Acceptable Levels of Risk and Decision-making Rodricks Some Attributes of Risk Influencing Decision-making by Public Health and Regulatory Officials

Transcription

1 American Journal of Epidemiology Copyright 2001 by the Johns Hopkins University Bloomberg School of Public Health All rights reserved Vol. 154, No. 12, Supplement Printed in U.S.A. Acceptable Levels of Risk and Decision-making Rodricks Some Attributes of Risk Influencing Decision-making by Public Health and Regulatory Officials Joseph V. Rodricks Epidemiologists are concerned with identifying and measuring health risks associated with environmental exposures of many types. Public health and regulatory policy-making relies upon such risk information, together with data from experimental studies, but also requires incorporation of decisions regarding the types and levels of risk that are acceptable in different circumstances. The topic of risk acceptability is a large, value-laden one, with more dimensions than can be adequately explored in a single presentation. Among the many questions that might be addressed are the following: Who has the authority or responsibility for deciding what risks are acceptable? How small does a risk have to be before it is considered acceptable? Why should any risk be acceptable if it is not understood and freely accepted by those upon whom it is imposed? How can acceptable risk decisions be made with due regard to the scientific limitations in our understanding of the risks? How should priorities for risk reduction be established? The list of questions could be significantly extended, and ones not yet thought of will no doubt arise in the future as new technologies come into use and social priorities change. Any systematic examination of these and related questions should perhaps begin with a discussion of the various ethical models that bear upon them. Thus, as Douglas and Wildavsky (1) have pointed out, viewpoints regarding the question of who should be given authority to make acceptable risk decisions will vary depending upon whether they derive from so-called elitist ethical models (i.e., decisions based upon complex technical matters should rest primarily in the hands of those who understand them); egalitarian models (i.e., all affected individuals should participate in a Received for publication November 22, 2000, and accepted for publication June 11, Abbreviations: EPA, US Environmental Protection Agency; FDA, US Food and Drug Administration; NRC, National Research Council; OSHA, Occupational Safety and Health Administration. From ENVIRON International Corporation, Arlington, VA.. Reprint requests to Dr. Joseph Rodricks, ENVIRON International Corporation, 4350 North Fairfax Drive, Arlington, VA ( jrodricks@environcorp.com). form of communal decision-making, because technical matters are never the whole story and may not even be important); or libertarian models (i.e., each individual should be given the appropriate information regarding risk, and then allowed to make a free choice). Further discussion of the bases for these three ethical models, and their full implications for the various questions regarding risk acceptability, is not within the scope of this presentation. The models are mentioned, however, as a reminder that risk acceptability questions cannot be fully and honestly discussed without the recognition that the answers we find for them will to a large extent rest upon the explicitly or implicitly held value systems from which they are approached (2). With respect to human health and safety risks arising from the environment (defined broadly here to include every type of activity, agent, or influence that is not innate or genetic), decision-makers from many institutions have been supplying practical answers to various risk acceptability questions for several decades now. This is exemplified in governmental decision-making based upon the many federal and state laws enacted since about mid-century that presumably reflect underlying social values. Of course, informal decision-making regarding risk acceptability has been going on among humans since we first walked the planet, and it is no great revelation that each of us in our daily lives make hundreds of risk acceptance decisions, most of them at a sub-conscious level. This presentation focuses specifically on the role of risk acceptability in decision-making concerning human exposures to hazardous chemicals. Most such decision-making has been undertaken in a regulatory or official public health context. The general lessons learned and some of the reasons many of the decisions provoke intense social conflicts are discussed. EARLY HISTORICAL PERSPECTIVE: THE 1940S TO THE MID 1970S Among the earliest efforts to promote the development of what were called safe limits on exposures to chemicals is one that arose in the 1940s from a voluntary action on the part of the American Conference of Governmental Industrial Hygienists who defined what came to be called Threshold Limit Values. These workplace exposure limits were seen at the time as based purely on scientific analysis. Similarly, early efforts to establish so-called Acceptable Daily Intakes for additives and pesticide residues in food were S7

2 S8 Rodricks considered to be objective, science-based activities. These early efforts were in the hands of scientists who understood the relative value of epidemiology and experimental data, and the problems of inter- and intraspecies variability. They extrapolated from the available data using safety factors (now called uncertainty factors ) to compensate for data deficiencies and to account for expected variabilities in toxic response (3). Early attempts to deal with chemical carcinogens were varied and unsystematic. The Food Additives Amendment of 1958, an amendment to the Federal Food, Drug and Cosmetic Act of 1938, included the Delaney Clause, a provision that prohibited the deliberate addition to food of any substance shown to be carcinogenic in humans or animals. The clause reflected Congressional reaction to testimony from government experts and policy-makers asserting that no safe level of exposure could be identified for carcinogens. During the 1950s and 1960s, however, efforts to deal with the increasing numbers of carcinogens that were being found in various environmental media were begun, but little consistency can be found in these early efforts. Banning a deliberately introduced food additive was, for example, relatively easy. On the other hand, deciding how to manage risks from carcinogens of natural or industrial origin that purportedly had no safe level of exposure, but that were contaminating water, air, or the food supply, proved to be so difficult that decision-making was often simply avoided. In some instances public health authorities turned to the analytical chemists for help. Thus, if a carcinogen could be detected in an environmental medium, it was unacceptable; if it could not be detected, no problem was thought to exist. This approach was not particularly helpful because it depended upon the implicit but unsupportable assumption that the dangers associated with carcinogens were somehow related to the detection limits of the analytical methods available to measure them (3). The first systematic methods to deal with carcinogens were introduced in the early 1970s by regulators, the US Food and Drug Administration (FDA) and the US Environmental Protection Agency (EPA). These agencies introduced the scientifically-based notion that low-dose carcinogenic risks could be estimated based upon extrapolation from high-dose, high-risk measurements, often made in experimental animals. These methodologies accounted for the no safe level orthodoxy by the inclusion of nothreshold models for the extrapolation to low dose. The concept of no safe level then began to be viewed in terms of risk; that is, carcinogens were assumed to pose non-zero risks at all non-zero exposures, and risks were assumed to increase as a linear function of dose. This recasting of the no safe level concept forced explicit recognition that no absolutely safe exposure level could be identified, but also led to the view that human health could be protected if estimated cancer risks were driven to some very low level. In their proposals and regulations, both the FDA and EPA recognized that decisions about insignificant risks (those that were in some sense acceptable ) were not scientific in nature, but involved value judgments (4). THE RED BOOK AND ITS CONSEQUENCES: 1983 TO THE PRESENT The now famous red-covered National Research Council (NRC) panel report, Risk Assessment in the Federal Government: Managing the Process (3) reviewed and evaluated the early historical developments and urged some reforms. It called for what we now commonly recognize as the fundamental distinctions and interrelations among research (the gathering of risk-related data and the accumulation of relevant knowledge), risk assessment (the evaluation of risk-related knowledge and information to characterize risk associated with specified circumstances), and risk management (the process of making decisions regarding the need for and extent of risk reduction, and the means for achieving and maintaining it). Application of this decision-making model has been the stated objective of regulatory and public health institutions since it was first proposed in 1983, and it is likely to remain so. It is by no means a perfect model, although many of its imputed limits reflect misrepresentations of some of the NRC panel s reasoning and conclusions. For example, it has been argued by many that risk assessment is itself valueladen, and is a potentially dangerous tool because these values are often disguised in scientific terms (5). This danger was recognized in the 1983 report as well as in an updated and substantially expanded report published by another NRC panel in 1994 (6). Both reports pointed out that few risk assessments can be completed without resorting to assumptions that are not completely supported by available scientific evidence. The NRC has urged the adoption of what have come to be termed default options to deal with such scientific uncertainties. It has proposed that such options be specified in the form of generally applicable guidelines for risk assessment, and that the same options be used in all risk assessments to avoid case-by-case manipulations of assumptions to achieve some premeditated outcome. Choices of such defaults involve what the NRC referred to as science policy in risk assessment (3, 6). The EPA has spelled out its preferred default options in such general guidelines for risk assessment and has also introduced the additional NRC proposal that the defaults be flexibly applied (7). This flexibility means that if data specific to a given risk situation are explicitly shown to be adequate, they may substitute for one or more of the generic default options. The default approach recognizes that science is incomplete and that practical decision-making requires the explicit imposition of science policy. While science policy cannot be avoided in the risk assessment process, it can at least be specified in advance and applied consistently. This approach to risk assessment has been challenged, however, on the grounds that the choice of defaults reflects the bias of regulators and public health officials toward excessive caution in the face of scientific uncertainty, and, therefore, makes it no different from the policy choices typically made in the risk management process (5). This debate remains unresolved. Although not discussed widely, all three components of the decision-making process (i.e., research, risk assessment, and risk management) actually entail both scientific issues

3 Acceptable Levels of Risk and Decision-making S9 TABLE 1. Comparison of types of questions addressed by science and policy in protecting people from risks of environmental origin Research Activity Scientific questions How should knowledge be acquired? What has been learned, and how well has it been established? Policy questions What knowledge should be acquired and who should pay? Risk assessment Risk management and policy considerations, as illustrated in table 1. Even within the world of research, for example, questions of how to acquire knowledge are different in kind from questions of what knowledge should be sought and of who should pay for the seeking. It is always essential to distinguish science (i.e., what is the world like?) from policy (i.e., how would we like it to be?) in all elements of decision-making. The two are inextricably linked, but understanding and accommodating their differences is necessary to achieve clarity regarding the appropriate decision-making models, institutions, and individuals. THE DECISION-MAKERS AND THEIR CHALLENGES Before discussing risk acceptability decisions, it is helpful to identify those who are responsible for such decisions. Four classes of decision-makers and the criteria they may apply are listed in table 2. Decision-makers include regulatory officials at agencies such as the EPA and FDA, public health officials at, for example, the Centers for Disease Control and Prevention and National Institutes of Health, and even corporate executives concerned with product liabilities (8). There are significant differences in how public health and regulatory institutions approach risk acceptability decisions (table 2). As used in this table, the term environmentalregulatory officials refers to those who have the authority to enforce laws under which various classes of environmental agents are regulated. In this context, environmental is used in the narrow sense of contaminants in the environmental media of food, air, water, and soil. In addition, many regulatory officials may view themselves as serving a public health function, but within the context of table 3, public health officials refers to individuals responsible for advising the community at large about threats to their health and safety, and encouraging the use of various means to reduce those threats. The public health officials approach typically involves the use of warnings, educational materials, and various How should acquired knowledge be interpreted to understand risks? What don t we know? What means are available to reduce risks? How effective? What costs? Will they introduce other risks? What risks should be assessed? What should be done when knowledge is lacking? How much risk reduction is necessary? What should influence the decision? What means for reducing risk should be adopted? incentives aimed at inducing people to change behaviors to quit smoking, to drink in moderation, to consume less fat and more fiber, to exercise, and to wear seat belts. Environmental-regulatory officials generally write enforceable rules that typically involve setting standards (i.e., limits on exposure to chemicals through various media). Given the respective missions of public health and environmental-regulatory officials, it is clear that attributes of the risks they are concerned with differ significantly, as shown in table 3. The concerns of the environmentalregulatory official are usually directed at poorly understood risks that are often small for most individuals, and usually beyond the detection power of epidemiologic studies. These small risks, however, may affect very large numbers of people who are often unaware that the risks exist. The public health official is usually concerned with large risks that are generally understood by affected individuals, and that are often subject to their personal control. In many cases the activities leading to these large risks are perceived by individuals as conferring positive, offsetting personal benefits. TABLE 2. Decision-makers and their criteria for judging acceptability Individuals Decision-maker Public health officials Environmental-regulatory officials Corporate officials Criteria used to judge acceptability Perceived personal benefits large relative to preceived risk Sufficient to improve public health (risks often within individual control) Sufficient to enforce legal requirements (risks usually not within individual control) Sufficient to avoid liabilities

4 S10 Rodricks TABLE 3. Some general attributes of risk influencing approach to risk management Risk attribute Degree of scientific certainty Understanding of risk by affected individuals Magnitude of risk to individual Risk within control of individual Accompanied by off-setting benefits Public health officials Usually high Often good Often large Usually Often perceived as large to affected individuals Environmental-regulatory officials Low Often poor Often small Usually not Usually small to affected individuals Thus, the missions of public health and environmentalregulatory officials are hindered in quite different ways. The environmental-regulatory official is faced with the dual challenges of convincing the public that small, poorly understood risks truly exist and are worth controlling. At the same time, they must convince that same public that it should be willing to accept some very small level of risk from activities or exposures from which it perceives no significant benefit. The public health official is faced with the difficult problem of convincing members of the public to alter behaviors they perceive as personally beneficial, even when the risks to which they are subject are reasonably well understood as significant. These challenges are, of course, further compounded by the resistance of those corporations that may be required to bear the costs of regulatory and public health initiatives. It is also apparent that there is sometimes conflict between public health and environmental-regulatory goals. The public health community often argues that enormous public and private resources are being devoted to manage the relatively insignificant risks that are the subject of many environmental regulations. In contrast, the environmentalregulatory community suspects that the traditional public health community is simply unaware of the magnitude of the threat to human health associated with the cumulative effect of large numbers of small individual risks, and is unwilling to take that threat seriously. Although seemingly operating in different spheres, public health officials and environmental regulators have in common some of the thought processes that guide the evaluation of evidence and the risk assessments that underlie their decisionmaking. The following definition of risk assessment is one that is sufficiently broad to encompass the activities and objectives of both communities. Recognizing this common definition may serve to bring the two communities closer together. Risk assessment is the analytical framework used to organize, evaluate, and characterize available knowledge and its associated uncertainties regarding the nature and magnitude of threats to human health arising from the environment, including both the natural world and every type of human influence on it. The results of risk assessments are used to guide policy decisions regarding the need to take actions to control or eliminate these threats so that human health is adequately protected. They are also used to identify the research needed to reduce uncertainties and, thereby, to improve understanding of these threats. Risk assessment is thus the instrument used to measure progress in understanding and managing every type of environmental threat to human health. RISK MANAGEMENT AND RISK ACCEPTABILITY Risk managers need to answer three basic questions: Which risks should be reduced?, and if so, By how much? and By what means? Except for the relatively trivial case in which complete prohibitions on exposure (i.e., complete ban on production and use of a substance) can be realized, the answer to the second of these three questions requires a decision about acceptable risks. If we examine the history of risk-based decision-making, several trends emerge. Different laws require risk managers to take into account different factors in their decision-making, such as considerations of technical feasibility or lost benefits. All environmental-regulatory laws require public health protection, but there is no single health protection standard. There has emerged from federal environmental-regulatory activities over the past two decades a body of risk decisions for chemical carcinogens. These decisions have been categorized into three action levels based on the magnitude of lifetime risk, as summarized in table 4: greater than one in 10,000 (10 4 ) (always regulated), >10 6 to 10 4 (accepted in many decisions), and <10 6 (generally considered negligible) (6). The table refers to general population risks rather than occupational risks. No single standard emerges in the several hundred specific decisions upon which the general categories of risk levels are based, although decisions made under statutes that include technical feasibility criteria tend to lead to acceptable risks at the higher end of the 10 4 to 10 6 range, and those made under statutes that are purely risk-based (e.g., federal laws governing pesticides added to food or indirect food additives) tend to fall closer to the lower end. What is interesting about these acceptable risk levels is that they are two to three orders of magnitude below the detection power of the most sensitive epidemiologic methods now available (6). The levels are mainly based upon extrapolations from high-dose animal studies and only occasionally upon epidemiologic data, because the latter usually fail to provide adequately quantitative dose-response information. Another matter of some interest is the lack of significant public health analysis supporting these risk acceptance goals. Perhaps they make sense, but it would seem useful to evaluate these environmental-regulatory decisions in some larger context, such as their actual value for reducing the overall cancer burden and their benefits relative to other types of public health approaches (1, 9). Occupational carcinogens have generally not been as stringently regulated. Although the Occupational Safety and Health Administration (OSHA) has promulgated fewer than

5 Acceptable Levels of Risk and Decision-making S11 TABLE 4. Historic precedents for risk decisions involving general population exposures to chemical carcinogens Excess lifetime risk* >10-4 >10-6 to 10-4 <10-6 Decision Always regulated Accepted in many decisions Generally considered negligible EXCEPT where the zero-risk standard of the Delaney Clause applies * Generally based on extrapolations from high-dose animal data using linear, no-threshold models (see (6)). two dozen decisions, most involve reductions to lifetime risks in the range of 10 3 to OSHA has used several types of arguments to defend these levels, but a discussion of these arguments is well beyond the scope of this presentation. The outcomes of the decisions, however, are consistent with a long history of different approaches to occupational and general population risks (9). It is widely accepted among regulators that noncarcinogenic chemicals that cause other types of health damage are not harmful unless a toxic threshold dose is exceeded; hence, these chemicals have been called threshold agents (3, 6). The estimation of threshold doses is a risk assessment activity. The population-based estimates of thresholds cannot, however, be known to be protective for every individual in a large population since they carry some degree of uncertainty of unknown magnitude, and the fraction of the population that might not be protected at the estimated threshold cannot be determined. Risk management for threshold agents has generally depended upon the criterion that exposures to the agent of concern should be limited to ensure that they do not exceed the estimated threshold. How well this stated risk-management goal has been achieved in practice has not been well examined. No clear generalizations can be made about acceptable risks in this area until an effort is made to evaluate human exposures to such threshold agents from all of their various sources, and to identify the fraction of the population having exposures consistently in excess of the estimated thresholds (8). CONFLICT: PUBLIC MISTRUST OF EXPERTS There are few regulatory and public health actions that do not generate some amount of conflict between different interests. Often these conflicts are over matters of cost, technical feasibility, and social disruption. There is, however, one often overlooked source of conflict that directly bears on the risk assessment component of the decision process: the sometimes vast differences in perspective between experts and the lay public regarding the nature and magnitude of the risk that is under scrutiny. Several theories have arisen to explain the possible sources of the various conflicts (2, 10); among them is the theory that the conflicts are in fact not about risk, but, rather, over differences in ideologies. If the experts assessment of risk does not coincide with the expectations and hopes of an interested segment of the lay public for certain actions to be taken, the interested group protests and refuses to accept the experts views. Other theories turn on the notion that the conflicts arise from lack of trust in science, scientists, and the institutions that support them, to say nothing about public suspicions regarding the motives of corporations. Both of these theories have explanatory power, but a third theory seems at least as well supported an explanation for the attitudes of the lay public. This theory holds that the expert-lay conflict arises from what has been called differences in risk perception (2, 10). Substantial social research supports the proposition that most people think about threats to their health and safety in ways that are quite different from the experts. Experts typically emphasize quantitative aspects of a risk, such as probabilities and safety margins, while most people think about risk in broader ways. They tend to focus on other attributes of risk, such as the extent to which it is voluntarily assumed, whether or not it arrives as a catastrophic event, the degree to which it is perceived to offer personal benefits, and the fearsomeness of the hazard to which the risk applies. Selected attributes of risk that have been found to affect perceptions are listed in table 5. Not included in this list is the probability that harm will occur, which is the principal interest of the expert. Many people seem uninterested in probabilities as a guide to threats to their health (10). The terms tolerated risk and non-tolerated risk characterize peoples reactions to these various attributes of a risk (table 5). Attributes of tolerated risks are those which tend to be more readily accepted, whereas attributes of risks that are not tolerated tend to be the ones that cause fear and outrage. Peoples toleration of a risk, or lack thereof, is generally unrelated to its magnitude. These attitudes do not seem to stem from some deep-seated irrationality limited to just a few, but seem to reflect some fairly common elements of human psychology. Efforts to improve public communication of risk information and decision-making need to accommodate an understanding of these perception issues because they are the source of much conflict. Risk communication will surely fail if public perceptions are ignored (10). Many other sources of, and reasons for, conflict arise in the public health and regulatory processes, but the risk perception phenomenon has a different impact on the challenges faced by public health and environmental-regulatory officials. Generally, public health efforts heavily focus on risks having the attributes that make them tolerable to many TABLE 5. Factors influencing toleration of risk (2, 9, 10) Tolerated risks Voluntarily assumed High personal benefit perceived Scientists agree Not catastrophic Natural origin Hazard not fearsome Common event Equitably distributed Non-tolerated risks Imposed by others No perceived personal benefit Scientists disagree Catastrophic Industrial origin Highly dreaded hazard Rare event Distribution inequitable

6 S12 Rodricks individuals, while environmental-regulatory efforts tend to focus on those that are not well tolerated (table 5). There are clear exceptions to these generalizations, but the trend is certainly in this direction. CONCLUSION: THE CHALLENGES AHEAD Perhaps the principal challenge facing the public health and environmental-regulatory communities is the need to find a common ground of action to improve public health. Sciencebased risk assessment, conceived as the principal tool for measuring the status of the public s health and assessing the relative importance of the various environmental influences on public health, should be the ground from which such action arises. Many have lost faith in science as a basis for public health improvement because it fails to provide all the desired answers exactly when they are wanted. Science-based risk assessment nevertheless remains the best available method for understanding threats to human health and safety, and it is the least likely to lead to error and self-deception. Continued reliance upon risk assessment as the best available guide to managing risks to health and safety, however, requires improvements in its conduct and in the manner of its presentation to risk managers and the public. Assessments are needed that avoid both reductionist, default-driven approaches and vague, semi-subjective approaches. Future risk assessments will make life more difficult for risk managers. Decision-makers will be required to delve more deeply into the content of risk assessments and to understand the relative scientific merits of the various characterizations of risk that might be consistent with the available data and knowledge. They will no longer be able to leave to risk assessors decisions regarding which risk estimates should be ignored because they are insufficiently certain. The risk assessor s role is to describe clearly the uncertainties associated with various characterizations of risk, but it is not to decide which characterizations are to be used for decision-making (5, 8). The issue of acceptable risk has been subjected to little public health evaluation. Economists and social scientists have devoted much more time to the issue. It is a public health matter of some importance. Individuals face a relatively small number of large risks of environmental origin, and a very large number of small risks. These include not only risks of chemical toxicity, but all types of risks that threaten human health and safety. No coherent, large-scale analysis of this entire pattern of risks to health and safety, and of the most effective means to manage it, has appeared in the public health literature. Piecemeal decisions continue to be made in substantial isolation from one another without the benefit of an analytical framework developed from the public health perspective. A coordinated effort to seek a common approach to science-based risk assessments, and to develop cohesive, public health-based criteria for risk management based on the results of applying such assessments, could perhaps unify and strengthen the efforts of public health and regulatory officials to protect human health from environmental threats of every type. REFERENCES 1. Douglas M, Wildavsky A. Risk and culture. Berkeley, CA: University of California Press, Margolis H. Dealing with risk: why the public and the experts disagree on environmental issues. Chicago, IL: University of Chicago Press, National Research Council, Committee on the Institutional Means for Assessment of Risks to Public Health. Risk assessment in the federal government: managing the process. Washington, DC: National Academy Press, 1983: Lowrance WW. Of acceptable risk: science and the determination of safety. Los Altos, CA: W. Kaufmann, Finkel AM. Confronting uncertainty in risk management: a guide for decision-makers. Washington, DC: Resources for the Future, National Research Council, Committee on Risk Assessment of Hazardous Air Pollutants. Science and judgment in risk assessment. Washington, DC: National Academy Press, 1994: US Environmental Protection Agency. Proposed guidelines for carcinogen risk assessment. (Notice). Federal Register 1996; 61: Rodricks JV. Risk assessment, the environment, and public health. Environ Health Perspect 1994;102: Rodricks JV. Calculated risks: the toxicity and human health hazards of chemicals in our environment. Cambridge, England: Cambridge University Press, Slovic P. Perceptions of risk: reflections on the psychometric paradigm. In: Krimsky S, Golding D, eds. Social theories of risk. Westport, CT: Praeger, 1992: