Daily consumption level (g/kg/day) Figure 6.2. Food consumption distribution.

level value and a food consumption level value from the available residue and food consumption data sets and multiply them together to yield an exposure level. This process would be repeated for a determined number of events (often thousands or tens of thousands), and the corresponding exposure levels would be combined to yield a distribution of daily exposures.

Probabilistic approaches utilize all of the data in both the residue and food consumption data sets rather than the single point estimates that are used in o : a :

99th percentile 99.9th percentile

Daily exposure (jig/kg/day) Figure 6.3. Probabilistic exposure distribution.

deterministic approaches. The corresponding distributions of exposure from the probabilistic approaches provide much more information than the exposure point estimates of the deterministic approaches. With a probabilistic approach, it is possible to estimate the median daily exposure level as well as levels corresponding to upper percentiles such as the 95th, 99th, or 99.9th percentiles (Fig. 6.3).

Although the results of acute probabilistic approaches may be significantly more useful than those obtained from deterministic methods, they also require far greater interpretation. In the case of the FQPA, the EPA is required to ensure that the levels of pesticides resulting in consumer exposure from dietary, water, and residential sources represent a "reasonable certainty of no harm." Under prior conventions, if an exposure point estimate generated from a deterministic approach were found to be below a level of toxicologically based concern such as a reference dose or acceptable daily intake, the criterion of "reasonable certainty of no harm" would likely be met. The situation is more complicated in the case of an exposure distribution developed from probabilistic methods because a science policy decision as to the desired level of population protection is required.

The current EPA approach calls for the "reasonable certainty of no harm" determination to apply if exposure to a pesticide at the 99.9th percentile for a population subgroup, as estimated by probabilistic analysis, is less than an accepted level of toxicological concern derived from the results of toxicology studies of the pesticide (EPA, 2000a). In cases where exposure at the 99.9th percentile exceeds this level, the EPA would generally conduct a sensitivity analysis to determine whether particular factors that serve to "drive" the exposure at the high end of the exposure distribution, such as high residue and/ or high consumption levels, are unusual and might represent artifacts of the data sets.

The accuracy of exposure estimates at the 99.9th percentile of exposure has frequently been questioned. A comprehensive paper by Chaisson et al. (1999) assesses the consequences of bias, error, and uncertainty in the upper percentiles of exposure distributions. The focus of the paper is on food consumption issues, and the authors contend that sample data invariably contain errors that bias the higher percentiles of exposure in a manner that overestimates, for example, the true 99.9th percentile. One source of error is the inaccuracy of dietary intake surveys that rely on interviews of participants to qualitatively and quantitatively recall their food consumption patterns over selected periods of time. Errors may also arise from insufficient sample sizes, improper weighing of sample data, and reliance on subpopulation data rather than population data to characterize the population. The authors recommend that risk assessments should be performed with the broadest population base in the analysis and that the choice of the "point of regulation," or highest percentile that is not dominated by overestimation bias and error, should be used.

Consumption estimates for specific foods should not be considered independently of consumption estimates for other foods because, in reality, high con sumption of a particular commodity on a given day might be compensated by little or no consumption of other commodities. If consumption of commodities is considered to be independent, this raises the unrealistic potential for consumption of relatively high amounts of several food items on a given day that might significantly exaggerate exposures at the upper percentiles.

The quality and availability of pesticide residue data also influence the accuracy of the upper percentiles of exposure developed through probabilistic approaches.

Pesticide residue samples are often taken as composite samples of commodities that represent several individual servings of the commodities. Research has demonstrated that single-serving size subsamples of the composite samples may be quite variable and raises the possibility that an individual consuming a single serving of the commodity might be exposed to a residue level significantly different than that of the composite sample or of other single-serving sub-samples. Andersson (2000), for example, reported that variability factors of 600, corresponding to the ratio of maximum to minimum residues found in related Swedish subsamples, were found for the insecticides methamidophos in peppers and monocrotophos in grapes. In another study, Harris (2000) indicated that the maximum levels of organophosphate insecticides in individual carrot roots could vary up to levels 25 times greater than those observed for composite samples and results from one subsample of plums showed residues present at 34 times the levels determined for the composite sample. It is clear that such residue variability issues may significantly impact the exposure findings for the upper percentile levels of exposure.

The cumulative risk provision of the FQPA requires assessment of all individual members of classes of pesticides possessing a common toxicological mechanism of action. Many of the individual members of a particular toxico-logically related class of pesticides may serve as substitutes or alternatives for each other; this means that their potential to be used in combination on the same commodity unit is quite remote. If residues of specific pesticides are considered to be independent of residues of substitute pesticides, the mathematical probability of their co-occurrence on the food item could lead to an exaggeration of their actual probability of co-occurrence. Compounding this issue is the fact that some composite samples include individual food items that that may have received different pesticide treatments than other food items in the composite sampling. As an example, the Agricultural Marketing Service of the U.S. Department of Agriculture ( USDA) provided comparisons between results of composite and single-serving analyses of pears obtained from its Pesticide Data Program (USDA, 2000). Composite pear samples indicated that as many as eight different pesticides (insecticides, herbicides, and fungicides) were detected while 36.4% of the samples contained residues of two pesticides, 18.1% contained residues of three, and 10.1% contained residues of four. In contrast, the maximum number of residues detected from single-serving samples was three (0.9%) and residues of two pesticides were detected only 19.5% of the time. Reliance on the composite samples could thus result in an exaggerated proba bility of co-occurrence of toxicologically related pesticides on single-serving amounts of specific commodities and could therefore exaggerate estimates at the upper percentiles of exposure for cumulative risk assessments.

Dose-Response Modeling

After the hazard identification step in risk assessment, a dose response evaluation is performed. This evaluation enables the relationships between the amount of human exposure to the chemical and the probability of adverse effects to be established.

Dose-response models vary based on the type of toxicological hazard that is being considered. Models evaluating the dose-response relationship for chemical carcinogens commonly assume that no threshold dose exists and that all levels of exposure to carcinogens provide at least some finite mathematical risk (Winter and Francis, 1997). For most noncarcinogenic effects, it is assumed that a threshold dose exists at low exposure levels, rendering the potential risks at these low levels of exposure to be considered as insignificant.

The evaluation of risks from chemical carcinogens in the diet is important, but the methods used for such evaluation, including deterministic exposure estimation and mathematical models to predict risks at low levels of exposure, are not receiving the current level of scientific and regulatory focus that exists for the probabilistic monitoring and dose-response evaluation for acute, non-cancer effects.

For noncancer hazards, it is usually believed that adverse health effects will not be observed until a minimum, or threshold, level of exposure is attained. This toxicity threshold tends to be theoretical and is practical only in relation to the effects that occur just above and just below the threshold dose. Toxicology studies are frequently conducted to identify the lowest dose level above the threshold at which adverse effects are noted (the lowest observed adverse effect level, or LOAEL) and the highest dose at which no adverse effects are noted (the no observed adverse effect level, or NOAEL). Limitations in the number of dose levels used in toxicology studies and statistical and biological limitations make it difficult to determine just how closely the LOAEL or NOAEL may approximate the "true" threshold; in the interest of prudence, the NOAEL is generally considered as a conservative estimate of the toxicity threshold (Winter and Francis, 1997).

NOAEL values can be determined for a variety of different toxicology end points and may vary dramatically among the different animal species tested. In most cases, the most sensitive toxicological effects (those occurring at the lowest levels of exposure in the most sensitive species) are considered and the corresponding NOAEL is selected.

It should be understood that NOAEL values are developed from toxicology studies using small homogeneous groups of laboratory animals and, as such, may not adequately represent toxicity thresholds for large and non-homogeneous human populations. In recognition of this fact, uncertainty fac tors (also commonly known as safety factors) are used to guide the animal-to-human extrapolation and to consider human variability. The most common uncertainty factor is 100 and is rationalized to provide a 10-fold uncertainty factor for the animal-to-human extrapolation (this assumes that humans may be 10 times more sensitive than the most sensitive animals studied) multiplied by an additional 10-fold uncertainty factor to account for human response variability (this assumes that some humans may be 10 times more sensitive than the "average" humans). In practice, overall uncertainty factors may range from 1 to 10,000 and the ultimate choice of uncertainty factors is influenced by the availability of human data, the quality of the animal toxicology data, and the nature, severity, and chronicity of the toxicological effect in question.

By dividing the NOAEL by the uncertainty factor chosen, it is possible to develop an estimate for the lowest level of toxicological concern for the chemical. Historically, this has been termed the acceptable daily intake (ADI) and is expressed as the amount of chemical exposure per amount of body weight per day. More recently, the EPA has replaced the ADI terminology with an analogous term, the reference dose, or RfD. This removes the inference of "acceptability" that may be plagued by the connotation of a nonscientific value judgment. In many parts of the world outside the U.S., the ADI terminology is still commonly used. For the purposes of this chapter, further references will be made to reference doses rather than to acceptable daily intakes.

Acute reference doses

The estimates of dietary risks posed by pesticide residues in food have typically focused on long-term (chronic) toxicity and have relied on deterministic methods to calculate exposure. A common approach used to assess chronic risks has been to assume that consumption of food items that may contain the pesticide in question is represented by the average daily intakes of the food items for a 70-year period and that the residue level point estimates frequently represent the maximum allowable residues on the food items or anticipated residues based on more realistic assumptions. The exposure estimate derived from this deterministic approach is compared with the chronic RfD to determine whether the exposure is sufficient to merit toxicological concern.

Acute risk assessments using deterministic methods take a similar approach but may use an upper percentile of food consumption rather than the average level and may consider the maximum detected residues or maximum allowable residues rather than the average anticipated residues. In the case of both chronic and acute risk assessments, however, the exposure estimates are frequently compared with the chronic RfD to determine the acceptability of the levels of exposure.

The development of probabilistic methods of exposure assessment, made possible by our improved computational capabilities and the regulatory requirements of the FQPA, may significantly limit the future use of deterministic methods to estimate acute dietary exposure to pesticides. Probabilistic methods may demonstrate instances in which exposure of a population subgroup at the upper end of the exposure distribution curve, such as the 99.9th percentile, may exceed the chronic RfD even though deterministic approaches might demonstrate that the point exposure estimate is at levels below the RfD. In such cases, regulatory actions may be taken to limit exposure even though the prior deterministic estimates of exposure suggested no cause for further regulation.

To improve the accuracy of acute dietary risk assessments for chemicals in food, it is critical that appropriate toxicological studies are used to determine an appropriate acute RfD. Unfortunately, toxicology studies used to determine RfDs have traditionally been conducted for chronic (lifetime) or repeated shorter-term dosing (28-90 days). Comparing single-day estimates of exposure with RfDs developed from longer exposure scenarios may exaggerate the probability of acute risks. This is particularly important in cases in which the pharmacokinetic factors such as absorption, distribution, biotransformation, and excretion of a chemical are known and demonstrate that continuous repeated exposure to the chemical may lead to greater concentrations of the chemical at the toxicity target site over time compared with a single-exposure scenario. Toxicological databases normally do contain the results of single dosing studies, but such studies typically involve high doses of the chemical and focus on animal lethality rather than determination of toxicity thresholds. Relatively few single dosing studies exist for pesticides that presently allow for determination of the acute NOAEL and subsequent acute RfD, and those that do exist frequently involve insufficient numbers and ranges of dose levels to accurately determine acute NOAELs.

A relatively recent regulatory decision made by the EPA demonstrates the need for accurate acute RfDs in the assessment of acute dietary risk. In August of 1999, the EPA severely limited the uses of the organophosphate insecticide methyl parathion, citing excessive dietary risk to infants and children (EPA, 1999). Preliminary assessments of methyl parathion dietary risk relied on an acute NOAEL considered to be as low as any used for organophosphate insecticides. This NOAEL was determined from a toxicological study containing a 300-fold difference between the LOAEL dose of 7.5 mg/kg/day and the NOAEL dose of 0.025 mg/kg/day, suggesting that the "true" NOAEL would be anywhere from 0.025 to 7.5 mg/kg/day and was more likely to be closer to the LOAEL based on toxicity comparisons of methyl parathion with other organophosphate insecticides, many of which are considered to be far more toxic than methyl parathion. The EPA's reliance on this exaggeratedly low NOAEL also led consumer (Consumers Union) and environmental (Environmental Working Group) organizations to perform their own risk assessments for methyl parathion in early 1999 that alleged that hundreds of thousands of U.S. children were routinely exposed to excessive levels of methyl parathion in their food (Wiles et al., 1999: Groth et al., 1999). These organizations demanded that the EPA take actions to restrict methyl parathion use.

The EPA's regulations were announced in early August 1999, immediately before an FQPA statutory deadline. Interestingly, eight days after announcing the decision, the EPA made available its revised methyl parathion risk assess ment for public comment. In the revised risk assessment, the EPA recognized the limitations of the toxicology study it had previously relied on to determine the acute NOAEL, and modified its acute NOAEL based on the results of a one-year repeated-dosing study in rats to 0.11 mg/kg/day. The EPA also acknowledged receipt of a new methyl parathion single-dosing study that indicated a NOAEL of 1 mg/kg/day. If this more accurate NOAEL were used to determine the acute RfD, the exposures at the upper 99.9th percentile for all population subgroups would have been below the acute RfD and no regulatory action would have been deemed necessary based on the methyl parathion dietary risk assessment. The curious timing of the regulatory decision (8 days before the public release of the revised risk assessment and immediately before FQPA statutory deadlines) suggests that political factors as well as scientific limitations influenced the regulatory decision (Winter, 2000). Subsequent revisions of the regulatory decision or the risk estimates with the most recent methyl parathion toxicology study seem unlikely.

In recognition of the potential inaccuracies that may arise when comparing single-day exposure estimates with RfDs derived from repeated-dosing studies, the Codex Committee on Pesticide Residues is developing guidelines for conduct and assessment of short-term toxicology studies used to derive appropriate acute RfDs (Herrman, 2000). Issues considered by this committee include how to determine which pesticides require acute RfDs, what the toxicological requirements to establish an acute RfD are, and what uncertainty factors are appropriate.

Infant and child susceptibility—the 10x factor

One of the most controversial provisions of the FQPA is the so-called "lOx factor" that requires consideration of the potentially greater susceptibility of infants and children to pesticides. According to Section 408(b)(2)(C)(ii)(II) of FQPA:

In the case of threshold effects, for purposes of clause (ii)(I) an additional tenfold margin of safety for the pesticide chemical residue and other sources of exposure shall be applied for infants and children to take into account potential pre- and post-natal toxicity and completeness of the data with respect to exposure and toxicity to infants and children. Notwithstanding such requirement for an additional margin of safety, the Administrator may use a different margin of safety for the pesticide chemical residue only if, on the basis of reliable data, such margin will be safe for infants and children.

The call for an additional uncertainty (safety) factor in the FQPA was derived from the National Research Council report (NRC, 1993) that investigated science and policy issues concerning pesticides in the diets of infants and children. The report concluded that current toxicology testing protocols might not be sufficient to address issues concerning toxicity and biotransformation of pesticides at early stages of development. A key recommendation of the report was that, because specific periods of infant vulnerability may exist during postnatal development, an uncertainty factor of up to 10-fold should be considered when either data exist to suggest evidence of greater postnatal developmental toxicity or data concerning child susceptibility are incomplete. Thus, in the cases in which data may be incomplete, there should be a presumption of greater toxicity to infants and children.

At the same time, the report also concluded that age-dependent differences in chemical lethality were usually less than one order of magnitude and usually varied no more than two- to threefold. Another finding of the report was that infants may be more sensitive to some chemicals at high doses than adults but may also be less sensitive to others.

Bruckner (1999), who served as a NRC committee member, maintains that the existing 10-fold interspecies uncertainty factor provides adequate protection of infants and children. A similar position is taken by Renwick et al. (1999), who argue that the use of the additional 10-fold factor for infants and children is not generally justified based on the existing usual 100-fold uncertainty factor used for the animal-to-human and the human-to-sensitive human extrapolations. Under certain circumstances, however, they contend that the lOx factor may be appropriate. These include cases in which 1) reproductive and developmental toxicity data are not available, 2) testing methods for assessing reproductive and developmental toxicity are inadequate, or 3) effects on neonatal and/or young animals are irreversible and severe.

The NRC report also indicated that there might be little difference in age-related human toxicological responses to chemicals after 6 months of age. In practice, the FQPA lOx factor is applied equally to infants and to children up to 12 years of age, even though it may be argued that the major toxicological differences would primarily affect only the infants. In the EPA's revised risk assessment for methyl parathion, which utilized the additional lOx factor in addition to the usual 100-fold uncertainty factor that covers inter- and intra-species variability, the population subgroup experiencing the greatest exposure was to children ages 1-6 (0.969 pg/kg/day at the 99.9th percentile). This exposure represented 8.8 times more exposure than was deemed acceptable (EPA, 1999). Elimination of the additional lOx factor for children ages 1-6 would result in acceptable levels of exposure at the 99.9th percentile. For infants, however, maintaining the lOx factor resulted in exposure (0.415 pg/kg/day at the 99.9th percentile) that was 3.8 times greater than the level considered acceptable.

Dose response considerations for cumulative risk assessments

The FQPA requires that risk assessments for pesticide residues be based on "available information concerning the cumulative effects on infants and children of such residues and other substances that have a common mechanism of toxicity."' This provision of the FQPA is derived from the NRC report finding that children may be exposed to residues of multiple pesticide residues that possess a common toxic effect and that such simultaneous exposures should be accounted for.

The NRC report suggested that such cumulative risk assessments could be conducted by assigning toxicity equivalence factors (TEFs) for each of the pesticides having a common toxicological mechanism (NRC, 1993). This practice was justified because a similar process had already been developed by the EPA to assess the risks from dioxins and dibenzofurans.

As an example, a member of a toxicologically related family of pesticides (presumably the member that was most widely studied) would be chosen as the reference chemical for the family. Comparisons of the potency of other family members to the reference chemical would yield the TEF. If the chemical in question were determined to be 2 times more potent than the reference chemical, the TEF would be 2; if the chemical in question were determined to be one-half as potent as the reference chemical, the TEF would be 0.5. Cumulative exposure to the class of pesticides studied could be determined by multiplying the actual level of each pesticide residue by its TEF and then adding results for each pesticide.

A specific example of this approach was provided in the NRC report (NRC, 1993). This example considered five organophosphate insecticides, all considered to possess a common mechanism of toxicity through cholinesterase enzyme inhibition, and their presence on eight foods and three juices that are common in the diets of infants and children. In this example, TEFs were determined relative to the insecticide chlorpyrifos and were based on comparisons of NOAEL values for cholinesterase enzyme inhibition.

As discussed above, the value chosen for the NOAEL (and therefore the RfD) is subject to the choices of experimental design such as dose levels, test species, and routes of exposure. Comparing such values to develop TEFs is therefore subject to great uncertainty. The NRC example relied on some NOAELs derived from animal studies and others derived from humans. Comparing RfDs among different chemicals to determine TEFs raises the potential for even more uncertainty because the RfDs are based on both the NOAELs and the choices of uncertainty factors used. Interestingly, the EPA has determined that the additional lOx factor should be retained for methyl parathion, whereas, its close chemical relative, ethyl parathion, which is identical in structure with the exception of two additional methylene groups, does not require the additional lOx factor. In cases in which large differences exist between the NOAEL and the LOAEL, the "true" NOAEL may not be well approximated by the experimentally determined NOAEL; such inaccuracies are magnified when NOAELs of different chemicals are compared to determine TEFs.

The EPA defines the "point of departure" as a point estimate of the dose or exposure level that is used to depart from the observed range of empirical response (or incidence) data for purpose of extrapolating risk to the human population (EPA, 2000b). Because the NOAEL represents a single arbitrary dose, the EPA prefers to use an "effective dose," essentially similar to a benchmark dose, that is associated with some designated level or percentage of response relative to the control or baseline level of response. The EPA suggests the adoption of a 10% effect level (EDio) as the standard default point of departure.

In addition to the TEF approach, the EPA also has considered a cumulative margin of exposure (MOE) approach. The MOE is calculated by dividing the point of departure (effective dose or, suboptimally, the NOAEL) by the expected or measured human exposure. A cumulative MOE approach would sum the MOEs of the individual pesticides possessing a common mechanism of toxicological action. This same type of process, deemed the total MOE, is advocated by Sielken (2000) as the preferred method for performing FQPA cumulative risk assessments.

Weight Loss All Star

Weight Loss All Star

Are you looking to lose weight or even just tone up? What is stopping you from having the body you want and help you feel great at the same time? I created Weight Loss All-Star for all those wanting to lose weight, and keep the weight off. I know how hard it is to do diets and stick with them, and get the motivation to get up and exercise.

Get My Free Ebook

Post a comment