Welcome to Oxygenie.com - The Power of Activated Oxygen

The report below has been reprinted from the original analysis conducted by the Consumers Union of the United States, Inc., Public Service Projects Department, Technical Division.

"Which Foods Have the Highest TI Values? Seven foods consistently had high or very high TI's each time tested: Fresh peaches (both domestic and imported); frozen and fresh winter squash grown in the U.S.; domestic and imported apples, grapes, spinach and pears; and U.S.-grown green beans. Among these, U.S. peaches and frozen winter squash had TI Values about 10-fold higher than even the other "high" scores..."

This Report is very long, and has been broken down into different sections:

Quick Links

|| Summary || Introduction to the USDA Pesticide Data Program || Methodology & Toxicity Index ||
|| Results and Discussion || Recommendations || Tables: Pesticide Contamination in Fruits & Vegetables ||

DO YOU KNOW WHAT YOU'RE EATING?
AN ANALYSIS OF U.S. GOVERNMENT DATA
ON PESTICIDE RESIDUES IN FOODS1

Methodology

You can't control pesticide use by the agricultural industry... ...but you can remove the pesticides! Food and Water Ozonator

A. Consumers Union's Analysis of the PDP Data

We obtained the results of the PDP pesticide residue analyses for the years 1994 through 1997. (The 1997 data are the most recent available, just released in January 1999.) The data are available to the public, and reports are published on the USDA web site, but most citizens are not familiar with the program or the data it has produced. In the four years we examined, the PDP tested over 27,000 samples in 27 different food categories.

Table 1 lists the foods and numbers of samples of each food tested each year. Sixteen of the tested foods were fresh fruits and vegetables; 8 were processed fruits and vegetables; milk, soybeans and wheat were also included. For 15 foods, at least 10 imported samples came from at least one foreign country, which was our "threshold" for examining imported samples as a separate category.

For the other foods, all or nearly all of the samples came from the U.S. Table 1 shows foods and countries they came from, where there were adequate samples. In nine cases (apple juice/Argentina, bananas/Central America, orange juice/Brazil, grapes, peaches and pears from Chile, green beans, tomatoes and winter squash from Mexico), the PDP tested at least 60 imported samples in one or more years. This large sample size provides an accurate picture of that food from that origin. For some of the other imported foods, though, sample numbers are probably too small to support precise estimates of their relative pesticide toxicity loading.

Our consultants, Karen Lutz and Chuck Benbrook, designed and built a large database program that we then used to analyze the USDA data. This database enabled us to examine the residue results in many ways--by year, food item, pesticide chemical, residue level, frequency of detection, country of origin, and by any combination of those parameters.
To the same database, we added information on the toxicity of every pesticide chemical detected in the PDP in the years we examined. The U.S. Environmental Protection Agency (EPA) has compiled toxicity data on all registered pesticides; we used EPA's most recent data.

The combination of USDA PDP residue data and EPA toxicity data enabled us to estimate the relative toxicity loading of pesticide residues in different foods. This integration of residue and toxicity data is an innovation by Consumers Union; we first applied this method last year in our report on Organically Grown Foods. We believe this method offers the scientific and regulatory communities, as well as consumers, a sound and useful way to compare the relative size of pesticide risks posed by different chemicals and combinations of chemicals in different foods.

B. CU's Toxicity Index

To compare the amounts of pesticide residues in different foods in a meaningful way, CU has developed a "Toxicity Index," or TI for short. The TI provides an integrated measure of the frequency of detection of residues in a food, the average levels of residues present, and the relative toxicity of the specific pesticides present.

This index is not a true measure of risk. Risk depends on how much a person eats of different foods and on characteristics of individual consumers like age, health status and other (non-food) exposures to pesticides. But the
TI depicts the relative amount of pesticide toxicity in different foods, and as such it provides a more robust index of relative risk than simple measures of residue frequencies or levels, unweighted for toxicity, can do.

To create our TI's, we first needed to calculate a TI for each pesticide chemical found in the foods the PDP tested. Then, using the chemicals' TI's and the PDP data, we computed TI's for the pesticide residues in each food.

(1) Toxicity of Different Pesticides

Pesticide toxicity has two components: Acute Toxicity, the propensity to cause immediately observable adverse effects at relatively high levels of exposure; and Chronic Toxicity, the propensity to cause long-term, delayed effects, following repeated lower-level exposure. Different pesticide active ingredients differ widely in both acute and chronic toxicity.
The standard toxicological measure of Acute Toxicity is the LD50, which is defined as the dose of a chemical that kills half of the exposed group of test animals. The smaller the LD50, the more toxic the chemical. Table 2 shows the LD50's of pesticide chemicals detected in foods in the PDP. The range from least toxic to most toxic of the listed pesticides by this measure of acute toxicity is over 5,000-fold.

To translate LD50's into an Acute Toxicity Index, (ATI), we took the inverse of the LD50 for each chemical (i.e., 1/LD50); this gives us an index in which larger numbers indicate greater toxicity. We multiplied the results by 100, to make the results whole numbers, instead of decimal fractions, while leaving the relative magnitude of the ATI's unchanged. Table 2 also shows the ATI for each pesticide chemical.

In summary: ATI = (1/LD50) x 100

The most widely used toxicological measure of Chronic Toxicity is the Reference Dose, or RfD. The RfD is derived by taking the highest dose level that had no observed adverse effect in test animals and dividing it by a "safety factor," typically 100. The result in theory represents a dose thought to be without appreciable risk to humans, although the uncertainties inherent in extrapolating from animals to humans and from high doses to lower doses must be acknowledged. As with the LD50, the more toxic a chemical is, the smaller its RfD. The EPA has published chronic RFD's for most registered
pesticides. Table 3 displays the RFD's of pesticides detected in the PDP.

Here too, there is a wide range (8,000-fold) from most to least toxic. CU has developed a Chronic Toxicity Index (CTI), which is based on the RfD, and also takes into account certain additional data on a chemical's toxicity. As with the Acute Toxicity Index, we used the inverse of the RfD, so that more toxic pesticides would have larger CTIs. RfD's typically are very small numbers, and the expression (1/RfD) yields results that range from about 6 to over 50,000. To express the results on a more manageable scale, we multiplied them by 0.1. We then added factors, where applicable, for endocrine disruption and carcinogenicity:

Endocrine disrupters: For pesticides listed as suspected endocrine disrupters by Colborn et al. (1993), the CTI was multiplied by a factor of 3. (I.e., CTI = (1/RfD) x 0.1 x 3.) Endocrine disruption is responsible for some of the most devastating documented effects of pesticides on wildlife, and as more research emerges, may well prove to be a very critical aspect of pesticides' impacts on human health. In our judgment, potential endocrine disruption is a more important aspect of a chemical's toxicity than even potential carcinogenicity, and our scoring scheme therefore gives it great weight.

Carcinogens: We incorporated a factor based on the U.S. EPA's classification of carcinogens and estimate of carcinogenic potency, or Q*. For those pesticides that have a Q* in EPA's database, we multiplied the Q* by 10 for pesticides classified by EPA as "known" or "probable" human carcinogens, and by 5 for those classified as "possible" human carcinogens. To put the results on a scale where they would comprise about one-third of the total when combined with the RfD-based index, we multiplied them by 50. This product was then added to the CTI. The effect of this additional factor is minor for pesticides that are very toxic in other ways (in which case, the RfD component of the CTI is dominant). For pesticides that have relatively low general toxicity but are carcinogenic, the carcinogen component of the CTI tends to dominate.
Table 3 also displays CU's Chronic Toxicity Index for pesticides detected in the PDP, and the factors used to calculate the CTIs.

(2) Calculating Toxicity Indices for Specific Foods

Using our ATI and CTI for each chemical, and the PDP data, we can compute a Toxicity Index (TI) for each category of food tested, based on the amounts of residues of different pesticides found in that food.

For example: The PDP tested 502 samples of U.S.-grown apples in 1996. The analysis detected residues of 37 different pesticide chemicals in those apple samples. The PDP data show us which chemicals were detected, how often (i.e., in how many of the 502 samples) each was detected, and at what levels they were detected in each sample. The PDP data provide this information on all the pesticide residues found in all 27 of the foods and in samples from each country of origin, in each year we examined. Overall, there are about 1,300 unique combinations of specific pesticides in specific foods from specific countries in specific years.

For each of those 1,300 combinations, we calculated frequency of detection (percent positive for the specific chemical) and the mean residue (the average residue level in the positive samples.) We then used those values, and the ATI and CTI for each individual pesticide, to compute an ATI and a CTI for each of the 1,300 combinations.

For example, for each of the 37 pesticides found in U.S. apples in 1996, we calculated an Acute Toxicity Index by multiplying the percent positive for a particular chemical in those apple samples, times the mean residue, times the chemical's ATI. We repeated the same process, using each chemical's CTI, to get the Chronic Toxicity Index for each pesticide found on the apples.

We repeated these steps for all 1,300 combinations of chemicals on foods from a given country in a given year. When this step was completed, we had 1,300 ATI scores and 1,300 CTI scores.

Before we could combine the ATI and CTI scores into a single TI for each individual chemical/food/country of origin/year combination, we had to convert them to the same scale. We standardized the two sets of numbers by converting them to a percent scale. Through this step, all but a few "outlier" values in the ATI and CTI data sets were expressed as numbers between 0 and 1002

After standardizing the ATI and CTI indices to the 100-point scale, we combined the indices for each individual food/country-of-origin/chemical/year combination into a single Toxicity Index using the formula TI = ATI + 2CTI. That is, we gave chronic toxicity twice as much weight as the acute toxicity component. We believe this weighting is appropriate for assessing
dietary exposure to pesticides.

To get a TI value for a given food/country-of-origin/year, we then added the TI values for all the pesticides detected in that specific category. For example, the TI for U.S. apples tested in 1996 is the sum of the TI's for the 37 individual pesticide chemicals found on those apples that year.

TI scores for all the food/country-of-origin/year combinations covered in our analysis are summarized in Table 4.

(3) Examining "Risk Drivers" For Specific Foods The overall TI for a particular food from a particular country in a particular year indicates the aggregate amount of pesticide toxicity that the food carries. The component TI's for the individual chemicals detected in that food indicate how much each pesticide contributes to the food's overall toxicity loading. In most cases, a small number of pesticide chemicals accounts for most of the toxicity loading. For example, U.S. apples tested in 1996 had a TI of 550, and 37 different pesticides contributed to that overall score. But just three--the insecticides methyl parathion and azinphos-methyl, and the fungicide diphenylamine--have a combined TI of 407, or 74 percent of the total TI. For U.S. fresh peaches, methyl parathion alone accounts for over 90 percent of the total TI in each of the three years tested.

We call pesticide residues that account individually for large fractions of a food's total toxicity loading risk drivers. Table 5 shows the component TI's of all the individual chemicals detected in each of the tested foods. This table shows which chemicals are risk drivers and which are minor factors in the overall TI's of different foods from different countries.

C. Some Data-Analysis Issues

The overall TI for any particular food category is the sum of a group of TI's for individual chemicals found in that food, and each chemical's TI, in turn, depends in part on the mean residue level of the chemical in samples of the food. Residue levels for individual chemicals can vary widely from sample to sample of a food, and an average residue may result from a wide range of different values. If the number of samples is small (as it is for some of the imported PDP food categories), one or two samples with an extremely high residue level or with a very toxic pesticide could skew the resulting TI score. When TI scores are determined by rare or somewhat random events, apparent differences might be due to chance, and not likely to represent what would be seen if one looked repeatedly at the same comparisons.

For example, U.S.-grown potatoes were analyzed in two years of the PDP. The TI in 1994 was 191; in 1995 it was 59. Did pest management on potato farms improve markedly? Probably not. Table 5 shows 688 samples of potatoes were tested in 1994, but only 36 of those samples were tested for dieldrin. Four of the 36 were positive for this very toxic insecticide, which
was banned in the 1970s, but persists in some soils. The TI value for dieldrin alone accounts for 73 percent of the total TI for potatoes in 1994. Dieldrin was not detected in any of 702 1995 samples. Two possibilities exist: The 1994 sampling may have overstated the presence of dieldrin in potatoes, or the 1995 sampling may have understated it. But it seems quite likely that, despite the large number of samples, the two years' data do not comparably represent the occurrence of dieldrin in potatoes, and that the large decline in TI's from one year to the next is a spurious difference, not a real change.

In conducting our analysis, we sought to determine whether any differences and trends in TI's shown in Table 4 might be due to chance, or to the random occurrence of certain rare, highly toxic residues. To reduce the likelihood of such skewing effects, we applied a "rule of 10" to the data. We excluded data for a food from a specific country in a given year if the PDP
tested fewer than 10 samples of that food/country/year combination. And, within larger data sets, we excluded from our TI calculations residue data for which less than 10 samples of a food/country/year were tested.

We also used a variety of other information at hand, such as USDA pesticide use data, to assess whether the patterns we saw in the residue data made sense. Our bottom line: We believe the differences shown in Table 4 are real. However, where sample size (see Table 1) is small, comparatively small differences (of 10-20 points or less on the TI scale) between scores for different foods are not very meaningful, statistically. Large differences, and scores based on large sample sizes, are not subject to this caveat.

Goto the Next Section: Results and Discussion

Bibliography and References

1 This report was compiled in February, 1999, by the Consumers Union of the United States, Inc. Public Service Projects Department, Technical Division
Edward Groth III, PhD, Project Director
Charles M. Benbrook, PhD, Consultant
Karen Lutz, MS, Consultant
The analysis was supported in part by the Pew Charitable Trusts, the Joyce Foundation and the W. Alton
Jones Foundation.

2The initial strategy in our standardization step was to make 100 the highest score on both the ATI and CTI scales, expressing all other values as a percent of the maximum. However, there are a few extreme values in each set. If we had simply used the highest score in each set as our divisor (i.e., fixed the top of the scale at 100), the rest of the values would have been compressed into a narrow range, e.g., about 0-6 on the 100-point scale for the CTI values. We addressed this problem by choosing a representative very high score as the divisor, and allowing a few outliers to have scores greater than 100. Less than 1 percent of the raw ATI and CTI scores exceeded 100.