The National Academy of Sciences (NAS) report Pesticides in the Diets of Infants and Children focused national attention on the process of assessing dietary risk from pesticides. A major conclusion of the report was that significant improvements in risk assessment are necessary if regulators are to ensure a safe food supply for all consumers, including sensitive subgroups such as infants and children. However, the report did not suggest that infants, children or adults are now at excessive risk from pesticides in the diet.

While risk assessments are developed to protect public health, they also have economic repercussions for consumers and the agricultural community. They are the basis for regulatory and legislative policies determining under which circumstances a particular pesticide's use is allowed. These, in turn, may influence the availability and cost of food items and may have consequences in foreign food production practices and the regulation of international trade, since exporters of food into the United States are subject to the same standards as domestic producers.

At the present time, risk assessment methods are complicated and controversial, and results of risk assessments may be subject to a variety of interpretations. This paper examines risk assessment and the potential impact of the NAS committee recommendations, suggests guidelines for the interpretation of risk assessments and discusses the scientific challenges and policy implications of the risk assessment process.

Risk assessment: an overview

The risk assessment process consists of hazard identification, dose-response assessment, exposure assessment and risk characterization (see fig. 1 and box, p. 23 ). The first two steps constitute the toxicological assessment phase, which generally consists of extrapolation of the results of animal toxicology studies to determine the potential for human effects at specific levels of exposure and evaluation of the quality of the toxicology data. The exposure assessment step requires estimates to be made for pesticide residue levels and food consumption rates. The final step, risk characterization, combines information from the toxicological assessment and exposure assessment phases to yield a risk estimate based on statistical probability. The accuracy of the risk estimate therefore depends upon the accuracy of the toxicological assessment and the exposure assessment, both of which present significant scientific difficulties.

Difficulties in toxicology

The basic challenge of the toxicological assessment is to identify the most important toxicological effects observed in animal studies and to determine levels of human exposure that would be required to cause the same types of effects. The premise is that the results of animal toxicology studies may be predictive of the potential effects in humans; this premise has been validated in many cases, although there are also several exceptions.

In most cases, few data exist from which to extrapolate the results of animal studies to potential human effects, and several conservative assumptions are made (such assumptions tend to increase the estimate of risk). In the case of noncarcinogenic effects, for example, uncertainty factors (also known as safety factors) are incorporated into the process. The highest dose of a chemical shown not to cause any effect (the no observable effect level, or NOEL) in animal studies is divided by an uncertainty factor to yield an acceptable level of human exposure (often referred to as the acceptable daily intake, ADI, or the reference dose, RfD). An uncertainty factor of 100 is most commonly used. The 100-fold uncertainty factor includes a 10-fold uncertainty factor to allow for differences in sensitivities between animals and humans multiplied by another 10-fold uncertainty factor to allow for susceptibility differences among individual human members of the population. The NAS report recommended that an additional uncertainty factor of up to 10-fold be used in some cases to account for sensitivity differences between infants or children and adults (for a summary of the NAS findings, see p. 15 ).

Carcinogens (cancer-causing chemicals) present additional challenges in the risk assessment process. In contrast to noncarcinogenic effects, it is often assumed that a threshold level of exposure to carcinogens does not exist. This assumption is the subject of intense scientific controversy. It derives in part from radiation biology studies and is based on the observation that many carcinogens are genotoxic — that is, they cause direct damage to genes that may lead to cancer. Therefore, any level of exposure to a carcinogen might present some calculable risk of cancer. The risk is still related to the dose, as it is with noncarcinogenic effects.

Since typical human exposure levels to potentially carcinogenic pesticides are often several thousand times lower than the doses that have produced cancer in animal studies, calculation of carcinogenic risks requires extrapolation of the results of high-dose animal studies to potential human effects at low levels of exposure. Scientists have adopted mathematical models to predict human effects based on animal effects.

Such models yield a value known as the cancer potency factor. Because a variety of different mathematical models exist, cancer potency factors vary widely depending upon the choice of the model and its assumptions. The most common model used is the linearized multistage model ( see p. 26 ). This model assumes that a cell which may be a target for a carcinogenic chemical goes through a specific number of different stages and that the probability of a “hit” on the cell leading to cancer is stage-specific. At very low levels of exposure, the relationship between exposure level and excess cancers is linear; any finite level of exposure leads to some numerical value of excess cancers (above the background level in the population).

The NAS report recommended use of the two-stage clonal expansion model, itself a recent interpretation of the multistage model. This is a biologically based model of carcinogenesis that takes into account the special physiological characteristics of infants and children, such as the rapid rate of cell division and tissue growth. Biochemical events in infants and children, such as enzymes induction at different stages of development, may affect their response to pesticide residues. In addition, the NAS report recommended development of “pharmacokinetic” models for infants and children — that is, models that describe the child's unique physiological patterns of residue uptake and distribution, allowing for factors such as lactation and development. The report also recommended use of the so-called “benchmark dose” in assessing risk to infants and children. The benchmark dose uses all available dose-response data, better reflects the slope of the dose-response curve and provides an explicit indication of risk at doses below the benchmark dose.

Exposure assessment

To use the information derived from the toxicology assessment phase in risk assessment, it must be combined with an estimate of the likely amount of human exposure, or dose. Two major questions need to be answered during the exposure assessment phase: (1) What are the levels of pesticide residues in various food items? and (2) How much of the various food items are eaten?

Residue estimates

Multiplication of the residue levels by food consumption values for an individual food yields an estimate of the exposure from the particular food; addition of exposures from all possible food items (including water) provides an estimate of total dietary pesticide exposure.

Typical approaches taken in the estimation of residue levels have ranged from those that simply assume that all residues are present at a predetermined level (for example, tolerance levels representing maximum legal residues) to more complex approaches based on actual measurements of residue levels at the time the food is ready to be consumed. A variety of intermediate techniques incorporating data on pesticide use and actual field residue levels may also be used. The choice of method used to estimate residue levels depends upon the availability and reliability of data and the degree of refinement necessary. In many cases, for example, assumptions of residues at the tolerance levels still lead to risks far below those considered to be significant and additional refinement of the residue levels is not warranted.

Some of the methods for assessing pesticide residue levels are shown in figure 2. Approaches requiring the lowest cost often yield the greatest amount of data but provide the greatest overestimation of residue levels. As refinements are made in the process to accommodate additional data, the accuracy of the residue estimates is improved, but the data are less available and the cost to produce additional data rises significantly.

Consumption estimates

The determination of the amount of consumption of particular food items by the population at large or by particular population subgroups is also important in the assessment of dietary pesticide exposure. In the United States, food consumption estimates are most commonly derived from the results of U.S. Department of Agriculture nationwide food consumption surveys. Such surveys involve tens of thousands of individuals in the 48 continental states and Hawaii, Alaska and Puerto Rico. Individuals are asked to describe the types and amounts of foods they consumed, both at home and away from home, for a defined 3-day period.

The adequacy of this approach has been called into question. A major limitation appears to be the sample size for certain subpopulation groups such as nursing infants, which was considered from the latest (1987–88) survey to be too small to confer statistical credibility. Also, the accuracy of surveys of this type is limited by the inability of those surveyed to accurately remember what they consumed during the specified period.

In addition to problems resulting from difficulties in acquiring accurate food consumption estimates from national surveys, individual variation in food consumption by members of the same population subgroup is also difficult to account for in the risk assessment process. This is particularly important in determining acute exposure to pesticides in the diet, since short-term consumption of particular foods may differ dramatically from average, long-term consumption.

To compensate for some of the inaccuracies in food consumption estimation, it is often suggested that average levels of consumption be replaced by consumption estimates of those consuming the greatest amounts of individual food items. This approach may overcome the possibility of underestimating the exposures to some members of the population but may also lead to unrealistic exposure estimates for the general population since it does not allow for compensation in the diet to reduce consumption of other foods.

More accurate estimates could be obtained using “risk distributions,” one of the chief recommendations of the NAS report. Risk distributions combine statistical data on the kinds and quantities of foods eaten by children with data on the pesticide residues on these foods. Risk distributions, which show a statistical distribution of risk across a population, would replace the current method of estimating risk from each individual pesticide using a single number to represent average exposure of the entire population. Risk distributions would help regulators view children in a variety of circumstances, those with above-average and below-average exposure.

Interpreting results

While risk assessment is limited in its ability to accurately calculate human risks from exposure to pesticide residues in the diet, it plays an important and necessary role in the pesticide regulatory process. The use of risk assessment allows regulators to prioritize risks, allocate finite resources, and base regulations upon consistent criteria. It is critical, however, that the limitations and inaccuracies of the risk assessment process be understood if results of risk assessments are to be properly interpreted.

Typical risk assessments rely upon conservative assumptions that increase the risk estimate over one determined using less conservative assumptions. This prudent approach has wide philosophical acceptance as society tends to agree that if errors are to be made, it is best to err on the side of safety. Thus, the risk estimates themselves may overstate actual human risks, as a deliberate public policy.

While the findings of the NAS committee may indicate that increased regulatory controls or more-refined risk assessment techniques are necessary, they should not be taken as indicators of immediate health risks. Conservative assumptions and uncertainty factors built into the process afford a margin of safety.

In the case of carcinogen risk assessment, care should be taken to avoid misinterpreting results through “body count” analyses in which risk estimates are multiplied by population numbers to suggest “actual” cases of human cancer. Significant uncertainty is factored into the process through the use of conservative mathematical models; the actual cancer risks may be significantly lower or even zero.

Scientific challenges

There is no doubt that significant scientific advances are needed to improve the accuracy and utility of risk assessments. It should be emphasized that toxicology itself is a young and evolving science and that risk assessment is only in its formative stage of development.

The recent NAS report made several suggestions for improving the accuracy of dietary pesticide risk assessment ( see p. 15 ). It is likely that adoption of many of the NAS recommendations could significantly improve the accuracy of the risk assessment process.

Over the past several years, much of the scientific community has maintained that improvements are particularly needed in the assessment of the risks from potentially carcinogenic pesticides. Carcinogen risk assessment receives a disproportionate level of attention and detracts from the assessment of other types of risks such as nervous system, developmental, immune system and reproductive effects ( see p. 17 ). This overemphasis stems primarily from the unique treatment that potential carcinogens receive concerning the lack of a toxicity threshold level and the use of particular mathematical models that incorporate nonthreshold behavior. While it is extremely difficult to determine experimentally whether carcinogens do or do not act in a threshold dependent manner, many view the assumption that carcinogens may not be subject to threshold doses to be overly conservative.

Another point of controversy surrounding carcinogen risk assessment involves toxicology testing protocols. Typically, massive levels of chemicals ( see p. 28 ) are given to animals throughout their lifetimes at the maximum tolerated dose (MTD), which is typically several orders of magnitude higher than levels humans would expect to receive. It has been argued that cancers produced at the MTD may result from cell death and errors in subsequent cell replication caused by the high doses and would not be expected at lower, more realistic doses. In addition, results of cancer studies are often equivocal, and many chemicals may be considered as carcinogens even though a majority of studies have been negative for carcinogenicity.

The conservative “nonthreshold” approach for carcinogens also ignores significant mechanistic developments in toxicology that have shown that chemicals may produce cancer by a variety of means. In several cases, chemicals classified as potential carcinogens have not been identified as genotoxic (mutagenic) agents; mounting evidence for many of these chemicals indicates that their carcinogenic effects may require exposure in excess of a toxicity threshold. Thus, it is likely that the use of a single mathematical model may not be applicable to all chemicals.

Regulatory, policy implications

Dietary pesticide risk assessments are commonly used to determine if pesticides may be registered for use on particular commodities. The Environmental Protection Agency (EPA), under provisions of the Federal Insecticide, Fungicide, and Rodenticide Act, has the authority to approve pesticides for use if the benefits of the intended uses are determined to outweigh the risks. The determination of dietary risks from pesticides is a critical component of EPA's overall risk assessment process, and may have significant effects upon pesticide use practices in the United States. It is important that the risk assessment decisions made by EPA be scientifically sound, because mandated changes in pesticide use have wide effects on the nation's agricultural economy, including food availability and cost. While public health should be government's first priority, policy makers also need to be cognizant of economic impacts on the nation's significant industries.

When introduced into the regulatory and policy arenas, the ultimate “number” generated from the risk assessment may take on exaggerated importance while appreciation of the imprecision and complexities of the risk assessment process tends to fade. The uncertainties of risk assessment do not lend themselves well to politics, and policies are often developed through use of a “bright-line” approach to determine, in absolute terms, the acceptability of specific risks. Such an inflexible approach may neglect the benefits of specific chemicals or the alternative risks from not using the chemicals. As an example, risk assessments have demonstrated that specific fungicides may present potential carcinogenic risks. Restrictions on fungicide uses would minimize these risks, but may lead to other risks, since fungicides control the production of naturally-occurring carcinogenic mycotoxins. More importantly, the inflexibility may not allow future scientific advances to be incorporated into risk assessment practices. As the scientific understanding of the mechanisms for chemical carcinogenesis increases, new mathematical models may allow the risks to be calculated more accurately.

Delaney Clause

As an example, a recent court decision upheld a strict interpretation of the Delaney Clause of the Federal Food, Drug, and Cosmetic Act; this action has forced EPA to begin cancellation procedures for several food-use pesticides ( see p. 30 ). While many uses of pesticides are not subject to regulation as food additives, several others are affected by the “zero-risk” standard of the Delaney Clause. Removal of these pesticide uses based upon statutory grounds will provide trivial health benefit to consumers but could cause significant effects upon domestic and international agriculture.

Significant effort is being made in Congress to replace the Delaney Clause with new legislation. In place of the “zero-risk” provision of Delaney, the concept of “negligible risk” is being considered. Different bills prescribe different methods for the determination of what constitutes a negligible risk and how risks should be calculated. Ironically, these risk assessment prescriptions may also be considered to be inflexible and may not allow future scientific improvements in risk assessment to be used. The legislation, if adopted, may appear to be as scientifically obsolete in 35 years as the Delaney Clause appears to be today. It is critical that legislation allow pesticide regulators to make use of the best available scientific practices rather than prescribed, outdated methods.

NAS recommendations

There is no doubt that incorporation of many of the NAS recommendations will improve the scientific basis for assessing human risks from pesticide residues in the diet. At the same time, however, major assumptions are still required and future risk assessments will still be limited in their accuracy. The costs associated with implementing many of the NAS recommendations need to be considered as well, since processes such as standardizing residue determination practices, improving food consumption estimates and incorporating earlier stages of animal development into toxicological testing protocols may involve significant expenditures. It is critical that such expenditures be balanced against possible health benefits that could be derived from implementation of such practices. At the present time, data indicate that for most pesticides, even under the exaggerated assumptions that they are always present at the maximum allowable levels, typical human exposure is well below established health criteria such as the ADI or RfD. Rather than employ state-of-the-art risk assessment techniques to all pesticides across-the-board, it seems prudent to initially use less sophisticated analyses to prioritize pesticide uses of greatest concern and reserve the more sophisticated models for the most important pesticides.

Priorities in perspective

While we need to improve the risk assessment process, and protect all subgroups of the population, we must keep pesticide residue concerns in proper perspective. The driving force behind pesticide legislation and regulation today seems to be food safety, although most would contend that environmental effects from pesticides and worker safety issues are much more important. In terms of overall food safety, pesticides are considered by the U.S. Food and Drug Administration as a fifth priority, of less concern than (1) microbiological contamination of foods, (2) nutritional imbalance, (3) environmental contaminants and (4) naturally-occurring toxins. Pesticide food safety issues should not be ignored, but should be considered at an appropriate level when precious resources are allocated and health policies are formulated.