![[logo] US EPA](../gif/logo_epaseal.gif){kind=logo}
March, 1998 Suggested Probabilistic Risk Assessment Methodology
Suggested Probabilistic Risk Assessment Methodology for Evaluating
Pesticides with A Common Mechanism of Toxicity:
Organophosphate Case Study Environmental Working Group -
January 1998
Submitted to the FIFRA Scientific Advisory Panel
for the meeting March 24-25, 1998
This paper details the methodology and findings of the January
1998 Environmental Working Group analysis of children's dietary
organophosphate exposure. This Monte Carlo risk assessment was based
on the latest government data available on food consumption, pesticide
residues in the food supply and organophosphate toxicity. No exposure
to OPs outside of the diet was modeled in this analysis, but could
easily be added to such a model if usable data were available. This
presentation to the Scientific Advisory Panel focuses on the methodology
with results provided for demonstration purposes.
You will need Adobe Reader to view some of the files on this page. See EPA's PDF page to learn more.
- Data Sources
- Food Consumption Data
- Pesticides in the Food Supply
- Toxicity Data
- Exposure Assessment
- Risk Assessment
- Sample Results
- High Risk Pesticides
- High Risk Foods
All data used in this risk assessment are from federal government sources, except for EWG data on commercial baby food which was added to supplement the FDA data. Food consumption data are from the USDA, pesticide residue data are from both the USDA and the FDA, and toxicity data are from the EPA.
The food consumption data used in this analysis are from the USDA Continuing Survey of Food Intakes by Individuals (CSFII) for the years, 1989, 1990, 1991, 1994 and 1995, the most recent years for which data are available. No data were collected in 1992 and 1993. The CSFII contains 3,695 coded foods and beverages reported as eaten by the survey population ranging from blueberry pie to scrambled eggs, potato chips to mint juleps. A total of 4,082 children between one month and five years of age were surveyed in the years 1989-91 and 1994-95 (Table 1, 43 KB). These 4,082 children provided one to three days of valid information each, for a total of 8,302 valid eating days.
CSFII survey participants are asked to complete a diary containing the amount (by weight) of each food eaten at each meal during the three non-consecutive days of the survey, and the weight, sex and date of birth of the person consuming that food. The information in the diaries was confirmed by telephone interview. These data allow age group analyses, as well as estimates of food consumption on a per kilogram of body weight per day basis, for each individual in the database.
The CSFII is a weighted, stratified sample of individuals that is designed to provide a representative picture of the dietary patterns of the U.S. population. Risk assessments and estimates of the number of children exceeding specific risk standards and safety margins are extrapolated from this data. However, EWG assessed the significance of incorporating sample weight adjustment into the risk calculation and found that it had little effect on the outcome. (see section on Risk Assessment)
Data on pesticides in the food supply that were used in this assessment were obtained from monitoring data collected by the USDA and the FDA. There are three major programs: the USDA Pesticide Data Program, the FDA Pesticide Surveillance and Monitoring Program and the FDA Total Diet Study (Table 2, 43 KB). Each has a specific purpose, and its own strengths and weaknesses.
The USDA Pesticide Data Program (PDP) was started in 1991 specifically to monitor pesticide levels in fruits and vegetables most commonly consumed by children. The purpose of the program was to supplement the FDA surveillance data with more accurate and statistically representative information on pesticides residues on fruits and vegetables heavily consumed by infants and children. PDP typically samples twelve to fourteen foods, mostly fresh fruits and vegetables, per year. Samples are collected to accurately reflect the percent contribution to the national food supply for a given crop by growing region and season. Samples are then washed, peeled, and cored to reflect normal food preparation and consumption practice for that fruit or vegetable. PDP residue testing uses highly sensitive analytical techniques that can detect trace residues in the 1 part per billion range or less, similar to the range of detection in the Total Diet Study (TDS). PDP takes 400-700 samples of each crop per year. More than 25,800 PDP samples from the years 1992 through 1995 were used in this analysis. These data were the data of first choice for assessing pesticides levels in fresh fruits and vegetable consumed by children.
The FDA Pesticide Surveillance and Monitoring Program enforces food tolerances established by the EPA. Because the monitoring is designed for regulatory enforcement purposes, as opposed to dietary exposure assessment, the data do not provide a strictly statistically representative picture of pesticides in the US food supply. This shortcoming, however, is largely offset by the sheer size of the database generated by the program and the fact that the program does sample food from all regions of the country and conducts analysis at laboratories located in nine different metropolitan areas. Between 12,000 and 16,000 samples of food are tested for pesticides each year, about half of which are imports. We analyzed all records from the FDA surveillance database from the years 1992 through 1995, which contained residue findings for 51,280 food samples.
Our analysis used only random "surveillance" samples. "Compliance" samples, which are specifically aimed at crops or growers with a known problem or history of violations, are not included in the analysis. Surveillance samples are typically taken at packing sheds, warehouses, or other central distribution points. They are not taken at retail points of sale or from grocery store shelves. Further, the samples are not washed or peeled prior to testing - e.g. the melon is tested with the rind, the banana is tested with the peel - so that the residue levels found tend to overstate the amount of pesticides consumed when the fruit is eaten. Because of these biases built into the FDA surveillance protocols, data from this program were used only as a last resort. And as discussed below, when they were used, a reduction factor was applied to the residues found on each sample, to better estimate actual exposure.
The second FDA pesticide residue monitoring program is the Total Diet Study (TDS). The TDS was started in the early 1960s to study the prevalence of radioactive fallout in the food supply as a result of atmospheric nuclear weapons testing. Today, the program tests 234 different foods four times a year for a host of contaminants. The 234 foods sampled are determined to be representative of the U.S. diet. The entire sample is purchased at grocery stores four times each year, one in each of four geographic regions of the country. This "market basket" covers a broad range of both processed (bottled, canned, frozen) and fresh foods including fresh fruits and vegetables, as well as baby food, dairy products, frozen meals, fresh meats, cereals and peanut butter, and prepared foods like pizza.
Prior to testing, the foods are prepared as they normally would be in the home. Bananas are peeled, tuna casserole is baked, rice is boiled, and the hamburger is grilled. The prepared food is then analyzed for pesticides and other toxic contaminants.
Unlike the FDA surveillance data, the TDS data are designed to provide a representative snapshot of contaminants in the U.S. diet. The biggest shortcoming of these data is that the sample sizes tend to small (4 samples of each food per year). The strength of the program is that it provides real world data that reflect pesticide residues very likely encountered by the average person. In addition, the TDS uses sensitive analytical techniques that can detect low level residues of pesticides in the range of 1 part per billion or less, a significant advantage over the FDA surveillance program which does not use analytical techniques with comparable sensitivity.
In 1997, under a Freedom of Information Act request, Environmental Working Group received TDS data for the years 1991 through 1996 in electronic form. These six years of data contain 4,520 food samples that were analyzed for pesticides. These data were a critical component of the pesticide residue data used in the sample assessment present in this document.
Additionally, 72 samples of commercial baby food sampled by EWG in 1995 as part of the report Pesticides in Baby Food were added to the pesticide residue database used in this analysis. This supplemented the 16 samples of each type of baby food in the Total Diet Study. The EWG data were found to be roughly similar in residue profile to the FDA data (e.g. the same OP pesticides are consistently found on the same products). However, the levels detected and frequency of detection were lower in most cases in the EWG data. If anything, adding the EWG baby food data lowered the overall estimates of risk from baby food consumption.
The toxicity of organophosphate insecticides is regulated on measurements of cholinesterase inhibition as a toxic effect or endpoint.. This assessment was conducted using the chronic RfDs. The chronic RfD was chosen as the appropriate measure of toxicity in this study only after our initial analysis revealed that on average, 88 percent of all children 5 and under were exposed to at least one OP each day. This in contrast to the EPA's practice of treating all OP exposures as acute (i.e. one-day exposures). Treatment of OP exposure as chronic is more appropriate given that nearly 90 percent of the study population is exposed each day to the pesticides being studied.
Specifically, the chronic reference dose is the agency's determination of an essentially safe daily dose of a pesticide, or in this case, the dose of OPs that will produce no adverse cholinesterase inhibition, expressed in milligrams of pesticide per kilogram of body weight per day (mg/ kg bw /day). The reference doses (RfDs), used in this report are the most recently calculated reference dose values used by EPA scientists, obtained directly from the EPA in January 1998.
All of these reference doses represent finalized agency decisions, except that for chlorpyrifos. The Reference Dose Work Group of the EPA Office of Pesticide Programs recently recommended that the reference dose for chlorpyrifos have an additional ten-fold safety factor added, per the requirements of the Food Quality Protection Act to protect infants and children from pesticides. The Reference Dose Work Group is the EPA pesticide program committee of scientists charged with making recommendations for pesticide safety standards under the FQPA. Generally, recommendations from this committee have been adopted as agency health standards.
The chronic reference dose is derived from any of a number of animal toxicity tests required by the EPA. These studies for OPs range from the 90-day rat study, to a two-year chronic feeding study. The species of test animals are most commonly rats or dogs. However, where acceptable data are available, studies in humans are used. For example, the RfD for chlorpyrifos is based on data from a study conducted on 16 adult male Dow Chemical employees in 1972. RfDs for pirimiphos-methyl, ethion and diazinon are also based on human data (Table 3, 44 KB).
Each year of CSFII data contained from 150 to 200 individuals per age group (one-year-olds, two-year-olds etc.). Each individual reported from one to three eating days that were validated by USDA. An eating day can be thought of as all the food reported eaten by one individual on one day. Only eating days with complete information and positive validation by USDA were used. The five years of CSFII data used in the report contained a total of 8,302 valid eating days for children age six months through five years. Age group cohorts were constructed by combining individuals of the same age from the five years of CSFII data used in the analysis.
Survey participants through five years of age reported eating about 3,695 different foods. Many of these different foods, however, are nearly identical versions of the same food. For example, orange juice drinks that would be considered different foods in the CSFII include unsweetened orange juice, orange juice with sugar, orange juice with calcium, orange juice from concentrate and fresh orange juice. For purposes of linking food consumption data with residue data in this report, these similar foods are considered the same food.
The goal of the exposure analysis was to produce the most accurate real world picture of pesticide exposure via the diet. To achieve that end the three residue databases described above were used in the analysis, in the following order of priority. For fruits and vegetables eaten raw, PDP data were used because the data represent residues after washing and peeling, and because samples are statistically reliable and representative of U.S. food consumption. For all other non-processed foods, FDA surveillance data were used. These data provide large sample sizes, but generally overstate residues at the time of consumption. To account for this, a residue reduction factor of from 25 to 95 percent was applied to all FDA surveillance data (Table 4, 41 KB). The reduction factors are based on actual reductions calculated by comparing detections of individual OPs on fruits and vegetables from USDA PDP (washed and peeled) to similar items in the FDA surveillance data (unpeeled, unwashed). In each case the OP found to have the greatest "washed-peeled" reduction served as the default.
For processed and cooked foods, data from the FDA Total Diet Study were used. The small sample sizes in the TDS created some concern that TDS data might overstate exposure to some OPs. For example, the six years of TDS data provide to EWG contained only 16 samples of wheat bread, but all of them were positive for OP residues. Using these 16 samples to represent the entire U.S. wheat bread supply might overstate OP exposure via wheat bread. On the other hand the residues in these products, while ubiquitous, were generally at low levels, and not likely by themselves to present great risk to any consuming individual.
To test the validity of the bread product residue findings, we examined OP residue data in all of the more than 600 samples of processed wheat products in the TDS. From pasta to pretzels, to wheat bread and wheat breakfast cereal, more than 99 percent of more than 600 samples tested for pesticides were positive for either chlorpyrifos, malathion, or both. This strongly suggests that the low level OP residues reported in the TDS for any single processed wheat product are very likely representative of the commodity as a whole. To increase the sample size for baby food, tests of baby food for pesticides commissioned by EWG in 1995 were added to TDS data from FDA. The results from both TDS and EWG were quite similar.
Of the 39 OP insecticides with a common mechanism of toxicity, only 13 were detected in the food supply. These 13 OP compounds, in turn, were found on 407 of the 3,695 foods reported eaten by children age five and under in the USDA survey.
More than 3,695 food items were reported eaten by children under age five in the CSFII. The federal pesticide residue databases used in this analysis contained directly comparable residue results for 561 of the 3,695 foods reported eaten by children six months through five years of age. These 561 foods account for 68 percent of the diet by weight for these children. Of these 561 foods, 407 were found to contain detectable levels of OP insecticides.
For purposes of predicting pesticide exposure, many of the 3,695 foods reported eaten by children can be considered the same food. For example, it is reasonable to assume that cooked carrots with fat, cooked carrots without fat, and cooked carrots (fat unspecified), are the same in terms of pesticide residues. Many other decisions were not that straight forward. Links between foods reported eaten, and residue findings were made as described below. As a general rule, foods were linked with residue values only when a direct match between the two foods was available. Any deviation from this rule is described below.
For fruits and vegetables eaten raw, food consumption values were matched first to data from PDP, when available. When PDP data were not available, the assessment used FDA surveillance data with a residue reduction factor applied. Frozen fruits and vegetables (not canned) that were not cooked before eating were assumed to have the same residue levels as fresh fruits and vegetables and the same residue values were applied. For fruits and vegetables eaten cooked, either from canned, fresh or frozen sources, residue values from the total diet study (TDS) were used.
For all other processed and cooked foods that were reported in the CSFII, TDS data were used when a direct match was available. For thousands of specific foods reported eaten by the population studied - cherries jubilee, pepperoni pizza, all soft drinks and many others - no direct matches were available in the residue files. These foods were not used in the analysis. For example, we did not attempt to match the pepperoni pizza consumption data with OP residue data from cheese pizza, because of uncertainty about the exact weight ratio of the foods that constitute each respective pizza. Likewise we did not match cherry pie residues with cherries jubilee consumption data, and so on for thousands of foods with no direct match. With sandwiches, consumption data was matched with residue data only when the sandwich consumption data was reported in its component parts that matched the residue data. For example, when a peanut butter and jelly sandwich was reported as X grams of bread, X grams of peanut butter and X grams of jelly, these consumption values were matched with corresponding test results from the TDS. When the sandwich was simply report as a peanut butter and jelly sandwich, it was not used.
The one technical exception to this rule was with wheat products in the form of pasta and bread. In this case residue data were available for white bread, wheat bread, macaroni and spaghetti. Children age five and under, however, reported eating many types of pasta (spaghetti, macaroni, lasagna noodles etc.) and many different types of bread (French bread, Italian bread, pita bread etc.). In this case, any wheat based bread or pasta was matched with the residue values from the most closely matched wheat based bread or pasta products in the TDS.
The exposure assessment was a Monte Carlo style probability distribution analysis designed to simulate real world dietary exposure to OP pesticides using the best available data. The analysis was modeled after that used by the National Research Council Committee on Pesticides in the Diets of Infants and Children (NRC 1993 pp. 297 through 307).
The program was run on a Power Mac 8100 using FoxPro software. A distribution of dietary OP exposure was simulated for 6 age groups from ages 6 months through 5 years old in grouping of 6-12 month olds, one-year-olds, two-year olds etc. Dietary exposure to OPs was analyzed in one day (24 hour) units as the food consumption data was grouped. The distribution was created by instructing the computer to identify a valid individual eating day in the database (person one, day one) and to match each food eaten by that individual on that day with a randomly selected residue result from all the samples for that food in the residue database described above. Total daily exposure to each of the thirteen individual OPs in the residue files was then calculated and converted to a mg/kg exposure value for each OP consumed, depending on the amount of the food consumed, the residue(s) found on the sample that was selected and the weight of the child.
For example, if a child ate 100 grams (a little under four ounces) of green beans and the green bean sample, randomly chosen from the residue database, had 1 part per million (ppm = mg/kg) of acephate, the program would calculate that 1 mg/kg acephate x 100 grams of beans = 0.1 mg of acephate on those beans. If the child weighed 10 kg, the dose of acephate that child got from those green beans would be 0.1 mg/10 kg = 0.01 milligram acephate per kilogram body weight .
For each of the valid eating days available for each age year, this process was repeated 2,000 times, to produce a distribution of three to four million individual exposure days, per age group, for each of the OP compounds.
Non-detects and traces
Food samples with no residue detected (non-detects or NDs) were included in this analysis assuming that a non-detection was equal to zero exposure. This assumption clearly underestimates real life exposure. EPA risk assessors generally use half the limit of quantification or some calculated residue value for those samples below the detection limits of the testing equipment. In addition, for any residue that was measured at the "trace" level, meaning below the level of quantification and above the limit of detection was also counted as zero exposure in our analysis.
Each of the two to three million individual exposure days generated for each age group contains a total mg/kg exposure value for each of the OPs for which residue data were available. For example, individual #2,789,450 might have eaten 0.3 mg of acephate, 0.04 mg of azinphos methyl and so on for all thirteen OPs found in food.
Conversion to Chlorpyrifos Equivalents
In order to compare the risks from chemicals with different RfDs, a Toxicity Equivalency Factor (TEF) approach was used. This approach normalizes the toxicity of chemicals to that of an index chemical so that direct comparisons can be made. For this assessment, chlorpyrifos was chosen as the index chemical partly because the 1993 National Research Council Committee on Pesticides in the Diets of Infants and Children used chlorpyrifos as the index and also because it is the most widely used OP nationwide. As explained below, any OP could have been chosen as the index with no effect on the final risk calculation.
To assess the risk of this exposure, an individual's total assorted OP exposure on any given day was converted to chlorpyrifos equivalents. To do this, a chlorpyrifos toxic equivalency factor (TEF) was applied to convert the mg/kg dose of each OP to the equivalent dose of chlorpyrifos. This TEF accounts for the difference between the reference dose of chlorpyrifos and the reference dose of any other OP.
A conversion factor for pesticide X would be calculated by dividing the reference dose for chlorpyrifos by the reference dose for pesticide X. For example, the proposed reference dose for chlorpyrifos is 0.0003 mg/kg and the reference dose for pesticide X was 0.0001 mg/kg. The conversion factor for pesticide X would be 0.0003/0.0001, or 3, meaning that pesticide X is estimated to be three times more potent than chlorpyrifos (chlorpyrifos TEFs are listed in Table 3, 44 KB).
To express the dose of pesticide X in chlorpyrifos equivalents, one would simply multiply the mg/kg dose of pesticide X by 3, and so on for all of the OPs. Total daily exposure is then calculated as the sum of chlorpyrifos equivalents for each OP, on any given day. An individual's total daily OP exposure, expressed in chlorpyrifos equivalents, can then be compared to the chlorpyrifos reference dose which is now a surrogate for a combined-OP RfD.
Risk estimates: population at risk
Estimating risk for the population as a whole requires an additional step. The simulated exposure data including the daily dose in chlorpyrifos equivalents for each individual are compiled to create a database of a population-wide range of exposure levels (daily doses). This range of exposure is then compared to the reference dose for the index chemical, in this case chlorpyrifos.
For the purposes of this analysis we assumed that there are 2 million 6-12 month olds and about 4 million children in each of the other age cohorts in the U.S. The latest US Census data indicates that there are over 19.2 million children under 5 years old (0-4 year olds) or roughly 3.8 million per age group.
To determine the number of children exceeding the RfD one simply derives the percentage of the simulated population that exceeds the RfD and multiplies that percentage by the real population. For example, if the estimated OP dose in 5 percent of simulated one-year olds exceeds the RfD: 5 percent of the real population of 4 million = 200, 000 one-year olds. Population-at-risk estimates could of course be refined further using Census figures for appropriate demographic, geographic or socio-economic sub-populations.
Sample weights
The CSFII is primarily aimed at gathering information on nutrition, malnutrition and dietary patterns and contains "sample weight" data for adjusting from the sampled population to the population at large for assessment of demographic variations. The CSFII sample population is weighted toward lower income families whose problems with malnutrition are more acute. It was determined that the net risk calculation result did not vary significantly when sample weights were incorporated (Table 5, 43 KB). This calculation could be easily added if, for example, the model were used to identify sensitive demographic sub-populations of children.
Comparison to the NRC method
This procedure differs slightly from the method used by the National Research Council Committee on Pesticides in the Diets of Infants and Children. The committee conducted a similar Monte Carlo analysis and converted exposure to chlorpyrifos equivalents using "no observable effect levels" (NOELs) instead of reference doses (RfD). An RfD, which is the functional equivalent of what EPA deems a "safe" daily dose of the pesticide, is derived by dividing the NOEL by a specified safety factor. Safety factors differ from pesticide to pesticide, depending on the quality of the data, the species of test animal in which the data is generated and the effects observed in critical studies.
Initially we employed a methodology similar to that used by the NRC committee. However, the results produced by this analysis (the number of children exposed to levels that exceed a specific safety margin) were entirely dependent on the safety factors applied to the various NOELs (Table 3, 44 KB). In essence, when NOELs are used to convert the toxicity of multiple pesticide exposures to a baseline compound, the number of children that exceed the reference dose changes, depending on the pesticide chosen as the baseline compound. For the results to be meaningful and unbiased, the estimated number of children exposed to OPs in food at levels that exceed a specific safety margin must be the same, regardless of the chemical chosen as the baseline pesticide. Basing the TEF on RfDs corrects this problem.
For example, the chlorpyrifos proposed RfD is based on a 100-fold uncertainty or safety factor applied to a NOEL from a study on humans, whereas methyl parathion is based on a 1,000-fold uncertainty factor applied to a NOEL from a study on rats. When conversions were based on the NOELs, using methyl parathion as the baseline pesticide put 9.1 percent of all one-year-olds over the RfD, whereas using chlorpyrifos as the baseline chemical put only 3.2 percent of these same one-year-olds over the RfD. In contrast, when the conversions are based on the RfDs, the analysis yields the same percentage of one-year-olds exposed to an unsafe dose of OPs on any given day (5.2 percent) regardless of the pesticide chosen as the baseline compound.
In terms of efficiency, this model allows easy recalculation of risk as RfDs are modified and updated over time. If an RfD is changed, one simply recalculates the TEF for that pesticide and runs the risk assessment program over again using the exposure data already in hand and the new toxicity data to arrive at a new range of "doses" in the selected population. In addition, exposure data from such sources as drinking water or residential residues could easily be added to this analysis should such data become available. The model could also be used to assess risks that vary demographically or socio-economically simply by incorporating that data.
Using this model to assess the combined risk to children from dietary sources of OPs, a number of interesting results can be derived. The flexibility provided by this methodology in interpreting the sources of exposures, both by chemical and commodity indicates that such an approach will be useful in risk management decisions.
The analysis reveals several important facts about dietary exposure to OP pesticides. Every day, nine out of ten American children from ages six months through 5 years are exposed to OP insecticides in the food they eat. The foods with the highest percentage of contamination are the processed grain products, particularly wheat products such as breads and pasta. Virtually 100 percent of all processed wheat products tested by the FDA are contaminated with low levels of chlorpyrifos, malathion, or both.
While almost always contaminated, the levels found on grain products are typically low. The foods with the most significant residues of OPs are apples, peaches, grapes and pears. The OPs that accounted for the highest risk to the population examined were methyl parathion, dimethoate, chlorpyrifos, pirimiphos methyl, and azinphos methyl.
Using reference doses (RfDs) updated by the EPA on January 14, 1998, we estimate that every day, more than 1.1 million children age six months through 5 years exceed the current safe daily dose of OPs, the chronic reference dose, set by the EPA. More than one hundred thousand of these children exceed the EPA safe dose by a factor of ten (Table 6, 44 KB). Only three of the reference doses, set by EPA for the 13 OP insecticides found in food, contain protections added since the 1996 Food Quality Protection Act to provide for the extra sensitivity of infants and children to the toxic effects of OPs.
There are 39 OP insecticides registered for use on food crops in the United States. Only a subset, however, are found at measurable levels on food. Our analysis of more than 80,000 samples of food inspected by the federal government for pesticide residues from 1991 through 1996, revealed that only 13 organophosphate insecticides were found in or on food above the limits of detection of the testing equipment used by the Food and Drug Administration and the U.S. Department of Agriculture during that period.
The highest risk OP compounds are methyl parathion, dimethoate, chlorpyrifos, pirimiphos methyl, and azinphos methyl which together account for about 90 percent of the risk from OP insecticides in the infant and child diet. Methyl parathion contributed 43-52 percent of the total OP risk for the age cohorts examined (ages 6 months through 5 years). Dimethoate accounted for 34 percent of the infant's (6-12 month olds) risk from OPs but only 7-12 percent of one to five year-olds' risk. Chlorpyrifos was responsible for 9-15 percent of the risk, pirimiphos methyl 5-17 percent and azinphos methyl 2-7 percent of the risk to children 5 years old and under (Figure 1, 45 KB).
Sources of exposure to OPs in the diet could be determined from this assessment. This aspect of the analysis would be particularly useful in identifying areas in which to focus resources for risk reduction.
The most contaminated foods
A significant percentage of several fresh fruits expose children to unsafe levels of OP insecticides. For example, our analysis indicates that one out of every four times (25 percent of the time) a child under six years of age eats a peach, he or she exceeds the EPA (adult) safe does of OPs. Thirteen (13) percent of the time a child of this age eats an apple, he or she exceeds the EPA reference dose of OPs (Table 7, 44 KB). Apples are followed by nectarines (12 percent), popcorn (8.5 percent), and pears at (7.5 percent).
Commercial Baby Food
OP compounds were found in a number of commercial baby food samples analyzed by the government. Environmental Working Group samples of commercial baby food taken in 1995 were added to the analysis to supplement the FDA data. The EWG samples had both lower residue levels and lower frequency of residues than the FDA data and therefore lowered the estimates of risk from baby food sources. Approximately ten percent of the time an infant between six and twelve months of age ate pear baby food, he or she exceeded the daily safe dose of OPs. Baby food apple juice and peaches caused this problem 5 and 2.4 percent of the time, respectively.
Foods that put the greatest number of children at risk
The foods that put the greatest number of children at risk are not usually those with the highest percentage of unsafe OP residues, but instead are those that have moderate to significant residues of OP, and are widely consumed in significant amounts.
Infants Six to Twelve Months
For infants six to twelve months of age, commercial baby food is the dominant source of unsafe levels of OP insecticides. OPs in baby food apple juice, pears, applesauce, and peaches expose about 77,000 infants each day, to unsafe levels of OP insecticides (Table 8, 43 KB).
One through Five Year Olds
Apples, apple juice, and apple sauce expose the most children age one through five years of age to unsafe levels of OPs. Greater than 50 percent of the children that eat an unsafe level of OPs each day, receive this unsafe dose from apple products alone. We estimate that each day, fresh, raw, apples expose more than 400,000 children age one through five to unsafe levels of OP insecticides (Table 8, 43 KB). When all apple products are included, the number jumps to 575,000 children per day.
Peaches and grapes are also significant source of unsafe OP exposure, putting 77,000 and 54,000 young children over the safe dose each day.
Foods that expose the most children to the most toxic dose of OPs
High consumption food items expose many children to levels of OPs that exceed safe levels by wide margins. OPs on washed apples, peaches, grapes, pears and pear baby food cause 85,000 children each day to exceed federal safety standards by a factor of ten or more (Table 9, 43 KB).
A small but real percentage of some foods contain residues of OP insecticides so high that virtually any consumption will put a child over the safe daily dose. For example, roughly 2 percent of the apples, grapes, raisins and pears have such a potent dose of OP insecticides that eating 10 grams of these fruits (the equivalent of two grapes) would cause the average 25 pound one-year-old to exceed the daily OP safety standard (Table 10, 43 KB). For peaches, 15 percent of 1,500 peaches tested had residues of OPs where 10 grams of consumption (about three bites) by the average sized one year old, would put that child over the safe daily exposure level for OPs. (All pesticide residues in the analysis were measured after the produce samples were washed and otherwise prepared for normal consumption by USDA technicians).
A child who eats an average amount of these foods is at high risk of exceeding the safe daily dose of OPs. For example, 26 percent of the time an average amount of peaches was eaten by a one year old, that single dose put that child over the safe daily dose of OPs. Apples put the average one year old apple eater over the safety standard for OPs 13 percent of the time, followed by pears and grapes at about 7 and 6 percent (Table 10, 43 KB).
A child who eats a lot of these foods is even more likely to exceed the safe dose of OPs. About 22 percent of the one-year-old apple eaters that eat at the 90th percentile of reported consumption (the equivalent of about one apple), exceed the safe daily dose of OP insecticides. Heavy (90th percentile) consumption of pears put 11 percent of the one-year-old pear eaters over the daily safety standard, followed grapes at 7.5 percent and raisins at 6.3 percent (Table 10, 43 KB).
Scientific Advisory Panel (SAP) March 1998 Meeting