Perspectives on Health Risk Assessments under Superfund

Stephen G. Zemba, Ph.D., P.E.

Introduction
Risk assessment is used to evaluate hazards associated with chemical exposure in settings such as toxic tort litigation, permitting of air pollution sources and waste disposal site characterization. This paper focuses on the longstanding use of risk assessment in the Superfund Program, in which it serves as a key element for decision-making in remedial investigations. Hundreds of Superfund risk assessments have been developed for hazardous waste sites according to guidance published by the U.S. EPA and other regulatory agencies. Despite consistent use of standard assumptions, methods, and procedures within a logical mathematical framework, risk assessment is a not a precise, well-understood process. The commodity-like appearance of risk assessments should not be construed as evidence of technical certainty. The underlying science of risk assessment is in some cases highly uncertain. Readers and users of risk assessments must understand these uncertainties and their implications, lest they be misled as to what is and is not known about risks to health associated with Superfund sites.

The standardization of risk assessment was in part a reaction to the variability that accompanied early practice, a consequence of scientific uncertainties about many aspects of contaminant fate, transport and toxicity, and the range of beliefs and styles among risk assessors. Little wonder that former U.S. EPA administrator William Ruckelshaus stated, “A risk assessment is like a captured spy: torture it enough and it will say anything.” As Ruckelshaus implies, results-oriented assessments were certainly possible, but given the lack of standard guidance, variability also stemmed from legitimate differences in analytical approaches.

This article reviews some basic elements of Superfund risk assessment and offers perspective regarding the details, strengths and weaknesses of the practice. The observations generally hold for risk assessments in other settings, but there are context-specific differences that affect the nature of application. For example, in the context of toxic tort litigation, a risk assessment may focus on the reconstruction of the chemical exposure experienced by specific people and attempt to incorporate as much realism as possible. A baseline Superfund risk assessment, however, is more general and hypothetical as it attempts to encompass all conceivable ways that individuals could contact contaminants, both present and future, sometimes evaluating wildly improbable events, so that remedial solutions protect health indefinitely.

Risk Assessment 101
Risk assessments generally address the potential effects of chemicals on human health and the environment. Procedures for assessing human health risks are better developed and standardized than those for assessing ecological risks, a consequence of historic emphases, political priorities and the enormous complexity of ecosystems. Compared to human health risk assessments, ecological risk analyses address a broader and more open-ended set of issues (e.g., identifying what species of plants and animals are important), so that problem formulation, not to mention problem solving, is much more challenging.

Gauging the degree of hazard presented by chemical contamination in the environment is quickly complicated by myriad assumptions, equations and calculations. Table upon table of numbers can make a risk assessment intimidating. However, the fundamental construct of a risk assessment is simple: a risk to health may be estimated when a person (or, in the case of ecological endpoints, an animal or plant) contacts any “toxic” chemical in any environment.

The term “conservative” is frequently used in risk assessment jargon as a synonym for “health protective,” or “erring on the side of safety.” Decisions, assumptions and parameters are typically selected to err in a manner likely to overestimate risks Conservatism is generally used to compensate for uncertainties, i.e, to provide a margin of safety regarding matters that we do not understand; and to account for variability in parameters that change from person to person, such as the rates at which people breath air or drink water.

Superfund risk assessment follows the classic risk assessment paradigm elucidated by the National Research Council in 1983:

• hazard identification involves the recognition of chemicals in the environment that, under existing or hypothetical conditions, might threaten health;
• exposure assessment marries measurements (or estimates) of pollutant concentrations in environmental media (e.g., air, soil) with rates at which these media may be contacted;
• dose-response (toxicity) assessment characterizes quantitatively for each chemical the type and degree of harm caused by different levels of exposure; and
• risk characterization integrates, in a comparative manner, the exposure and dose-response assessments to provide numerical estimates of risk.

Some observations regarding each of these elements are discussed in turn.

Hazard Identification
Recognition of potential hazards is an open-ended process. There are thousands of chemicals present in any sample of air, water, soil or other environmental medium, but in practice the list of chemicals formally addressed in a risk assessment is limited to no more (and usually far fewer) than about 100 “standard” industrial pollutants. The determination of what chemicals receive consideration in a risk assessment entails many considerations, most importantly our collective knowledge of chemical toxicology. Risk assessors tend to look for chemicals that have been demonstrated in some context to be toxic. Also, we tend to look for chemicals that can be quantified with reasonably available analytical techniques. Obviously, these tendencies are related, in that techniques are developed specifically to find chemicals of concern.

The standard lists of chemicals have been developed over the years, and lists of Priority Pollutants become tied to published analytical methods. By definition, there will always be chemicals present in samples that are not quantified. Hopefully, historic practice and research have identified the majority of chemicals that merit investigation. However, new chemicals of concern will almost certainly be identified, as progress continues in the fields of analytical chemistry and toxicology, improving our ability to detect ever lower concentrations and our understanding of chemical interactions with living things.

Contaminants that are present at background levels in the environment (such as most metals) present sometimes challenging risk management issues. Most regulatory programs allow for exclusion of chemicals present at background levels, even though risks associated with them might be larger than those for release-related contaminants. Demonstration that contaminant levels are the result of natural background, however, can be a difficult task, requiring sampling of nearby uncontaminated locations (if possible) and well thought-out statistical comparisons.

Evaluation of data quality (quality assurance) is a necessary but frequently overlooked element of hazard identification. Environmental sampling and analysis is not simple and can be plagued by errors and uncertainties. Quality Assurance Project Plans (QAPPs) are a standard part of remedial investigations, and the U.S. EPA’s implementation of the Data Quality Act highlights the importance of data quality. Important elements include analysis of field and laboratory blanks to check for the presence of spurious detections, and measures such as spike recovery and duplicate analysis to examine the accuracy and consistency of laboratory methods.

Exposure Assessment
Exposure assessment attempts to evaluate all of the ways in which a person or environmental receptor might contact a chemical in the environment. An individual exposure pathway describes the specific means whereby a receptor might contact a contaminant, and the degree or intensity at which the contact might take place. Typically, exposure pathways are first assessed for completeness, i.e., a demonstration that contact with a contaminant is in fact possible, even if unlikely. A well-documented risk assessment references assumptions and reasoning for each exposure pathway. Explanations for dismissing potential exposure pathways can be equally as important as the stated assumptions for fully-evaluated routes.

Under the Superfund program, potential exposure is quantified for a hypothetical individual, generic in identity, who may be representative of some or no members of the actual population. A typical baseline risk assessment considers both present and future conditions. In each case, worst-case assumptions are usually set forth, as risk assessments are designed to be all-inclusive. Thus, a risk assessment typically assumes that a resident or a trespasser will contact contamination at a site irrespective of the likelihood that such contact might occur, and that such contact will be experienced at high-end levels. In some cases, the degree of conservatism is extreme. For example, worst-case future-use scenarios frequently assume that a property will be redeveloped for residential use without physical or institutional measures to reduce exposure to pollutants. While this scenario may be useful for determining the implications of residential development, it might provide no information regarding the risks associated with actual or reasonably likely site use.

Exposure assessment calculations, which yield estimates of an individual’s chemical dose, often combine multiple assumptions of varying scientific certainty and likelihood. For example, consider the standard analysis that assumes a young child ingests contaminated soil repeatedly over time. The amount of incidental soil ingested by a child is not a well-known quantity, and thus there is a considerable degree of inherent uncertainty in its value. Exposure to the contaminated soil is assumed to occur relatively frequently over appropriate seasons, thus discounting the possibility that incidentally ingested soil might derive from multiple origins. Barring soil-specific testing, chemicals are assumed to be absorbed by the gastrointestinal tract at the same rates they are absorbed from foods, even though many chemicals are known to bind more tightly to soils. In addition, the soil is assumed to contain higher-than-average chemical concentrations, as Superfund risk guidance demands the use of high-end estimates of mean (average) concentrations, or, in many cases, the maximum values detected.

Not all exposure assumptions reflect conservative, high-end values. Some are assigned average, or typical, values. For example, body weight, used in the denominator of dose estimation, is usually assigned at an average value for a given age class. The net effect of mixing high-end and average exposure parameters likely results in a conservatively estimated doses, meaning that predicted exposure levels will be higher than are typical for the average individual. It is generally impossible to gauge accurately the degree of conservativeness, although some sense may be provided by a probabilistic analysis that accounts for uncertainties and variabilities in parameters.

Various exposure pathways require differing numbers of exposure assumptions depending on their complexity and directness. Detailed fate-and-transport modeling may be needed in cases where it is impractical (or impossible) to measure contaminant concentrations in the media of interest (e.g., modeling might be used to estimate chemical concentrations in the indoor air of a hypothetical home built atop contaminated soil or groundwater, accounting for possible volatilization and subsurface transport).

Dose-Response Assessment
Chemicals are evaluated along two lines with respect to human health. First, the likelihood that a chemical causes cancer is evaluated through a qualitative weight-of-evidence approach that considers studies in humans and laboratory animals. Uncertainties in these interpretations tend to be resolved on the side of caution such that it is assumed that a chemical does or could cause cancer in humans, even when evidence is conflicting or equivocal. Carcinogens are assumed capable of contributing to cancer at any level of exposure, i.e., there is no absolutely safe dose. If a chemical is identified as a (potential) carcinogen, the U.S. EPA estimates its potency, or the increase in risk of cancer per a unit of dose. Generally, the methods are designed to estimate carcinogenic potencies on the high side.

Non-cancer health effects represent the second category of harm examined in a risk assessment. In contrast to the approach for carcinogens, a safe dose, or threshold level of effect, is assumed to exist for a chemical that causes (or may cause) non-cancer health problems. The U.S. EPA reviews the body of toxicologic data for a given chemical to determine the potential adverse effects on humans, and the most sensitive health effect is generally selected as the basis for a Reference Dose, defined as a level of exposure that can be safely tolerated for long-term, repeated exposure. Again, data from both human and animal studies are considered. For most chemicals, animal studies are more readily available and reliable, due to the control of extraneous variables. Typically, the lowest level of exposure observed to cause the critical effect in the laboratory study (termed a LOAEL), or the highest level of exposure observed not to elicit the effect (termed a NOAEL), is used as the starting point in the derivation of a Reference Dose. If from an animal study, the LOAEL or NOAEL is usually adjusted using physiologic parameters appropriate to humans. Various safety factors are applied to the adjusted LOAEL or NOAEL to produce the Reference Dose, accounting for possibilities such as humans might be more sensitive to the chemical than the animals studied in the laboratory, animals might be more sensitive at doses lower than the high ones that are tested, and other factors. Overall safety factors of 1,000-fold and greater are not unusual. All safety factors err on the side of caution, even though uncertainties in extrapolation could go in either direction (e.g., a chemical could in fact be less toxic to humans than to animals, in contradiction to the standard assumption that humans are more sensitive). Incorporation of multiple safety factors yields a very low (perhaps negligible) possibility of any adverse health effects at exposure levels lower than the Reference Dose.

Risk Characterization
Quantitative estimates of potential risks to human health are calculated by combining data from the exposure and dose-response assessments. Incremental cancer risk (i.e., the chance of getting cancer above the risk from causes not related to site contamination) is calculated by multiplying the lifetime-averaged doses of those chemicals thought to possibly cause cancer by their respective potency values. Incremental cancer risks from various chemicals are generally added together under the assumptions that chemicals do not interact with each other and that the metric of interest is the overall risk of contracting cancer.

It is important to recognize that risks are predicted for an individual. Risks to populations can be constructed by multiplying the individual risk by the population size, but in doing so the assumption is made that each member of the entire population receives the degree of exposure embodied in the individual risk estimate. Given the conservative nature of risk estimates, this is rarely a valid assumption.

Hazard ratios are used to characterize non-cancer risk, calculated for an individual chemical as the estimated rate of exposure (dose) divided by the Reference Dose. A hazard ratio of one occurs when the exposure level equals the Reference Dose. Hazard ratios smaller than one indicate potential exposure levels smaller than the Reference Dose and imply no appreciable risk of adverse health effects. Hazard ratios greater than one may also be safe, even though the estimated exposure level exceeds the Reference Dose, because the Reference Dose may be predicated on a very large safety factor. A screening-level Hazard Index is frequently calculated as the sum of the hazard ratios for all of the chemicals of concern. If the Hazard Index falls below one, potential non-cancer adverse health effects are usually assumed to be acceptably small. Regulatory agencies sometimes apply a target Hazard Index smaller than one to allow for background exposure to chemicals of concern.

Relative Risk
It is sometimes useful to compare risk estimates to actuarial values and risks from everyday activities. The literature contains many such comparisons (a good source is Risk-Benefit Analysis, by R. Wilson and E.A.C. Crouch, Harvard University Press, 2001). Cancer risk provides a direct and illustrative comparison.

The U.S. EPA’s range of acceptable incremental cancer risk for Superfund sites is 1 to 100 chances per 1,000,000 over the course of an individual’s lifetime. This level of risk is considerably smaller than everyday risks faced by people from a variety of sources. The chances of getting cancer in one’s lifetime are about 1 in 2 for men and 1 in 3 for women. Expressed as the number of chances per 1,000,000, the comparative risks of getting cancer, and the additional chances of getting cancer that correspond to the Superfund acceptable risk range are:

• Chances per 1,000,000
  • Lifetime chance of getting cancer from everyday living (men): 500,000
  • Lifetime chance of getting cancer from everyday living (women): 333,333
  • Additional chance of getting cancer allowed by Superfund range: 1 to 100

Thus, one could expect only a small increase in actual one’s risk of cancer due to the increment allowed in the Superfund Program. As calculated in a risk assessment, this increment is likely overestimated in magnitude (perhaps by considerable degrees) due to deliberate conservatism in estimates of both exposure rates (doses) and potencies.

Suggestions for Reviewing Risk Assessments
Even given the ability, most people could not devote sufficient time and resources to gain a comprehensive understanding of a risk assessment. Notwithstanding, there are a variety of techniques that can be used to distill the essential points of a risk assessment, check it for inconsistencies and omissions, and aid in its interpretation. The following suggestions are offered for consideration.

Scrutinize Critical Risk Estimates

Although a risk assessment may contain dozens of chemicals, numerous pathways and hundreds of calculations, usually its conclusions are affected by only a few. Tables and discussion in the text typically enable one to identify the critical chemicals and pathways that dominate the risk estimates. Limiting the focus of review to these elements can make a difficult task manageable. Care should be taken, however, to consider risk estimates relevant to one’s goals. If concerns center on imminent risks, risk estimates based on potential future land uses may be irrelevant.

Compare with Background

Generic background concentrations of contaminants are available for almost all environmental media, and often local measurements can be obtained from databases and agencies. Some sources of data might include the regional office of the U.S. EPA, the U.S. Geological Survey, agricultural offices, water supply districts, fish and wildlife agencies, state environmental and conservation departments, and county-level air pollution bureaus. Care must be taken to judge the appropriateness of background data not collected specifically in conjunction with a site investigation, but often such data can be valuable. As examples, background water quality in a river can provide important perspective regarding surface measurements of contaminants identified in a site-influenced tributary stream. Also, contaminant levels in fish in related or similar water bodies may be available for comparison.

Compare with Screening-Level Estimates

U.S. EPA Regions 3 and 9 provide tables of Risk-Based Concentrations for a large number of chemical contaminants. Values are available on-line and are periodically updated. Risk-Based Concentrations reflect contaminant levels in soil and groundwater that correspond to regulatory target risk levels (e.g., an incremental cancer risk limit of 1 in a 1,000,000) based on simple, conservative exposure pathways such as incidental soil ingestion and groundwater use for domestic consumption. With appropriate scaling of target risk levels, Risk-Based Concentrations can be compared to site-specific concentrations of chemicals to provide a rough sense of the magnitude of risks that would be anticipated in the risk assessment, and hence provide a gross check for errors and help to identify unusual features in the risk assessment.

Look for Missing and Unusual Elements

The standardization of Superfund risk assessments provides the expectation of considerable uniformity in their content and format. Individuals new to the Superfund risk assessment process can develop some background from basic guidance documents and directives (e.g., the 1991 OSWER Directive 9285.6-03 contains a concise descriptive summary of the scenarios typically found in a Superfund risk assessment). Upon review of even a few risk assessments, one gains a sense of the typical elements, and perhaps even becomes familiar with specific values and assumptions. One should anticipate finding features and considerations common to Superfund risk assessments in terms of the pathways, receptors, and scenarios considered. Increased experience will likely bring familiarity with specific parameters and assumptions, and perhaps expectations regarding the nature and magnitude of risk estimates given site-specific contaminant levels. A checklist of expectations, either formal or internal, can facilitate review.

In reviewing a risk assessment, two complementary approaches are to identify elements of a typical risk assessment that have not been included, and to identify elements that are not typically found. Note that there are any variety of possible reasons from departing from typical procedures. As examples, there may be unusual site-specific considerations, or guidance may have been modified. The reasons for departure from a typical baseline risk assessment should be sought and judged in terms of their reasonableness and applicability.

Ensure that Data Appear Robust, Sufficient and Appropriate to Support Conclusions

Individuals with an innate statistical sense can sometimes identify uncertainties related to sampling data. To be robust, exposure point concentrations should be based on a reasonably large number of samples. Calculations based on only a few samples may produce considerable uncertainty. Unusually high or low values might deserve exploration. Moreover, data should be managed and used intelligently. In many cases it makes sense to differentiate surficial soil contamination from subsurface samples, especially if the latter are to remain inaccessible. As another example, site-wide averages may be inappropriate indicators of exposure point concentrations if redevelopment plans for the site call for mixed uses across the site.

Match Reality and Likelihood to Purposes and Needs

The potentially hypothetical nature of a baseline risk assessment should always be kept in mind. Frequently, realistic exposure scenarios constitute only a portion of the pathways considered in a baseline risk assessment. The reality and likelihood of exposure pathways, especially when viewed in conjunction with institutional controls, should always be considered when evaluating potential risks. Thus, risk assessment review should be focused according to needs and goals regarding the specific exposure scenarios of relevance.

Provide Direction to Consultants

Consultants are frequently employed to review and critique many aspects of Superfund remedial investigations, including baseline risk assessments. Carte blanche examination of a risk assessment can yield many different levels of review. Thus, responsible parties should provide consultants with a sense of their goals, and provide clear direction regarding the required nature and scope of review.

Exercise Common Sense

Experience is a discriminating judge, and one should never underestimate one’s ability to “smell a skunk.” If some aspect of a risk assessment appears unusual, there is a good chance that an error has been made, or that a high degree of uncertainty is present. Extended discussions in a risk assessment are sometimes indicative of contentious matters that demand justification, and may be debatable on technical grounds. Critical examination of conclusions and the uncertainty section can provide valuable clues for interpreting the a risk assessment’s implications.

Stephen G. Zemba (zemba@cambridgeenvironmental.com) is a risk assessor with Cambridge Environmental, Inc. in Cambridge, Massachusetts. Laura C. Green and Sarah R. Armstrong of Cambridge Environmental Inc. also contributed to this article.