Carcinogenic risk assessment in the United States and Great Britain: the case of Aldrin/Dieldrin.

The question is posed: why were two pesticides, Aldrin and Dieldrin, judged to be carcinogenic in the US but not in Britain when the same evidence was available to the public authorities in both countries? No single cause is identified; rather, a variety of mutually reinforcing factors account for the decisions by the two public authorities: the uncertainty of the scientific evidence; the application of different standards of carcinogenicity associated with different social and scientific commitments; the government agencies with primary responsibility for the decisions; the way in which pesticides are regulated; and several cultural and economic considerations. The case study illustrates the analytical inadequacy of the fact-value distinction, and the tendency of decision-makersto justify their decisions by -ecourse to science. It also supports the view that the traditional relationship between science and public policy is being redefined in complex, technical areas of decision-making like risk assessment. Carcinogenic Risk Assessment in the United States and Great Britain: The Case of AldrinIDieldrin Brendan Gillespie, Dave Eva and Ron Johnston The nature and scale of the hazards associated with new technologies have required governments to make difficult decisions about the form of control they should exercise over new products and processes. Although governments have received advice from a large and growing body of scientists, this has not always simplified the decision-making process nor rendered it more rational and objective. In this paper, we try to explain why two governments reached contradictory conclusions to the same problem on the basis of the same scientific evidence. More specifically, we will try to explain why two chemical pesticides, Aldrin and Dieldrin (A/D), were judged to be carcinogenic in the US but not in Britain when the same data were available to public authorities in both countries. Social Studies of Science (SAGE, London and Beverly Hills), Vol. 9 (1979),265301 266 Social Studies of Science The issue here is not one of differing assessments of risk or of balancing risks and benefits. Rather, what requires explanation is the different regulatory statuses assigned to the same products by the US and British governments. The observation that scientific experts disagree and oppose each others' views is hardly new -witness the appearance of a sizeable literature on the subject.' However, much of the literature incorporates weaknesses which inhibit our understanding of the role played by science in risk determination and, more generally, in controversies among experts in matters of public policy. Firstly, much of it focuses on one country, usually the US, whilst assuming that the roles of scientific advisers, and the problems which their use entails for governments, are everywhere the same. The comparative approach adopted in this paper should offset this, as well as a second, def i~iency:~ namely, the roles of science and scientists are generally analyzed in terms of universalistic categories which obscure the effects of different political and cultural settings on the type of science providing the basis for policy decisions. One frequently used variant holds that science provides the facts, and their evaluation from divergent value and ideological perspectives results in contrary interpretations. Conflicts among scientists are then explained in terms of the political views, bias or irrationality of one or more of the disputants, and little attention is paid to the content of the conflicting positions. This approach has merit to the extent that it accepts that scientists become integrally involved in political conflicts. Nevertheless, two weaknesses severely limit its explanatory power, and call into question the fact-value distinction on which it is based. Firstly, it does not provide an accurate description of actual controversies. Nowotny, for instance, has argued that the traditional division of labour between science and society is, in practice, breaking down in complex technical spheres of decision-making like risk as~essment.~ Secondly, the essentially positivist view of science, on which the distinction is based, is overly restrictive in limiting explanations of scientific controversies to factors 'extrinsic' to science: values, bias, and the like. When science is examined as a form of organized, intellectual production, a much more complex relationship between scientific concepts, theories and methodologies, on the one hand, and ideological and value commitments, on the other, emerges, which also allows explanations of controversies in terms of factors 'intrinsic' to scientific de~elopment .~ Gillespie, Eva and Johnston: Carcinogenicity in US and UK 267 Through a cross-national analysis of the contradictory A/D decision, we hope to set the pertinent features of the science-public policy relationship into sharper focus than before. We begin our analysis by looking at the differing interpretation of the evidence about A/D provided for the British and US authorities. Significant differences are found to exist in the criteria used to infer carcinogenicity, in the type of scientists providing the authoritative interpretations, and in the choices with which the respective public authorities were faced. This examination of the provision of scientific advice is complemented by an analysis of the factors affecting its reception. We identify those institutional, legal, cultural and economic features of the decision-making contexts which made them more or less receptive to the advice they were proffered. BRITISH AND US DECISIONS5 Aldrin and Dieldrin are two organochlorine pesticides that were widely used in agriculture in the 1960s. Workers applying A/D and consumers eating residues in and on treated crops were exposed to shortand long-term toxic hazards. (Aldrin rapidly degrades to Dieldrin in plants and animals, so when we discuss the hazards of A/D we are really talking about the hazards associated with Dieldrin.) The carcinogenic risk of A/D had been reviewed in a variety of national and international settings since the mid 1 9 5 0 ~ . ~ While several expressed uncertainty and requested more evidence, none of the reviews had been prepared to declare A/D carcinogenic. None, that is, until September 1974, when the US Environmental Protection Agency (EPA) found that A/D posed an unacceptable carcinogenic hazard.' Not surprisingly, the decision was viewed as controversial in the US, and was criticized in other countries. EPA had initiated administrative hearings in August 1973 to determine whether A/D's registrations should be cancelled that is, effectively banned.8 These hearings had been in progress for a year when EPA held more hearings to determine whether A/D's registrations should be suspended that is, their sale temporarily banned. Congress had provided EPA with these statutory alternatives to ensure that the agency could eliminate 'unreasonable' risks that might arise in the course of lengthy cancellation hearings. The two decisions are administratively distinct, and a decision to Social Studies of Science suspend does not automatically prejudge the cancellation decision. It was during the suspension hearing that the question of A/D's carcinogenicity came to the fore and, in fact, dominated the proceedings. An environmental group, the Environmental Defense Fund (EDF), had petitioned EPA for A/D's cancellation and suspension. Their case was supported, and effectively prosecuted, by EPA's Office of General Counsel (OGC). Shell, A/D's manufacturer, opposed these groups and was supported by the Department of Agriculture (USDA). The case was argued before Administrative Law Judge Perlman who was paid by EPA but independent of the agency for career advancement and t e n ~ r e . ~ He recommended that A/D be suspended because of the carcinogenic risk the chemicals posed, and this was endorsed by EPA's Administrator, Russell Train. When the decision was upheld in the Appeals Court, Shell withdrew from the cancellation hearings.1° The following month, in October 1974, a group of British experts reviewed EPA's decision. They concluded 'that there was no reason to recommend any change in the UK action on Aldrin and Dieldrin as decided at the time of the 1969 reviews of organochlorine pesticides'." No new evidence had been produced in the interim and the British experts had frequently discussed the evidence on A/D with Shell toxicologists over the years.12 The British experts were members of a Panel of the Advisory Committee on Pesticides and Other Toxic Chemicals. The Committee is an expert body that advises U K government departments on the hazards associated with pesticides. The departments then negotiate what action to take with manufacturers of the product in question. The negotiations are conducted within the framework of a voluntary (that is, non-statutory) agreement between government and industry the Pesticides Safety Precautions Scheme (PSPS).I3 It is quite clear, then, that the experts and decision-makers in PSPS and EPA reviewed the same experimental evidence of A/D's possible carcinogenicity and came to contradictory conclusions. Furthermore, although each group of experts and decision-makers was aware of A/D's status in the other decision-making forum, this did not change their conclusion. We therefore have a genuine paradox to explain. Gillespie, Eva and Johnston: Carcinogenicity in US and UK 269 LIMITATIONS OF TOXICOLOGY The first possible explanation of the different decisions is the uncertainty in determining carcinogenic risk. Toxicology, the field most directly concerned with the harmful effects of chemicals, has developed a fragmentary and incomplete understanding of the cancer-causing process.14 It is conceivable, then, that a number of competing and equally plausible interpretations of the same evidence could co-exist. The toxicologists' uncertain knowledge of cancer might make it difficult to choose between rival interpretations, or to completely rule one out. Although this argument could be applied to all chemicals tested for carcinogenicity, it seems particularly relevant in this case. Firstly, other chemicals have been shown to be more definitely carcinogenic than A/D.I5 Secondly, even those scientists who thought AID to be carcinogenic regarded them as considerably less potent than other known carcinogens.I6 Thirdly, Shell had spent $10 million producing evidence of the hazards associated with A/D, so there was a much better data base for these than for most other chemicals.'' Briefly, the evidence was of three types: 1) Epidemiology that is, correlations between incidence of human cancer and exposure to particular substances. Although ingestion of A/D-contaminated foodstuffs had resulted in British and US consumers storing Dieldrin in their body fat, there were no control populations which could be used to evaluate the effects of these exposures. The best available epidemiological data derived from the medical records of workers employed in the production of A/D.I8 About 1,000workers had been occupationally exposed, but only 69 of this pre-selected group had been exposed for 10 years or more. Since the latency period for cancer can be 20, 30 or even 40 years, it was difficult to know what significance to attribute to the two cases of cancer that had occurred in this group by 1974. Finally, Shell argued that indirect epidemiological evidence was available. Their scientists believed that humans metabolized A/D analogously to the drug phenobarbitone. They argued, therefore, that epidemiological data on phenobarbitone were relevant to the determination of A/D's carcinogenic hazard. 2) Animal models. These were experiments in which specific doses of A/D were administered to populations of laboratory animals, and the incidence of cancer tumours in exposed populations were compared with those in controls. The significance of Social Studies of Science results from these experiments for humans is highly controversial. By 1969, experiments with A/D in mice, rats, dogs and monkeys had been performed. 3) Biochemical tests. In recent years, a variety of so-called 'quick tests' have been developed to predict carcinogenic hazards more cheaply and quickly than animal tests.19 Shell scientists developed several in vitro tests which they used on A/D. Essentially, they tested for certain biochemical properties of chemicals that Shell claimed were correlated with carcinogenic activity. Clearly, there are general problems in determining carcinogenicity, and these were exacerbated by the particular biological properties of A/D. The uncertainty gave rise to the possibility of different interpretations and hence different decisions. Even so, uncertainty cannot explain why the British and US decisions went one way rather than the other; that is, it cannot explain why A/D were judged to be carcinogenic in the US but not in Britain, and vice versa. Since the uncertainty of the evidence was common to both EPA and PSPS, it cannot be a sufficient explanation of their difference. EVALUATION OF THE EVIDENCE The different interpretations of the A/D evidence could have arisen in at least two ways, depending on the degree of consensus that existed over the standards for inferring carcinogenic risk. If there were consensus over the standards, the explanation of the differing interpretations would be different than if there were competing standards of carcinogenicity. Thus, the way in which the two decisions were reached is important to our overall argument; clarification of just how the decisions differed must logically precede any account of why they differed. We will outline the arguments that Shell and EPA presented to Judge Perlman in the first part of this section, and the evaluation of the A/D data by British experts in the second. (A) US Evaluation At the suspension hearing, the Shell lawyers argued that a scientific approach requires all the evidence to be examined and that the results should be r ep rod~c ib l e .~~ In a similar vein, the Director of Shell's Tunstall Laboratories argued that carcinogenicity could be inferred only when five criteria were met: Gillespie, Eva and Johnston: Carcinogenicity in US and UK 27 1 1 . The exposed animals experience a higher incidence of tumors. 2. Tumors develop in more than one species. 3 . The development of these tumors can be proven to be compound-related. 4. The animals have proven to be an adequate model for extrapolating to man. 5. Human data is available proving at least one incidence of cancer." Shell conceded the first criterion and accepted that five experiments Davis and Fitzhugh (1962), Davis (unpublished), McDonald et al. (unpublished), Walker et al. (1973) and Thorpe and Walker (1973) -demonstrated that Dieldrin had increased the incidence of liver tumours in mice.22 Shell then argued that this finding could not be extrapolated to humans because A/D did not meet the other four criteria. Evidence was produced by several Shell witnesses that the mouse liver was not a valid predictor of human cancer (Criterion 4). They argued that Dieldrin had not induced tumours at sites other than the liver, and that the induction of hepatomas (liver tumours) in mice was highly dependent on a variety of genetic and environmental factors. More critically, Shell witnesses argued that the response of the mouse liver was unique and quite different from other species. Thus, they argued that evidence from rat, monkey and dog studies failed to indicate a carcinogenic response (Criterion 2). Next, they advanced a five stage process whereby liver tumours were formed, and argued that whilst all five stages occurred in mice, they did not occur in humans. This latter argument was quite subtle, and drew on evidence from a variety of sources. The indirect epidemiological evidence, the results from the in vitro biochemical experiments and the medical histories of occupationally-exposed workers, were all marshalled to support an argument for a mechanism which, Shell argued, showed how A/D did not induce human cancer (Criterion 3). Finally, Shell argued that the direct epidemiological evidence indicated that A/D did not meet their fifth criterion: the demonstration of one A/D-induced human cancer. The EPA case against A/D was quite different in form. Their lawyers argued that a number of principles had been wellestablished in the scientific community for assessing carcinogenic hazards, and that on this basis A/D must be considered carcinogen~.?~ Nine cancer principles were presented in the A/D case. The most relevant ones were as follows: 272 Social Studies of Science 1. A carcinogen is any agent which increases tumor induction in man or animals. 2. Well established criteria exist for distinguishing between benign and malignant tumors; however, even the induction of benign tumors is sufficient to characterize a chemical as a carcinogen. . . 7 . The concept of a 'threshold' exposure level for a carcinogenic agent has no practical significance because there is no valid method for establishing such a level. 8. A carcinogenic agent may be identified through analysis of tumor induction results with laboratory animals exposed to the agent, or on a post hoc basis by properly conducted epidemiological studies. 9. Any substance which produces tumors in animals must be considered a carcinogenic hazard to man if the results were achieved according to the established parameters of a valid carcinogenesis test.24 We may demonstrate that EPA was using different standards than Shell by considering the agency's response to Shell's five criteria. EPA's witnesses had re-examined the pathological and statistical analysis of the animal data, and it was this which helped elicit Shell's concession of their first criterion. With this established, EPA strongly objected to Shell's 'two species' criterion (Criterion 2). They argued that the negative evidence from the rat, dog, and monkey studies was the result of poorly designed experiments. Further, the uncertainties surrounding the use of animal models and the dictates of 'prudent policy,' meant that, for them, positive evidence should supercede negative results.25 In this case, however, evidence on the mouse derived from five different strains, and was therefore sufficient, in their opinion, to indict A/D as carcinogens. Although EPA could hardly disagree with Shell that the induction of tumours in test animals be 'compound related' (Criterion 3), there was scope for disagreement on the meaning of this term. EPA witnesses insisted that they were not required to produce causal mechanisms, as Shell had demanded. Indeed, Epstein argued that such a requirement 'would define away the entire field of chemical EPA's ninth cancer ca r c inogene~ i s . ' ~~ principle declared that a positive result from a 'valid carcinogenesis test' was sufficient to consider that chemical carcinogenic. Since A/D fulfilled this condition, EPA concluded that they were carcinogens. There were also objections to the mechanism Shell had presented. Farber argued that it is evident that many chemicals require metabolic conversion to active derivatives before they can initiate the development of cancer. However, the specifics of the metabolic processes which result in cancer in various test animals Gillespie, Eva and Johnston: Carcinogenicity in US and UK 273 are not clear, to say nothing of the metabolic processes in man. No one as yet can draw any valid correlation between a particular pattern of metabolism and the induction of cancer in any species, and any judgments concerning carcinogenicity or lack thereof based on metabolic patterns have no scientific basis at this