Current Approaches to Risk Assessment
The first part of the report examines the background and current practices of risk assessment consistent with the paradigm first codified in the 1983 NRC report Risk Assessment in the Federal Government: Managing the Process, often known as the Red Book (See Figure I-1). Chapter 2 of this report discusses the historical, social, and regulatory contexts of quantitative risk assessment. Chapters 3, 4, and 5 describe the Environmental Protection Agency's approach in applying the Red Book paradigm for risk assessment. As shown in Figure I-2, assessing human-health risks associated with a pollutant requires analysis of three elements: the source of the pollutant, the transport of the pollutant into the environment (air, water, land, and food), and the intake of the pollutant by people who might suffer adverse health effects either soon after exposure or later. Scientists and engineers take four basic interrelated steps to evaluate the potential health impact on people who are exposed to a hazardous air pollutant: emission characterization, exposure assessment, toxicity assessment, and risk characterization. In emission characterization, the chemical's identity and the magnitude of its emissions are determined. Exposure assessment includes how the pollutant moves from a source through the environment (transport) until it is converted to other substances (fate) or comes into contact with humans. In assessment of toxicity, the specific forms of toxicity that can be caused by the pollutant and the conditions under which these forms of toxicity might appear in exposed humans are evaluated. In risk characterization, the results of the analysis are described. These steps are described in detail in Chapters 3, 4, and 5.
The increase in the sophistication of the field of risk assessment since the Red Book requires risk assessors to have the ability to recognize and address fully such cross-cutting issues as uncertainty, variability, and aggregation, in
addition to having a more overarching view of the practice of risk assessment. Therefore, the committee supplements the Red Book paradigm with a second approachone that is less fragmented (and hence more holistic), less linear and more interactive, and, most important, one organized not according to discipline or function, but according to the recurring conceptual issues that cut across all the stages of risk assessment. These cross-cutting issues are described in Part II of this report.
Risk Assessment and Its Social and Regulatory Contexts
This chapter provides an overview of the origins and uses of quantitative risk assessment and the problems associated with it. Historical perspective is offered to aid understanding of how a method infused with so much uncertainty has still come to be seen by many as useful. Some attention is devoted to the important questions of how risk assessment has been used in decision-making and whether its use has improved decisions. The issues of public acceptance of the method and the degree to which decisions based on it are seen to provide adequate protection of the public health are also addressed. This chapter lists the major criticisms of risk assessment and the ways in which its results have been used, thus providing the justification for the selection of issues discussed in the succeeding chapters.
This section briefly discusses some basic definitions and concepts concerning human-health risk assessment, its content, and its relationships to research and to decision-making. The definitions and concepts were first systematically formulated by a National Research Council committee in a report issued in 1983, Risk Assessment in the Federal Government: Managing the Process. The Red Book had a major influence on the practice of risk assessment and will be discussed extensively in this section of the report.
What is Risk Assessment?
Human-health risk assessment entails the evaluation of scientific information on the hazardous properties of environmental agents and on the extent of
human exposure to those agents. The product of the evaluation is a statement regarding the probability that populations so exposed will be harmed, and to what degree. The probability may be expressed quantitatively or in relatively qualitative ways. There are other types of risk assessment that use similar processes but are outside the scope of this report, e.g., the risk assessment of the relative safety of a bridge.
Chemical hazards come in many forms. Some substances are radioactive, some explosive, some highly flammable. The particular hazard of concern here is chemical toxicity, including but not limited to carcinogenicity. Risk assessments can be carried out for any form of chemical toxicity. Risk assessment can be qualitative or quantitative. Many of the issues covered in this report concern quantitative expressions of risk.
How Is Risk Assessment Conducted?
The 1983 NRC report described a four-step analytic process for human-health risk assessment. A substance leaves a source (e.g., an industrial facility), moves through an environmental medium (e.g., the air), and results in an exposure (people breathe the air containing the chemical). The exposure creates a dose in the exposed people (the amount of the chemical entering the body, which may be expressed in any of several ways), and the magnitude, duration, and timing of the dose determine the extent to which the toxic properties of the chemical are realized in exposed people (the risk). This model is captured in the following analytic steps:
Not every risk assessment encompasses all four steps. Risk assessment sometimes consists only of a hazard assessment designed to evaluate the potential of a substance to cause human health effects. Regulators sometimes take the additional step of ranking the potency of a number of chemicalswhat is known as hazard ranking. Sometimes potency information is combined with exposure data to produce a risk ranking. These techniques all use some, but not all, of the four steps of the quantitative risk-assessment process.
Much of this report is devoted to the technical contents of the four steps of the process, because therein lie the issues that affect the reliability, utility, and credibility of risk-assessment outcomes. One important feature of those steps, however, needs to be emphasized here.
The 1983 NRC committee recognized that completion of the four steps rests on many judgments for which a scientific consensus has not been established. Risk assessors might be faced with several scientifically plausible approaches (e.g., choosing the most reliable dose-response model for extrapolation beyond the range of observable effects) with no definitive basis for distinguishing among them. The earlier committee pointed out that selection of a particular approach under such circumstances involves what it called a science-policy choice. Science-policy choices are distinct from the policy choices associated with ultimate decision-making, as will be seen below. The science-policy choices that regulatory agencies make in carrying out risk assessments have considerable influence on the results and are the focus of much that follows in this report.
What is the Relationship Between Risk Assessment and Research?
Although the conduct of a risk assessment involves research of a kind, it is primarily a process of gathering and evaluating extant data and imposing science-policy choices. Risk assessment draws on research in epidemiology, toxicology, statistics, pathology, molecular biology, biochemistry, analytical chemistry, exposure modeling, dosimetry, and other disciplines; to the extent that it attempts to capture and take into account uncertainties, it also draws on the research efforts of decision analysts.
Risk assessment, at least in theory, can influence research directions. Because, at its best, risk assessment provides a highly organized profile of the
current state of knowledge of particular issues and systematically elucidates scientific uncertainties, it can provide valuable guidance to research scientists regarding the types of data that can most effectively improve understanding. Little effort seems to have been made to use risk assessments in this way, although the Office of Technology Assessment has recently completed a study that describes the role of risk assessment in guiding research (OTA, 1993).
What is the Relationship Between Risk Assessment and Regulatory Decision-Making?
Risk management is the term used to describe the process by which risk-assessment results are integrated with other information to make decisions about the need for, method of, and extent of risk reduction. Policy considerations derived largely from statutory requirements dictate the extent to which risk information is used in decision-making and the extent to which other factorssuch as technical feasibility, cost, and offsetting benefitsplay a role.
Some statutes seem not to permit risk-assessment results to play a substantial role; they stress reductions of exposure to the ''lowest technically feasible level" and usually require the best available technology. Proponents of such technology-based approaches often argue that they facilitate more rapid regulatory action and are especially suitable for making large and relatively inexpensive "first-cut" emission reductions. Proponents of quantitative risk assessment argue that such approaches are blind to the possibility that the risks remaining after application of such technology might still be unreasonably large or, in other situations, that they have been pushed to unnecessarily low values. As amended in 1990, Section 112 of the Clean Air Act gives quantitative risk-assessment results a secondary but still important role relative to technology-based controls.
What Is a Default Option?
EPA's guidelines set forth "default options." These are generic approaches, based on general scientific knowledge and policy judgment, that are applied to various elements of the risk assessment process when specific scientific information is not available. For instance, ambient doses of contaminants in humans are generally far lower than the doses that produce tumors in animals in controlled studies. The guidelines advise that, in assessing the magnitude of cancer risk to humans from low doses of a chemical based on the results of a high-dose experiment, "in the absence of adequate information to the contrary, the linearized multistage procedure will be employed" (EPA, 1986a, 1987a); that is, cancer risk in humans exposed to low doses will be estimated mathematically by using high-dose data and a curve-fitting procedure to extrapolate to low doses. Departure from the guideline is allowed if there is "adequate evidence" that the mechanism through which the substance is carcinogenic is more consistent with a
different model; for instance, that there is a threshold below which a substance will not cause a risk. Thus, the guideline amounts to a "default" that guides a decision-maker in the absence of evidence to the contrary; in effect, it assigns the burden of persuasion to those wishing to show that the linearized multistage procedure should not be used. Similar guidelines cover such important issues as the calculation of effective dose, the consideration of benign tumors, and the procedure for scaling animal-test results to estimates of potency in humans. In the absence of information on some critical point in a risk assessment, default procedures seem essential. The question, then, is not whether to use defaults, but which defaults are most appropriate for a specific task and when it is appropriate to use an alternative to a default.
It is helpful to provide a brief historical perspective on the origins and evolution of risk assessment, so that some of the reasons that led to the use of the technique can be seen. The review is divided into two main parts, with an intervening section devoted to the NRC study of 1983 that was so influential in the developments of the last decade.
Early Efforts to Establish Safe Limits of Exposure to Toxic Substances
About 50 years ago, toxicologists began to study the problem of establishing limits on exposures to hazardous substances that would protect human health. The early efforts began in the 1940s in connection with concerns about occupational exposures to chemicals and about residues of pesticides in foods. Toxicologists were guided by the principle that all substances could become harmful under some conditions of exposurewhen the so-called threshold dose was exceededbut that human health could be protected as long as those exposure conditions were avoided. Threshold doses were recognized to vary widely among chemicals, but as long as human exposures were limited to subthreshold doses, no injury to health would be expected. The threshold hypothesis thus involved rejection of the simplistic view that the world is divided into toxic and nontoxic substances and acceptance of the principle that, for all chemicals, there were ranges of exposure that were toxic and ranges that were not. The threshold hypothesis was based on both empirical observations and basic concepts of biologythat every organism, including the human, has the capacity to adapt to or otherwise tolerate some exposure to any substance and that the harmful effects of a substance would become manifest only when exposure exceeded that capacity. Even at that early stage, there were questions about whether carcinogens always had thresholds, but otherwise the threshold concept became widely accepted.
Although there was widespread acceptance of the threshold hypothesis (ex-
cept among scientists working in genetics and in chemical carcinogenesis) (NRC, 1986), it was not apparent how the threshold dose was to be estimated for a large and diverse human population whose members have different thresholds of susceptibility. Experts in occupational health tended to rely heavily on observations of short-term toxicity in highly exposed workers and established acceptable exposure limits (the most prominent of which were the so-called threshold limit values, TLVs, first published by the American Conference of Governmental Industrial Hygienists in the 1950s) that were below the exposures that produced observable toxic effects. In the early 1950s, two Food and Drug Administration (FDA) scientists, O.G. Fitzhugh and A. Lehman, proposed a procedure for setting acceptable limits, which became known as acceptable daily intakes (ADIs), for dietary pesticide residues and food additives. Their procedure was based on the threshold hypothesis and first involved identification of a chemical's no-observed-effect level (NOEL) from the set of chronic animal-toxicity data in which the animals responded to the lowest dose testedthe "most sensitive" indication of the chemical's toxicity. Several response levels are characterized by acronyms. The first is the "no-observed-effect-level," NOEL. Earlier this was called the no-observable-effect level. Observable was changed to observed to be more in keeping with actual data ("observed"), rather than a rather vague potential "observable," which might be related to the size and sensitivity of the experiment. What is not observable in a small experiment might be easily observed in a large experiment. The word adverse was added to NOEL, making it NOAEL and making it clearer that adverse effects were of concern. The LOEL and LOAEL have a similar genesis and currently refer to the "lowest-observed-adverse-effect level''the lowest dose at which an adverse effect was seen.
Fitzhugh and Lehman cited data suggesting that "average" human sensitivities might be up to 10 times those of laboratory animals and that some members of a large and diverse human population might be up to 10 times more sensitive than the "average" person. Thus came into use the safety factor of 100. The experimental NOEL was divided by 100 to arrive at a chemical-specific ADI. If human exposure was limited to daily amounts less than the ADI, then no toxicity was to be expected. In fact, Fitzhugh and Lehman, and later other authors and expert groups, including the World Health Organization, did not claim that an ADI arrived at in this fashion was risk-free, but only that it carried "reasonable certainty of no harm." No attempt was made to estimate the probability of harm. A variation of the safety-factor approach, often called margin of safety, is the estimate of the ratio of the NOEL to actual exposures. A judgment is made as to whether that ratio is acceptable. This margin-of-safety approach seems to be most common for substances already in general use, and in practice is often associated with lower ratios of NOEL to exposure than those based on safety factors.
The use of safety factors to establish ADIs was also recommended by various NRC committees (NRC, 1970, 1977, 1986) and adopted by the Joint Food
and Agriculture Organization and World Health Organization expert committees on food additives (FAO/WHO, 1982) and pesticide residues (FAO/WHO, 1965).
Although it has since been modified in several minor ways, the basic procedure for setting limits on human exposures to chemicals in air, water, and food persists to this day. The threshold hypothesis has been criticized as inadequate to account for some toxic effects, and it has not been accepted by regulators as applicable to carcinogens, but it remains a cornerstone of other regulatory and public-health risk assessments. Section 112 of EPA's authority for regulating toxic air pollutants envisions a safety-factor approach for some kinds of risk assessment.
The Problem of Carcinogens
Not only is cancer a much-feared set of diseases, but public and scientific concerns about cancer-inducing chemicals in the environment have centered on the possibility that such substances might act through nonthreshold mechanisms; that is, that exposure to even one molecule of a carcinogen is associated with a small but non-zero increased risk of tumor induction. This possibility served as the basis for modern dose-response models, which were developed initially from observations of radiation-induced cancer. These models came into wide use and were promoted by the National Research Council's series of reports entitled Biological Effects of Ionizing Radiation and later incorporated into the regulatory decision-making of the Nuclear Regulatory Commission. Perhaps the earliest legislative acknowledgment of the possibility that chemical carcinogens might act in the same way came in the form of the "Delaney clause" of the Food Additive Amendments of 1958. Following the suggestions set forth by several FDA and National Cancer Institute (NCI) officials, Congress stipulated that no additive that concentrates in food during processing or is added to food during or after processing may be allowed in the food supply if it is found to be carcinogenic in animals. The basis for the Delaney clause was that it is not possible to specify a safe human exposure to a carcinogen in the same sense that a safe intake of a substance acting through threshold mechanisms could be identified.
Through the 1960s and into the early 1970s, toxicologists avoided the problem of identifying "acceptable" intakes of carcinogens. Where it was possible, regulators simply prohibited introduction of carcinogens into commerce. But where banning was difficult or even infeasiblefor example, for environmental contaminants that were byproducts of manufacturing and energy productionchoosing a maximal permissible human exposure, and acceptance of some risk. Limits were sometimes based on some concept of technical feasibility. The problem with such a criterion for setting limits was that it provided little confidence that human health was being adequately protected or, conversely, that risks were not being forced to unnecessarily low levels. In many cases, carcinogenic pollutants were simply ignored (NRC, 1983a).
Those approaches to the problem of regulatory exposure to environmental carcinogens became problematic in the face of two trends. First, government and industrial testing for carcinogenicity began to increase rapidly during the late 1960s; during the 1970s, regulators had to begin to deal with large numbers of newly identified carcinogens that were found among the many commercial products introduced after World War II. Second, analytic chemists became able to identify carcinogens in the environment at lower and lower concentrations. It became clear during the early to middle 1970s that a systematic approach to regulating carcinogens was needed.
Several authors had published methods for quantifying low-dose risks associated with chemical carcinogen exposure in the 1960s and 1970s, and regulatory agenciesFDA and EPA in particularbegan adopting some of the methods in the middle 1970s. EPA, for example, estimated low-dose risks associated with several carcinogenic pesticides and relied in part on its assessments in actions to cancel or limit their registrations. FDA began using low-dose risk estimation to deal with so-called indirect food additives and some food contaminants that proved to be carcinogenic. The Occupational Safety and Health Administration (OSHA) at first rejected the use of risk quantification as it mounted a major effort during the late 1970s to regulate occupational carcinogens, because it believed that the statute under which it operated did not permit the use of risk assessment. But a Supreme Court decision regarding the agency's efforts to establish a permissible exposure limit for benzene caused OSHA to incorporate risk quantification (see below).
Those trends of the 1970s toward increasing the use of risk assessment in carcinogen regulation caused several regulatory agencies, working together as the Interagency Regulatory Liaison Group (IRLG), to develop and publicize a set of guidelines for the conduct of risk assessments (IRLG, 1979). The guidelines were said by the agencies to specify a common approach to risk assessment. No commitment was made by the agencies to use the methods for all possible carcinogens in all classes of regulated products, but, to the extent that an agency decided to use risk assessment, its approach would be that specified in the IRLG guidelines. The agencies also noted that the guidelines did not include an approach to what later came to be called risk management; such issues were said to remain the prerogative of the individual agencies.
The IRLG guidelines embodied several important scientific principles that originated in efforts of the WHO International Agency for Cancer Research (IARC) (IARC, 1972, 1982), NCI (Shubik, 1977), and the federal regulatory agencies 9FDA, 1971; Albert et al., 1977; OSHA, 1982). Among them were principles concerning the appropriate uses of epidemiologic and animal data in identifying potential human carcinogens and the extrapolation of such data to humans. The IRLG guidelines did not explicitly incorporate the "default options" language described earlier (that came only after the 1983 NRC report), but
it is clear that they do include science-policy choices (e.g., the generic adoption of a linearized, no-threshold model for carcinogen dose-response assessment).
By the early 1980s, risk assessment had begun to take on considerable importance within the regulatory agencies and to capture the attention of regulated industries. One important impetus to the development of risk-assessment techniques was the Supreme Court's decision in Industrial Union Department, AFLCIO v. American Petroleum Institute, 448 U.S. 607 (1980), the "Benzene" decision. That decision struck down the OSHA standard for exposure to benzene in the workplace. The standard was based on OSHA's policy of trying to reduce concentrations of carcinogens in the workplace as far as technological possible without consideration of whether existing concentrations posed a significant risk to health. There was no opinion for the majority of the Supreme Court in Benzene, but four justices concluded that, under the Occupational Safety and Health Act, OSHA could regulate only if it found that benzene posed a significant risk of harm. Although the plurality did not define significant risk of harm and stressed that the magnitude of the risk need not be determined precisely, the decision strongly signaled that some form of quantitative risk assessment was necessary as a prelude to deciding whether a risk was large enough to deserve regulation.
Under those circumstances, Congress instructed FDA to arrange for the National Research Council in 1981 to undertake a study of federal efforts to use risk assessment.
NRC Study Of Risk Assessment In The Federal Government
In 1983, NRC was asked to issue recommendations regarding the scientific basis of risk assessment and the institutional arrangements under which it was being conducted and used. In particular, NRC's charge involved a close examination of the possibility that risk assessment might be conducted by a separate, centralized scientific body that would serve all the relevant agencies. It was proposed that such an arrangement might reduce the influence of policy-makers on the conduct of risk assessment, so that there would be minimal opportunities for the results of risk assessments to be manipulated to meet predetermined policy objectives.
The NRC committee drew extensively on the earlier work of EPA, FDA, OSHA, IARC, and NCI, and much of its effort was directed at a synthesis of scientific principles and concepts first elucidated by these agencies. The NRC study did not, however, recommend specific methods for the conduct of risk assessment.
The risk assessment framework and specific definitions of risk assessment and its component steps from the 1983 NRC report have been widely adopted.
Many of the recommendations from the 1983 report have been implemented by EPA and other regulatory agencies. Two of the major recommendations of the committee, summarized below, are particularly relevant to this report:
The NRC committee did not specify and particular methodologic approach to risk assessment, nor did it address the issue of which default options should be used by regulatory agencies. It did, however, note that provisions should be made for continuing review of the science underlying the guidelines and of the basis of the default options incorporated in them.
Events After Release Of The 1983 NRC Report
The Office of Science and Technology Policy (OSTP) brought together scientists from the regulatory agencies, the National Institutes of Health, and other federal agencies and, in 1985, issued a comprehensive review of the scientific basis of risk assessment of chemical carcinogens. The OSTP review adopted the framework for risk assessment proposed by the NRC committee and provided the individual regulatory agencies a basis for developing the type of guidelines recommended by that committee.
Alone among federal agencies, EPA adopted a set of guidelines for carcinogen risk assessment in 1986, as recommended by NRC. The EPA guidelines specify default options, note the distinction between risk assessment and risk management, and otherwise meet NRC's and OSTP's recommendations. EPA has issued guidelines for assessing risks associated with several other adverse health effects of toxic substances (without the benefit of OSTP review of the underlying science) and for the conduct of human exposure assessments. Beginning in 1984, it initiated work and published guidelines for evaluating mutage-
nicity, developmental toxicity, effects of chemical mixtures, and human exposure (EPA, 1986a, 1987a). It later published proposed guidelines on female reproductive risk (EPA, 1988a), male reproductive risk (EPA, 1988b), and exposure-related measurements (EPA, 1988c). Final, revised guidelines on developmental toxicity were published in 1991 (EPA, 1991a). The agency is now in the process of issuing revised guidelines on cancer risk assessment and has issued revised guidelines for the assessment of human exposures (EPA, 1992a).
Increasing activity at the state level was first indicated by California's publication in 1985 of Guidelines for Chemical Carcinogen Risk Assessments and Their Scientific Rationale (CDHS, 1985). The purpose of the guidelines was "to clarify internal procedures which risk assessment staff of the California Department of Health Services will usually use to deal with certain decision points which are characteristic of most risk assessments." The authors went on to state why guidelines were thought necessary, in language consistent with earlier statements of IRLG (1979), NRC (1983a), OSTP (1985), and EPA (1987a):
These California guidelines, while in harmony with recent federal statements on carcinogenic risk assessment, are more specific and practical. The Department of Health Services' staff believe that there are important advantages to the announcement of such flexible nonregulatory guidelines. First, the publishing of guidelines increases the likelihood of consistency in risk assessment among agencies and decreases the time spent repeatedly arguing risk assessment policy for each separate substance. Second, announcing guidelines in advance makes it clear that one is not tailoring risk assessment to justify some predetermined risk management decision. Third, specific guidelines allow the regulated community to predict what emissions, food residues, or other exposures are apt to be of public health concern. Fourth, the publication and discussion of these guidelines should make the process more understandable to risk managers who have to make decisions that depend in part on risk assessment determinations.
The NRC, OSTP, EPA, and California documents were produced during a time in which the uses of risk assessment to guide regulatory decision-making were expanding rapidly. Particularly important was EPA's adoption of risk assessment as a guide to decisions at Superfund and other hazardous-waste sites, including those covered by the Resource Conservation and Recovery Act (RCRA).
The agency also extended the uses of risk assessment to decisions regarding pesticide residues in food, carcinogenic contaminants of drinking-water supplies, industrial emissions of carcinogens to surface waters, and industrial chemicals subject to regulation under the Toxic Substances Control Act (TSCA). Risk-management approaches varied according to the specific legal requirements applicable to the sources of carcinogen exposure, but the EPA guidelines were intended to ensure that the agency's approach to risk assessment was uniform across the various programs.
Uses Of Risk Assessment In The Regulation Of Hazardous Air Pollutants
Section 112 of the Clean Air Act, as originally adopted in the Clean Air Act Amendments of 1970, required EPA to set emissions standards for hazardous air pollutants so as to protect public health with an "ample margin of safety." EPA was slow in carrying out that mandate, listing only eight chemicals as hazardous air pollutants in 20 years.1Standards were issued for only seven (there was no standard for coke ovens), and the standards that were issued covered only some of the sources that emit these pollutants. One major reason was the ambiguity of "ample margin of safety." Many commentators long thought that that term barred EPA from considering costs; EPA might well have to set a standard of zero for any pollutant for which no threshold could be defined (i.e., virtually all carcinogens).
That interpretation of the act (originally developed well before 1987), however, was unanimously rejected by the District of Columbia Circuit court in Natural Resources Defense Council v. EPA (824 F.2d 1146 [en banc] [D.C.Cir. 1987]). At the same time, the Court of Appeals also rejected EPA's position that it could use technologic or economic feasibility as the primary basis for standard-setting under Section 112. Instead, the court held that EPA had first to determine what concentration was "safe"i.e., represented an acceptable degree of riskand had then to select a margin of safety necessary to incorporate the uncertainties in scientific knowledge. In the latter step, but not the former, technological feasibility could be taken into account. In accordance with the plurality opinion in the Supreme Court's Benzene decision, the circuit court also held that EPA's standards did not have to eliminate all risk.
As in the Benzene case, the court did not define any particular method for EPA to use in determining what risks are acceptable. On remand, the agency, after taking comment on a number of possibilities, decided that it could not use any single metric as a measure of whether a risk is acceptable. Instead, it adopted a general presumption that a lifetime excess risk of cancer of approximately one in 10,000 (10-4) for the most exposed person would constitute acceptable risk and that the margin of safety should reduce the risk for the greatest possible number of persons to an individual lifetime excess risk no higher than one in 1 million (10-6). Such factors as incidence (e.g., the number of possible new cases of a disease in a population), the distribution of risks, and uncertainties would be taken into account in applying those benchmarks. The agency approach thus put primary emphasis on estimating individual lifetime risks through quantitative risk assessment.
Congress lessened the role of quantitative risk assessment for air-pollution regulation by rewriting Section 112 in Title III of the 1990 amendments. Congress defined 189 chemicals as hazardous (subject to possible deletion) and required technology-based controls on sources of those chemicals, as well as any
others that might be added to the list by EPA. Sources that emit hazardous air pollutants will be regulated in two stages. In the first, technology-based emissions standards will be imposed. Each major source (defined, generally, as a stationary source having the potential to emit 10 tons per year of a single hazardous air pollutant or 25 tons per year of a combination of hazardous air pollutants) must meet an emission standard based on using the maximum available control technology (MACT) as defined by standards to be issued by EPA. Smaller sources, known as area sources, must meet emissions standards based on using generally available control technology.
Section 112 defines some contexts in which quantitative risk assessment will remain important. First, quantitative risk assessment will be relevant in determining which categories of sources will not be subject to technology-based regulation; EPA may delete a source category from regulation if no source in the category poses a risk of greater than 10-6 to the "individual most exposed to emissions." Even here, judging from the use of the word "may," EPA is not required to make the deletion; thus, the results of the quantitative risk assessment need not be decisive.
Quantitative risk assessment has a greater, but still limited, role in the second stage of standard-setting under Section 112(f), the "residual-risk" stage. That section requires EPA to set standards that protect public health with an ample margin of safety if it concludes that the first stage of technology-based standard-setting has not done so. Second-stage standards must be set for a category of "major sources" if the first stage allows a residual risk of greater than 10-6 to the individual most exposed to emissions. This requirement might seem a wholesale adoption of risk management based on the maximally exposed person, but two points must be noted. First, the 10-6 criterion for standard-setting need only be an upper-limit screening device. EPA is free, if it chooses, to set second-stage standards for source categories posing lesser risks. Second, the actual second-stage standard need not be expressed in terms of quantitative risk. Section 112(f)(2) authorizes EPA to continue the 10-4/10-6 approach described earlier, but it does not require the agency to do so. Instead, any methods is acceptable that comports with NRDC v. EPA's requirement that the standards provide an "ample margin of safety" in addition to reducing risk to a level judged acceptable by EPA.
Such techniques as hazard assessment, hazard ranking, and risk ranking (discussed above), and in some cases quantitative risk assessment, can also play a role in the agency's decisions on questions such as these:
Noncancer Risk Associated With Hazardous Air Pollutants
The current EPA approach to risk assessment for noncancer hazards posed by hazardous air pollutants, refined in several ways, is conceptually similar to the traditional approach to threshold agents described earlier. The agency identifies a so-called inhalation reference concentration (RfC). An RfC is defined by EPA as "an estimate (with uncertainty) of the concentration that is likely to be without appreciable risk of deleterious effects to the exposed population after continuous, lifetime exposure" (EPA, 1992b). RfCs are derived from chemical-specific toxicity data. The latter are used to identify the most sensitive indicator of a chemical's toxicity and the so-called no-observed-adverse-effect level (NOAEL) for that indicator effect. If the NOAEL is derived from an animal study, as is typically the case, it can be converted to a human equivalent concentration by taking into account species differences in respiratory physiology. Uncertainty factors, whose magnitudes depend on the nature of the toxic effect and the quantity and quality of the data on which the NOAEL is based, are applied to the human-equivalent NOAEL to estimate the RfC. That procedure is used for all forms of toxic hazard except carcinogenicity. The use of RfCs depends on the assumption that toxic effects will not occur until a threshold dose is exceeded (EPA, 1992b).
Another important provision of Title III of the 1990 Amendments was the requirement that environmental effects be included in the evaluation of a risk associated with a pollutant. An adverse environmental effect is defined in Section 112(a)(7) of the act as "any significant and widespread adverse effect, which may reasonably be anticipated, to wildlife, aquatic life, or other natural resources, including adverse impacts on populations of endangered species or significant degradation of environmental quality over broad areas." Appendix III of EPA's Unfinished Business report (EPA, 1987b) found that airborne toxic substances, toxic substances in surface waters, and pesticides and herbicides were in the second highest category of relative risk in the ecological and welfare catego-
ries. Of particular concern in this report was the transport by air and water of toxic substances (heavy metals and organics) that accumulate in ecological food chains. Such bioaccumulation has impacts on both ecological resources and the use by humans of specific ecological populations (e.g., fish consumption). Ecological risk assessment is not discussed in this report except to the extent that bioaccumulation affects the health of people who eat and drink contaminated ecological resources, but is discussed in another recent NRC report entitled Issues in Risk Assessment (NRC, 1993a).
Public Criticism Of Conduct And Uses Of Risk Assessment
The development of risk-assessment methods and their expanding uses in the federal and state regulation of hazardous substances have been carefully scrutinized by interested parties in the regulated industries, environmental organizations, and academic institutions. That scrutiny has led to frequent and sharp criticisms of the methods used for assessing risk and of ways in which the results of risk assessment have been used to guide decision-making. The criticisms have not been directed solely at the use of risk assessment in regulation of hazardous air pollutants, but rather cover a range of uses.
We cite here some of the criticisms that have appeared in the literature or that have otherwise been presented to the committee, because they help to define the issues reviewed in this report. We emphasize that our citation of these criticisms does not mean that we believe them to be valid. Nor is the order of their listing meant to suggest our opinion regarding their possible importance.
Criticisms Pertaining to Conduct of Risk Assessment
Criticisms Pertaining to the Relationship Between Risk Assessment and Risk Management
Are any of those criticisms justified? If so, what responses can be made to them? Can improvements be made? If so, how will they affect the conduct of risk assessment and the use of risk-assessment results in regulatory decision-making? These and related issues are the primary focus of Chapters 6-12 of this report.
1. The chemicals listed as hazardous air pollutants under the National Standards for Hazardous Air Pollutants (NESHAP) (with the date of public notice): asbestos (3/71); benzene (6/77); beryllium (3/71); coke-oven emissions (9/84); inorganic arsenic (6/80); mercury (3/71); radionuclides (12/79); and vinyl chloride (12/75).