For many years, the U.S. Environmental Protection Agency ("EPA") has been criticized for adopting overly conservative human health risk assessment policies. In particular, EPA's Office of Solid Waste and Emergency Response ("OSWER") consciously adopts very conservative assumptions that overstate (or at least assure that EPA will not understate) the human health risks posed by sites contaminated with industrial chemicals. This article describes the human health and ecological risk assessment policies that EPA has adopted to address such sites.

EPA risk assessment policies are used to determine whether and to what degree to require cleanup under the Comprehensive Environmental Response, Compensation, and Liability Act ("CERCLA," commonly known as the "Superfund" law) and the Resource Conservation and Recovery Act ("RCRA"). Section I of this article provides an introduction to the origins and nature of EPA's risk assessment policies under Superfund and RCRA. Section II provides an overview of the key concepts for cancer and noncancer human health risk assessment and traces the history of these policies. Section III describes recent EPA policies that might lead to more realistic human health risk assessments. Section IV describes how EPA may soon adopt conservative risk assessment policies for assessing environmental risks. Before turning to EPA's policies, the article provides some recent background to risk assessment issues.

EPA's risk assessment policies substantially affect decisionmaking under various statutes and may affect political debates over the merits of federal environmental laws. For example, EPA recently drafted a study of the first twenty years of implementing the federal Clean Air Act. The draft study concluded that the benefits exceeded the costs by $22.5 trillion, a factor of seventy. But altering just a few of the human health risk assessment policies employed by EPA would reduce the calculated benefits by a factor of ten. In 1996, moreover, EPA proposed to: (1) revise its general policies for risk assessments that address carcinogenic chemicals; and (2) establish general policies for ecological risk assessments.

In 1996, Congress repealed in part the 1958 Delaney Clause, which banned even trace residues of carcinogenic pesticide chemicals in processed foods. But Congress also required EPA to establish new risk assessment policies for pesticides. The new policies will consider: (1) aggregate exposures to particular chemicals; and (2) different eating patterns and higher sensitivity of infants and children, including cumulative risks from pesticides and other chemicals having similar biological activity.

Congress also amended the Safe Drinking Water Act to revise the standard for EPA to set maximum contaminant levels ("MCLs") for drinking water. Congress required that EPA: (1) set MCLs to maximize benefits at a cost justified by the benefits; and (2) specify in its MCL risk assessments population risks, upper-bound or lower-bound estimates, significant uncertainties, and effects on sensitive populations. The risk assessment policies developed by EPA to implement these specific statutory mandates may in the future be extended to other regulatory programs.

In 1976, Congress quietly enacted the hazardous waste provisions of RCRA, focusing primarily on developing a regulatory program to avoid contamination of land in the future. In 1980, Congress enacted CERCLA to address sites already contaminated with industrial chemicals. EPA contemporaneously adopted regulations to implement the hazardous waste provisions of RCRA. In 1984, Congress amended RCRA, requiring hazardous waste treatment, storage, or disposal facilities to undertake "corrective action" regarding releases of hazardous constituents from solid waste management units.

During this time, Congress, EPA, and the public perceived serious risks of cancer from exposure to contaminated sites. Some scientists assumed that exposure to chemicals in the environment caused between sixty and ninety percent of all human cancers. The Administrator of EPA also believed that exposure to such chemicals posed substantial cancer risks. Current information, however, suggests that all environmental "pollutants could contribute to 2 percent of fatal cancers, mainly of the lung and bladder." Risks posed by contaminated sites likely represent an extremely small percentage of overall human cancer risks. To implement the provisions of CERCLA and RCRA requiring cleanup of contaminated sites, EPA had to develop policies to assess and to manage the risks. EPA adopted conservative methods and assumptions, particularly in regard to the potential for human exposure to hazardous substances and to the hazardous constituents of wastes that are present at contaminated sites. The conservative methods and assumptions adopted by EPA create the appearance of significant risks. These apparently significant risks engender public fears that are hard to quiet, even if more realistic risk assessments are developed and presented.

For example, EPA conservatively assumes that industrial chemicals are carcinogenic to humans based on their potential to induce tumors in laboratory animals. The animals are fed extremely high concentrations of the chemicals and EPA includes in its evaluations tumors that may rarely develop into cancer. EPA then derives quantitative estimates of risk from the animal tumor data. EPA conservatively adopts "upper bound" or "upper confidence limit" values. These high-end estimates account for limited or poor animal data, for biological differences between humans and animals, and for the varying sensitivity among human populations and animal species.

At Superfund and RCRA sites, humans are normally exposed to substantially lower concentrations of chemicals than are fed to laboratory animals. EPA therefore must apply mathematical models to extrapolate risks at low-concentration exposures from the risk estimates derived from the high-concentration animal studies. This is referred to as "dose-response assessment." EPA's dose-response models conservatively extrapolate risks by assuming that there is no "safe" threshold of exposure to carcinogens and that the risks increase in linear proportion to the exposure concentration. These models are "inconsistent with a variety of secondary mechanisms of carcinogenesis now identified."

EPA also has adopted conservative policies to address noncancer health risks. Again, risk data are derived from animal studies. But here, EPA uses the data to establish a "safe" concentration level for noncancer risks. EPA then applies to these levels conservative "safety factors" (sometimes exceeding a factor of one thousand) to account for inadequate data, for biological differences, and for varying sensitivity. These conservative safety factors may be applied without regard to the severity of the toxicological effects of concern.

Next, EPA must estimate the potential for humans to come into contact with chemicals that are present at particular Superfund and RCRA sites. EPA conservatively assumes that normal children and adults will have access to contaminated industrial property and will indirectly ingest large quantities of the most contaminated soils 350 days a year for 30 (and previously for 70) years. Just one of these assumptions, the amount of soil ingested per day, may overstate better estimates of risk by at least a factor of two and as much as a factor of twenty-five. When multiplied together to assess the risks, EPA's "worst-case," "single-point" exposure assumptions may compare incomparable data. The risk estimates thereby derived may exceed intended conservative upper-bound estimates by orders of magnitude and may overstate "best-estimate" or "probabilistic" measures of risk by a factor of thousands or even millions. In the soil ingestion example, the estimates of risk may be reduced by a factor of 1,167 just by using more reasonable single-point estimates of the concentration of chemicals in soil, the rate of soil ingestion, and the frequency and duration of exposure.

Further, EPA excludes from its risk assessments any consideration of institutional controls, fences, and other measures that currently exist or that might be employed. These measures may in reality eliminate or substantially reduce the estimated risks. But even without regard to such measures, the vast majority of EPA's cleanup decisions are based on potential future risks posed by residential land uses that would require dramatic changes in housing patterns.

Unlike for human health risk assessment, EPA has yet to develop significant guidance to address the risks posed to other organisms and to ecosystems. Consequently, EPA has developed ecological risk assessment policies on a site-by-site basis. The ecological risk assessments conducted at Superfund and RCRA sites have been highly inconsistent.

HUMAN CARCINOGEN RISK ASSESSMENT POLICIES

The Nature Of Cancer

Cancer is defined as "any of various malignant [tumors] marked by the proliferation of [relatively undifferentiated] cells that tend to invade surrounding tissue and [transmit] to new body sites." Until recently, the cellular mechanisms of genetic mutation and of cancer formation have been poorly understood. Inherited physiological traits play a role in cancer development, in part by providing to subsequent generations defective genes responsible for cellular systems that prevent or repair damage to the genetic material involved in cancer. Aging and other cellular processes also contribute to cancer. Finally, exposure to external agents, such as infectious microorganisms and industrial chemicals, may cause genetic mutations or may affect cellular expression.

The Origins Of EPA's Carcinogen Risk Assessment Policies

Quantitative carcinogen risk assessment traces its origins to epidemiological studies of Japanese survivors of American atomic bombs detonated at the end of World War II. In the 1950s, the Atomic Energy Commission began to assess the carcinogenic potential of ionizing radiation from use of nuclear power reactors.

In the early 1960s, the U.S. Food and Drug Administration ("FDA") began to address the carcinogenic potential of agricultural chemicals in order to evaluate the risks of cancer from the presence of those chemicals in food. In the 1970s, the FDA began to perform more sophisticated assessments, attempted to address limitations on analytic detection capabilities, and ultimately settled on a policy that a lifetime risk of developing cancer of one-in-one-million, or 10-6, represented an "insignificant carcinogenic risk."

In the early 1970s, EPA developed seventeen simplistic "cancer principles" for use in proceedings to suspend three widely used pesticides. Following trenchant criticism of these principles by the scientific and medical community, EPA and other federal agencies began to develop more sophisticated methods of assessing cancer risks. An integrated federal approach to the science of risk assessment was initiated in 1979 and a unified approach was adopted by the National Research Council ("the NRC") in 1983. Additional analysis of the science of risk assessment and recommendations for risk assessment policy were prepared in 1985 by the Office of Science and Technology Policy in the White House.

The NRC identified four discrete steps required for quantitative risk assessment. Each step itself typically represents dozens of different data evaluations and policy assumptions. The first step is hazard identification, or determining the contaminants that may be present in sufficient quantities to pose a hazard to humans or to other receptors. The second step is dose-response assessment, or evaluating the relationship between exposure to contaminants at particular concentrations and the incidence of cancer or other adverse effects. The third step is exposure assessment, or determining the likelihood of actual exposure to particular concentrations of contaminants. The fourth and final step is risk characterization, or specifying the likelihood that cancer or other adverse effects will result, by combining the dose-response and exposure assessments and by assessing the strength of the evidence and the uncertainties of assumptions.

EPA's Framework For Carcinogen Risk Assessment Under Superfund

By 1986, EPA adopted the general framework established by the NRC. For Superfund sites, however, EPA originally developed a slightly different process for assessing the "baseline" human health risks. EPA addressed hazards posed by the many chemicals that might be present by selecting indicator chemicals having the greatest relative toxicity, as determined from animal studies. The highest concentrations of these chemicals were then measured from soil samples or predicted at "human exposure points." Human exposure points were identified pursuant to assumptions regarding the environmental fate and transport of chemicals and the potential for human contact. EPA then compared these concentrations to applicable or relevant and appropriate requirements ("ARARs") in order to determine if cleanup was required and to characterize the risks. Although the 1986 policy included risk management principles in the baseline risk assessments, EPA later separated these conceptually distinct roles.

To characterize the risks, EPA focused on the additional lifetime risk of developing cancer. Further, EPA has required estimates of chemical concentrations at "the location of highest individual exposure to the general public," rather than attempting to model exposures that might be expected for the general population. EPA also has required use of simplistic, "reasonable worst case" fate and transport models to predict exposure point concentrations at particular locations, because of uncertainties in data and methods and because of the high costs associated with more sophisticated fate and transport modeling. In 1989, EPA changed its terminology and redefined its reasonable worst case exposure assumptions as the "reasonable maximum exposure (RME) expected to occur under both current and future land-use conditions. The reasonable maximum exposure is defined here as the highest exposure that is reasonably expected to occur at a site." Nevertheless, by 1991 EPA required that these conservative single-point measures be multiplied, making the assessed risks even more conservative.

EPA risk managers were to require cleanup when conservative, single-point, deterministic estimates at human exposure points ranged from 10-4 to 10-6 (originally 10-7) increased lifetime risk of cancer. Although early guidance discouraged EPA risk assessors from assuming improbable future land uses, later policy encouraged EPA risk assessors to assume future residential land uses. In order to reject remedies that would impose substantial and unjustified costs, EPA thus has been required to recharacterize as "unreasonable" land use assumptions that it should never have considered "reasonable" in the first place.

EPA's Framework For Carcinogen Risk Assessment Under RCRA

Since 1982, EPA has required corrective action to address hazardous constituents released to ground water from certain "regulated units" at hazardous waste treatment, storage, or disposal facilities ("TSD facilities"). Corrective action is required when hazardous constituents are detected above background levels in ground water monitoring wells. TSD facilities must reduce the concentrations of hazardous constituents to meet a "ground water protection standard" ("GWPS") established by EPA (or an authorized state) at the downgradient boundary of the regulated unit. The GWPS is set at: (1) background levels; (2) any applicable "maximum contaminant level" (or "MCL," i.e., "safe" level for human drinking water ingestion established under the Safe Drinking Water Act); or (3) "alternate concentration limits" ("ACLs"). ACLs are developed through site-specific risk assessments.

Similarly, EPA allowed TSD facilities to "clean close" many of their regulated units, i.e., to remove all waste residues and contaminated soils in order to avoid additional requirements to maintain the units after closure. For clean closure, EPA ultimately required TSD facilities to demonstrate the lack of health risks by assuming human exposures at or within the waste unit boundary. If EPA-approved national regulatory "action levels" did not exist for the hazardous constituent of concern, EPA required clean-closing owners and operators to achieve background levels or to submit constituent-specific toxicity data and site-specific risk assessments. EPA also rejected the use of risk assessments based on institutional controls that might prevent exposures. Recently, EPA indicated that it will accept site-specific CERCLA risk assessments that calculate a point of exposure beyond the waste unit boundary for purposes of determining clean closure under RCRA.

To implement the 1984 requirements for corrective action to address releases from solid waste management units, EPA required limited, site-specific risk assessments to determine whether "interim measures" should be adopted to reduce risks. TSD facilities were required to analyze: (1) the sources of hazardous constituents; (2) the environmental fate of the constituents; and (3) the potential receptors exposed to the constituents. EPA also suggested using existing regulatory protection standards where they were available during the RCRA "Facility Investigation" ("RFI") phase, but allowed use of ACLs where available or development of site-specific risk assessments if national standards were not applicable.

Following the RFI phase, EPA required studies of "Corrective Measures" ("CMSs"), i.e., analysis of alternative remedies. Originally, a CMS was required to include environmental and human health risk assessments. But EPA subsequently limited the need to perform site-specific risk assessments. Unless EPA believed that site-specific exposure conditions posed a threat, owners and operators were required to prepare a CMS when hazardous constituent concentrations exceeded generic risk "screening" levels, or "action levels." Nevertheless, EPA has allowed owners and operators to submit site-specific risk assessments in order to establish that a CMS is not required. Action levels are established for corrective action using existing regulatory standards or by performing conservative assessment of risks consistent with EPA's generic carcinogen and human health risk assessment guidelines. In 1990, EPA proposed to establish action levels for soils by assuming residential land uses and by calculating exposures to contaminated soils by assuming long-term, 365-days per year consumption of soil or dermal absorption. These action levels also applied to buried soils that do not pose ingestion risks but may contaminate ground water. In contrast, EPA proposed to use action levels for surface water contaminated by constituents from contaminated sites by reference to ecological risk management levels established under the Clean Water Act.

Finally, in contrast to its risk management policies under Superfund, EPA proposed to allow the use of "conditional" remedies. Facilities could maintain or install institutional controls, and subsequent risk assessments and risk management decisions could based on non-residential land uses posing substantially lower potential for exposure. Subsequent to the proposal, EPA developed a "stabilization" effort to address risks at RCRA sites. By taking interim measures and other actions, EPA would reduce the need for governmental oversight and delay or ultimately avoid the need for a site-specific CMS. EPA has recently adopted a "parity policy" to achieve consistency of risk management decisionmaking with CERCLA. Under that policy, EPA will consider the use of tiered, site-specific risk assessment approaches for RCRA sites.

NONCANCER HUMAN HEALTH RISK ASSESSMENT POLICIES

The Nature Of Noncancer Human Health Risks

Chemicals may produce other forms of damage than cancer. In particular, chemicals may exhibit neurotoxicity, reproductive toxicity, developmental toxicity, immunotoxicity, and organ-specific toxicity. The assessed risks of developing cancer typically exceed the risks of other toxic effects by a factor of one hundred to one thousand. Thus, noncancer risks are normally addressed only when a chemical is thought not to be a carcinogen.

The Origins of EPA's Noncancer Human Health Risk Assessment Policies

Except in regard to cancer, toxicologists traditionally have assumed that chemicals exhibit "threshold" health effects, i.e., that exposure to low concentrations are safe and do not increase the risks of adverse effects. As a result, EPA has adopted the concept of "reference dose" ("RfD"), which is an "acceptable daily intake" ("ADI") that poses no increased risk of adverse effects, measured within some range of statistical uncertainty.

The reference dose is typically set by extrapolating to humans from data demonstrating functional or structural impairment of animals. If animal studies establish concentrations that do not result in such adverse effects (the "no observed adverse effect level" or "NOAEL"), the highest such value is used. If animal studies establish concentrations that result in adverse effects, the lowest such value is used (the "lowest observed adverse effect level" or "LOAEL"). Finally, a "safety factor" is applied to account for differences of metabolic and toxicological processes among species and for variable sensitivity in humans, and an added safety factor is applied when data are very limited. In the 1950s, the FDA adopted a one hundred-fold safety factor to account for variability. In the late 1970s, the National Academy of Sciences recommended application of an additional "modifying factor" of up to ten when the data are of poor quality. In 1980, EPA adopted this conservative approach for its water quality criteria, adding an extra safety margin of two to account for limitation of data to a single species. EPA and other government agencies often apply these conservative safety factors without regard to the seriousness of the adverse effects that result, which may range from reversible skin irritation to death.

EPA's Framework For Noncancer Human Health Risk Assessment Under Superfund and RCRA

EPA developed guidelines for various toxicological effects in humans based on qualitative and quantitative extrapolation from animal data and use of uncertainty factors and margins of safety. EPA applied these data at Superfund sites, and often used them to identify "indicator chemicals" for the toxicological effects of concern. In the absence of data regarding interactions of chemical mixtures, EPA recommended adopting an additive approach for each chemical to each toxicological concern. EPA thus defined a "hazard index" ("HI"), which summed the fractional contributions of each chemical's exposure to the acceptable levels calculated from the reference dose and margins of safety. If the hazard index exceeds unity, HI 1, then the chemicals present a risk above acceptable threshold levels. The degree to which the sum exceeds unity measures the relative degree to which the exposures exceed acceptable levels. For RCRA corrective action, EPA simply compares predicted concentrations of hazardous constituents to action levels derived from RfDs.

RECENT EPA EFFORTS TO IMPROVE EXPOSURE ASSESSMENT AND RISK CHARACTERIZATION

EPA's Revised Guidelines for Exposure Assessment

In 1992, EPA issued revised Guidelines for Exposure Assessment. These guidelines present in detail the general principles of exposure assessment that should be considered, but are typically ignored in practice. For example, EPA recommended use of population-based estimates and cautioned against calculation of expected "cases" by simplistic multiplication of average population values when health effects reflect nonlinear dose-responses. EPA also recommended use of more than one type of risk assessment measure. Importantly, EPA did not attempt to alter existing risk assessment practices under Superfund or RCRA, even though those practices are in substantial tension with the population risk assessment principles identified for exposure assessment.

EPA's Guidance on Risk Characterization

In 1992, EPA also provided additional guidance that candidly acknowledged that single-point risk estimates omit significant information, particularly regarding uncertainty, that may be developed during the risk assessment process. EPA thus recommended that risk assessors acknowledge the degree of confidence, uncertainties, and conservatism of assumptions in risk assessments.

In 1995, EPA's Science Policy Council ("SPC") revised the 1992 risk characterization guidance. The 1995 guidance weakens the earlier commitment to description of population risks: "for site-specific assessments, the utility and appropriateness of population risk estimates will be determined based on the available data and program guidance." The 1995 risk characterization guidance also provides support for the Superfund and RCRA programs to continue to estimate risks from the mere potential for residential exposures. These risks will appear significant, even though detailed epidemiological data may later prove that they are insubstantial. These cumulative policy changes remove the practical impact of the many useful recommendations made in 1992, which might have substantially altered risk assessment practices at Superfund and RCRA sites.

EPA's Soil Screening Guidance

In 1996, EPA issued a "Soil Screening Guidance" for use in identifying or eliminating areas of concern for Superfund sites and as action levels for RCRA sites. The purpose of the Soil Screening Guidance was to eliminate from consideration the many sites that do not warrant detailed risk assessments and remedies. Unfortunately, the Soil Screening Guidance is likely to remove few if any sites from consideration. It applies on a generic basis the conservative policies developed to address particular sites. Further, it adds new conservative risk assessment policies and recommends site-specific risk assessments to determine whether use of the Soil Screening Guidance levels is appropriate.

RECENT EPA EFFORTS TO IMPROVE HAZARD IDENTIFICATION AND DOSE RESPONSE ASSESSMENT POLICIES

The newest proposal to revise EPA's carcinogen guidelines, if adopted, would reduce many of the conservative biases of EPA's current carcinogen hazard identification and dose-response policies. Conversely, the proposal would make existing policies more conservative in one important respect. Without substantial discussion, EPA has suggested eliminating the requirement to document statistical significance of human cancer potential when evaluating epidemiological studies.

The proposed guidelines would evaluate human carcinogenic potential from animal data based on substantially more flexible considerations. For dose-response assessment, scaling from animal data to human data would rely more on toxicokinetic data and would apply new default factors when toxicokinetic data are inadequate. The use of toxicokinetic data also would permit development of case-specific dose-response models, avoiding resort to the linearized, multistage model currently employed. Significantly, the revised guidelines would permit nonlinear low-dose extrapolation for modeling nonthreshold effects levels when data support "margin-of-exposure" estimates. Whether the margin-of-exposure approach will result in substantially less stringent dose-response assessment, however, will depend on the safety factors ultimately adopted.

The proposed guidelines would represent another step in the right direction for overall risk assessment policy but would need to be implemented correctly and adopted for use at Superfund and RCRA sites. Until that time, existing conservative hazard identification and dose-response assessment policies will continue to apply.

IV. Ecological Risk Assessment Policy Under Superfund and RCRA

The science and policy of ecological risk assessment are still in their infancy. In part, this is because ecological risk assessment is much more complex than human health risk assessment. As a result, EPA has inconsistently developed and applied ecological risk assessment policies for use at particular Superfund and RCRA sites. Although EPA is developing guidance that may make ecological risk assessment policy more consistent, it also is likely to make it more conservative.

The Nature Of Ecological Effects

Chemical, physical, or biological agents may cause a wide variety of ecological effects, such as mutagenicity, carcinogenicity, teratogenicity, reproductive toxicity, and adverse changes to growth, reproduction rate, behavior, demography, population density, age structure, survivorship, etc. The agents creating these adverse effects are typically called "stressors." Stressors may adversely affect: (1) individual organisms of a wide variety of species of mammals, birds, fish, invertebrates, or plants; (2) populations of individual species; (3) a community composed of the populations of these species; (4) environmental resources on which individual organisms, populations, or community depend; and (5) the ecosystem composed of the community and resources. The assimilative capacities of receptor organisms and of ecosystems may vary and populations of individual species typically undergo natural cycles of growth and decline. As a result, it may be extremely difficult to determine whether and to what extent stressors will have adverse effects and when and for how long those effects may last.

Because of practical and scientific limitations, ecological risk assessments cannot consider all of the potential effects of stressors. Ecological risk assessments must identify "assessment endpoints" and "measurement endpoints." Assessment endpoints represent the ecological characteristic of concern for risk management, which may include "indicators" of broader effects, e.g., the viability of a population of a species of fish located within a particular ecosystem in response to the potential chronic toxicity of a particular stressor. If assessment endpoints cannot be measured directly, measurement endpoints must be selected to represent the direct or indirect effects of stressors on the assessment endpoints.

Measurement endpoints are typically derived from "conceptual models" of ecosystems, i.e., working hypotheses regarding the effects of exposure to the stressor in a particular ecosystem. The conceptual models, in turn, reflect numerous underlying assumptions and choices among competing scientific and statistical models for organism, population, community, and ecosystem interactions with stressors. Predicted concentrations of the stressor also may be derived, using additional modeling assumptions regarding environmental fate and transport. Safety factors may be used to address cross-species sensitivity differences and transfer of data from acute to chronic exposures. As with carcinogen and noncancer human health risk assessment, substantial disagreement may exist over the default assumptions to be applied in the face of inadequate data and scientific understanding.

The Origins of Ecological Risk Assessment Policies

Lacking guidance, EPA ecological risk assessors have adopted widely varying assessment approaches at particular Superfund and RCRA sites. Prior to 1992, EPA acknowledged that "there is no OSWER policy or guidance for selecting for a particular site the appropriate receptors and ecological endpoints of concern; background values and/or reference sampling locations; approaches, techniques, and level of effort; and contaminants of concern. There is also no OSWER policy or guidance for the use of existing toxicity data, interpretation of survey and media toxicity test results, and presentation and discussion of results in site-specific documents."

In 1992, EPA adopted for quantitative ecological risk assessment a modified, three-stage version of the approach recommended by the National Research Council for carcinogen risk assessment. The additional guidance, however, fails significantly to limit discretion regarding when to perform assessments, what type of assessments and level of effort to apply, and which endpoints to measure. Instead, EPA has suggested that risk assessors consult with Biological Technical Assistance Groups ("BTAGs") to determine when particular ecological assessment methods are appropriate.

Emerging Conservative Details of Ecological Risk Assessment Policy Under Superfund and RCRA

In January 1996, EPA recommended the use of "appropriately conservative benchmarks" to screen for contamination at Superfund and RCRA sites that "should not result in significant adverse effects to ecological receptors." EPA has not, however, defined the degree of conservatism that is appropriate or the level of ecological effects that are significant. In May 1996, EPA indicated that development of "facility-specific [screening] action levels to account for ecological risk issues" may be appropriate for RCRA corrective action sites.

In September 1996, EPA proposed to adopt guidelines for ecological risk assessment that would enshrine the three-stage approach previously developed by EPA. "Because of their broad scope, the [p]roposed [g]uidelines do not provide detailed guidance in specific areas nor are they highly prescriptive." Instead, EPA offices and regions will retain substantial flexibility to develop their own guidance, which potentially will be inconsistent and likely will be incomplete.

In particular, EPA notes that the scope and complexity of ecological risk assessments should depend on the degree to which uncertainty may be tolerated. The proposed guidelines thus recommend use of "tiered" approaches, or performing iterative steps with progressively higher investment of resources and more complete decision criteria. "[S]imple models that err on the side of conservatism may be used first, followed by more elaborate models providing more realistic estimates." The screening function served by use of benchmarks is one example of such a tiered approach.

As EPA recognizes in the proposed guidelines, use of multiple safety factors or combined, upper confidence limit values for ecological risk assessment will lead to extremely conservative and potentially unjustified risk estimates. Multiplication of conservative, single-point upper confidence limit values in exposure assessments would add many additional orders of magnitude to the risk estimates. In the unlikely event that EPA risk assessors possess the resources to develop probabilistic analyses that consider all uncertainties, the high-end estimates may so greatly exceed central tendency estimates that the risk assessment may be essentially meaningless. Given the tremendous range of uncertainties, however, EPA may adopt conservative risk management assumptions to protect against the remote possibility of ecological risks.

As for carcinogen risk assessment, use of conservative screening benchmarks and conservative ecological risk assessment assumptions are likely to require costly investigation of sites posing little real risk. The public is likely to be skeptical of more detailed, subsequent risk assessments that do not confirm initial, screening characterizations. The likely result will be the selection of more stringent, less cost-effective, and often unnecessary remedies.

Although it has gained substantial experience with risk assessment, EPA continues to apply very conservative policies to assess the risks at Superfund and RCRA sites. These conservative policies overstate the risks actually posed at many sites. Further, the Superfund and RCRA policies provide risk managers with minimal and less useful information than would result from application of the more general risk assessment policies adopted by EPA. The result is to spend large amounts of public and private funds developing risk assessments that provide little information and that fail to prioritize the real risks that are presented.

In the 104th Congress, numerous bills were introduced to reform EPA's risk assessment policies. Although Congress enacted laws addressing pesticide and drinking water risk assessment practices, the prospect for further legislative reform of risk assessment policies is highly uncertain. Unless EPA acts to remove the historic conservatism from its Superfund and RCRA policies, substantial time and money will continue to be devoted to estimating significant risks at contaminated sites that in reality pose little human health concern. Further, EPA appears likely to transfer its historic conservatism at Superfund and RCRA sites from human health risk assessment to ecological risk assessment. Again, the public and risk managers are likely to focus on the risks estimated by these policies, resulting in costly and unnecessary remedies.