from laboratory ani- mals to man. Environmental Health Perspectives 13, 43-50. Dobson, BL. (1978). How toxic is tritium? Relevance of high-dose results and gamma-ray data to evaluating low-level, chronic exposure. Environmental Health Perspectives :22, 145-147. Doll, R. (1971). The age distribution of cancer: Implications for models of carcinogenesis. Journal of the Royal Statistical Society, Series A 37, 133-166. Druckery, H. (1967). Quantitative aspects of chemical carcinogenesis. In Potential Carcinogenic Hazards from Drugs (Evaluation of Risks). R. Tru- haut VICC Monograph Series, Vol. 7, Springer-Verlag, New York, 60-78. . Emmelot, P. and Schrerer, E. (1977). Multi-liit kinetics of tumor formation, with special reference to experimental liver and human lung carcinogenesis. Cancer Research 37, 1702-1708. Gehring, and Blau, GE. (1977). Mechanisms of carcinogenesis: Dose response. Journal of Environmental Pathology and Toxicology 1, 163-179. Cehring, Watanabe, RC. and Park, C.N. (1978). Resolution of dose response toxicity data for chemicals requiring metabolic activation: Exam- ple?vinyl chloride. Toxicology and Applied Pharmacology 44, 581-591. Cehring, P.J., \Vatanabe, RC. and Park, ON. (1979). Risk of angiosarcoma in workers exposed to vinyl chloride as predicted from studies in rats. Toxi- cology and Applied Pharmacology 49, 15-21. Cehring, Watanabe, RC. and Young, JD. (1977). The relevance of dose dependent phannacokinetics in the assessment of carcinogenic hazard of chemicals. In Origins of Human Cancer (Book A: Incidence of Cancer in Humans). H.H. Iliatt, JD. Watson and Winston (eds), Cold Spring Harbour Laboratory, Cold Spring ?arbour, 187-203. Gillette, (1976). Application of phannacokinctic principles in the extrapo- lation of animal data to humans. Clinical Toxicology 9, 709-722. Gillette, LR. (1977). Kinetics of reactive metabolites and covalent binding in vivo and in oitro. In Biologically Reactive Intermediates. oilow, JJ. Kocsis, R. Snyder and H. \"ainio (eds), Plenum Press, New York, 25-41. . Vol. 3 radiation car- for radiation 3, 64-78. 51: assessment. model for the war 4, 93-112. relics and Bio- lab-oratory ani- zh-dose results Environmental for models of A 37, 133-166. cinogenesis. In Risks). It. Tru? ag, New York. nnor formation. carcinogenesrs. l0g8n6515: Dose ngy I, 163-179. olution of dose titration: Exam- 14, 581-591. 3f angiosarcoma es in rats. Toxi- he relevance of inogenic hazard lence of Cancer 5.), Cold Spring 5 in the extrapo- 9-722. covalent binding :63. DJ. Follow, \?ew York, 25-41. . . . No. 1 Carcinogenic Risk Assessment 99 Hoe], DC. (1977). Some problems in'low-dose extrapolation. In Origins of Human Cancer (Book C: Human Risk Assessment). 11.11. Iliatt, Watson and LA. (eds), Cold Spring Ilarbour Laboratory, Cold Spring Harbour, 1391?1396. Jones, 11.13. and Crendon, A. (1975). Environmental factors in the origin of cancer and estimation of possible hazard to man. Food and Cosmetics Toxi- cology 13, 251?268. Kellerer, AM. and Rossi, 11.11. (1971). RBE and the primary mechanism of radiation action. Radiation Research 47, 15-34. chihues, P. and Cooper, H.K. (1976). Repair excision of alkylated bases from DNA in vivo. Oncology 33, 86-88. Kolbye, A. (1976). Cancer in humans: Exposures and responses in a real world. Oncology 33, 90-100. Kon, 5.11. (1978). Underestimation of chronic toxicities of food additives and chemicals: The bias of a phantom rule. Medical Hypotheses 4, 324-339. Krasovskii, (1976). Extrapolation of experimental data from animals to man. Environmental Health Perspectives 13, 51-58. Krocs, R. (1979). Animal data, interpretation and consequences. In Environ- mental Carcinogenesis. Occurrence, Hisl: Evaluation and Mechanisms. P. Emmelot and E. Kriek (eds), New York, 287-302. Lepowski, W. (1978). Extrapolation of carcinogenesis data. Environmental Health Perspectives 22, 173-181. Littlefield, N., Cueto, C., Davis, A. and Mediock, K. (1975). Chronic dose- response studies in mice fed 2-AAF. Journal of Toxicology and Environ- mental l-lealth 1, 25-37. Maugh, T.H. (1978). Chemical carcinogens: How dangerous are low doses? Science 202, 37-41. Discussion (by Hooper, N., Harris, R. and Amos, 13.; and Schneiderman, MA.) Science 203, 603. Discussion (by Reitz, It, Quasi, Watanabc, P. and Cehring, P.) Science 205, 1206-1208. National Academy of Sciences Safe Drinking ?Water Committee/Risk Subcom- mittee (1979). Problems of risk estimation. National Academy of Sciences, Washington, DC. Nelson, N. (1978). Comments on extrapolation of cancer response from high dose to low dose. Environmental Health Perspectives 22, 93-95. Pochin, BE. (1978). Assumption of linearity in doseoeffect relationships. Environmental Health Perspectives 22, 103-105. Port, R., Schmahl, D. and Wahrendorf, J. (1976). Some examples of dose- response studies in chemical carcinogenesis. Oncology 33, 66?71. Fall, D.P. (1979). The role of laboratory animal studies in estimating carcino- genic risks for man. In Carcinogcnic Risk Strategies for Intervention. IARC Scientific Publications No. 25, Lyon, 179?189. Reitz, R.II., Quasi, Schumann, A.M., Watanabe, RC. and Cehring, (1980). Non-linear pharmacokinctic parameters need to be considered in meanings? u. ?gla . dish-AI; -. 100 Toxrc SUBSTANCES JOURNAL Vol. 3 high dose/low dose extrapolation. In Quantitative Aspects of Risk Assess- ment in Clmnical Carcinogenesis, Archives of Toxicology, Supplement 3, 79-94. - Beitz, 11.11., Quast, J.F., Stott, W.T., Watanabe, P.C. and Cehring, PJ. (1979). Pharmacolcinetics and macromolecular effects of chloroform in rats and mice: Implications for carcinogenic risk estimation. In ?later Chlorination. Environmental impact aml Health Effects, Volume 3. R.L. Jolley, WA. Brungs, 11.13. Cumming and V.A. Jacobs Ann Arbor Science, Ann Arbor, Michigan, 983-993. Reitz, R.H., ??atanabe, P.C., McKenna, 31.1., Quast, and Cehring, PJ. (1980). Effects of vinylidene chloride on DNA and DNA repair in the rat and mouse: A comparative study with Toxi- cology and Applied Pharmacology 52, 357-370. Rossi, H.H. (1977 The effects of small doses of ionizing radiation. Radiation Research 71, 1-8. Rossi, RH. and Kellerer, AM. (1972). Radiation carcinogenesis in low doses. Science 175, 200-202. Salsburg. D.S. (1979). The effects of lifetime feeding studies on the patterns of senile lesions in mice and rats. Drag and Chemical Toxicology 3, 1-33. Staffa, and Mehlman, BLA. (eds) (1979). Innovations in Cancer Risk Assessment (E001 Study). Pathotox Publishers, Park Forest South, Illinois. \Valsh, PJ. and Richmond, C.B. (eds.) (1981). Health Risk Analysis. Franklin Institute Press, Philadelphia. In press. Watanabe, P.C., Reitz, R.H., Schumann, A.M., McKenna. Quast, and Cehring, (1979). Implications of the mechanisms of tumorigenicity for'risk assessment. In The Scientific Basis of Toxicity Assessmem. H. Witschi Elsevier/North Holland, New York, 69-89. Watanabe, P.C., Young, JD. and Cehring, PJ. (1977). The importance of non-linear (dose-dependent) pharmacokineties in hazard assessment. Jour- nal of Environmental Pathology and Toxicology 1, 147-159. WHO Scientific Group (1974). Assessment of the carcinogenicity and muta- genicity of chemicals. Technical Report No. 546, World Health Organiza- tion, Ceneva. 3.4 Optimal Designs Chambers, EA. and Cox, DE. (1967). Discrimination between alternative binary response models. Biometrika 54, 573-58. Farmer, Uhler, and Haley, (1977). An unbalanced experimental design for dose response studies. Journal of Environmental Pathology and Toxicology 1, 293-299. Heel, RC. and Jennrich, R.I. (1979). Optimal designs for dose response ex- periments in cancer research. Biometrika 66, 307?316. ?3 .-. .qn ..-- . . Vol. 3 rf Risk Assess- Supplement 3, g, (1979). in rats and Chlorination. . Iolley, W.A. Science, Ann Cehring, id DNA repair 'osamine. Toxi- _tion. Radiation in low doses. on the patterns cology 3, 1-33. in Cancer Risk South, Illinois. ialysis. Franklin Quasi, tumorigenicity :ent. H. Witschi .- importance of :sessment. Jour- icity and muta- ealth Organiza- veen alternative zed experimental Pathology and use response exCarcinogenic Risk Assessment 101 Kalish, L.A. and Rosenberger. (1978). Optimal designs for the estima- tion of the logistic function. Technical Report No. 33, State University, University Park, Krewski, D. and Kovar, J. (1980). Low dose extrapolation under single param? eter dose response models. Conmumicalions in Statistics, Series B. In press. Meeker, W.Q. and Hahn, (1977). optimum over-stress tests to estimate the survival probability at a condition with a low expected failure probability. chlmomctrics 19, 381-399. Wong, S.C. (1979). Design for low dose extrapolation of carcinogenicity data. Technical Report No. 48, Stanford University, Stanford, California. 4. Regulatory Considerations Burton, 1. and Whyte, A. (1979). Environmental Risk Assessment. John Wiley Sons, New York. Calkins, D.R., Dixon, R.L., Gerber, C.B., Zarin, D. and Omenn, GS. (1980). Identification, characterization, and control of potential human carcinogens: A framework for federal decision making. Journal of the National Cancer Institute 64, 169-176. Citizens? Commission on Science, Law and the Food Supply (1974). Current ethical considerations in the determination of acceptable risk with regard to food and food additives. Citizens? Commission on Science, Law and the Food Supply, New York. Citizens? Commission on Science, Law and the Food Supply on recent symposia and public conferences which considered some aspects of the social and ethical implications of benefit-risk decision making. Citi- zens' Commission on Science, Law and the Food Supply, New York. Cohen, BL. and Lee, LS. (1979). A catalog of risks. Health Physics 36, 707-722. Cooper, RM. (1978). The role of regulatory agencies in risk-benefit decision- making. Food, Drug, Cosmetic Law Journal 33, 755-773. Cornfield, 1., Rai, K., and Van Ryzin, J. (1979). Procedures for assessing risk at low levels of exposure. Archives of Toxicology, Supplement 3 (Quanti- tative Aspects of Risk Assessment in Chemical Carcinogens), 295-303. Coulston, F. (1979). Regulatory Aspects of Carcinogenesis and Food Additives: The Delaney Clause. Academic Press, New York. Darby, (1973). Acceptaiile risk and practical safety. Philosophy in the decision-making process. Journal of the American Medical Association 224, 1165-1168. Darby, VVJ. (1976). Benefit-risk decision making and food safety. In Food, Man and Society. D. Walcher, N. and H. Barnett (eds), Plenum Publishing, New York, 128-149:- Darby, (1976). Balancing the benefits and risks of the application of science to agriculture and food production. In Nutrition and Agricultural . . . . . . .. A Als?lm man' .a A A-J- 102 Toxic SUBSTANCES IOUBNAL Vol. 3 1 Development. 3. Scrimshaw and M. B?har (eds), Plenum Publishing, New York, 155-188. Darby, (1980). The nature of benefits. Nutrition Reviews 38, 37-44. rm? ?howl-ISM Food and Drug Administration Advisory Committee on Protocols for Safety Evaluation (1971). Panel on carcinogenesis report on cancer testing in the safety evaluation of food additives. Toxicology and Applied Pharmacology 20, 419-438. Food Safety Council (1980). Proposed System for Food Safety Assessment. Food Safety Council, 1725 Street, Washington, D.C. 20006. Food Safety Council (1980). Principles and Processes for Making Food Safety . . 3 Decisions. Food Safety Council, 1725 Street, N.W., Washington, DC. 3 i 5 20006. i 1? 4, Food Technologists? Expert Panel on Food Safety and Nutrition (1978). The 5 risk/benefit concept as applied to food. Institute of Food Technologists, 221 North La Salle Street, Chicago, Illinois 60601. Gail, M. (1975). Measuring the benefit of reduced exposure to environmental carcinogens. Journal of Chronic Diseases 28, 135-147. I Cori, 0.13. (1980). The regulation of carcinogenic hazards. Science 208, 256-261. 1 - Grocery Manufacturers of America, Food Safety Council and The National Consumers Symposium (1979). Developing public policy for food safety. 1 Food Technology 33, 43-67. I 1 Hammond, EC. and Selikoff, (eds) (1979). Public control of environ? mental health hazards. Annals of the New York Academy of Sciences 329, 14105. Heel, D.C., Gaylor, D., R., Saffiotti, U. and Schneiderman, MA. (1975). Estimation of risk of irreversible delayed toxicity. Journal of Toxi- cology and Environmental Health 1, 133-151. _w ??up Hunter, W.C. and Crowley, JJ. (1979). Hazardous substances, the environ- ment and public health: A statistical overview. Environmental Health Per- spectives 32, 241-254. Interagency Regulatory Liaison Group (1979). Scientific bases for identifica- tion of potential carcinogens and estimation of risks. Journal of the National Cancer Institute 63, 241-268. Kates, RJV. (1978). Risk assessment of environmental hazard. John Wiley Sons, New York. Lowrance, W.W. (1976). Of acceptable risk: Science and the determination of safety. William Kaufmann, Los Altos, California. Maugh, (1978). Chemical carcinogens: The scientific basis for regulation. Science 201, 1200-1205. Miller, S.A. (1978). Risk/benefit, no effect levels and Delaney: Is the mes- sage getting through? Journal of Food Technology 32, 93-96 Vol. 3 dishing, New 33, 37?4-1; '15 for Safety esting in the ?harmacology Assessment. 3. 20006. Food Safety ungton, DC. (1978). The Jologists, 2:21 nvironmental Science 208, the National food safety. 1 of environ- Sciences 329, Lennan, BLA. moi of Toxi- the environ- Hcalth Per- or idea tifica- the National 0111] \Viley 6r tetermination or regulation. Is the mes- m, . kt". .IW1 . .xLuAs-nnr I- . .II No. 1 Carcinogenic Risk Assessment 103 Miller, S.A. (1930). The new metaphysics. Nutrition Reviews 38, 53-6?1. National Academy of Engineering Committee on Public Engineering Policy (1971). Perspectives on benefit-risk decision making. National Academy of Engineering, \Vashington, D.C. National Academy of Sciences Committee for a Study on Saccharin and Food Safety Policy (1978). Sac'charin: Technical assessment of risk and benefits. National Academy of Sciences, Washington, D.C. National Academy of Sciences Committee for a Study on Saccharin and Food Safety Policy (1979). Food safety policy. Scientific and societal considera- tions. National Academy of Sciences, \Vashington, DC. National Academy of Sciences Executive Committee (1975). Pest control: An assessment of present and alternative technologies Vol. 1. Contemptn'ary pest control practices and prospects. National Academy of Sciences, Wash- ington, DC. National Academy of Sciences Safe Drinking ?later Committee (1977). Drink- ing water and health. National Academy of Sciences, \Vashington, DC. National Academy of Sciences Safe Drinking Water Committee/Risk Subcom- mittee (1979). Problems of risk estimation. National Academy of Sciences, Washington, DC. National Cancer Advisory Board Subcommittee on Environmental Carcino- genesis (1977). General criteria for assessing the evidence for carcinogen- icity of chemical substances. Journal of the National Cancer Institute 58, 461-465. Newill, V.A. (1974). Regulatory decision making: The scientist's role. Journal of the Washington Academy of Sciences 64, 31-47. Oller, VV.L., Cairns, T., Bowman, M.C. and Fishbcin, L. (1980). A toxico- logical risk assessment procedure: A proposal for a surveillance index for hazardous chemicals. Archie-cs of Environmental Contamination and Toxi- cology 9, 483-490. Otway, H. and Palmer, P. (1976). Risk assessment. Futures 8, 122-134. Ramsey, Park, C.N., Ott, MC. and Cohring, PJ. (1978). Carcinogenic risk assessment: Ethylene dibromide. Toxicology and Applied Pharmacol- ogy 47, 411-414. Regulatory Council (1979). Regulation of chemical carcinogens. Regulatory Council, Washington, DC. Reitz, R.H., Gehring, and Park, ON. (1978). Carcinogenic risk estima- tion for chloroform: An alternative to procedures. Food and Cosmetics Toxicology 16, 511-514. Rhein, R.W. and Marion, L. (1977 . The Saccharin Controversy: A Guide for Consumers. Monarch Press, ew York. Rowe, W.D. (1977). The Anatomy of Risk. John Wiley (it Sons, New York. Schmidt, A. (1974). The benefit-risk equation. FDA Consumer 8, 27-31. v? j?q: lath-tantra?law?4L a; 104 TOXIC SUBSTANCES JOURNAL Vol. 3 Starr, C. (1969). Social benefit versus technological risk. Science 165, 1232- 1238. Starr, C. and Whipple, C. (1980). Risks of risk decisions. Science 208, 111-1- 1119. Stumpf, 5.13. (1978). Social aspects of risk/benefit analysis of food supply. Journal of Food Technology 32, 68-69. U.S. Congress, Office of Technology Assessment (1977). Cancer Testing Tech- nology and Saccliarin. Office of Technology Assessment, Washington, D.C. 20510. U.S. Congress. Office of Technology Assessment (1979). Environmental Cori- taminanis in Food. Office of Technology Assessment, \Vashington, D.C. 20510. U.S. Department of Health, Education and Welfare (1979). Chemical corn- pounds in food producing animals. Federal Register 44, 17070-17114. U.S. Department of Health, Education and \Velfare, Environmental Protection Agency (1976). Health risk and economic impact assessment of suspected carcinogens. Federal Register 41, 21402?91406. U.S. Department of Health, Education and Welfare, Environmental Protec? tion Agency (1979). Water quality criteria; Availability. Federal Register 44, 56627-56657. U.S. Department of Health, Education and Welfare, Food and Drug Admin? istration (1977). Food producing animals: Criteria and procedures for evaluating assays for carcinogenic residues. Federal Register 42, 10412-10137. U.S. Department of Health, Education and Welfare, Food and Drug Admin- istration (1977). Indirect additives: Polymers. Federal Register 42, 48528- 48544. Walsh, PJ. and Richmond, CR. (eds.) (1981). Health Risk Analysis. Franklin Institute Press, Philadelphia. In press. WHO Scientific Group (1974). Assessment of the carcinogenicity and muta- genicity of chemicals. Technical Report No. 546, World Health Organiza- tion, Geneva. VVodicka, v.0. (1980). Risk and responsibility. Nutrition Reviews 38, 45-52. NOTE 1. Single copies of the reports on screening bioassays conducted by NCI may be requested from the Office of Cancer Communications, National Can? cer Institute, Building 31, Room 10A21, National Institute of Health, Bethesda, Maryland 20014?that; tcial use some if?. 19). thus ?subvert the possibility of in- formed consent" (36. p. Ill). ?Prior gen- CuthL?llI? or "presumptive consent" ., p. have been proposed to deal this ethical problem. Recombinant DNA research makes it at least theoretically possible to combine the genetic characteristics of plant and mammal. to produce a ?plamrnal? or a ?mant.? We need to find a balance be- tween possibly inadvertently producing the means to cause catastrophe to man- kind. and potentially high bene?cial de- velopments. The genetic splicing of re- combinant DNA technology has already been used to transfer the rat gene for in- sulin production to bacteria (27). This development has the potentially high bene?cial consequence of making pos? sible massive commercial production of human (instead of other Species) insulin for diabetics. It also has. in the eyes of some. the possibility of catastrophe should insulin-producing bacteria get out of the laboratory into the body of a hu- man. to multiply and throw the person into insulin shock. One argument is that knowledge is power. and if we do not acquire the knowledge. other countries will. Re- Oember that in World War II the other ..1de was also working on an A-hornb. If 'we acquire the knowledge. we can also acquire the means to control the knowl? edge. if we do not. the comrols may be in other hands. These. too. are ethical considerations. As Jonas notes (28). generally there is something cxoerimental because tenta- tive about every individual'treatment. beginning With the itself. He would be a poor doctor ?ho \tould not learn front every case for the of future patients. and a poor member of the profession ?ho would not make any new insights gained from his treatments available to the profession at large. ln-stuw. we recogni/c that acquir- ing. new information-while.retaining old ethics demands adherence to the funda- mental rulc that a person should not be subjected to avoidable risk of death or physical harm unless be freely and in- telligently consents. The problem is to balance rights against bene?ts with re- spect for human dignity in the quest for the cure of human diseases. References and Notes I. S. J. Reiser. A. J. Dyck. W. J. Curran. Eds? Erhic: in Press. Cambridge. 2- W. A. Sci. Am. 236 (No. 6). 100 (I977). 3. B. Barb-er er al.. rm Human Subjects (Russell Sage Foundation. New York. 1973}. 4. G. J. Annas. The Right; of Hospital Patients. American Civil Liberties Union Handbook (Avon. New York. 1975). 5. V. Herbert. in Proceedings. lt?ettem Henri- sphere Nutrition Congress JV. P. L. White and N. Selvey. Eds. {Publishing Sciences. Acton. Mass? l975). pp. 84?91. 6. IE. E. Conn. in fruit-mm Nunuulh Occurring in Food: (National Academy of Sciences. Wash- ington. D.C.. I973). pp. 299-3?; J. Lewis. J. Med. 127. 55 (1977): Fed. chisr. 42. 3971.8 [1977). 7. H. K. Beecher. J. Am. Med. Assoc. 159. 1602 ?955); {lawsuit and the Sttrilit's (Little. llrown, nostvn. 1970). R. J. W.Tultey.$cirnce 9. J. P. Gilbert. ll. Mcl?cek. F. Musteller. mitt. p. H. C. Black. Law Dictionary. Revised {51. Paul. Minnesota. ed. 4. I963). In "United States of America v. Articles of Food and Drug Consisting of. . . apricot kernels. . . . . Civrl No. 7742-285 (US. District Court. Eastern District of Wisconsin. 29 July I977), Judge Reynolds closed a Iactrile factory after Carcinogenic Risk Assessment Man is exposed to a variety of natural and substances that are known 0 be harmful to experimental animals in Oiglt doses and consequently are under suspicion of being harmful to humans in low ones. Exposure to many of these - Substances. particularly those involving involuntary exposure through food. wa? ter. air. or the workplace is subject to 18 NOVEM ill-ZR I977 Jerome Corn?eld regulation by governmental agencies. In some instances the bene?ts conferred by a suspected substance can be achieved by other safe substances in equally satis- factory ways. in which case the most ap- propriate regulatory action is an outright ban. no regard being given to the strength of the suspicion. But irt many cases important benefits are lost if the holding as :t Finding of Fact tltal. "Anecdotal and testimonial evident t: to. to (tires or ("acts of treatments on cantr-r victims as described he lay ur possession all Hi). (If Ht but who are no: seicrttitie training and experience as experts in the field of cancer therapy. is not probative or substantial evidence of the safety and ctticacy of cancer treatments." Judge Reynolds held as 3 Corn of Law. "The testimony of lay witnesses as to the existence of cancer and the safety and cflicacy ?fan alleged cancer treatment based on their personal experience with the treatment is entitled to no weight and is therefore in- admisuble as irrelevant and non-probatiVe evi- dence." As preCedents for this Conclusion of Law. Judge Reynolds cited United States v. Horsey Cancer Clinic. l?irl Fed Rep. 2nd ser. 273 l5th Cir. CL. 1953). United States \K'tcr, 251 Fed. Rep.. 2nd ser. (51h Ctr. CL. Ithll. and liedCral Rules of Evidenc: 40]. 403.. 40.1. and 701. II. J. Cortt?cld?cirnce 193.693ll977). 13. J. A. Robinson. Law Rr't'. 76. 43 (1975). I3. V. Mik? and R. A. Good. Science 198. 677 U977). 14. L. A. Altman.N. EnglJ. Med. 286. 346(1972). 15. V. Trims- Assoc. Am. Physicians 75. 307 (I963). lb. Ant. J. Clin. 23. 555 (I975). I7. L. Thomas. Science I98. 675 (1977]. l8. 5. Barrett and 0. Knight. lids..Tne ?with Rob- bers: Him- to Protect Your Money and Your Life (Stickley. Philadelphia. 1976). J. W. Miner. J. Sci. 9. I ?Well: California v. Phillips. 75 Calif. Rep. 720 ?909). CCI?TlOl?df?l denied. 3% US. 102] (I970). l9. 8. l?l7. A bill to tepulate activities involving re- combinant acid. 19 May I977 (introduced by Senator Edward Kennedy. D- Mass); Newsletter. Fed. Am. Soc. Exp. Biol? Ill (No. S). 2 {1977). 20. ll. D. Davis. ll. S. Chem. Eng. News. 30 Ma 1977. pp. 26-43. The latest evtdencc is that cars recombinant DNA research may be greatly exaggerated H. Abelson. Science ?977); W. Gaylin. N. Enid. J. Med. 201. has N. Kinder tearing-t ltcfutc the Subcommittee on ?with. 11.5. Senate. on S. 1217 (1977). 22. T. M. Hurting: Catt. Rep. 7 (No. 2). 180977); I). Callahan. ibid.. p. 20; K. Dis- Hm!? 25. 23. C. Cohen. N. Engl. J. Med. 296. 1203 U977). 24. R. (ioldstein.il-id.. rt. IZth. 25. G. Robinson and A. Surg. 22. 209ll?J77l. 26. S. Milprim. Hastings Cent. Rep. 7 (No. 2). 19 U977). 27. A. Ullrich Science 156 l3l3 N. Wade. ibid. 197. l342 I977). 23. ll. Jonas. irt Ethita' in Mi'dt'r'itte. 5.1. Raiser. A. J. [qr-k. w. J. Curran. Lids. (MIT Press. Cam? bridge. Mass? I977). agent is banned. and the magnitude of the risk must then be balanced against the benefit conferred. The risk may be of such magnitude that banning is appropri- ate even in the face of the bene?tsthe levels to which hu- mans are exposed that a ban is not con- sidered appropriate. Risk assessment is therefore an essential component of reg? ulatory decisions. It is also a particularly appropriate topic for consideration be- cause of the mixture of statistical. scien- ti?c. and public policy considerations that it presents. The problem of risk as- sessment is the same formally. no matter what the route of exposure. but since much of the exposure is by way of food. I will con?ne my discussion to that topic. The author is professor of statistics at George Washington University. Washington. DC. 20052. 693 -v?ur- . . .1, eta-sari? I. r- v?r 'u?w 1 .4 .4 .-. train..- .4- l. iceplual llases for Safety 'aluatiun A substance may be carcinogenic or noncarcinogenic and independently it may appear in food as an additive. resi- due. natural contaminant. or migrant (I). )ii t: Trninct. d~ There is one {311th pt L) ?o it? ditives presenting. a potential carcinogen- ic risk and another governing the other three categories. Carcinogenic additives are governed by the Delaney clause that says (2): . . . no additive shall be deemed safe if it is found to induce cancer when ingested by man or animal. or if it is found. after tests which are appropriate for the evaluation of the safe? ty of food additives. to induce cancer in man or animal. . . . Such additives are therefore proscribed without regard to the magnitude of the risk at human use levels and without re- gard to possible bene?ts. in contrast. food containing a potential carcinogen. but which does not enter the food supply as an additive. for example. peanuts con- taining a?atoxin. is subject to regulation but is not automatically proscribed. It ?rst involves lite concept of a "tin tili- SL?tved effect icsel? and the use of salcly factors and is the standard tox? icologic pIoCetlurc both here and abroad (3). A Iccent document flout the linviv routnental l?rotet lion Agency (lil'A) states. ?the NOEL is delined to be the level (quantity) of a substance adminis- li??ed-tuagrottp of experimental animals at which those effects observed or meas- ured at high levels are absent and at which no significant dill'erences between the group of animals exposed to the quantity and an unexposed group of con- trol animals maintained under identiCal conditions is produced" (4). The ADI is obtained by dividing the NOEL by IUD. the rule of thumb being that man may be tenfold more sensitive than the experi- mental animal used and that there may be in addition a tenfold variation in sensi- tivity among individuals. The second set of procedures. devel- oped without explicit consideration of the first. are intended to apply speci?cal- ly to carcinogenic responses. They in- volve extrapolation downward from ob- served effects to a risk level deemed vir- Surnmary. Carcinogenic risk assessment involves a mixture of statistical, scientific. and public policy considerations. Concepts in.current use, such as "no observed el- fect levels" and "virtual safety," and the problems in implementing them by means of dose-respon .3 models. particularly the probit-log dose and linear models are re- viewed. The upper limits to risk provided by some conservative procedures are incon- sistent with coherent balancing of risks and benefits. A common basis to the dose- response curves describing both carcinogenic and noncarcinogenic effects is to be found in deactivating reactions. A simpli?ed model in which a toxic substance is acti- vated and deactivated in separate and simultaneous reactions is presented and the dose response curve implied by the model is deduced. This curve has the general lorm ol a hockey stick, with the striking part ?at or nearly ?at until the dose adminis- tered saturates the deactivation system, after which the probability ot a response rises rapidly. Such a curve describes the Bryan?Shimkin in- cidence dose response curve as well as the probit log-dose mod at. The concept of a saturation dose is relevant to risk assessments for carcinogenic and noncarcinogenic substances alike. may not be sold it? levels of the substance exceed a certain amount. termed a toler- ance. but can be sold if the amount pres- ent is below the tolerance. Similarly. any agent presenting a potential non- carcinogenic hazard. whether it appears in food as an additive or in other ways. is acceptable as long as its concentration does not exceed its tolerance. For all but the potentially carcinogenic food additives covered by the Delaney clause. therefore. the determination of an acceptable daily intake (ADI). or al- ternatively. of the risl. corresponding to a given exposure. is an i nportant part of the regulatory decision-making. Two dif- ferent procedures have been developed for making such determinations. The 694 tually safe. the extrapolation involving use of an assumed mathematical model. All models express the probability of a lifetime response. P. as a function of dosage. [for example. and differ only with respect to the choice of the They all assume the ab. sence of a threshold. that is. that if pro-- portion of control animals respond._ that 110) only for equal to zero. and that for any nonzero DJID) p. If safety is defined as zero elevation over control risk. then these models require that any nonZero dosage be deemed un- safe. exactly as in the Delaney clause. With this definition of safety. foods con- taining carcinogenic residues below the limit of analytic detectability. and hence not from foods cott- laining; no residues. must be deemed tltl- sale. leaving: the regulatory agencies with an impossible enforcement prob- lem. The concept of virtual safely. in- troduced by Mantel and Bryan (5). has provided a way out of this dilemma and has been adopted by the Food and Drug: Administration (FDA) A dose. is said to be virtually safe if ,where P.I is some near-zero quantity such as the Mantel?Bryan proposal. or the Value adOpted by the FDA. The virtually safe dose (VSD). concep- tually equivalent to the ADI of the tradi- tional toxicologic procedure is then corn- puted as The calculation in- volved thus requires that we determine the diSposable constants of the assumed function.f. from observations in the ob- servable range, and extrapolate down to the unobservable response. to deter- mine the V80. The Prohit-Log Dose Model The ?rnajor problems with this ap? proach are: The choice of function has a major clTec: on the V50. more than according to the FDA Ad- visory Committee on Safety Evaluation (ii) Such functions often cannot be distinguished from each other in the ob- servable range. No ?rm scienti?c basis now exists for choosing among them. The use of probit-log dose function for description of carcinogenic dose-re- sponse relations was introduced by Bry? an and Shimkin in their classic study of the three carcinogenic hydrocarbons. dibenzanthracene. and benzopyrene (3). That function is n+0 I030 where a and {3 are disposable constants. whose values are determined from ex- perimental observations according to any one ofa variety of possible statistical methods (9). it" the probability of a back- ground response is and response to background and substance are indepen- denL ll?Plle} (2) The argument leading to the probil function is entirely a statistical one. Each experimental animal is considered characterized at the time of the experi- ment by a toieranCC. such that any dose above it will induce cancer and any be- low it will not. Because ofnormal biolog- ical variation not all animals will have SCIENCE. VOL. I93 i i? 'of the log tolerance distribution and a de- -. and at any given those animals with toler- 0w will respond. 1f the tlistri- ution ofttilcrances is assumed to be log- normal. then the probit-log dose function results. with the probit slope. B. being tc reciprocal of the standard deviation pending in a simple way on the mean as well The major problem with the use of the probit-log dose model for extrapolation is that the normal distribution may not provide as reliable a description in the tails of the distribution as it does in the central part. particularly if one goes as far out as or 10". One does not ex- pect to see 6-inch-tall men or human liv- ers in the picogram range. despite the lognormality of human height or liver weight in the central part ofthe curve. If the same argument can be applied to the distribution of tolerances, the probit . function obtained from the obsewable range would overestimate the probability of a reSponse at low doses. On the other hand. human tolerance distributions could be more variable than those of in- bred strains oflaboratory animals. and to allow for this (but not the possible trun- cation ofthe tolerance distribution above zero), Mantel and Bryan proposed downward extrapolation using an ?arbi- arily low" slope of unity. the rationale "being that all observed probit slopes at the time of the proposal exceeded that value This allowance, which is the conceptual equivalent of the standard toxicologic allowance of tenfold for hu- man variation could also have been achieved by using some fraction. such as one-half. of the observed slope for ex- trapolation. thus preserving some con- tact with the observed dose-response re? lation. The failure of the Mantel-Bryan procedure to give any weight to the ob- served slope can be considered a weak- ness. and the assumption that the toler- ance distribution starts at zero lacks ob- servational support. Low Dose Linearity The probit-log dose extrapolation. even with the Mantel-Bryan modi?ca- tion. has been criticized (12) as in- su?ieiently conservative on the grounds that the extrapolated probability ap- proaches zero with decreasing dose Grime rapidly than any polynomial func- i on of dose. and in particular more rap- idly than a linear function of dose. and hence may overestimate probabilities at low doses. Thus. if at dose the propor- tion responding is 2.5 percent above con- I8 NOVEMBER 1977 trol. a 10?? elevation above control will occur at (lust: under a linear?cx- trupolatiott. at dose Di?blfl under an ex- trapolation with a probit-log dust: slope of unity. and dose DHU with a probit slope M179. The low dose linearity as? 5untption. which thus clearly leads to very much lower VSD. isjustitied in two Lllfi?CTi?luhmwS. Became carcinogen? esis is tive" assumptions are required to pro- tect the public safety. (ii) Carcinogenesis is well enough understood to make the low dose linearity assumption :1 scientif- ically reasonable one. The one-hit model provides one pos? sible scientific rationale for low dose lin- ean'ty (13). Whether or not a hit occurs is considered a chance event. If the proba- bilities of a hit on many exposures are constant and independent. then the Pois- son distribution, which is applicable. im- plies that the probability of one or more hits. and hence of an observable con- sequence, is given by ftD) expt- in) o) where til) is the expected number of hits at dosage D. The function (liq. 3) is also a dose-response curve. which at low re? sponse levels. say less than 10 percent. is essentially linear in with slope a. The probabilistc component in this dose-re- sponse curve arises not from variation in susceptibility of organisms. as in the pro- bit-lug dose model. but from whether or not a hit occurs. In some experiments in carcinogenesis the one-hit model provides a satisfactory description of the dose-response relation in the observable range (H). but since" the one?hit model and the probit-log dose model with slopes of 1.5 to 2.0 are not readily distinguishable this by itself provides no evidence on linearity at low doses. Much larger numbers at low doses are required to distinguish be- tween these models. Such numbers are provided by certain kinds of epidemio- logic studies. particularly in cigarette smoking (l6) and allatoxin (l7) and these do appear to exhibit low dose linearity. But errors itt reporting dose in the case of smoking. or variations around the av- erage consumption for the village in the case of a?atoxin. will distort any true convex dose-response curve in the direc- tion of linearity so that the epi- demiologic evidence is inconclusive on this point (18). Large numbers at low doses are also available in experiments with radiation-induced carcinogenesis. but there is considerable dispute as to whether these demonstrate linearity Furthermore. the applicability to chem- ical carcinogenesis of even a generally accepted demonstration in radiation car- cinogenesis would remain in doubt. Low do~.~e linearity has been found in muta- genesis with chemical carcinogens in bacterial systems (.70) but not in other cell systems 22). Mutagenesis is gen- erally regarded as an early step in the carcinogenic process. but as we shall ar- gue in more detail shonly. the linearity of all other steps can by no means be safely assunied. There is also a well? based mathematical argument due to ct al. (23} which says that if enr- cinogens in the environment and a newly introduced one are additive. the effect of low doses of the newly introduced must be linear. But the additivity as- sumption is a major one that lacks exper- imental support (24). Thus. none of the above arguments for low dose linearity, singly or in com- bination. can be regarded as convincing. and the scienti?c reasonableness of the hypothesis remains in doubt. Another possible justification for it does emerge in my later discussion. however. On ?Conservative" Procedures The other argument for the low dose linearity assumption. its conservatism in the face of scienti?c ignorance. raises phil050phical rather than scienti?c is- sues. and is more dif?cult to discoss. The same issue of conservatism arises in oth- er ways than as a justi?cation for low- dose linearity. Thus it is ?conservative" to use upper confidence limits on the es- timated VSD rather than the V50 them- selves. and Mantel and Bryan (5) and Hartley and Sielken have proposed this. Similarly. ifan agent is carcinogenic in one species another, it is ?conservative" to assume that the more sensitive sex in the most sensitive species best describes man. In transferring the VSD from the most sen- sitive species to man the dosage can be expressed on a body weight or a dietary concentration basis. the latter being more "conservative" since it leads to about a 15-fold lower VSD for man when the experimental animal is the mouse. Although comparison of animal and hu- man results on the same compounds sup- ports the body weight conversion (14). the concentration conversion continues to be used (6). apparently because of its greater conservatism. One problem with the conservatism argument is that there is no place at which it can stop. Use of the observed probit slope is almost always less ?con- servative? than the Mantel-Bryan slope of unity. which is in turn less ?con? 695 I. spa 4+1" - u. . trim-122314 . l?h' illwillihle .UA- we" than assuming linearity. tich is in turn less "conservative" than assuming: linearity with a 2:10 which last is consistent with the Delaney clause. Similarly. why stop at Using the mast sensitive species. the most sensi- tive strain within species. and the more sensitive sex? Why not use only the most sensitive individual animals. thus obtain- ing IOO percent incidence at each dose level? Or why stop at the upper 95 or 99 percent con?dence limits when an un- countable in?nity of more "conserva- tive" choices remain? In practice. of course. people do stop. but then it be- comes dif?cult to understand what they mean by being conservative. A more fundamental problem. no mat- ter where one stops. is that "con? servative" risk assessments distort the cost-bene?t analysis since an exaggerat? ed estimate of risk cannot be balanced against a sober analysis of bene?t. The principles of decision-making under un? certainty are well?known, and. although dif?cult to apply in practice. leave no doubt as to the inappropriateness of ?conservative" risk assessments in deci- sion-making. since they indicate that ex- pected risks and expected bene?ts. rath~ er than upper limits. are required (25). Thus. the appropriateness of con?dence limits for general decisirm-ntaking pur- poses was questioned by Savage more than 20 years ago (26). and developments have strengthened his ar- gument. These principles also establish the necessity of separating scienti?c as- sessments. such as the assignment of probabilities. from value judgments. such as assignment of utilities to the con- sequences ofdecisions. It appears. how- ever. that conservative assessments em? body value judgments in a not clearly identi?able way (in the use. for example. of upper rather than lower con?dence limits on risk) and should not be imposed on the decision-maker who balances risks and bene?ts in the guise of ?con- servative" mathematical assumptions. An example will perhaps illuminate this general argument. Consider a single Observed Ito-effect level. which for sim- plicity we regard as consisting of 0 ani- mals affected out of observed. The "conservative" analysis remarks that such a result is consistent with the exist? ence of a small probability of risk, say and regards the upper con?dence limit as the proper quanti?cation of this remark. From a decision point of view the appropriate eitpression of the uncer- tainty is the expected risk. given the ob- servation 0 out ofn (27). lfa uniform pri- or distribution is assumed. the expected risk is known to be l/(n 2). that is. if} 696 fut it I and ill?? for it till). 'I'htts. l/(n l- 2) the appropriatt' llayesian risk assessment at the no olIsct vt'd level fora uniform pt ior disttilnttion. ll'. Uttc lltt: cslitt'iitlt?d t'isli the upper con?dence limit for sotne st:- lected value of the confidence coef?- cient. Ir- a. namely it?. this is equivalent- to selecting a prior distribu- tion with expectation above 1 - that is. above 0.99 for n: 2 and 1. Such a prior distribution would rarely correspond to anyone?s real be? liefs and seems to reflect instead a con- cealed value judgment. that is. assign? ment of great weight to the risk. without regard to possible bene?t. Such value judgments can not and should not be avoided. but they should be made esplic- it and not introduced in so intellectually muddled a way that no one knows which are the facts and which the judgments. Carcinogenic and Noncarcinogenic Toxicity A reason often assigned for the use of different safety evaluation ptocedures for carcinogenic and noncarcinogenic substances is that the classical tox- icological procedures are useful for as- sessing acute. reversible. nonprogres- sive effects not for the chronic. pro- gressive. and irreversible effects of car- cinogenesis (38). A corollary to this view is that while pottulation thresholds may exist and can be estimated for non- carcinogenic substances. no such esti- mation is possible for carcinogens. since even one molecule might be sul?cient to initiate the process. This distinction may not be as clear-cut as it seems at first sight if one considers such toxic sub- stances as lead or such disease states as and atherosclerosis. But more fundamentally it is not clear why chronic. progressive. irreversible effects must lead to qualitatively different dose- response curves from those found for acute. short-term reversible effects. since from the point of view of establish- ing ADI it is only the dose-response curve that matters. An alternative to this dualistic view regards dose-response curves for carcinogenic and noncar- cinogenic substances alike as re?ecting the saturation of protective biological mechanisms. such as detoxi?cation and DNA repair. and considers differences on the ante-chronic or reversible-irre- versible axes as simply re?ecting de- tailed differences in the kinetics of the re? actions involved [see also and Thus. eVen if one molecule of carcinogen is suf?cient. the carcinogen will often be a metabolite of the compound adminis- tered. and. as Miller and Miller (.t'fll [ml it. much of the llUsC of the compound ad- ministered ?tnay dissipated in deacti- vation reactions" so that it may be im- possible for one molecule of the adminis- tered compound to lead to one molecule of carcinogenic metabolite. A mathemat- ical analysis of this argument is of inter- est in view of its implications for the de- velopment of more rationally based safe- ty evaluation procedures for carcinogen- ic and noncarcinogenic agents alike. It need h. rdly be emphasized that the only objective is to explore the qualitative im- plications for safety evaluation of some reasonable biological assumptions. and not the development ofa universal kinet- ic model for all toxic reactions. A Simple Kinetic Model Two reactions are considered here. In the ?rst. moles of toxic substance com- bine with 3 moles of free substrate to form x_ moles of an activated complex. which in turn reversibly disassociates in? to toxic substance and substrate. For? ward and back reactions are governed by 'rate constants I: and The amount of activated complex formed in the forward reaction is thus Ads and the amount lost in the hack reaction is LI. and since in the steady state these two amounts are equal we have lids Lx 0 (4) In the second reaction (.1 moles of free toxic substance are simultaneously deactivated by moles of free deactiva- tor to form moles of toxin?deactivator complex, which in turn reversibly dis- associates into toxic agent and deactiva- tor, the governing rate constants being k* and E. In the steady state we thus al? so have led: kfy The total moles of toxic substance in the system is .r y. which we denote by the total moles of substrate is x. which we denote by the total moles of deactivator is which we denote by T. leading to the balance equa- tions D=d+x+y T=t+y (6) Denoting k_/k by and Iii/k? by It? we rewrite Eqs. 4 and 5. using Eq(8) SCIENCE. VOL. 195 I ?Slope Fig. l. Dose-response relation Under a model of irreversible deactivation. where from Eq. 6. .r mintD.5) and miniD.T). The probability of a toxic reaction in an organism exposed to the toxic sub- stance.P. is considered proportional (29) to amount of activated complex x. and since S. we take as the con- stant of prOportionality l/S. so that 1/5 (9) We thus embody in the model the as? sumption that even one molecule of: could lead to a toxic reaction. We are in- terested in the relation between and de?ned by Eqs. 7. 8. and 9. this relation yielding the dose?response curve implied by the system. The quantities S. T. K. and are parameters and .r and vari- les to be eliminated. A direct solution yields an explicit re? lation between and but more insight results from ?rst considering an implicit solution for for the case 0. that is. the case in which there is no back deactivation reaction. it is then easily veri?ed that Eqs. 7. 8. and 9 are satis?ed by the following solution: For y=D 0030 andfor 927"" v=T (10b) Since .P for all 1: T. the dose-re- sponse curve yielded by this solution has a threshold at T. but for I) it increases with decreasing slope from 0 to 1. Such an asymmetric sig- moid curve is shown in Fig. 1. Although D. S. and in Eq. 10 are ex- pressed in moles. and is dimension- rss. multiplication of numerator and de- nominator by any constant needed to convert moles to any other unit. such as milligrams. leaves the form of the rela? tionship unaffected and converts the pa- rameters and to milligrams and to IE NOVEM ill-LR milligrams per mole. To see whether liq. 10 provides a satisfactory description of an observed tlt?tsc-tcspottsc curve. it is therefore suf?cient to estimate and and to compare the values of with those computed from liq. is done in Table for the lifetime tumor incidences in mice subcutmtcotrly in- doses of as reported by Bryan and Shim- kin The parameters 3. T. and were Estimated. somewhat crudely. by equat? ing the yielded by Eq. 10 with the yielded by the ?tted probit curve at val- ues ofP 0.05. 0.50. and 0.95. It will be observed that the description provided by Eq. 10 is satisfactory despite its no- glect of possible animal to animal varia- tion in the values of S. T. and K. This might be considered scarcely surprising. given the three disposable constants in Eq. 10, but there is no a priori reason why the values estimated for these con- stants must be positive. as the model re- quires. and as they are in this case. Thus. ifthe model is lobe rejected. it cannot be because of failure to describe the meth- ylcholanlhrene dose?response curve. Since we are here interested in qualita- tive implications. and not detailed statis- tical procedures. 1 shall not consider the modi?cations of Eq. 10 required to allow for animal variation in parameter values. although it is clear that without them the possibility of negative estimates of S. T. . and cannot be exoluded. An Expanded Kinetic Model Not all deactivating reactions are irre- versible. in the hydrocyanic acid-thio? sulfate-cytochrome oxidasc system. for example. a back deactivation reaction exists. which although quite slow com- pared to the forward reaction should Fig. 2. Model ola two-step chain of protective reactiom. not be disregarded. To see the eti'ect of such a reaction we now relax the as- sumption that in liq. Sis equal to ac- ro. A simple. though perhaps inelegant. way to investigate this more general case is to note that for small D. S?szandT?yaT (11) in which case liqs. 7 and 8 become linear ins and y. leading immediately to the so- lution *5 for 5 702) +Klt+ Thus. for small I), is linear in D. with slope approaching zero as ap- proaches 7cm. l-?or we use Eqs. 7 and 9 to ol-?ain -- SP (13) for T. where. by eliminating - from Eqs. 7 and 8 we find K*(l - PH (14) for I) T. a reversible deactivation reac? tiv -p ies IOW-dose linearity. lfthe iti? ra for the system are known. the slip-.- of the linear portion can be calcu- lated from Eq. 12. bui for a very low Table 1. Comparison of observed and estimated lifetime tumor incidence at various doses. D. or for Eq. 10 and probit-log dose curve. {r32} Observed Eq. 10? Probit-log (Jo-vet 1 20/30 0.9964 1.0000 0.5 0.9926 0.9997 0.25 21/21 0.9343 0.9962 0.125 21/21 0.9644 0.9?33 0.062 0.9044 0.8808 0.031 1320 0.6955 0.6680 0.0156 6/13 0.3636 0.3783 0.00?3 0.1490 0.1457 0.0039 0H9 0.0333 0-0360 0.002866: 0 0.0165 0.00195 0.19 0 0.0055 0.00093 0'41 0 5.0 10? 0.00024 ??79 0 8.9 10'" . -. ins calculated from the equation {set - (.0390-31 - .0025!? ?Oi-?5 \aturatiun dose. 697 . $171,347 . . . q- ,l 5-1? I ??fg Jami-r. v?v 4i"; .- 21. A e' ease of the hydroesanic acid- cytochrome system. it ?ill be negligible int v. :en the iti- neties are not known and' the slope must be inferred from the observed dose-re? sponse curve. the threshold argument provided by liq. l0 can no longer be sus- tained. That argument assumes. how- the Nit-step steady state system ol- lilii- 3 to a one-step system for lies the ?rst too parts of liq. l7 and no I) remains in step I. lint lrom liq. l3 is greatest for K, (J. in which case it re- duces to liq. l3. thus demonstrating that in that Last- ll hulls- ever. one step between the toxin a?d'thMa; existence of an additional step organism?s toxic reaction, while in fact numerous steps may exist {29. 30). it is an interesting mathematical exer- cise to deduce the consequences of as- suminga chain ofn reactions. each ofthc type previously considered. It is suf- ?cient. however. to consider the case 2 in order to demonstrate that ir- reversible deactivation in any one step in the' chain leads to a threshold for the function and (ii) that even in the absence of irreversible deactivation. the addition of a step reduces the low- dose slope. In what follows. the ?rst step can be thought of as a detoxi?cation re- action and the second as DNA repair. The model is shown in Fig. 2. The bal- ance equations arel'i .VI .12: I: (15) so that miniD. S.) y, 5 min?). T.) x, 5 mint?. 5.. )3 mintl). 3.. The equations of the system are: (xi I: Tel] (xi 1-: 0 ."ell [Tl yr] 7" 0 (K.. S, (ll reduces the low dUSe slope. These results are easily extended to cow er an ?1 Step system. Thus. for a chain ofn protective reac- tions the resulting dose-response curve is shaped much like a hockey stick. with the striking part ?at or nearly ?at and the handle rising steeply once the protective mechanisms are saturated. Gehring and Blau (29). using a kinetic system not un- like the present one. eight simultaneous reactions. typical rate constants selected from the literature, and a computer simu- lation found exactly such a curve with a saturation dose of about It)" molelkg. Discussion This analysis establishes that even if carcinogenesis is an irreversible one-hit phenomenon between the ultimate car- cinogen and DNA. an assumption em- bodied in Eq. 9. the existence of a 110 ef- fect or threshold level for the carcino? genic compound administered is not pre- cluded. Whether such lcVels do or do not exist depends on the presence of at least one irreversible protective reaction. but there seems no present reason for believ- ing that all carcinogenic processes are characterized by the absence of such re- actions and all or most noncurcin- Nils: 12) 0 ogenic pro- - 0 cesses by arr/52??) their pres- ence. We where K. and Then. for the case 0.KT 2 0 and min(5.. Eqs. 17 are satis?that it for all 0 mintsl. Ll. Simi- larly. for the case U. tt. and T.. Eqs. are satis?that 0 for all I) 1: 7.. Thus irreversibility in either of the steps leads to a threshold. To find the solution for Ni. 0. we generalize the argument used in the one- step case starting with Eq. ll. thus mak- ing Eqs. l7 linear in 5.. and and ?nd that698 have thus found no analytic basis for the sharp distinction drawn between the two classes oftoxic reaction by present safety evaluation procedures. Other bases. per- haps in the realm of value judgments. may exist. but they do not appear to have been made explicit. Although many observed dose-re- sponse curves are consistent with the existence of thresholds. for example. the data in Table i. no ?nite set of dose-re- Sponse observations could establish this. All present safety evaluation proce- dures. whether involving the use of or of some favored nort- threshold dose- response func- tion with a "virtually safe" iewL must la.- regarded as Inathetnatieal formalisms whose eut- respondenee with the realities of low. dose ell'et'ts is. and may loot: remain. largely cottieettu'al. ltut retaliatory de. cisions must be made. and formalisms with more theoretical or experimental support. or both. should be preferred to those with less. in this spirit I suggest that another formalism. with at least as much claim to reality as those now in use. is pt?UVided by the saturation dose de?ned by Eq. it]. and that the devel- Opment of ef?cient statistical procedures for commuting the posterior expectation of the saturation dose may ?nd a place in safety evaluation procedures. Reference; and Notes I. Residues include such substances as pesticides left on crops and from beef: natural contaminants include the albumin found in groundnut: and corn: migrants include such compounds as those from packatung materials. for example. vinyl chloride or aerylunitrilc. . Code 348(c)(3). . Sec Code of Federal Regulations. title 21. sect. IZI.5: ll. L. Oscr. Fond Carmel. Toxicol. 7.415 4. Attachment to letter to Chairman. EPA Envi- ronmental Health Advisory Committee from As- sistant Administrator of the EPA. Of?ce of Wa- ter and Hazardous Substances. l9 April I977. 5. N. Mantel and R. Bryan. J. Natl. Cancer Inst. 27. 455 (196?. 6. FM. Regisr. 41?. (No. 35}, Ill-ll! (22 February with. 7. Food and Drug Administration Advisory Com- mittee on Protocols for Safety Evaluation. Panel on Carcinogenesis: "Report on cancer testing. in the safety evaluation of food additives and pesti- cides. Torit?ol. ?ripl. 20. 419(1971). 8. W. It. Bryan and M. B. ShimltinJ. Natl. Can- cer inst. J. 503 U943). 9. J. liinncy. Probir Analysis: A Treatment of the Sigmnid Response Curve (Cambridge Univ. Press. London. 1952}. It). An alternative. equally descriptive. route to the probit vlog dose function is provided by the vtork of ii. Diuckery [in Potential Carcinogenic Hit:- artis- jirrmi Drugs. Evaluation of Riskr. R. Tru- haut. Ed. (Union Internationale Contra Ie Can- cer Monogr. Ser. No. 7. Springer-Verlag. New York. and of R- E. Albert and H. Icr Symp. Ser. CONFJZUSUS (1973)}. In studying the dose dependence of time to tumor for a number of carcinogenic agents Dnickery noted that they were all characterized by the empirical relation Dr' a constant. where is median time to tumor and is a constant varying between I and 4. Albert and extend- ed Drucltery's work by considering varioas sta? tistical distributions of time to tumor. in particu- lar the lognormal. It is easy to Show. however. that the relation in combination with the assumption ofa iognormal distribution of time to tumor implies that the probability of a tumor at dose level 0 by any time. T. such as the life span ofthc experimental animal. is given by Eq. I withn in a simple way on rr. T. :1an the two constants of [he distribution of time to tumor. The probit-log dose model therefore gives a more appropriate expression of time to tumor than is sometimes realized. II. It is ltnown that measurement errors. although normal in the central part of the distribution, tend to occur with above normal frequency in the tails Ill. Je?'reys. Hit-arr oj'f'mbrihiliry (Cla- rendon Press. Oxford. cd. 1. sect. 5.77.] and this makes lower slopes in the tails seem more plansiblc to some. even though the rele- vance of the distribution of measurement errors to that of tolerance is questionable. l2. H. (J. Hartley and R. L. Siclken. Jr.. Biometrics 33. I (1917}. 13. A hit is any event necessary for .1: production of an observable consequence. The event might be a bacterium landing on an agar plate or a num- ma ray penetrating a ehrommome. the con- sequence being the growth of a colony or a chromosome breakage. in a one-hit model a HH SCIENCE. VOL. I93 to the con? Academy of h?cwtact?s. (norm! I'mr'tir'rt i'rmpi't ti (National Asiatic-my i-l' Sun-trees. Washingmm 197)]. vol. I. 5. P630. 9, 33? 16. H. A. Kuhn, in Ann?. Cancer Inst. .lfonogrCarlhorg. J. Van Ryzin. Pro- u'eding'r of the first International 7ortcuiogi' Congrats. in press. Ill An unpublished Calculation by J. Van Ito ap- plying the continuous ntultt-htt generalization of the one-hit model (if) to the Born smoking-lung Cancer results [see {loll leads to an estimate of 0.7 hit. A less than one-hit model, for which there appears to be no biological ts Cd?l)? e\pl.iined on the repurttnit errnr hypothe- sis. \u that this result adds empirical suppon to that hypothesis as a theoretical explanation of the epidemiolngie results. l9 ll. Wollc. 196. I?ll? Ill. 1 McCamt and ll, N. Attics. in Urcupurionui U. Safliotti and l. K. Wagoner. Eds . published in Ann. Aeud. Sn. 5 (Win). B. E. Matter and Grauwiler. Murat. Res. 23. 339(l97-ll; D. J. Kilian Hui? 4-1. 22. A. W. li?lt,l?f oi. In preparation. 1'3. K. 5. Comp. D. G. lloel. C. H. Langley. R. Pelu. Cancer Re?l?. 36, 297.1 [1976). To distinguish between independent and addi- a. 0 live ellecls of two or more .u-rntr. one has to male large numbers of uhsen.:lwns. use of the cell systems Int tun-slut dim-punt ruutatdenests \tlt?fly and seems promising. Summaries in! ul'etper- in can be Canada. ?to listing of for and (Maul) Will; U. Urrc?uiugy 3.3. 73 J. C. Arum. in ('hi'rttiral ('arrimwmru'r, 0, Ts'o and A. Di- - (neither, New York. H72). part 25. 'lhe is clearly in- appropriate since it is not conservative tn foregn bene?ts?which may be more than monetary. as in the antimalarial ctfect of ?lhe strict Bayesian decision procedure. which requires as- signment of prior probabilities to all the possible Scienti?c hypotheses. utilities to all the possible consequences. the computation of an expected utility for each possible decision, and the selec- tion of the decision with maximum expected may be well beyond the capacity of any scienti?cally. legally. or politically onenlcd de- cision-maker short of l?lato's philompher hing. cVen though it is the only coherent one [see I). V. Lindley. Malling Decisions (Wiley. New York. and H. in Environmen- tal Health. Quantitative Methods. A. Whitt- more. lid. (Society for Industrial and Applied Mathematics. Philadelphia. 1977)]. 26. L. J. Savage. Foundations omeristie: (Wiley. New York. I954). The Code of the Scientist and Its Relationship to Ethics Several decades ago. Paul Valery, po- et and essayist. declared (I): Never has humanity known so much power and so much confusion. so much worry and so much play. so much knowledge and so much uncertainty. to equal measure does now an- guish. now futility. command the hours of our days. These words were undoubtedly appro- priate when Vale?ry gave pen to them. Yet today they are perhaps even more apposite. indeed. they seem to apply to three distinct spheres of human action. In the ?rst place. we live in a time in which the industrialized countries are experiencing unparalleled technological development. in large part the fruit of science. However, the bene?ts of new technologies are distributed in a grossly alanced manner. not only within indi- industrialized countries. also among all the nations ofthe world. Oven crowding and environmental degradation are already significantly reducing the quality of life in the developed nations lh? NOVEMBER l977 Andrei Coumand and give stark evidence oftheir inability to confront the problems of the future and its planning. Excess population and famine are on the increase in some re- gions. while in others there are those who enjoy material goods and leisure as never before. ln :1 word, our inability to regulate the processes of cultural and technological development poses a grave threat to our ability to achieve a decent and Immune future. - in the second place, as we all are aware. the trends toward nationalism. and its opposite, multinational indus- trialization. are growing. Many will agree with me that if a universal world order of some type is not achieved by agreement based upon reason and eco- nomic justice. the prospect is that it may be imposed by force. And third. science is now in a state of siege. Those who in the past have praised its contributions to human un- derstanding and material well-being are now questioning many facets of the sci- . .44.. mm? )7 This assumes that uliltty is linear in prohilnlity (if lls'x. a accept il?lc Hung-pup" ihc small probabilitles of com. um. hill of doubt- lul and. Cl tUllif-C. ludicrous 35 the approaches unity. 28, Summit") Heron: Urtrn'tt're Witter and Health. a fcporl of the Safe Drinking Water Committee. Adtisor} Center on Toxicolo_.y, Assembly of Life Sciences (National Research CouneiL Na- tiotm.? Academy of Sciences. Washington. I977). 29. Gehting and G. E. Blau.Pmn-edingr ofrhe First International Toxicology Congress. in press. 30. J. A. Miller and E. C. Millet]. Natl. Cancer Inst. 47 [1971). S. Goodman and A. Gilman. The Pharmera. logical Basis of Therapeutics (Macmillan. New York. l'JiUl. 32. This model thus explains low-dose linearity for mutagenicity in some cellular systems and non- linearity for carcinogenicity in mammals by the ulster-cc of deactivating mechanisms. such as detoxi?cation by the liter (r in the latter, but not in the former IT 0). A striking ex- ample ol'this is given by Hsie who studied the mutagenic ell'ects of ethyl methanesulfonate in Chinese hamster ovary cells both in vitro and in ?ve. The dose-response curve for the former was linear. but for the latter showed an apparent threshold. 33. The preparation of this manuscript was support- ed by Mil grant l?l9l. enti?c enterprise. Some even go so far as to ask whether it does not contain the po- tential for destroying civilization. in this article 1 center my discussion on something which shall argue is com- mon to each of these problems. namely. the operating and ethical code ofthe sci- entist. First. I discuss some aspects of the situation of scientists and the possi- bilities for preserving the norms of sci- ence. Second, I deal with an even broad- er question: could there be a relationship between the ethical stance of the scien- tist. qua scientist, and the problem of fostering humane socioeconomic devel- opment? In turn. these re?ections will prepare the way for a brief examination ofthe possible relationship between sci- ence and a uni?ed world-order of some type. Formulation ofan Operational and Ethical Code ofthe Scientist Scientists have developed character~ istic rules of procedure that help to pro- duce the intended outcome of their activ? ity, which is certi?ed knowledge. These rules also guide the conduct ofindividual investigators toward each other in their capacity as scientists. [it 19-12. Robert Merton formulated these rules as the The author is professor of medicine emeritus and special lecturer. Columbia University College of Physicians and Surgeons. New York 10032 and chairman of the Editorial Advisory Board of Man and .?lhu?ieine: The Journal of Value: and Ethics in Health Care. This article is the edited text of an ad- dress delitcred un 27 May 1917. at the Memorial- Sloan-Ketlen?ng Center's "Symposium on Medical Ethics: Statistics and Ethics." held at Rockefeller Universitynit- L. '1 ll"! r1 1. vvw-F'T' .: I .. i-t rvw'wi- to ii . . .f?1t? 4? ?i rail f7} l'fl.? .: i New Scientist 18 August 1977 . ?ne-mama Ecme??eme The ?Cchreshoiti eontroversy Almost every week yet another chemical is found to cause cancer in animal tests. But the doses given in animal assays are so high that the chemical may induce cancer simply because it overwhelms detoxi?cation mechanisms which protect humans from the smaller doses we meet in real life Has our increasing reliance on chemicals created an increased risk of cancer? This is rapidly be- coming the central question in public health. Our anxiety about ?the threat from chemicals" is bolstered by the fact that some chemicals have caused cancer in people exposed to them at work?~vinyl chloride, benzidine, bis(chloromethyl)ether, arsenic, chromates and nickel. But these and other chemi~ cals cause only a small, albeit still important, proportion of the cancer in Western society. John Higginson. director of the International Agency for Research in Cancer, Lyon, estimates that only 1 to 3 per cent of all tumours in indus- trial countries are caused by exposure to chemicals at work. Included in this estimate are cancers caused by exposure to asbestos dust, which may have been res- ponsible for more occupational cancer than all chemical carcinogens combined. Furthermore, occupational cancer is not a reliable indicator of the risk chemicals present to society as a whole, since ?doses? received at work are generally many times greater than outside the plant. Indeed, occupational exposure is decreasing as technology provides better means of control. So much so that most occupational cancers occurring now are the delayed rcsul . of past, rather than current, practices. Anxiety about cancer is reinforced by the fact that its incidence is apparently rising. This increase renders us gullible targets for those who wish, for whatever reason, to blame cancer on the ?nasty? chemicals made by man. Statements such as ?60 to 90 per cent of human cancer is caused by environmental agents? are naively inter- preted by many as an indictment of man-made chemical pollutants. Most peeple do not realise that smoking causes about 30 per cent of deaths from cancer, and 50 per cent of cancer in women and 30 per cent in men are related to nutritional factors. If there has been any rise in the incidence of cancer in recent years it has been caused by cigarette smoking, excessive drinking and ?overnutrition?. But because cancer is a disease of old age, the rise may be largely a result of the increasing proportion of old people in Westem populations. Indeed, the American Cancer Society reports a slight decrease in the age- adjusted incidence of cancer over the past 25 years, rather than an increase. Despite these arguments. there are those who have exploited public anxiety to further their own cause. Thus. an apparent 5-2 per cent increase in US cancer mortality during the ?rst 7 months of IQTS over that of the previous year make instant headlines. A member of the US Con- gress launched an inquiry to determine whether the increase was due to the growing use of man-made chemicals. Later, Dr Leonard Chiazze showed that the apparent increase was. in fact, a result of the improper use of statistics (Chiazze et at, JAMA, 1976, 2310). Although the anticipated environmental time bomb never went off, headlines correcting the earlier impetumis and entirely wrong predations never appeared. Such exploitation o' legitimate concern has been helped by the waiter of reports of animal experiment "hat incrimi- nate chemicals as carcinogens. The rate of 'ese revela? tions has increased steadily: the ?chemical carcinogen of Dr Perry Gehring is director of toxicology research at Dow Laboratories. Michigan [nnrarn Finn THIS CHEMICAL CARCINOGEN the year" has become ?the chemical carcinogen of the month". Now that the US National Cancer Institute is living up to its commitment to publish reports of its annual bio-assays of over 200 chemicals at a rate of three a Week, ?the chemical carcinogen of the week? is almost with us. Unfortunately, the willingness to tell the public that a chemical causes cancer in animals is not matched by an equivalent desire to make clear the inherent limitations of animal tests to assess carcinogenicity. For example, it is often not stated that the daily doses given to animals in such tests are frequently many times greater than those to which people are exposed. in many cases. the doses were so large that many animals Suffered severe tissue and organ damage that would have demanded hospital treatment had the subject been a human being. Indeed. for some chemicals, the doses have been large enough to kill a portion of the animals during the study. More fundamentally, animal tests are based on the assump- tion that an increased rate of cancer in a small poiiulation of animals is substantive. or even conclusive, evidence that much smaller doses are a measurable risk to man. Such an assumption [lies in the face of substantial evi- dence that the damage to tissue and stress caused by large doses predisposes both animals and man to cancer. There evidence that large doses of many chemi- felm the body's ?defence" mechanisms that chemical. There is also evidence that cancer discernible only in animals given doses large enough to cause major alterations in their biochemistry. At the heart of this problem of interpreting data from .xperiments in animals, is the dose-response curve (a typical curve is shown in Figure 1). Response is not the degree of an individual?s reaction to a chemical, but is the proportion of a population that suffers a speci?c effect. The solid part of the curve shows, not unexpectedly, that the percentage of a population su?'ering an eli'ect in- creases with increasing doses. This is the form of the dose-reSponse curve when the response of individuals within a population is distributed normally, a characteris- tic feature of most pharmacological or toxicological reac- tions to chemicals. The concept of a dose Flu-ports: 40" _30" 20- 1' 111 QQWB 1 Increasing dose or exposure elow which no response will occur?has been accepted for most pharmacological and toxicological reactions. After ?nding the dose that produces no adverse response in animals, it has been common practice to assume that 0-1 to 00002 of that dose will be a safe level of exposure for man. For chemical carcinogens, however. the ?threshold" concept is not universally accepted. Some believe that can- cer can be induced after the reaction of a single chemical ~molecule with a single critical receptor of DNA (or per- haps other macromolecules) in a single celL Once such a reaction has occurred, this theory holds, the cell is pro- grammed irreversibly and there is thus a ?nite probability of initiating a new line of cells growing in a disorganised manner, thus producing a tumour. This mechanism pre- dicts that there will be a ?nite incidence of cancer in a population no matter how low the dose of carcinogen. Indeed, the incidence of cancer will be a linear function of dose rather than a linear function of the logarithm of the dose as depicted in Figure 1. Unfortunately, the boundaries of this concept are beyond experimental resolution. In Figure 1 the dilemma is repre- sented by the boxed portion in the lower left corner. The question is whether the incidence of cancer quickly reaches zero as the dose is decreased, as predicted by the "thres- hold" theory (dot/dash curve) or whether the incidence will decrease as predicted by extrapolation of the known experimental dose-response data (short-dash curve) or, deed, whether the incidence may be greater than pre- ??ted by such extrapolation (long-dash curve). The resolution of this question is dit?cult, if not impos- sible. To decide which curve is correct would mean _col- lecting data on the response to low doses of the chemical concerned. But to do this?to put data in the boxed section of the curve?would mean measuring low incidences of cancer in animal pepulations, say 0 to 15 per cent. But these low incidences can only be distinguished with any degree of statistical certainty from the ?natural" rate of cancer in animals by trials involving impossibly large num- bers of both control and treated animals. Thus only rela? tivcly high doses can, in practice, yield statistically signifi- cant data. But in many cases such high doses produce cancer simply because their very quantity overwhelms the biochemical pathways that would detoxify smaller, more realistic, doses. To illustrate why this is true, refer to Figure 2 which shows the typical chemical processes that Reactive. Reactive intermediate ctissue Chemical - . . mum-el bound cmoge mecbolire irreversib!y cndfor . to mocro- Cancer molecule _Excrered - Detoxified . FigureZ determine the fate of a chemical in the body. Each arrow represents a ?rst-order rate process. The dominant pro- cess governing the fate of the chemical may be different for different species, different chemicals, and frequently, for different doses of the same chemical. Interfering with excretion or detoxi?cation will enhance covalent binding of a reactive, electrophilic metabolite to macromolecules. For example, furosamide (a diuretic) is excreted pre- dominantly intact in the urine when low doses are given to patients. When high doses are administered, renal clear- ance is overwhelmed causing a disproportionate increase in the formation of toxic metabolites which react co- valently with macromolecules. In the body, bromobenzene is transformed in the liver to the chemically reactive 3,4? bromobenzene epoxide. This molecule is detoxi?ed en- zymatically by conjugation with glutathione. However, if large doses are administered, glutathione is depleted and the reactive metabolite reacts instead with macromole- cules. The barrels in Figure 3 illustrate how dose may alter the fate of a chemical. As more and more ?uid (putative Figure 3 IVCW screnust ll} [\llgUSI It]! I nwgcu) flows into the harm-l, elimination via the. lower - it (excretion and "normal" metabolism) is overwhelmed, resulting in disproportionate increases in the amount of fluid in the barrel and/or elimination through the upper slit (excretion via other routes, metabolism via other path- ways and, in some cases, reactions with macromolecules). For many chemicals, excretion, activation, and detoxi?- cation are active transport or enzymatic reactions. Rather than being best portrayed by ?rst-order rate processes, these reactions are more accurately described by Michaelis- Menten kinetics or dose-dependent kinetics in accordance with the equation 5C 5t In this equation 8C/6t is the rate of change in the concern- tration of the chemical at time t, is the concentration of chemical at time t, is the maximum rate of the pro- cess, and Km, the Michaelis constant. is equal to the con- centration of the chemical at which the rate of the process is equal to one-half There are two important limiting cases revealed by this equation. When the concentration, that is the dose, is small compared to classical ?rst-order kinetics apply: 5C Vac 5! Km However, when the dose is large, the rate becomes constant or zero order: In this case. the process has been overwhelmed and no longer functions ef?ciently. It is likely that this phenome- non is the rule rather than the exception for chemicals possessing a low order of toxicity (which allows the administration of high doses without killing the animal). Figure 4 is a diagram depicting more fully the fate of a chemical in the body. and shows processes governed by Michaelis?Menten kinetics. The reactions of the reactive metabolite with macromolecules (both nongenetic and genetic) are, however, ?rst~order rate processes. Replica- tion of genetic material. with the reactive metabolite bound covalently to it, is naively assumed to be a ?rst?order rate process giving rise to replicated genetic material with a ?nite probability. albeit very small, of programming the cell to become cancerous. Genetic material with reactive metabolite bound to it can. however, be repaired and there is evidence that. like detoxification and excretion, the repair process can be overwhelmed. Hence this pro- cess is assumed to be governed by Michaelis-Menten kinetics. Bll'ldil'lg to non- genehc mare-rial . . 'rltefm-rijiafe Repaired . . rc- Chemical 95?th Bezcrwe irrecersiciy -pah" generic I microcode boundlro generic material mar-mat of I. .. . not can do!? ifica ion repucchon Extra-ted inscrive chemical menbolire . generic material I Possible cancer I Figure 4 This model can be quanti?ed as a series of differential equations. Integrating these equations means that dose can be plotted against nongenetic material bound Co- valently with reactive metabolite, genetic materia- bound covalently with reactive metabolite and then rc- paircil, and genetic material hound covalently with re- active metabolite but not repaired. Figure 5 shows the results. Normalised quantity in10? 10? 10" 10" to" 10* 10? Dose From these curves it is clear that there is a dispropor- tionate increase in the amount of reactive metabolite bound to genetic material which escapes repair (curve as dose increases. The formation of such material is prob- ably associated with chemical carcinogenesis. The highly conservative model demonstrates why excessive doses of some chemicals cause discernible increases in cancer. while smaller doses may have no effect. It illustrates clearly that data generated from studies of animals given large doses cannot be extrapolated to predict the incidence of cancer caused by exposure to smaller amounts of the same chemi- cal. Such extrapolation assumes that the fate of the chemi- cal does not change with dose?an assumption which re- quires the curve of unrepaired "damaged" genetic material to be a straight line parallel to the abscissa. Clearly it is not. The model does not establish that exposure to chemicals found to be carcinogenic to animals at high doses is abso- lutely safe. ?Absolute safety"-is not only unrealistic but ludicrous. For example, ?overnutrition? causes more human cancer than any other identi?able cause. Extra- polating those data, and introducing preventive measures consistent with absolute safety, would certainly reduce drastically the incidence of cancer?starvation would soon predominate! I do not propose that we should relax our efforts to assess the hazards of chemicals or to control exposure to them. However, the hunt for causes of human cancer must be seen in perspective. No assessment of the safety of chemicals can guarantee their absolute safety. But an approach based on the concepts of the model I have out- lined will increase the reliability of assessment. It will not guarantee that some uniquely sensitive person {who may also be exposed to small amounts of other chemicals) will not develop cancer. It is high time that those scientists dedicated to ing the potential hazards of chemicals rededicate them- selves to being a part of the solution. Hopefully. their theories of carcinogenesis will come to terms with the fact that only a small amount of human cancer is caused by chemicals. The effect of impetuous, indeed frankly dis- honest. attempts to link cancer with chemicals will, if un- checked, become a nemesis rather than a saviour of our society. El . of the ingested shot . age and sex of the hitd. position and volume of the umed 1 orth America several efforts have een made to develop a nontoxic shot. deadly. three approaches have been l) developing a disintegrahle lead shut which would fragment in water and thus become unavailable. 2) coating the pellets to prevent the absorption of lead. and 3) replacing the lead with a less toxic metal or alloy (20). A soft steel shot has recently been pro- posed as a substitute for lead shot in the United States (II). It has been found to have some disadvantages. however. cg some gun damage and its low effec- tiveness in bagging ducks has caused con? troversy (23). A solution to the problem is . perhaps the Canadian lead-iron pellets. ln- itial toxicity tests with 50 percent lead- iron pellets have shown a reduced mortal- ity in mallards compared to commercial lead pellets at the same level (22). A pos- sible reason for the difference is the antagonistic effect of the metals in the lead-iron pellets Relerences and Notes: I. Locke. Baglcy. Journal of ll'ildlt?' Managerial-at 253 . Bellrose. Bulletin Natural History Surrey 27. . l. Westemeier. lt'ilrmt Built-tin 78. 471 (I966). . Campbell. Journal ofll?t'ldlt'ji- illunagt-nmtt I4. 243 (I950). . Hunter. Rosco. ("ulilimtia Fish and Game SI. 20? {1965). QC Jones. Journal nfll't'ldl?i?- .?llanagentent 3. 353 lulu-l Id '15 939). Locke. Bagley. [but .21. SIS 8. Anderson. A Skaptason. Fahcy. Cl Henny. 5 Bureau Fisheries and ll?ild- life. Resource Publication 119. 30 ?974}. 9. Lavery. Emu L18 l0. (3 Del Bono. Braca. .?trt?an Pathology 2. I95 ?973). ll. Hoffmann. Torre at We 2. U960). l2. Olney. Bulletin British (Jmitltulngirts' Club 80. 35 (1960). ES. Muller. Dartsk "I'll?1973. 4 ?973). H. Erne. Borg. Ecological Research Bulletin 5. 3i ?969}. IS. 5 Carney. US Fish and tl?ildlifr Sen-ire. Spe- cial Scienti?c Report. Wildlife 32. I964). If). Pirkola. Simmer: Krista 20. 35 (I968) (for mallard] and preliminary key tt'or teal}. 17. A Joensen has kindly demonstrated the refer- ence collection and age and sex criteria used at Game Biology Station. Kalo. Ronde. Denmark. See also A Danish rthame Biology 9. No. (I974). 13. The con?dence limits (9551) from the binomial distribution table in Diem and lids. Scl- entt?c Tables Geigy A G. Basic. pp l9. Trot-martian: of tht? First Inn-ma- tt'nnal ?Symposium. Smuf?'r?s Riverfront Town-L Si Louis. Missouri. February 4-6. 1975 (International Symposium Head- quarters. Box 66 300. Chicago. Illinois 60 666. USA) I63. 20. Locke. ltby. Bagley. Bulletin Wildlife Disease Association J. I43 21. .IPRogers. ?pf" 19. no. 22. Ferret. Association of Game. Fish and Conservation Carnatisrium'rs 63. 87 ?911). I. Kozicky. Up I it t9. 163. ?e acknowledge all the hunters who aided in the collection of the birds. and Mr F. Andersson for \aluahle technical assistance. The study was sponsored by the Research (?omntilte of the tionnl Swedish Protection Board and by the Suetlish Asmciation. 25. Received January 7. AMINO. I977 N-Nitroecdiothanolamine in a Grindino Fluid Concentrate After Storage mtrr?.? In'wand Mom'- By P-A Zingmark and Rappe, Department of Organic Chemistry. University at Umea, 8-901 87 Umea, Sweden (Present address for P-A Zingmark: The Labor Inspectorate, Magasinsgatan 1, 8-902 47 Umea, Sweden) The formation of N-nitrosamines from secondary and tertiary amines with nitrite in acidic solution is well known. The acid catalyzes the reaction by forming nitrous acid, which then participates in the N-nitrosation. The reaction should proceed very slowly under alkaline conditions because of the low concentration of nitrous acid. Neverthe-_ less, analyses of grinding fluid concentrates containing . . triethanolamine, nitrite and water show that small amounts of N-nitrosodiethanolamine are formed during storage even though the pH values of the fluids are between 10.2 and 11.4. The fluids we tested had been stored for five to seven months. and shat- ed a N~nittosamine concentration ranging from 400 to 800 ppm. These analyses clearly show that the mixing of nitrite and an amine in a technical product should always be regarded with some suspicion as to N-nittosamine formation, especially if the product is stored. N-nitrosodiethanolamine has been shown to induce cancer in rats, even though high doses are required. N?Nitrosamines are believed to be an im- portant group of human carcinogens (sec. for instance. reference I). They can easily be formed from secondary and tertiary amines and nitrite in a reaction catalyzed by acid. The acid serves to generate ni- trous acid. which then gives dini- trogentrioxide (N203). which is consi- dered to be the nitrosating species. at least under mildly acidic conditions {see refer- ence 2). Nitrous acid is a weak acid (pKa 3.4) and in a water solution there is an equilibrium between nitrite ion and ni- trous acid. This means that in an aqueous solution of sodium nitrite there is always a small fraction of nitrous acid present; the more basic the solution. the smaller the fraction. Consequently. the nitrosation proceeds best at acidic pH. lt'the reaction time is long enough. however. the forma- tion of small amounts of N-nitrosamines should be possible even at basic val- ues. N-nitrosations have been performed experimentally in basic solutions. these were catalyzed by formaldehyde. which completely changes the mechanism of the reaction (3). We have recently shown (4) that N-nitt'osodiethanolamine is readily formed in acidic solutions from the com- ponents of a grinding fluid concentrate used in the metal manufacturing industry. The concentrate contains only tricthanolamine. sodium nitrite. water and some coloring matter and is used as a 2 percent solution in water. The triethanolamine is of a technical quality, containing 85 percent trielhanolamine and l5 percent of a mixture of di- and monoethanolamine. We have analyzed the grinding fluids used in our previous work (-1) for N-nitrosodiethanolantine after storage for 5?7 months at room temperature in glass or plastic bottles. These fluids differed only in the amount oftriethanolamine. We have also analyzed a grinding fluid con- centrate made from 984th percent lriethanolamine stored for 14 days. and a stored aqueous mixture of tri- ethanolamine and sodium nitrite. The latter llttid has the same composition as one of the commercial products. but was prepared from analytically pure trielhanolatninc (pm 237 1.1 ed in 1?0 nil water in one-necked ?ask. Urea (Merck's was added in a ten-l'olil excess river the nitrite and the ?ask was cooled in an iccwatcr bath: The was then adjusted to It). rising ll('l. When gas evolution had ceased. the mixture was kept at room temperature for it) minutes and then neutralized 7) with a 30 per~ cent potassium hydroxide solution. Dur- ing the whole experiment the mixture was stirred with a magnetic bar. The pH was measured with a Metrohm EA 125 corri- bined glass electrode connected to a Radiometer pli meter 25. 'l'hereafter. tilt) pl of the neutral water solution was mixed with 1.4 ml of buffer pH 7 (Merck?s l?uf? ferlbsung. 0.03 phosphate buffer) in a l?cm quartz cell and irradiated at room temperature for 30 minutes with a Blak- Ray B-lOt) A UV lamp (Ultra-Violet Pro- ducts. Inc). The distance from the from glass of the lamp to the quartz cell was approximately 10 cm. 0.5 ml of GrieSs' reagent was added to the solution and af- ter IS minutes the absorbance was meas- ured at 520 nanometers. using a Cecil CE 303 grating spectrophotometer. As a ref? erence. we used 1th ,ul of the non- irradiated neutral water solution in L4 ml of buffer pH 7 mixed with 0.5 ml of Griess' reagent. The amount of nitrite formed by photolysing the nitrosarnine was given by a calibration curve. where the absorbances of pure sodium nitrite solutions of known concentrations were plotted against the concentration of nitrite. The reproducibility of the photo- lysis experiments was very good. The yield of the photolysis was determined by using a solution of pure N- nitrosodiethanolamine in pH 7 buffer. the concentration of which was determined ?rst by UV spectrometry and then by ir- radiation for 30 minutes as described above. The yield was 47 percent. We have previously shown (4) that the nitrosamine formed in these grinding ?uids is N?nitrosodiethanolamine. Blank tests of freshly prepared grinding fluids as described earlier (4) reVealed no nitrosodiethanolamine. but the same blank tests with the stored ?uids gave gas chromatograms which indicate the pres- ence of this nitrosamine. in order to ex- clude the possibility of nitrosamine forma- tion during the acidification of the grinding ?uid?urea solution we did the following experiment: ?Technical" triethanolamiae v.23 prepared by mixing tri? and diethanolamine in proportion 85:]5 From this amine mixture we prepared aqueous solutions with the same compositions as the commercial products. Analyses of these mixtures for nitrosamine directly after their prepara- tion showed that no nitrosamine had been formed. In a series of blank tests. the compo- nents of the grinding ?uid were separately in exactly the same way at the grinding ?uid itself. Tests were thus performed with sodium nitrite. mono-. di-. and triethanolamine. respectively in con- centrations Corresponding to those in the grinding ?uid concentrate. No colored IIK {lsil v\ l\.Il 1 RESULTS AND DISCUSSION in the first step of our analyses. excess nitrite is destroyed by with urea at pH l. The reaction is Very rapid and alter 4 minutes no nitrite can be detected with nitrite test strips (Merck's Mcrcico- 'l'o expel the gaseous reaction products (carbon dioxide and nitrogen). the solution is stirred for H) minutes at room temperature. The second step is based on the well-known photolysis of N-nitrosamines. in which the nitroso group is split off using long-wave UV ir- radiation 340 nm) and is then quanti?ed as nitrite ion after forming a Colored diazo compound with the (iriess? reagent. This method has been described by Daiber and l-?rcussman (Ti) and Sander (6) and has been modi?ed for automatic determination of nitrosamines by Fan and 'I'annenbaum (7). When a solution of sodium nitrite is acidified. the resulting nitrous acid might in part disproportionate into nitrate ion and nitric oxide. If the spectrum of the UV lamp contains short wave UV 280 i am). the nitrate ion can be reduced tor nitrite ion. leading to erroneous results. Therefore. the nitrite blank test is essen- tial in assuring that experimental condi- tions are such that the nitrite formed dur- ing irradiation is really generated from nitrosamine(s) in the solution and not from other sources. By varying the irradi- ation time we found maximum yield after 30 minutes. The results of our analyses are given in Table 1. Hold A contains 20 percent triethanolamine and fluid 8 l2 percent. the sodium nitrite concentration being the same (20 percent) in the two products. Our own mixture (D) has the same com- position as flttid except for the coloring matter. which we haVe excluded. The triethanolamine used in (D) is a pro analyst quality (Fisher Certi?ed Reagent). The product was prepared in order to find out if nitrosarnine formation is due only to the impurities of dicthanolamine in the technical products as diethanolamine is more readily nitrosated than triethanolamine. As can be seen from Table I. this was not entirely the case. Our results clearly show that the stor- age of products containing amines and nitrite can lead to the formation of small amounts of N-nitrosamines even under alkaline conditions. A storage period of 5?7 months is quite normal between the mixing of the grinding ?uid concentrate and its use. Furthermore. the diluted grinding ?uid is circulated for a long time. up to one year. During this time the nitro- sation reaction cart still proceed. retarded in part due to dilution but also accelerated due to a more favorable (pH of a 2 percent solution in water i~ round 9). The net effect of these tnutually opposing factors is not knoun. People working close to a machine here this type of pro- duct is being list: .Eght inhale and partly swallow the mist it is formed around the machine. This ltt:\l may contain not only the prerequisites tor nitrosainine sis in the stomach but also the preformed these products. Grinding Storage tluid period concentration concentrate pH (months) (ppm) it 4 5 400 A 11.4 7 800 10 6 7 600 10.8 0.5 50 10.2 6 200 1) The same composmon as A but prepared from high purity technical tnethanotamine (93?100 percent). 2) Prepared from triethanotamine pro analysi. nitrosamine itself. The maximum yield of N-nitrosodiethanolamine -in the tested products is 0.5 percent of the theoretical. which certainly is a low yield. Neverthe- less. the formation of small amounts of carcinogenic substances should never be neglected. N-nitrosodiethanolaminc is a weak liver carcinogen in the rat (8). Converting to triethanolamine of high purity will diminish the problem but can .-not eliminate it. as can be seen from Table l. The only sure way to elimin_a_te N-nttrosamtne Tormauon dunno storaoe must be to eliminate sodium nitrite from References and Notes: I. Arabia 5- 67 (I976). 2. A Turney. A Wright. Chemical Reviews 59. 497 3. Keefer. Roller. Sr?r?mre WI. 1245 (1973). 4. P-A Zingmark. Rappe. (?it I. 80. 5. Daiber. P'eussman. Zritsr'hri? ?ir- rmalytisrhe Chrmit- -06. 344 [1964). 6. .l Sander. HoppesSry-lrr's. frir physiologist'he Chi-mic 343. 8?2 (1967). 7. T-Y Fan. 5 Tannenbaum. Journal of Agricultural and Food Chemistry 19. 1267 8. Druckrey. Preussmann. lvankovic. fur Krebs?arsrlumg 69. ID) ?967). 9. A grant from the Swedish Work Environment Fund (No. 75M) is gratefully acknowledged. 10. Received August 37. 1976. 11. Note added in proof. While this report was in press another report with similar results has been submitted to and published in Science 196. 70 (I977) by Fan et at.