Tobacco Documents Online

REGULATORY IMPACT ANALYSIS PROJECT, INC. 0 CHOICES IN RISK ASSESSMENT THE ROLE OF SCIENCE POLICY IN THE ENVIRONMENTAL RISK MANAGEMENT PROCESS Prepared for Sandia National Laboratories Sponsored by The U.S. Department of Energy Office of Environmental Management and Off ce of Environment, Safety and Health

• • EXECUTIVE SUMMARY What Is Science Policy? In the context of this report, "science policy issues" are the gaps and uncertainties in scientific knowledge and data that arise in the assessment of risks to human health and the environment associated with exposure to substances, conditions, activities, and sites. "Science policy decisions" are the policy choices made to bridge such gaps and uncertainties. Science policy decisions are vital to the regulatory risk assessment and management processes. Science policy decisions enable regulators to justify the costs of regulatory programs in terms of estimated health and environmental risk reductions. Default Assumptions Default assumptions are science policy decisions that are applied automatically when certain science policy issues arise. Examples of science policy issues and the corresponding default assumptions are presented in Table ES• 1. Default assumptions are perceived-and criticized-by some as being conservative. There are others who criticize them for insufficient protectiveness. The selection of default assumptions generally is driven by the policy decision to avoid underestimating potential risks. Given the frequent use of quantitative risk assessment in health and environmental regulation, for any individual science policy issue, use of a default assumption may be the most practical option for getting the work done. Departures from default assumptions have been rare in the past, but alternate assumptions have been adopted in limited cases. Attempts to depart from default assumptions in future risk assessments may invite increased scrutiny, which could result in a reluctance to consider or adopt alternatives based on new scientific information. Continued reliance on default assumptions can be problematic in two scenarios: l. Multiple conservative science policy decisions, known as "compounded conservatism," may result in inconsistent or unduly biased decisions; and 2. Whether or not compounded conservatism results, policy makers, risk managers, the media, and the public are often unaware of: a. The gaps and uncertainties in scientific knowledge and data used in conducting a risk assessment; b. The policy-based default assumptions that are used to bridge these gaps and uncertainties; and c. The extent to which default assumptions may determine the outcome of the risk assessment.

vw CHOICES IN RISK ASSESSMENT Table ES•1. Basic Science Policy Issues and Default Assumptions • Science Policy Issue In the absence of adequate human data, what is the relevance of animal bioassay data to the estimation of human risk? Is the occurrence of benign tumors in experimental animals relevant to estimating human cancer risk? When both positive and nonpositive cancer incidence data exist, should the nonpositive data be used for quantitative risk assessment purposes? Default Science Policy Assumption A substance that is carcinogenic to animals is also a human carcinogen. Benign tumors are combined with malignant tumors in animals to establish carcinogenic potential in humans. In the presence of positive data, nonpositive data do not indicate safety and should not be used in quantitative risk assessment. • What is the relevance of data from animal bioassays conducted with MTD protocols to estimating potential human risk? Which animal species should be used to represent humans in terms of carcinogenic response? When predicting human health risk on the basis of animal data, how should mechanistic variations between species be taken into account? Data indicate that ingestion of a substance may be associated with cancer. If inhalation exposures are of concern, what is the relevance of the ingestion data to the assessment of risk? The available data do not demonstrate the absence or existence of a threshold for carcinogenesis. Data indicate a dose-response relationship at high doses, but few or no data concerning the dose-response relationship at lower levels exist. If data on human exposure are unavailable for a particular substance or site, how can exposures be estimated for purposes of quantitative risk assessment? Carcinogenic effects observed at the MTD in animals are predictive of effects in humans at much lower doses. The animal species exhibiting the greatest sensitivity is the most appropriate for risk assessment. Differences between species in mechanisms of carcinogenicity are not taken into account when extrapolating data from one species to another. A carcinogen by one route of exposure is a carcinogen by any other route of exposure. There is no nonzero dose below which an increase in cancer risk does not occur. The dose-response relationship is linear at low doses. Chosen values for exposure variables are upper-bound point estimates which, when taken together, do not result in unrealistic exposure estimates.

EXECUTIVE SUMMARY ix • • Current scientific knowledge cannot determine which default assumptions are correct. Science may never be able to answer certain questions that transcend the capabilities of the scientific method. These "trans-scientific" questions include: "What is the shape of the dose-response curve at low doses?" and "Do thresholds for carcinogens exist?" Therefore, it is likely that policy-based default assumptions will always be necessary in risk assessment. However, continued reliance on default assumptions in all cases represents and promotes the stasis of science and risk assessment, and research will continue to identify plausible alternatives for default assumptions. Pressure to incorporate alternative assumptions in risk assessment is increasing. Alternatives to the Default Assumptions Plausible alternatives to the default assumptions are available in many, specific instances. In most cases, justification of an alternative relies on chemical- and species- specific data and arguments. Consequently, it is unlikely that any default assumption will be completely replaced. A justifiable alternative may be identified for a class of chemicals, but at present there is no universally justifiable and acceptable alternative to any of the default assumptions. Replacement of default assumptions will occur only after sufficient research and data have indicated that an alternative is more likely to be correct than the default. The alternative must also still be protective of public health. Thus, in the near future, research on alternatives will be limited in impact and will likely result only in incremental changes in the risk assessment process. The Comprehensive Methodology developed by the American Industrial Health Council represents a potential revolution in the way risk assessments are conducted. Some believe that, if combined with physiologically based pharmacokinetic models and distributional exposure assessments, this methodology could be a dramatic improvement over current risk assessment methods. Full and complete incorporation of all uncertainty and variability would be achieved, and exposures and risks would be expressed in terms of probabilistic distributions. Regulatory decision makers would be presented with complete probabilistic descriptions of the ranges of expected exposures and risks, rather than point estimates. Probabilistic distributions would enable decision makers to consider the likelihood that various exposure and risk estimates will occur and determine explicitly the appropriate degree of conservatism in regulations. This would allow for a degree of separation of risk assessment and risk management, as advocated by the National Research Council, that cannot currently be achieved. Such a change in environmental regulatory decision making within federal agencies will require a commitment to the need for such a change as well as a commitment to funding the required research. If regulatory agencies indicate a willingness to evaluate and incorporate alternatives to default assumptions in regulatory risk assessments, the regulated community will have an incentive to conduct the necessary research. In the end, all parties likely will benefit as knowledge of mechanisms of carcinogenesis and understanding of the hazards posed by environmental contaminants is increased.

• • x CHOICES IN RISK ASSESSMENT Case Studies on Science Policy Fluoride in drinking water Fluoride has been artificially added to drinking water as a public health measure to reduce the incidence of dental caries since the U.S. Public Health Service (PHS) first identified an optimal fluoride concentration in 1943. Nevertheless, communities have debated the relative benefits of reduced dental caries and improved oral health versus the potential risks of adverse health effects on teeth and bones. The public health community has long held that the benefits of fluoridation far outweigh any potential risks. Potential long-term health effects, however, are poorly understood. Recent animal studies associating increased cancer risk with fluoridated drinking water provoked renewed concern. Epidemiologic studies have not conclusively established an association between fluoride and bone cancer risk in humans. The major science policy issue considered in this case study is the evaluation of fluoride as to its potential to cause cancer in humans. Two recent reviews of the available animal and human data concluded that there is no evidence that fluoride is a human carcinogen. The Environmental Protection Agency (EPA) subsequently announced that the existing fluoride drinking water standards would not be revised. Given the benefits of fluoridation and the ease with which cosmetic and potentially adverse effects can be minimized or avoided, it would have been imprudent to suggest a change in the regulation of fluoride in drinking water on the basis of inconclusive evidence of carcinogenicity in male rats. A possible additional motivation behind the decision not to change the regulatory standard for fluoride in drinking water might have been fear of the tumult that would have ensued in the public health community and in the public at large if fluoride were judged to be carcinogenic. Classification of fluoride as a possible human carcinogen could have critically damaged the credibility of the PHS, which has aggressively promoted fluoridation for fifty years. Asbestos in consumer products Due to its durability and heat-resistant properties, asbestos has been used in a variety of consumer products since the late nineteenth century. Concern about asbestosis and lung cancer associated with asbestos exposure has grown throughout the twentieth century. Asbestos is regarded by EPA and the International Agency for Research on Cancer (IARC) as a known human carcinogen. New uses of asbestos have been banned in the United States since the 1980s. Considerable resources have been devoted to removing asbestos from public buildings, especially schools. In response to growing concerns about adverse health effects associated with asbestos, EPA promulgated a ban on the manufacture, importation, processing, and distribution of existing consumer products containing asbestos in 1989. The ban was remanded by the U.S. Court of Appeals for the Fifth Circuit in 1991 because EPA had not sufficiently justified the ban and had not fulfilled the requirements of the Toxic Substances Control Act (TSCA). The science policy issues supporting the court's decision were EPA's inadequate consideration of risks to health and safety posed by potential asbestos substitute

EXECUTIVE SUMMARY Xi products and EPA's use of analogous exposures to estimate benefits of the ban without provisions for public review and comment. EPA has yet to take further action. If the regulation of asbestos-containing products is revisited, special attention should be paid to substitute risk issues. A risk analysis supporting a proposed regulation is not complete unless the full consequences of the regulation are evaluated. Unleaded gasoline Automobiles and other motor vehicles are a widely recognized source of significant air pollution. Pollutants of concern associated with motor vehicles include lead, hydrocarbons, carbon monoxide, and nitrogen oxides. Unleaded gasoline, which was originally required for use with catalytic converter-equipped cars, has been on the market for more than twenty years. Continuing attention to reducing air pollution has resulted in increasingly stringent exhaust emissions requirements on automobiles and the elimination of leaded gasoline as a fuel. This case study focuses on unleaded gasoline, which has been associated with kidney tumors in male rats. The central science policy issue is determining the relevance of a particular type of kidney tumor in male rats to human cancer risk assessment. As a default science policy decision, cancer in animals is assumed to be predictive of carcinogenic effects in humans. EPA scientists evaluated mechanistic data and determined that the kidney tumors observed only in male rats exposed to unleaded gasoline were of no relevance to potential human cancer risk. If unleaded gasoline had been implicated as a rodent carcinogen, and subsequently suspected of being a human carcinogen, significant upheaval concerning the use of unleaded gasoline could have ensued, potentially damaging the credibility of EPA which has promoted the use of unleaded gasoline over the last twenty years. Used oil More than 1 billion gallons of used oil are generated each year in the United States. Used oil contains a variety of toxic and carcinogenic substances and can therefore pose a threat to human health and the environment, especially when improperly managed or disposed. In developing the Hazardous Waste Management System mandated under the Resource Conservation and Recovery Act (RCRA), EPA had to decide whether or not to designate used oil as a hazardous waste. All hazardous wastes must be managed under strict standards in Subtitle C of RCRA. EPA evaluated and re-ev.aluated the data and requirements of RCRA and other statutes and changed its position several times. As a matter of science policy, EPA determined that the hazards posed by used oil did not meet the criteria for hazardous waste listing under RCRA. Litigation ensued when petitioners questioned the validity of an EPA proposal not to list used oil on the basis that the resulting stigma would have negative effects on recycling. Eventually, EPA fulfilled the RCRA mandate and fostered recycling by instituting special management standards for used oil but not listing it as hazardous waste.

s1k xii CHOICES IN RISK ASSESSMENT • i Trichloroethylene Trichloroethylene (TCE) has long been used in a variety of industries and is therefore a groundwater contaminant at numerous sites. Superfund law and policy require that contaminants in certain groundwater aquifers be cleaned up to drinking water standards. Remediation of contaminated groundwater often drives the cost and duration of Superfund site cleanups. EPA classifies TCE as a probable human carcinogen. However, evidence is gathering that TCE is either not carcinogenic or not as carcinogenic in humans as once thought. Incorporation of alternative science policy decisions could result in less stringent drinking water standards for TCE. The standards applied to groundwater cleanup would also be less stringent, which would reduce remediation costs but not public health protection. Workplace indoor air quality The Occupational Safety and Health Administration (OSHA) recently proposed to regulate indoor air quality (IAQ) in the workplace. Improved IAQ is intuitively desirable, but scientific data concerning IAQ are sparse. The lack of data limits OSHA's ability to assess the health risks posed to workers by poor IAQ. The proposed IAQ regulation includes a ban on workplace smoking, except in specially designated and separately ventilated areas, as well as measures designed to address other indoor air contaminants. This case study examines the science policy in OSHA's risk assessment for environmental tobacco smoke (ETS) and the estimated costs and benefits of the proposed smoking ban. Despite a relatively large database of information on human lung cancer risk from exposure to ETS, OSHA had to make a number of science policy decisions to conduct the quantitative risk assessment necessary to justify the proposed smoking ban. Although the estimated costs of the proposed smoking ban appear to be relatively low and the estimated benefits appear to be relatively high, the costs may be incomplete and the benefits may be substantially overstated. The information database for the remainder of the proposed rule for IAQ is not nearly as extensive as that for ETS, and the associated science policy decisions are likely to be more tenuous than those for ETS. Because the estimated costs of the portion of the proposed IAQ rule not addressing smoking are very high, the proposed science policy decisions are even more questionable. Toxics Release Inventory The Emergency Planning and Community Right-to-Know Act of 1986 (EPCRA) requires industrial facilities to report their releases and transfers of toxic chemicals listed on the Toxics Release Inventory (TRI). The purpose of such reporting is to provide communities with information concerning routine local releases and transfers of toxic chemicals. TRI reporting is not intended to reduce or restrict routine or permitted releases of and exposures to chemicals and does not directly reduce health risks. Congress established the initial list of chemicals subject to TRI reporting, but EPA is authorized to add chemicals to and delete chemicals from the TRI. The criteria for listing chemicals on the TRI are expressly stated in EPCRA, but their broad wording

EXECUTIVE SUMMARY xiii • • requires EPA to exercise judgment in determining whether a chemical is toxic. Decisions to label chemicals as toxic depend on science policy decisions. This case study 'focuses on EPA's recent proposal to add another 313 chemicals to the TRI and provides insight on how EPA currently makes science policy decisions in the context of TRI reporting. Radon in drinking water I The 1986 Amendments to the Safe Drinking Water Act (SDWA) required EPA to regulate eighty-three contaminants, including radon, by June 1989. On July 18, 1991, EPA proposed a drinking water standard for radon of 300 picocuries per liter (pCi/L). The proposed standard is based on the capability of available technology to reduce radon levels in water to less than 100 pCi/L and on detection limits for radon in water. Final drinking water standards for radon have not yet been promulgated due to the controversial nature of the proposal and continued congressional and Science Advisory Board involvement. EPA is under a court-ordered deadline to issue a final standard for radon by April 30, 1995. The assessment of risks associated with exposures to radon in drinking water is highly uncertain. Relevant science policy issues addressed in this case study include the assumption of low-dose linearity for risk extrapolation, the use of surrogate data to estimate risks of nonlung cancers associated with ingestion of radon in drinking water, and the choice of assumed values for exposure variables used in the quantitative exposure and risk assessment. EPA did not use maximally conservative estimates and approaches in calculating the risk attributable to radon in drinking water. Had typical default assumptions been used, the estimated benefits of adopting a standard of 300 pCi/L would have been greater. The uncertainty regarding the risk assessment is illustrated by alternative assumptions, which if used would reduce the published risk estimates by a factor of ten or more. The SDWA does not allow for the consideration of exposures and risks from other sources. Thus, despite widespread dismay that EPA is proposing to devote considerable resources to addressing a small portion of the total potential risk due to radon, EPA is subject to an antiquated, media-specific law that effectively precludes multimedia approaches and relative risk considerations. Conclusions Many risks to human health and the environment are "unprovable." Some risks to human health and the environment are provable. Provable risks can be measured or observed directly and include actuarial risks such as those associated with highway or air travel accidents. In contrast, other risks-such as those associated with low-doses of radiation or exposure to chemicals in the environment-are often too small to be measured or observed directly with existing scientific methods and available resources. Additionally, specific health and environmental effects are often difficult to attribute to specific causes because other competing causes cannot be excluded with reasonable certainty. Such risks are unprovable. However, the fact that a risk is unprovable does not mean that it does not exist. Provable risks can be calculated,

xiv CHOICES IN RISK ASSESSMENT • • whereas unprovable risks can only be estimated through the risk assessment process. Although unprovable risks may be estimated and expressed in probabilistic terms, they are at best educated guesses and do not constitute knowledge or uncontroverted fact. In other words, the ability to produce a numerical estimate of an unprovable risk does not mean that the risk is proven. Science policy issues are unavoidable in, and science policy decisions are essential to, the regulatory risk assessment process. Risks are unprovable because of significant gaps and uncertainties in scientific knowledge, data, and method. When risk assessment is used to estimate unprovable risks, these gaps and uncertainties become science policy issues. Both risk assessors and risk managers make science policy decisions in order to bridge the gaps and uncertainties. Thus, science policy decisions enable the estimation of unprovable risks. Science policy decisions, particularly when compounded, lead to conservative risk assessment results. By design, many science policy decisions lead to risk assessment results that are more likely to overstate than to understate risks. In other words, compensation for the lack of knowledge in the risk assessment process is intended to be protective of public health. Risk assessment results are even less likely to underestimate risk when, as is generally the case, a series of conservative science policy decisions is involved. There is nothing wrong with such science policy decisions and risk assessments unless the nature and extent of the science policy decisions made are not fully disclosed to policy makers, risk managers, the media, and the public. The existence and extent of science policy in risk assessment are rarely fully and fairly disclosed The numerical results of risk assessments tend to be emphasized while discussions of the role of science policy in generating the risk assessment results tend to be de- emphasized. For example, given that many risks are unprovable, there is some probability that, in fact, they are zero. For unprovable risks, science policy decisions enable the estimation of nonzero risks. However, this fact rarely, if ever, is clearly presented in a risk assessment. The lack of disclosure causes risk assessment results to be communicated essentially as fact. Such communication is misleading. Lack of full and fair disclosure of the role of science policy in risk assessment is not the fault of regulators alone. Media communication of risk information tends to omit discussions of science policy because such discussions: (1) do not fit into sound bites; (2) tend to detract from the sensationalism of the risk information; or (3) are not simple to communicate, and subtleties are lost.

EXECUTIVE SUMMARY xv • • Science policy decisions are responsible for regulatory programs and regulatory impacts that are justified on the basis of risk assessment. For regulatory activities and programs that involve or depend upon risk assessment, the science policy decisions made generally determine the existence, extent, and continued credibility of the regulatory activities and programs. As illustrated by the case studies in this report, science policy decisions have been instrumental in determining that: . Used oil should not be classified as a hazardous waste subject to regulation under Subtitle C of the Resource Conservation and Recovery Act; . Unleaded gasoline is not carcinogenic to humans; . Fluoridated drinking water is not carcinogenic, and drinking water should continue to be fluoridated as a public health measure; and . Commercial uses of asbestos could be banned under the Toxic Substances Control Act. In the future, science policy decisions will be used to help determine whether: . Glass wool, food additives, nitrogen dioxide, sulfur dioxide, nitrate ion, phosphorus compounds, and other chemicals will be added to the Toxics Release Inventory; . ' Workplace indoor air quality will be regulated; . Drinking water standards for radon will be made more stringent; and • Remediation of Superfund sites contaminated with trichloroethylene will continue to be as stringent as currently required. As in the risk assessment process, science policy and other assumptions play a significant role in the estimation of benefits and costs associated with regulatory programs. When risks can only be estimated, the benefits of regulatory programs to reduce those risks also can only be estimated, are not verifiable, and depend on science policy-based assumptions. Similarly, cost assessments often depend on assumptions, are uncertain; and cannot constitute uncontroverted fact. An important distinction between estimates of costs and benefits is in the certainty of their existence. Because it is not possible to prove with certainty the existence of unprovable risks, the existence of benefits from regulatory programs also cannot be proven. In contrast, while there is uncertainty involved in cost assessments, such uncertainty is associated with the magnitude of the estimated costs, not their existence.