Debate and Commentary
The U.S. environmental protection agency's examination of its risk assessment principles and practices: A brief perspective from the regulated community
In March 2004, the U.S. Environmental Protection Agency (U.S. EPA) published a review of its risk assessment principles and practices. In examining this review, we find a broad, but not particularly deep, review by the authors and areas where the U.S. EPA should strive to further improve the risk assessment process. Recommended areas of improvement include the use of less conservative default assumptions, calculating risk based on populations instead of individuals, and fostering research on multiple stressors and their impacts on estimating risks to ecological receptors.
In many developed countries over the last decade, risk assessment (RA) has become an important approach used by the regulatory and regulated communities to formally collect and analyze public health, ecological, and environmental information. The RA process and resulting output—the estimation of risks to human health or the environment—then become key elements in risk management (RM) decisions and actions (Presidential/Congressional Commission 1997; Stahl et al. 2001). In the United States, the U.S. Environmental Protection Agency (U.S. EPA) has developed a substantial amount of RA guidance for the protection of human health and the environment. Typically these guidelines are implemented through various U.S. EPA program offices (Air, Water, Waste). The application of RA over a nearly 20-year time period has lead to a wide assortment of principles and policies, which were either explicitly or implicitly embedded in a vast array of U.S. EPA regulatory decisions. Over the past 2 years (2002–2004), the U.S. EPA has been examining the application of science in its RA guidance, first at the request of the newly established EPA Science Advisor (early 2002) and subsequently when the Office of Management and Budget (OMB) requested comments (early 2003) on how the federal government uses risk assessment, and particularly the role of precaution in decision making (OMB 2003).
In March 2004, the U.S. EPA published a systematic examination of its RA principles and practices, along with recommendations for improving the process within the agency (USEPA 2004). Given the enormous amount of information on which to draw, the authors noted that an exhaustive examination of every principle, practice, or area of focus within the agency was not possible. The examination, however, is broad and includes sections on human and ecological risk assessments, site- and chemical-specific assessments, and agency-wide recommendations. This paper provides a perspective from a member of the regulated community on the U.S. EPA's analysis and recommendations and includes additional suggestions on areas for improvement. It is not an exhaustive review of the examination, nor is it intended to represent any “official” or consensus position of the regulated community.
We acknowledge from the outset that U.S. EPA's state of practice of RA has evolved dramatically over the past two decades. For example, we have found the use of probabilistic methods, providing a broad range of exposure or hazard values, is slowly replacing simple estimates using the hazard quotient approach. Probabilistic methods may not be needed in every RA, but they can improve the RA by providing a wider and perhaps more realistic view of possible outcomes of exposures and/or effects. The greater use of this tool is occurring both in RA for human health and ecological receptors. We have also seen the increasing use of mechanistic toxicological data to help explain how substances exert their toxicity and why there may be differences in response between laboratory test species (rodents) and humans. This information also can improve the RA. If the mechanism is known whereby a substance causes toxicity in rodents, and the mechanism is not present in humans, then the risk assessor is able to adjust parameters in the RA accordingly, rather than assume that the effect in rodents is predictive of the effect in humans.
We believe that many of the criticisms that have been leveled at the agency related to RA often have less to do with the content of the U.S. EPA's guidance and more to do with its interpretation and application in a given situation. In our experience, there have been instances where the agency has failed to “walk its own talk” as embodied in the prevailing guidance. Sometimes this lack of “consistency” occurs at the level of an individual risk assessor who may favor one interpretation of guidance and apply it without exception, either as a result of their formal training or experience within the agency. In a wider review of U.S. EPA practice, the chemical industry through its trade association, the American Chemistry Council, provided a substantial number of examples where RA implementation has fallen short in this aspect (ACC 2003). We discuss this latter point further in the context of the frequent reliance on default or overly conservative assumptions within the RA.
Also, we acknowledged that there is a degree of flexibility built into U. S. EPA guidance, and in some cases the flexibility is well used by agency risk assessors. Unfortunately, the use of the inherent flexibility is inconsistent across the agency and sometimes ignored in favor of undue “precaution” in RA implementation. We have seen this occur most frequently in contaminated site assessments focused on estimating risks to human health. Interestingly, this conservatism appears most often where the local community is actively involved with the site and where RA is viewed by the public or other groups as a tool to justify leaving contamination in soils, sediments, or groundwater (Foreman 2000). Public involvement is an important element in site clean-up but should not cause risk assessors and risk managers to shy from using the full range allowed in RA guidance and policy. We believe that flexibility in the conduct of an RA is important to both the regulated and regulatory communities because neither guidance nor policy can anticipate all of the possible situations that might be faced by risk assessors. Rigid guidance and unbending policies invariably conflict with the diversity of situations that can arise, and where applied without careful scientific debate lead to poorly informed RAs and potentially to inappropriate RM decisions.
Because the regulated community has experienced this conservatism frequently despite flexible guidance being in place, an attempt was made to remedy the chasm between guidance and practice through a series of workshops in 1997 (U.S. EPA/American Industrial Health Council, Risk Assessment Demonstration Workshops). Focused on RA for protecting human health, the workshops brought together the U.S. EPA's program staff with RA practitioners from both the regulatory and regulated communities. The workshop illustrated that the major problem was not the guidance, but how individual RA practitioners chose to implement or interpret the guidance.
With that as a backdrop, we will look at some specific areas of RA with respect to their evolution, current state, and the potential for continued improvement.
RISK ASSESSMENT AND PUBLIC AND ENVIRONMENTAL HEALTH PROTECTION
As a general matter, the U.S. EPA's examination is long on retrospective of its risk principles and practices and short on the practical impacts. Since the practical impacts are the basis for much of the regulated community's discontent with the U.S. EPA's past and current approach to RA, the examination appears defensive and does little to provide meaningful fixes.
Indeed, one cannot help but note that the authors appear convinced that the U.S. EPA is doing an excellent job in terms of meeting regulatory requirements and being protective of health:
… EPA seeks to adequately protect public and environmental health by ensuring that risk is not likely to be underestimated …
… In general, EPA uses defaults that guard against underestimating risk while also being scientifically plausible given existing uncertainty …
… since EPA is a health and environmental protective agency, EPA's policy is that risk assessments should not knowingly underestimate or grossly overestimate risks. This policy position prompts risk assessments to take a more “protective” stance given the underlying uncertainty with the risk estimates generated …
However, the examination takes an idealistic view of the preference for not underestimating risk, which, in effect, is a euphemism for overestimating risk. Although recognizing that this preference has repercussions, the authors do not recognize the impact that the preference has as a practical matter. The OMB report (OMB 2003) and that of the Executive Office of the President (EOP 1991) detailed clearly that overestimating risks is not warranted in most cases and should not be an overarching principle of conducting RA in the federal government. As discussed in ACC (2003), the Executive Office of the President noted
EPA risk assessments must not intermingle important policy judgments within the scientific assessment of risk. Rather, the choice of an appropriate margin of safety should remain the province of responsible risk-management officials, and should not be preempted through biased risk assessments.
Additional specific recommendations similar to the above were provided in the reports that, had the U.S. EPA implemented them, would have addressed a number of the problem areas identified by the regulated community.
In general, U.S. EPA guidance has a major trickle-down effect for most states and many regional offices that do not have a cadre (or even a single resource) of risk assessment staff. At a practical level, no matter the amount of flexibility inherent in guidance, risks are typically overestimated. Overestimating one type of risk can increase other risks that may not be a focus of the RA (OMB 2003). The regulation of drinking water disinfection by-products is an example of where overestimating risks can lead to poor public policy. In an effort to reduce the potential cancer risks associated with chlorinating (disinfecting) drinking water, the public may face increased exposure to viable microbial agents that cause intestinal and other diseases. As a general matter, overly conservative RAs can lead to overly conservative or inappropriate RM decisions, an inability to distinguish between high-priority and low-priority risks, and for the risk assessors involved, the prospect of having to continually defend the highly conservative nature of their work.
Overall, little effort was made in the examination to address the issue of the conservative nature of the U.S. EPA's risk assessments. On this issue, the agency maintains that it takes a reasonable approach as noted above: “… guarding against underestimating risks while also being scientifically plausible.” We disagree with the U.S. EPA since the outcome (as explained in comments provided to OMB) is often unreasonable (OMB 2003). The OMB report noted that
When federal decision-makers decide the appropriate level of precaution in a specific decision, they need to consider the extent of precaution that is embedded in the methods and assumptions used in the risk assessment. They may also need to consider other factors such as technological and economic feasibility, or more holistic benefit-cost balancing, including considerations of countervailing risks, depending on the overall objective of statutory requirements to protect the public and the environment, and improve societal welfare. Critical to the application of precaution is deciding when additional information is needed before making a final regulatory decision or revising a previous regulatory decision.
The examination makes the point that apparent inconsistencies in the interpretation of guidance can stem from differences in statutory language. Although this is true, the use of conservative defaults are more of a concern and, as the examination suggested when discussing the National Research Council's report of 1994, is the likely basis for many of the differences of opinion about the U.S. EPA's risk assessment practices. This concern and the debate about the adequate level of conservatism continue today. However, with much more experience in the field, one would expect that there be a greater reliance on the maximum use of scientific information in selecting defaults as opposed to the reliance on “plausible conservatism.”
Inherent in how the U.S. EPA chooses defaults are implicit science policies. Although the document attempts to make these clear and advocates for transparent presentation of these in risk assessments, transparency does not change the effect that these policy choices have on eventual risk management decisions. We suggest that the U.S. EPA needs to rez-examine these policy choices in light of the OMB report (2003) and strive to have RAs that balance “precaution” with other interests. The result should be RAs that are balanced, use all available data, and help focus risk assessors and risk managers on identifying those risks of greatest concern to humans and ecological receptors.
UNCERTAINTY AND VARIABILITY
Again, the examination presents a good retrospective on the issues of uncertainty and variability. Although this is useful, an examination of this magnitude (agency-wide) should have taken a more critical approach. Instead, the examination seems to hide behind the framing policy arguments to make science policy choices that center around conservative defaults that “… are unlikely to underestimate risk,” and indeed overestimate risk in light of uncertainty.
In fairness, agency policy in recent years has been moving away from a single point determination to one that more appropriately looks at the range of risks (at least high-end and central tendency estimates), and the agency has guidelines on the use of probabilistic methods. In practice, many of these guidelines are still ignored in favor of the single point high-end estimate as the decision-making tool. For example, it is still typical for RAs to utilize the highest concentration of a chemical contaminant found in media of interest (soil, groundwater, air) as the single point for estimating human exposure. The examination places emphasis on differentiating between screening analyses, bounding estimates, and more complex (presumably more chemical- or site-specific) evaluations. When conducting RAs for contaminated sites, we have found that it is not a trivial effort to get approval from the agency for making alternate (to default) assumptions. In some cases, individual risk assessors are uncomfortable using anything other than the highest concentration point estimate. In such a situation, using the highest concentration may lead the risk manager to believe the exposure to unacceptable levels of a contaminant is more widespread than is actually the case, which may lead to potentially unwarranted risk management actions. Notwithstanding the examination's elaborate evaluation of “compounding conservatism,” it is our experience in contaminated site assessments that the use of anything other than the most conservative U.S. EPA-recommended default is very difficult to change. Having to face the use of conservative default assumptions is a situation that many regulated entities encounter sooner or later. Often regulated entities expend a considerable amount of resources in futile attempts to convince agency personnel, despite the admonitions provided by OMB (2003). This type of situation is also complicated by the differences in application of agency policy and practice among the various regional offices and, more importantly, among the various program offices and individual risk assessors.
We believe that the agency should take a more proactive stance in having risk assessors and risk managers use all of the tools that are available to communicate uncertainty and variability associated with risk assessment results. Despite the examination's discussion of this issue, we do not believe that in actual practice the full range of tools are applied across the agency. We believe that the U.S. EPA should, at a minimum, require risk assessors to report central tendency as well as high-end risk estimates, provide a transparent disclosure of the policy choices in the assumption, and advocate greater utilization of distributional analysis. As the U.S. EPA notes,
Further, probabilistic analysis may avoid some potential problems of upwardly biased estimates of risk from combining uncertainty/variability input variables in a deterministic risk assessment. (By “upwardly biased,” it is meant that, on average, there is more of a tendency for the estimate to be higher than the “true” risk, rather than lower. In individual cases, however, it is most certainly possible that the final, overall risk estimates will underestimate “true” risk for a particular risk assessment.)
We agree that the use of probabilistic analysis may alleviate some of the problems with upwardly biased estimates, but a more holistic application of this method along with greater reduction in the use of default assumptions will be necessary to affect meaningful change.
INFORMATION GAPS IN HEALTH ASSESSMENTS: USE OF DEFAULT AND EXTRAPOLATION ASSUMPTIONS
The examination provides an interesting history and rationale for the use of animal data to develop human criteria. The examination rightfully acknowledges a number of areas of significant uncertainty in the process of extrapolating from animal to human conditions. First, the levels of exposure that are generally encountered in human environmental settings are far below those encountered in laboratory animal settings. Second, there are often considerable toxicokinetic and toxicodynamic differences between the animal test species and humans. Third, this is a situation where one is attempting to predict responses in large human populations based on observations in a limited number of test animals.
With all of the aforementioned limitations, the current animal-testing paradigm has generally served the risk assessment process well. Certainly one could point to instances where animal testing has failed to identify a potentially significant effect in humans. Thankfully, these cases are rare. More typically, criteria developed using the current animal extrapolation procedures overpredict human toxicity. That is to say, the existing approach that incorporates a significant margin of safety in extrapolating from animal models to humans is protective. Indeed, one could argue that the lack of health effects in humans attributable to exposures below these regulatory-derived criteria is evidence of the protectiveness of the current approaches. In addition, as more biomarker studies of exposure—but more importantly biomarkers of adverse effect—are completed, the results are likely to provide such evidence.
Over the years, the U.S. EPA has incorporated into its guidance documents several approaches that should help to reduce uncertainty in the dose-response aspect of risk assessment and bring our understandings of human and animal toxicology closer into alignment. Specifically, U.S. EPA guidance endorses the use of pharmacokinetic modeling as a basis for deviating from standard default uncertainty factors used to develop human toxicity criteria from animal benchmarks. This is an area of science with enormous promise that should be aggressively pursued. Unfortunately, only a handful of compounds has engendered sufficient scientific interest to warrant the levels of human and fiscal resources necessary to develop a predictive physiologically based toxicokinetic model in humans.
Another area in which the U.S. EPA has recently provided guidance on dose-response assessment is the endorsement of benchmark dose levels as points of departure for the development of toxicity criteria (USEPA 1995b). The use of benchmark dose levels makes full use of the toxicity testing data set, as opposed to relying on single dose levels such as a no or lowest observable adverse effect level (NOAEL or LOAEL), for the basis of the criteria. Although the agency has clearly articulated a preference for the use of a benchmark dose level approach, the behavior of some in the regulatory community (including at the regional and state level) has yet to catch up with this position, as some risk assessors retain a strong preference for the use of NOAELs as a point of departure in criteria development. This is an area where the science is sufficiently mature to allow its full incorporation into regulatory decision-making.
We agree with the agency's conclusion that there are a number of areas in cancer dose-response assessment that can be improved. The U.S. EPA's Draft Final Cancer Risk Assessment Guidelines prescribed a number of these areas (USEPA 1999a). Specifically the use of nonlinear extrapolation methods in those cases where a chemical is shown to be nongenotoxic should be promoted. Other efforts to harmonize cancer and noncancer dose-response assessments based more on linear versus nonlinear responses, rather than threshold versus nonthreshold approaches, should be encouraged.
Another noteworthy area that the U.S. EPA examines relative to dose-response assessment is the potential for application of probabilistic methods in toxicity criteria development. The entire area of probabilistic assessment has become well established in the exposure assessment arena (USEPA 2001). The U.S. EPA rightfully acknowledges its utility in hazard assessment in this examination. In our opinion, the primary benefit of probabilistic assessments is the use of all available data in the evaluation. This approach leads to a more robust scientific process versus the value judgment inherent in choosing a single point estimate. The U.S. EPA should promote the use of probabilistic methods in the development of toxicity criteria wherever possible subject to the availability of sufficient data for particular compounds.
The final aspect of hazard assessment that the U.S. EPA examines in this review is the Integrated Risk Information System (IRIS) process by which many of the established human criteria are developed and memorialized. Over the years, IRIS has been a bit of a “whipping boy,” drawing criticism for the imperfection of the method it was trying to implement and for the process by which such implementation was performed. One highlight of the examination is the U.S. EPA's acknowledgment of the need to increase the transparency of the IRIS process and to provide an opportunity for constructive input as toxicity data are assembled, analyzed, and peer reviewed. We encourage the U.S. EPA to open the IRIS process to greater external review and input.
Site- and chemical-specific assessments
Site assessments in the United States, whether under the Resource Conservation Recovery Act (RCRA) or the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA), have proven to be one of the more important areas of RA practice. As the examination notes, a great many of the large-scale assessments (in terms of geography and demographics) result from contaminated site assessments. Although not discussed explicitly in the examination, it is in these situations that the greatest diversity is found among the regional U.S. EPA offices. One example is the larger number of mining sites evaluated in the western United States compared with the eastern United States. Risk assessors in the western United States familiar with evaluating mining sites take a much less conservative view of metals in soils, and their potential risks to humans and ecological receptors, compared with risk assessors that have not conducted mine site evaluations. This type of diversity in practice and experience base among the regional offices can be problematic for companies with operations across the United States. Simply put, what may be acceptable in a RA for one U.S. EPA region is not for another. These differences extend down to the state level, where policy differences even between adjacent states on issues ranging from the toxicity characteristics of a compound to exposure assumptions can lead to major differences in what are considered acceptable levels in environmental media (e.g., soil, groundwater, air). This further complicates the development of the RA and requires a company to be cognizant of the wide array of requirements and regional preferences that are found across the country. However, having a grasp of the mechanics of the specific requirement in a state or region is only part of the problem. It is particularly difficult to rationalize to stakeholders in a local community why a company can “clean up” a contaminated site (or media) to a different level under what appears to be the same conditions. This latter point can become a focus of media and local government attention and, in its worst case, totally derail the risk assessment and remedial processes.
Regulatory decisions need to be based on the best scientific data, and these scientific underpinnings need to be transparent. Although this is a subject much discussed, in practice decisions largely default to the “precautionary principle.” Although the precautionary principle is an important component to consider, it should not be the overriding criterion. We charge the regulatory community (at all levels, including federal, regional, and state levels) to adhere to the U.S. EPA's basic tenet of making decisions based on the best available science and to recognize the significant potential misuse of resources (time, funds, personnel) inherent in a default precautionary principle position.
On the positive side, we appreciate and applaud the flexibility provided in some of the U.S. EPA's guidance to allow different sites to be addressed according to the site-specific conditions. As an overarching principle, we encourage the U.S. EPA (i.e., both at the federal and regional level) to harmonize its RA practice in such a way that retains flexibility for site-specific circumstances. More importantly, this flexibility needs to be practiced and the regulatory community (including federal, regional, and state) should be held accountable for policy decisions that limit such flexibility.
Chemical-specific RA and ecological risk assessments (ERAs) are mostly a result of registration requirements under the Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA). In this case, there is a plethora of environmental fate and effects tests that must be conducted that become the database on which the RA and ERA is based. In this situation, the agency seems more and more inclined to take a conservative approach to the RA and ERA despite the large data sets. We believe this conservatism stems more from the public and media scrutiny given to the use of chemicals to protect the human food supply than it does from a need for greater scientific rigor. The benefit of using these substances is that the food supply is protected against disease and pests. With that protection comes risks, leading individuals and society to trade-offs (risks vs benefits) that can trigger substantive scientific and emotional debate. Over the last 10 years manufacturers of crop protection chemicals have found it more and more expensive to register these products, largely because of the expense involved with the testing requirements. The testing requirements are a necessary element of regulatory oversight; however, there needs to be greater flexibility in what hazard and fate data are acceptable and how a RA or ERA is developed for the registration process.
The history of ERA within the U.S. EPA stems from a recent and evolving practice. As noted in the examination, ERA was not a formal approach within the U.S. EPA until the mid 1980s and more precisely not until the publication of the U.S. EPA Framework for Ecological Risk Assessment in 1992 (USEPA 1992).
Although not addressed in great detail in the examination, the issue of what to protect and at what level only recently found its way into agency guidance and reports (USEPA 1995a, 1997b; Barton 2001). The difficulties faced by ecological risk assessors within the agency in the past stemmed as much from the lack of technical tools as it did from the lack of clear, specific guidance from Congress and others on risk management. As noted earlier, agency scientists conducting human health risk assessments did not face this lack of guidance, as protection goals provided by Congress were specific (e.g., 1 × 10−6 for cancer endpoints), although not mandatory, in all situations. Also not discussed in the examination is that ecological risk assessors in the agency today are less challenged by the lack of tools than by the lack of clear (or consistent) guidance from the public and elected officials on what receptors are important to protect and at what level. Today, ecological risk assessors have the benefit of a peer-reviewed, widely utilized framework for ERA. In addition, ecological risk assessors have software and similar tools that allow them to conduct probabilistic analysis, make estimates of population-level risks, and employ sophisticated habitat and geographical information system programs that readily display the areas under assessment.
Over time, the vast majority of ERAs conducted by the U.S. EPA have been in three main categories: premanufacturing notices (PMNs) under the Toxic Substances Control Act (15 U.S.C. 2601–2692), chemical or pesticide registrations under FIFRA, and contaminated waste sites under RCRA or CERCLA (CENR 1999). Of these, numerically, there tended to be more ERAs directed at PMNs and contaminated waste sites than for any other application. ERAs for PMNs were more or less simple applications of the hazard quotient approach, or, due to the lack of large data sets and time constraints, simple applications of structure-activity relationships. Using these simplified approaches is not a significant problem since the majority of companies applying for the PMN wish to know fairly quickly whether or not the application is approved. As they enter the application process, companies know that numerous extrapolations and default assumptions will be used in the RA. Where the PMN is not approved for lack of data, a company has the choice to generate the data or simply drop the application. Ultimately, it is a business decision, and not a scientific one, whether to fund the generation of data, which may explain why most private-sector risk assessors are not more engaged in scientific debate with agency scientists on the issue of assessments for PMNs.
As the examination notes, over the years the ERAs for contaminated site assessments, particularly under CERCLA and RCRA, have been some of the most fertile areas for the evolution of the science and practice of ERA within the agency. With the publication of ERA methods for Superfund (USEPA 1997a), much of the practice was memorialized in guidance that received substantial review by the regulated community and other stakeholders. Looking back, one of the bright spots in the development of ERA practice and policy for contaminated sites has been the willingness of agency scientists in the Superfund and other program offices to engage scientists from academia, industry, and elsewhere in healthy scientific debates. This certainly improved the ERA guidance and its implementation across contaminated site assessments. Having this published guidance, along with the ERA Framework and the U.S. EPA general guidelines for ERA (USEPA 1998), provided ERA with a wider recognition and credibility across the program offices. This recognition, wider use, and credibility also appeared among the private sector (Pittinger et al. 2001) where ERA was becoming a new, useful tool for evaluating products and emissions.
Our experience has been that the agency places much more emphasis and gives greater credence to the ERA, particularly for contaminated waste sites, than in most other areas. We believe this stems from the greater application of ERA at waste sites than before, the wider publication of waste site ecological evaluations in respected scientific journals (i.e., Environmental Toxicology and Chemistry), and the frequent interactions at a scientific level between ERA practitioners in the U.S. EPA and the regulated community.
Although not highlighted in the examination, the agency scientists in Superfund developed and published a set of ecologically based risk management principles for the program (U.S. EPA 1999b), a bold and welcome step. For the first time, risk assessors and decision-makers in the agency and the regulated community clearly understood what was important to protect. This set of principles has given greater relevance to ERA and afforded risk managers within the agency and the regulated community clear expectations on how to manage risks to ecological receptors at contaminated waste sites.
The SETAC Pellston-style workshops have been exceptionally valuable for the scientific interactions and debates among ERA practitioners in the U.S. EPA, the regulated community, academia, and environmental organizations. Workshops whose genesis was the open discussions at SETAC annual meetings and other venues include those that addressed multiple stressors; ecological risks to wetlands; endocrine disruption in wildlife; the use of sediment quality guidelines; and most recently, the valuation of ecological resources. Recommendations from these workshops have had an important influence on ecological risk assessment practitioners in the public and private sector and will continue to do so in the future.
The recognition that chemical and other stressors do not act singly has been another important step forward in the ERA process (Foran and Ferenc 1999), although this too was not highlighted in any great detail within the examination. Multiple stressors can be a combination of chemicals (or a chemical mixture), but more specifically are the suite of chemical, physical, and biological stressors encountered by ecological receptors on a daily basis. It is in this area that ecological risk assessors are lacking adequate tools with which to estimate risks from disparate stressors. We believe this is an important area in need of research and development, and we encourage the U.S. EPA to address it in future program funding.
The heart of the RA is the risk characterization, the step in which the exposure and effects (or hazard) data are integrated. The examination describes how the U.S. EPA characterizes risks from carcinogens and noncarcinogens. Specifically, it describes the concept of high-end exposures and the goal to be protective of sensitive individuals within a given exposed population. In most regulatory contexts, the U.S. EPA bases its decision-making on what constitutes acceptable risk to humans using calculations specific to an individual, rather than a population.
The focus on individual- instead of population-level risks is not well discussed in the examination. In fact, the examination is conspicuously silent on population-based risk assessment. Because of the agency's predisposition to the use of risk descriptors that focus on the individual, the concept of population-based risk characterization is often overlooked. This focus on the individual should not be applied to all risk assessments and risk management decisions in all cases. In fact, there are instances, including elements of the Residual Risk Program in the Clean Air Act, where the U.S. EPA has a mandate to consider the risk to populations as part of its decision-making. Population-based descriptions of risk can take many forms, from simple descriptions of the number of individuals exceeding a given risk level to actual calculations of disease incidence within the exposed population. Focusing a risk assessment on the population rather than the individual is controversial, as it can be difficult to argue against the need to protect individual humans. Nevertheless, the fundamental benefit of a population-based description of risk is that it acknowledges the legitimacy of the exposed population size in risk characterization. This aspect of risk characterization should not be forgotten, despite its controversial nature, and is conspicuous by its absence in the U.S. EPA examination.
The examination details a number of recommendations that may improve risk assessment practice within the agency. We do not believe these recommendations will, in and of themselves, solve all of the problems we have experienced. The examination is a good first-step at self-reflection, but it would have greater credibility had it been conducted by scientists outside of the agency. We encourage the U.S. EPA to consider this point in determining how best to address the issues raised in the examination.
We have seen that many of the important advances in RA have resulted from taking on challenges that required better science, tools, and ultimately applications thereof. For example, the desire to improve exposure assessments for humans and ecological receptors lead directly to the development of methods and guidance for probabilistic techniques. The ability to more fully characterize exposure and hazard is valuable to scientists and policy makers faced with protecting diverse populations of humans and ecological receptors. We have stated earlier that this is an area in need of additional research and development.
The more recent concern about chemicals that can disrupt or modify endocrine functions in humans and ecological receptors has lead to improved physiological, biological, and genetic probes (tools), as well as the guidance that goes along with the applications of the tools. Similarly, we anticipate the development of new tools and guidance to address the next frontier of risk assessment that deals with cumulative or integrated exposures (i.e., from multiple chemicals, multiple pathways, and across time).
Knowing that the above challenges have fostered improvements to the science, we propose the following specific recommendations for improving the RA process:
Promote the use of mode-of-action as a consideration when extrapolating from animals to humans;
Encourage the use of predictive physiologically based toxicokinetic modeling as a way to move from the current default uncertainty factors used in extrapolating from animal data to human criteria;
Promote the use of benchmark dose level methods to establish points of departure for human criteria development;
Foster the use of probabilistic methods in dose-response assessment and sponsor its research and development;
Follow up on the work conducted through SETAC-sponsored workshops on multiple stressors (Foran and Ferenc 1999) and sponsor research to close this information gap;
Explore methods to incorporate economic valuation techniques, public desires, etc., into the ERA and the risk management decision;
Work to reduce the high variability in the application of RA guidance among regions and individuals;
Include specific consideration of exposed population size in risk characterization—the number of people exposed to a given risk level does matter; and
Develop tools and guidance on cumulative and integrated risk assessments.
We welcome the opportunity to review the agency's examination and look forward to fruitful scientific discussions on these issues with risk assessment practitioners within and outside of the U.S. EPA.
Disclaimer—The opinions expressed in this article are those of the authors.