Characterization of risk is the final step in health risk assessment. This chapter discusses the methods used by the Environmental Protection Agency (EPA) to characterize the public-health risk associated with an emission source. In risk characterization, the assessor takes the exposure information from the exposure-assessment stage (discussed in Chapter 3) and combines it with information from the dose-response assessment stage (discussed in Chapter 4) to determine the likelihood that an emission could cause harm to nearby individuals and populations. The results of this risk characterization are then communicated to the risk manager with an overall assessment of the quality of the information in that analysis. The goal of risk characterization is to provide an understanding of the type and magnitude of an adverse effect that a particular chemical or emission could cause under particular circumstances. The risk manager then makes decisions on the basis of the public-health impact as determined by the risk characterization and other criteria outlined in the appropriate statute.
The elements of risk characterization are discussed here on the basis of several EPA documents, including EPA's Risk Assessment Guidelines of 1986 (EPA, 1987a); Guidelines for Exposure Assessment (EPA, 1992a); a memorandum from Henry Habicht II, deputy administrator of EPA, dated February 26, 1992 (EPA, 1992c) (see Appendix B) (known hereafter as the ''risk-characterization memorandum"); and Risk Assessment Guidance for Superfund (EPA, 1989a) (the "Superfund document").
Elements Of Risk Characterization
EPA's risk-characterization step has four elements: generation of a quantitative estimate of risk, qualitative description of uncertainty, presentation of the risk estimate, and communication of the results of risk analysis.
Quantitative Estimates of Risk
To determine the likelihood of an adverse effect in an exposed population, quantitative information on exposurei.e., the dose (determined from the analysis in Chapter 3)is combined with information on the dose-response relationship (determined from the analysis in Chapter 4). This process is different for carcinogens and for noncarcinogens. For noncarcinogens, the dose estimate is divided by the RfD to obtain a hazard index. If the hazard index is less than 1, the chemical exposure under consideration is regarded as unlikely to lead to adverse health effects. If the hazard index is greater than 1, adverse health effects are more likely and some remedial action is called for. The hazard index is thus not an actual measure of risk; it is a benchmark that can be used to estimate the likelihood of risk.
For carcinogens, excess lifetime risk is calculated by multiplying the dose estimate by a potency factor. The result is a value that represents an upper bound on the probability that lifetime exposure to an agent, under the specified conditions of exposure, will lead to excess cancer risk. This value is usually expressed as a population risk, such as 1 × 10-6, which means that no more than one in 1 million exposed persons is expected to develop cancer. Risk estimates obtained in this way are not scientific estimates of actual cancer risk; they are upper bounds on actual cancer risk that are useful to regulators for setting priorities and for setting exposure limits.
When exposure to more than one agent occurs simultaneously, the cancer risk estimates obtained for each agent can be combined in an additive manner for each route of exposure. Hazard indexes for noncarcinogens may be combined when the agents of concern elicit similar end points of toxicity.
Sometimes, this risk-characterization technique is used to estimate an upper bound on excess lifetime cancer risk to exposed individuals, instead of populations. EPA's Guidelines for Exposure Assessment (EPA, 1992a) (not yet implemented) lists some of the questions that should be answered when considering individual versus population risk. These questions are stated by EPA as follows:
Description of Uncertainty
Analysis of the uncertainty associated with a health risk estimate involves each step of the risk-assessment process: it brings together the uncertainty in emissions and exposure estimates with that of the toxicity dose-response assessment. Table 5-1 lists the uncertainty issues to be addressed at each step of a health risk assessment. Uncertainty analysis can take place at the time of each of those analyses, but because it affects the eventual risk estimate, it is considered part of the final step of risk assessmentrisk characterization.
Several recent documents illustrate EPA's current approach to the analysis of uncertainty associated with health risk assessment, including the Superfund document (EPA, 1989a), the background information document for NESHAPS for radionuclides (EPA, 1989b), the Guidelines for Exposure Assessment (EPA, 1992a), and the risk-characterization memorandum (Appendix B).
Superfund Risk-Assessment Guidance
The Superfund document provides guidance to EPA and other government employees and contractors who are risk assessors, risk-assessment reviewers, remedial project managers, or risk managers involved in Superfund-site cleanup.
(Table continues on following page.)
Section 8.4 of the document "discusses practical approaches to assessing uncertainty in Superfund site risk assessments and describes ways to present key information bearing on the level of confidence in quantitative risk estimates for a site." The document considers three categories of uncertainty associated with site risk assessments: selection of substances, toxicity values, and exposure assessments. Table 5-2 is EPA's uncertainty checklist for Superfund-site risk assessments. Risk assessors are to use the checklist to ensure that they describe adequately the uncertainty in a risk assessment. The document indicates that, although the uncertainty associated with each variable in a risk assessment would ideally be associated with the final risk estimate, a more practical approach is to describe qualitatively how the uncertainties might be magnified or the estimates of risk biased because of the risk models used. This document is being updated.
Uncertainty Analysis for Radionuclide Risk
EPA undertook a more comprehensive, integrated, quantitative approach to uncertainty characterization in the background document for its environmental impact statement on the National Emission Standards for Hazardous Air Pollutants (NESHAPS) for radionuclides (EPA, 1989b). This document includes an extensive presentation of estimates of fatal cancer risks associated with exposure to radionuclides. The estimates were "intended to be reasonable best estimates of risk; that is, to not significantly underestimate or overestimate risks and be of
(Table continues on following page.)
sufficient accuracy to support decisionmaking" (EPA, 1989b). One chapter of the document, however, provides a detailed analysis of uncertainties in the calculated risks that was undertaken by EPA's Office of Radiation Programs for four selected exposure sites, such as a uranium-mill tailings pile in Washington and an elemental-phosphorus plant in Idaho. The stated reason for the uncertainty analysis was that "quantitative uncertainty analysis can provide results that indicate the likelihood of realizing different risk levels across the range of uncertainty. This type of information is very useful for incorporating acceptable and reasonable confidence levels into decisions" (EPA, 1989b).
The EPA uncertainty analysis for radionuclide risks focused on "parameter uncertainty," because it was felt that other sources of uncertainty involving alternative or additional exposure pathways and risk-model structures were "not readily amenable to explicit analysis" (EPA, 1989b). Parameter uncertainties were first modeled as particular probability distributions for each parameter involved in four key components of the radionuclide risk assessments: source terms, atmospheric-dispersion factors, environmental-transport and radionuclide-uptake factors, and risk-conversion (that is, radionuclide-potency) factors. All the distributions pertaining to exposure-related factors were intended to model uncertainty in factor values characteristic of a maximally exposed person. All the distributions pertaining to uptake-related factors were intended to model uncertainty in factor values characteristic of an average individual, except in a set of separate corresponding analyses in which census-based interindividual variabili-
ty in home-residence time was incorporated into the analysis, where it was computationally treated as an uncertain parameter.
Monte Carlo methods were used to propagate uncertainty within contamination-uptake-risk models for calculating radionuclide-specific, increased lifetime risks of fatal cancer to an otherwise typical person who is maximally exposed over a lifetime (70 years) or over some shorter period sampled randomly from the distribution used to characterize home-residence time. The resulting characterization obtained for uncertainty in estimated total increased fatal-cancer risk associated with potential maximal exposure to all radionuclides for an exposure scenario involving a uranium-mill tailings pile is shown in Figure 5-1. The horizontal axis in that figure represents increased risk multiplied by 3.5 × 10-6, which is the geometric mean of the distribution (shown as the solid curve) of risk to an individual maximally exposed for 70 years. (Normalization to the geometric mean value was done simply because all the risk distributions obtained were very close to lognormal.)
The vertical axis in Figure 5-1 represents cumulative probability expressed as a percentage, that is, the probability that the true (but certain) risk is less than or equal to a given, corresponding particular risk value shown on the horizontal axis. The solid horizontal line in the figure corresponds to cumulative probability equal to 50%. The dashed curve in the figure represents estimated risk accounting for less-than-lifetime home residence. In commenting on the substantial difference between the solid and dashed curves for the four types of exposure scenarios considered in this uncertainty analysis, EPA concluded that "it is clear … that many moves are to nearby locations," that "we do not believe that including a factor for exposure duration improves the assessment of maximum individual risk," and that "improper application of such a factor can easily lead to erroneous conclusions regarding uncertainties in the risk assessment" (EPA, 1989b).
Presentation of Risk Estimates
Several methods can be used to display health risk estimates. Some of the terms used most often are listed in Table 5-2. The definitions are from the new 1992 exposure guidelines (EPA, 1992a). Any combination of them can be used to display the risk estimate to either the risk manager or the public. The choice of descriptors is often based on legal mandates. In general, the display includes a table indicating the risk estimated for the exposed population by route of exposure.
1992 Exposure-Assessment Guidelines
EPA's 1992 Guidelines for Exposure Assessment shows a clear presentation of hazard-identification, dose-response, and exposure-assessment information that might be useful in future risk assessments. Risk assessors are to examine the judgments made during the process, the constraints of available data, and the state of knowledge. According to EPA, the risk characterization should include (EPA, 1992a)
The risk-characterization summary should highlight the key points of each step of the risk-assessment process.
EPA is in transition on risk characterization. Besides the exposure guidelines described above, the risk-characterization memorandum (Appendix B) provides guidance on risk characterization and uncertainty analysis for EPA risk managers and risk assessors. The memorandum
addresses a problem that affects public perception regarding the reliability of EPA's scientific assessments and related regulatory decisions… Significant information is often omitted as the results of the assessment are passed along in the decision-making process. … Often, when risk information is presented to
the ultimate decision-maker and to the public, the results have been boiled down to a point estimate of risk. Such "short hand" approaches to risk assessment do not fully convey the range of information considered and used in developing the assessment. In short, informative risk characterizations clarified the scientific basis for EPA decisions, while numbers alone do not give a true picture of the assessment.
A statement attached to the memorandum from the Risk Assessment Council, made up of EPA senior managers, emphasized the following principles:
In implementing that guidance, EPA staff should:
The risk-characterization memorandum goes through each step of risk assessment and outlines the questions to be answered. These are shown in Table 5-1, which suggests several issues that should be addressed to describe the information in each step fully.
Communication of Risk
Risk communication consists of two parts: communication between the risk assessor and the risk manager and communication between the risk-assessment management team and the public. The risk manager often receives the individual and population risk estimates (generally point estimates but occasionally ranges of these estimates) with only a qualitative description of the uncertainties in each. The general public often receives much less informationonly the point estimate or range (without a description of the uncertainty) and the risk manager's decisionalthough far more is available from published sources or on request. In most regulatory situations, the manager's decision and supporting information are published in the Federal Register. In addition, extensive background documents that discuss the risk analysis in much more depth are often available to the public. The public is generally given an opportunity to comment within 30-60 days on the analysis and resulting decision. EPA may adjust a risk assessment on the basis of public comments.