The National Research Council report, "Science and Judgment in Risk Assessment" (
1
), mandated in the 1990 Clean Air Act Amendments, contains a comprehensive analysis of the state of the science of cancer risk assessment and ways in which it could be used to improve decision-making at the U.S. Environmental Protection Agency. The report reflects congressional concerns about the evolving state of risk assessment, the uncertainty inherent in risk estimates based on theoretical models, the divergence of opinion within the scientific community on the merits of the underlying scientific evidence, and the large economic costs at stake in regulating the 189 air pollutants in question. In light of these concerns, Congress requested the EPA to engage the National Academy of Sciences in a review of 1) the methodology used by the agency to determine the carcinogenic risks associated with exposure to the chemicals listed, 2) improvements in such methodology, 3) the techniques used for estimating and describing the carcinogenic potency of hazardous air pollutants, 4) the methods used for estimating exposure to such agents, and 5) the methods for assessing the risks of adverse health effects other than cancer, for which safe thresholds of exposure may not exist. The act also requested that EPA comply with the academy's methodological recommendations or explain in the Federal Register why it failed to do so.
The review of risk assessment methodology called for in the act, carried out under the aegis of the National Research Council by a 25-member committee, resulted in the Science and Judgment report, a 600-page publication containing some 70 specific recommendations for improving EPA's policies and procedures for assessing risks to human health. The report covers a wide variety of topics, ranging from short-term methodological refinements to long-term research strategies, but its major recommendations focus on the following cross-cutting questions, which pertain to virtually every step in the risk assessment process:
* Default options: Are the default options used by EPA to guide its risk assessments in the face of scientific uncertainty adequately justified, and does the agency have clear and consistent principles for departing from their use?
* Data needs: Are the data currently available to EPA sufficient to enable it to generate risk assessments that are adequately protective of public health and that are also defensible scientifically?
* Validation: Has EPA adequately validated its risk assessment methods and shown them to be consistent with currently available scientific information?
* Uncertainty: Has EPA taken sufficient account of the need to consider, describe, and explain the inevitable uncertainties inherent in its risk assessments?
* Variability: Has EPA adequately considered the extensive variation among individuals in exposure to toxic agents and in susceptibility to injury from such exposure?
* Aggregation: Have the possibilities of interactions among pollutants and of multiple pathways of exposure been adequately addressed by EPA in assessing risks to human health?
In response to each of these questions, the committee's major findings and recommendations are summarized below.
Default options
. As recognized in the 1983 National Research Council report, "Risk Assessment in the Federal Government: Managing the Process" (
2
), inference guidelines are useful, if not necessary, in selecting among the different policy choices that must be made when assessing risks in the face of scientific uncertainty. Given this fact, the committee considered EPA to have acted reasonably in formulating default options for the purpose, and it recommended that the agency continue to use default options when dealing with uncertain models and mechanisms in risk assessment. In addition, however, the committee recommended that EPA improve its use of defaults by 1) explicitly identifying each default option when used, 2) reevaluating the options regularly and clearly stating the scientific and policy basis for each, 3) giving greater formality to the criteria for departing from default options to provide enhanced guidance to the public and to lessen the possibility of ad hoc, undocumented departures that could detract from EPA's credibility, and 4) continuing to make the fullest possible use of peer review, by means of the Science Advisory Board, other expert groups, and workshops, to ensure that its risk assessment decisions have access to the best science available through full public discussion and participation by the scientific community.
Validation.
In view of the fact that the uncertainty and predictive accuracy of the various methods and models used by EPA in characterizing emissions, identifying hazardous agents, and evaluating dose-response relationships are not adequately known or explained in many instances, the committee recommended that EPA: 1) more rigorously establish the uncertainty and predictive accuracy of the models and data it uses; 2) develop guidelines for the amount and quality of emission information required for particular risk assessments and for assessing human exposures; 3) evaluate critically the models used for the purpose, with attention to such state-of-the-art techniques as stochastic dispersion models; 4) use bounding estimates for screening assessments to determine the need for further levels of analysis (e.g., the development of distributions of exposure based on acual measurments, modeling, or both); 5) continue to explore and, when appropriate, to incorporate into such assessments pharmacokinetic models of the relation between exposure and biologically effective dose; 6) develop biologically based quantitative methods for assessing the risks of adverse effects other than cancer that may result from exposure to chemicals, taking into account relevant causal mechanisms, as well as possible interindividual variations in susceptibility; and 7) continue to use, as a risk-characterization potency metric, upper-bound estimates of the probability of developing cancer from a lifetime of exposure, but supplementing this metric whenever possible with other descriptions of cancer potency that more adequately reflect the uncertainty inherent in the estimates.
Data needs
. Because EPA lacks the exposure and toxicity data needed to assess the health risks associated with all 189 chemicals identified as potentially hazardous air pollutants in the 1990 amendments to the Clean Air Act, and because it has yet to define how it will determine the types, quantities, and quality of data needed, the committee recommended that EPA: 1) compile an inventory of the chemical toxicological, clinical, and epidemiological literature on each of the 189 chemicals in question; 2) screen the chemicals by an iterative type of strategy to set priorities for assessing health risks, identifying data gaps, and developing incentives for expediting the generation of data by other concerned parties, public and private; 3) consider mobile and indoor sources of hazardous air pollutants, as well as stationary sources, as the former may be more important in some cases than the latter; 4) consider explicitly all direct and indirect routes of exposure to such pollutants, including ingestion and dermal absorption; and 5) develop a two-part scheme for classifying evidence on carcinogenicity that would incorporate a narrative evaluation as well as a simple classification, each of which would address the strength and relevance of the evidence.
Variability
. Although the health risks from a given pollutant may vary appreciably among individuals and among populations, the committee noted that such variability has generally received little consideration in EPA's risk assessments, in part because of limitations in the availability of pertinent data. To remedy this situation, the committee recommended that EPA and other federal agencies sponsor research into the causes and extent of interindividual variations in susceptibility to cancer and other diseases and that the results of such research be used to refine estimates of risks to individuals and population groups. In addition, EPA was recommended to: 1) adopt a default assumption for interindividual differences in susceptibility; 2) increase its efforts to validate or improve the default assumption that humans on average are similar in susceptibility to one another and to the most sensitive laboratory animals tested; 3) include in its guidelines, data permitting, a clearly stated default assumption of nonthreshold, low-dose linearity for genetic effects to enable estimates of the genetic risks to future generations from chemical exposures; 4) assess risks to infants and children specifically whenever their risks appear likely to exceed those of adults; and 5) distinguish rigorously between individual variability and other sources of uncertainty in each component of risk assessment.
Uncertainty
. Unless each of the various sources of uncertainty typically inherent in a risk assessment is adequately identified and explained, a decision-maker acting on the assessment may not know the extent of conservatism, if any, provided by the assessment. The upper-bound point estimates of risk typically computed by EPA, for example, do not convey the degree of uncertainty inherent therein. To explicate the uncertainty in EPA's risk assessments and to help inform the agency and the public more fully of the extent of conservatism embedded in such assessments, the committee recommended that EPA: 1) subject its assessments to formal uncertainty analysis, which could show where additional research might resolve major uncertainties; 2) consider in its risk assessments the limits of relevant scientific knowledge, as well the need to identify and minimize errors of underestimation or overestimation; 3) develop guidelines for quantifying and communicating the uncertainties inherent in each step of the risk assessment process; and 4) inform risk managers of the uncertainties inherent in risk assessments, and not merely give them a single point estimate or range of numbers for characterizing a risk.
Aggregation
. Because pollutants emitted into the air may be deposited on other exposure media, such as water, soil, and vegetation, and because EPA had not yet indicated whether it would consider multiple routes of exposure under the 1990 amendments, the committee recommended that the agency: 1) consider using appropriate statistical procedures for aggregating cancer risks from exposure to multiple compounds and from exposure to chemicals via multiple media and routes; and 2) in analyzing animal bioassay data, estimate and sum the cancer potencies for each individual tumor type.
Risk communication
. Because certain expressions of probability are subjective, and because the reliability of a single number or range of numbers is subject to misinterpretation, it was recommended that risk managers be given descriptive as well as mathematical characterizations of risk.
An iterative approach
. In view of the fact that EPA does not have sufficient resources and data to perform a full-scale risk assessment on each of the 189 chemicals listed in the 1990 amendments and that no such assessment may actually be needed in many instances, the committee recommended that EPA seek to develop a tiered, iterative assessment strategy. Such a strategy would call for the use of relatively inexpensive screening techniques for each chemical at the outset and would proceed stepwise to more resource-intensive methods only insofar as further analysis was warranted by health, economic, or other concerns, thus enabling the most cost-effective use of limited resources, prioritizing needs for further research, and providing incentives for regulated parties to undertake assessments independently.
It will be evident from the foregoing summary that the report endorses EPA's overall approach for assessing risks to human health as fundamentally sound, it will also be apparent that the report suggests ways to improve nearly every step of the agency's risk assessment process. Although generated in response to the mandate of the 1990 Clean Air Act amendments, the report presents recommendations that are applicable in virtually every risk assessment context. If acted upon, therefore, the recommendations could lead to major gains in the cost effectiveness and reliability of risk assessment efforts not only throughout EPA but in other agencies as well. Furthermore, to the extent that many of the recommendations identify needs and opportunities for additional research, they represent invitations and challenges to the scientific community at large.
Arthur C. Upton
Former Director
Institute of Environmental Medicine
New York University Medical Center
REFERENCES
1. National Research Council. Science and judgment in risk assessment. Report of the Committee on Risk Assessment of Hazardous Air Pollutants. Washington, DC:National Academy Press, 1994.
2. National Research Council. Risk assessment in federal government: managing the process. Washington, DC:National Academy Press, 1983.
Last Update: June 18, 1998