Regulating Pesticides in Food: The Delaney Paradox (1987)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Est~ates of Dietary Oncogenic Risks INTRODUCTION To provide some perspective on the potential impact of alternative policies for setting tolerances, this chapter assesses the estimated dietary oncogenic risks associated with 28 out of 53 pesticides that the EPA has identified as oncogenic or potentially oncogenic. The purposes of this exercise are to gain a sense of the magnitude and distribution of current dietary risks and crops associated with oncogenic pesticides and to establish an estimate from which to measure the direction and magnitude of changes in dietary risk and pesticide use that could result from different policies for setting tolerances for oncogenic pesticide residues in food. All risk estimates in this report are limited to oncogenic risks from residues of currently registered pesticides in or on food. The study focuses on the potential impact of the Delaney Clause on agricultural innovation and the public's dietary oncogenic risk. Oncogenic risks associated with exposure to residues in drinking water or other sources are not included. The risks of other chronic health effects are not examined. The committee has confined its review to risks from herbi- cides, insecticides, and fungicides that the EPA has found to be oncogenic. Plant growth regulators, rodenticides, and other types of pesticides are not considered. A number of analyses were performed on the selected pesticides. The most important analyses are examinations of the distribution of dietary risks by (1) the type of pesticide (insecticide, herbicide, or fungicide), (2) 45

46 REGULATING PESTICIDES IN FOOD type of tolerance (processed versus raw food), (3) crop, and (4) the year in which a tolerance was granted. Each analysis is presented as part of the characterization of estimated oncogenic risk. The committee wishes to make clear that emphasis should not be placed on specific risk estimates associated with particular pesticides, groups of tolerances, or food types. The analysis is subject to a wide range of uncertainty, even though based on state-of-the-art data. In developing these estimates, the committee used data that the EPA provided and followed the agency's risk assessment procedures as closely as possible. Basic questions addressed include · How many and what percentage of all pesticides used on food are currently thought by the EPA to be oncogens? · How is the risk from these pesticides distributed across crops and among types of pesticides? · How is the risk distributed by age of tolerance? · What portion of risk is associated with section 408 raw agricultural commodity tolerances in contrast to section 409 processed-food toler- ances? Pesticide Use Patterns in the United States The benefits of pesticide use are not examined in rigorous fashion in this report, nor are they considered in the process of making most decisions on tolerances. The committee lacked the time and resources to perform detailed benefit assessments for all oncogenic pesticides. Instead of benefit analyses, use and sales data are given for various pesticides and crops. This information is presented in terms of the number of acres treated with a pesticide, the pounds applied and annual expenditures. The portion of herbicide and fungicide use accounted for by oncogens is described in Tables 3-1 and 3-2. A more detailed analysis of the benefits associated with oncogenic pesticides used on eight selected crops is presented in the next chapter. To appreciate the potential impact of the Delaney Clause, one should note the percentage of all pesticide use that is accounted for by oncogenic herbicides, insecticides, and fungicides. Approximately 480 million pounds of herbicides are used annually in the United States. Of these, about 300 million pounds are agents that the EPA presumes to be oncogenic or for which positive oncogenicity data are currently under review by the agency (see Table 3-1~. These agents account for about 50 to 60 percent of all expenditures on herbicides in U.S. agriculture. In 1985, these expenditures added up to about $1.4 billion of the $2.7 billion spent on all herbicide products. "Not all oncogenic herbicides are

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 47 TABLE 3-1 Agricultural Use Information for Selected Oncogenic Herbicides All Herbicide Pounds Applied Pounds Applied Herbicidea (millions) (%) Alachlor (Lasso) 85 18 Tr~fluralin (Treflan) 39 8 Metolachlor (Dual) 38 8 Glyphosate (Roundup) 8 8 Linuron (Lorox) 7 1.5 Paraquat (Gramoxone) 2.8 1.5 Oryzalin (Surflan) 1.7 0.6 Acifluorfen (Blazer) 1.4 0.3 Subtotal 182.9 38.2 Atrazineb 79 17 2,4-Db 39 8 Total 300.9 63.3 aThe names of the biggest-selling pesticide brands are listed next to the appropriate chemical compounds to serve as examples in this table and in those following. bThese compounds are not on the list of oncogens the EPA made available to the committee (see the discussion of the Waxman list on pp. 5~51). After this correspondence, however, the EPA received data that show positive results for oncogenicity. The EPA has not officially characterized these compounds as oncogenic, but it is significant for the purposes of this report that they induced tumors when tested on animals. Also, the EPA may classify these compounds as oncogenic in the future. These compounds are not included in any risk estimates contained in this report. SOURCE: U.S. Department of Agriculture, 1984, Inputs: Outlook and Situation Report, No. IDS-6, Washington, D.C.: U.S. Government Printing Once; Gianessi, L. P., 1986, A National Pesticide Usage Data Base, Washington, D.C.: Resources for the Future, photocopy; and unpublished data from the EPA for the years 1981 through 1985. considered in the analyses presented in this report. Specifically, data indicating oncogenicity for the herbicides atrazine and 2,4-D were re- ceived by the EPA after the committee's analysis. These pesticides are included here and in Table 3-1 to indicate the potential impact of the Delaney Clause. Atrazine and 2,4-D are not treated as oncogens for any subsequent analysis presented in this study.) In terms of pounds applied, the percentage of oncogenic insecticides is relatively small. This is primarily because two oncogenic synthetic pyrethroid insecticides, permethrin and cypermethrin, are applied at very low rates per acre. The percentage of all acre treatments by oncogenic insecticides is higher, however. (One acre treatment is defined as one acre to which one pesticide has been applied one time.) Presumed oncogens

48 REGULATING PESTICIDES IN FOOD TABLE 3-2 Fungicide Use for 10 Major U.S. Food Commodities Fungicide Use Levela Oncogenic Planted Acres Total Total Oncogenic Acre Treated with Treated Fungicide Expendituresb TreatmentsC Fungicidesa Acresa Expendituresa Crop (%) (%) (%) (million) (million) Potatoes 91 80 55 3.2 16.4 Peanuts 83 85 81 6.6 38.3 Apples 53 59 78 3.2 23.5 Tomatoes 52 49 60 2.6 14.6 Plums, prunes 50 49 48 0.1 1.8 Chemes 49 47 80 0.4 3.8 Peaches 38 37 79 1.0 8.2 Almonds 27 26 78 0.7 11.5 All citrus 17 8 72 2.9 29.0 aThis includes organic and inorganic compounds. bThis is the sales value of oncogenic compounds as a percent of total fungicide sales on that crop. It includes expenditures on inorganic compounds such as copper and sulfur. CThis is expressed as the percentage of total fungicide acre treatments on that crop. It includes acre treatments with inorganic compounds. SOURCE: Webb, S.E.H., 1981, Preliminary Data: Pesticide Use on Selected Deciduous Fruits in the United States, 1978, Economic Research Service Staff Report No. AGES810626, Washington, D.C.: U.S. Department of Agriculture; Parks, J. R., 1983, Pesticide Use on Fall Potatoes in the United States, 1979, Economic Research Service Staff Report No. AGES830113, Washington, D.C.: U.S. Department of Agriculture; Ferguson, W. L., 1984, 1979 Pesticide Use on Vegetables in Five Regions, Springfield, Va.: National Technical Information Service; and unpublished data from the EPA for the years 1981 through 1985. make up between 35 and 50 percent of all insecticide acre treatments and expenditures.2 In comparison, fungicides include the highest percentage of oncogenic compounds. Table 3-2 describes fungicide use on 10 major crops. About 90 percent of all agricultural fungicides show positive results in oncogenicity bioassays. These oncogenic fungicides represent from 70 million to 75 million of the 80 million pounds of all fungicides applied annually in the United States.3 Pesticide Use Data Pesticide use patterns in U.S. agriculture and thus pesticide residues in food are changeable. In any growing season, economic factors can alter which pesticides are used on a given crop in a given area. The price

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 49 of the crop might be up or down, affecting how much growers are willing to spend for a certain amount of pest control. Weather and soil conditions can preclude or command certain treatments. The presence or absence of a given pest affects pesticide use. The emergence of pest resistance to previously applied pesticides can lead to rapid shifts in pesticide use patterns. Government acreage reduction programs and other policies alter crop- and land-use patterns, which thereby affect pesticide use. Pesticide use patterns also vary widely across major crops. Nearly all cultivated cropland in the United States is treated annually with at least one herbicide. About 15 percent of these acres receive a treatment with a fungicide. Some crops do not depend greatly on any pesticide. This is particularly true of improved pasture and hay, small grains, and some orchard crops. Virtually all perishable fresh fruits and vegetables, on the other hand, depend heavily on pesticides. Some are treated a dozen or more times each year with six or more different active ingredients. Farmers spent about $5 billion on pesticides in 1984. These costs represent a little more than 21 percent of farm expenditures for manufac- tured products such as seed, fertilizer, electricity, fuels, and oils. Pesti- cides accounted for only 4 percent of all production costs, however. Hired-labor costs were twice as much as pesticide costs; interest on debt and depreciation costs were five times as much.4 Problems in Estimating Current Risk The analytical methods involved in estimating oncogenic risks from pesticide residues in food presume resolution of complex technical and policy issues. The risk assessment methodology currently used by the EPA is guided by a set of standard procedures. These procedures are modified on an ad hoc basis when the situation warrants. In each analysis, the committee adopted what it understood to be the EPA's current methodology. The committee recognizes, however, that many key ele- ments of the agency's risk assessment procedures are under review.5 Choosing one set of assumptions can have profound implications for the resulting estimates. For example, depending on how the agency establishes average expected residue levels in food, the calculation of exposure to pesticide residues in a given foodstuff can yield risk estimates that vary by orders of magnitude. Assumptions of how and when to aggregate risks from a pesticide used on a variety of crops will also influence risk estimates. A pesticide's oncogenicity potency factor, called a Q star or Q* (see the boxed article "The 'Q Star' " on pp. 54-55), can also vary by orders of magnitude. This variation depends on such factors as whether a surface area or body weight correction is made in extrapo-

50 REGULATING PESTICIDES IN FOOD rating risks from rodents to humans, whether malignant and benign tumors are combined, and what extrapolation model is used. The EPA generally follows a conservative policy in estimating risk. The agency tries to make necessary assumptions in a way that minimizes the chance of underestimating risks. The result is that these risk estimates probably overstate true oncogenic risk. The substitution of more refined information on exposure to residues, or the potency of the oncogen at low doses, could alter risk estimates substantially. This report only notes the importance of these assumptions and underlying issues; it does not offer guidance on how to solve the problems associated with them. The EPA provided all the data used to establish the committee's estimates of current dietary risk. The committee made no adjustments in the EPA's data. In certain cases, the committee used the data in new analyses to understand the theoretical impact of different regulatory standards and methods of calculating risks and benefits. Although estimated oncogenic risks generally are presented in a quan- titative fashion, a wide margin of uncertainty surrounds nearly all of the numbers. With this in mind, the reader should focus on general patterns of risk distribution and how key parameters change when policy alterna- tives are assessed in Chapter 4, not on specific point estimates of risk. DESCRIPTION OF THE DATA BASE AND THE ANALYTICAL METHOD An estimate of a chemical's oncogenic potential generally is derived from the results of chronic feeding bioassays, which typically involve rats or mice. The committee was not charged with the task, nor did it have the expertise or resources, to review the EPA's toxicological data for the purpose of making case-by-case determinations of oncogenic potential. For this analysis, the committee adopted the list of 53 pesticides that the EPA preliminarily has determined to have oncogenic potential. The EPA transmitted this list to Congressman Henry Waxman on October 21, 1985. The pesticides on the list are presented in Table 3-3. As the EPA receives and analyzes additional oncogenicity data, some active ingredients on the list may be removed and others added. The committee did not guess how many currently untested pesticides will be oncogenic. Although more oncogenic pesticides will be found, the committee cannot say which ones or how many. In this report, pesticides that the EPA has characterized as suspect oncogens are treated as oncogens, even though a final judgment on oncogenicity may not have been reached on the basis of available data. This approach parallels EPA policy. Once the EPA determines that a pesticide has oncogenic potential, even on a preliminary basis, the

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 5 ~ pesticide is treated as an oncogen for regulatory purposes.6 For consis- tency in the following chapters, the committee treats all chemicals on the Waxman list as oncogenic compounds. In such cases the EPA usually does not approve new food tolerances for these pesticides until a thorough risk/benefit assessment of all current uses is completed The Universe of Oncogenic Pesticides In Chapter 2, the committee discussed some of the uncertainties surrounding the determination of a pesticide's potential to induce cancer. Of the 289 pesticides identified for this study as the universe of pesticides used on foods, the EPA found 53 active ingredients oncogenic or potentially oncogenic. This figure represents about 18 percent of all pesticides used on foods. Unfortunately, the data supporting many of these pesticides are incomplete. For some, particularly certain insecti- cides, most registered uses on foods have been canceled. The committee did not assess the quality or completeness of the oncogenicity data supporting these 289 pesticides. The EPA's registration standards program is designed for this purpose, however. Data supporting 115 pesticides registered for use on foods, most of which were registered before 1980, have been assessed under the program. (New active ingre- dient registrations generally require two valid oncogenicity studies and are rarely subject to a registration standard.) Of the 115 older pesticides, only 23 percent fully satisfied the oncogenicity data requirement; 41 percent had some oncogenicity data on file but did not fully satisfy the EPA's current oncogenicity data requirements; and 36 percent had no adequate oncogenicity data on file. Oncogenic risk is estimated by multiplying human exposure to pesti- cides by the Q*. The agency supplied the committee with potency factors for 30 of the 53 oncogenic pesticides currently used on food. The committee used 28 of these potency factors in generating the estimates of oncogenic risk. The number and percentage of oncogenic pesticides with available potency factors are shown in Table 3-4. Two chemicals for which the EPA has calculated potency factors, daminozide and asulam, are excluded from the committee's analysis. Daminozide, a plant growth regulator, is not characterized as a herbicide, insecticide, or fungicide; asulam has no food tolerances. The Q~'s and the food consumption and tolerance information in the EPA's Tolerance Assessment System (TAS) form the principal components of the risk calculations in this report. Table 3-4 illustrates that the committee derived its risk estimates from a roughly equivalent percentage of currently used oncogenic insecticides, fungicides, and herbicides. The portion of oncogenic active ingredients analyzed ranges from 63 percent for insecticides to 79 percent for

52 REGULATING PESTICIDES IN FOOD TABLE 3-3 Potentially Oncogenic Pesticides Identified by the EPA Year First Volume of Use Active Ingredient Tolerance (pounds active Major Crop (common/trade name) Granted Type ingredient/year)a Uses Acephateb (Orthene) 1972 Insecticide 1,900,000 Citrus Acifluorfen (Blazer) 1980 Herbicide 1,400,000 Soybeans Alachlorb (Lasso) 1969 Herbicide 85,000,000 Corn, soybeans Amitraz (Beam) 1968 Insecticide 50,000 Cattle Arsenic acid NA Herbicide NA Cotton Asulam 1975 Herbicide NA Sugar cane Azinphos-methylb 1956 Insecticide 2,500,000 Peaches, pome (Guthion) fruits Benomylb (Benlate) 1972 Fungicide 2,000,000 Citrus, rice, soybeans, stone fruits Calcium arsenate NA Insecticide NA Stone fruits Captafolb (Difolatan) 1959 Fungicide 6,000,000 Apples, cherries, tomatoes Captanb 1955 Fungicide 10,000,000 Almonds, apples, peaches, seeds Chlordimeformb 1968 Insecticide NA Cotton (Galecron) Chlorobenzilate 1956 Insecticide/ 1,500,000 Citrus acaricide Chlorothalonilb 1961 Fungicide 6,000,000 Fruits, peanuts, (Bravo) vegetables Copper arsenate 1971 Insecticide NA Vegetable crops Cypermethrinb 1984 Insecticide 600,000 Cotton (Ammo, Cymbush) ~ , , . Cyromazine~ 1984 Insecticide NA Poultry (Larvadex) Daminozide (Alar) 1967 Growth 875,000 Apples, peanuts regulator Diallate 1969 Herbicide 500,000 Sugar beets Diclofop methylb 1980 Herbicide 1,200,000 Soybeans (Hoelon) Dicofol (Kelthane) 1955 Insecticide/ 2,500,000 Citrus, cotton acaricide Ethalfluralinb 1982 Herbicide NA Soybeans (Sonalan) Ethylene oxide NA Bactericide NA Spices, walnuts Folpetb 1955 Fungicide 1,500,000 Cherries, fruits, vegetables Fosetyl Alb (Aliette) 1983 Fungicide NA Pineapples Glyphosateb 1976 Herbicide 8,000,000 Hays, orchard (Roundup) crops Lead arsenate 1955 Insecticide NA Apples, orchard crops Lindane 1955 Insecticide 200,000 Avocados, pecans Linuronb (Lorox) 1966 Herbicide 7,000,000 Soybeans Maleic hydrazide 1955 Growth 300,000 Onions, potatoes regulator Mancozebb 1962 Fungicide 16,000,000 Fruits, small (Dithane M-45) grains, vegetables

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 53 TABLE 3-3 Continued Year First Volume of Use Active Ingredient Tolerance (pounds active Major Crop (common/trade name) Granted Type ingredient/year)a Uses Manebb 1957 Fungicide 10,000,000 Fruits, small grains, vegetables Methanearsonic acid NA Herbicide 4,000,000 Cotton Methomyl (Lannate) 1963 Insecticide 4,500,000 Citrus, cotton, vegetables Metiramb 1967 Fungicide 1,000,000 Fruits, small grains, vegetables Metolachlorb (Dual) 1976 Herbicide 38,100,000 Corn, soybeans O-Phenylphenolb 1955 Fungicide 200,000 Citrus, orchard crops Oryzalinb (Surflan) 1974 Herbicide 1,700,000 Soybeans, vineyards Oxadiazonb (Ronstar) 1977 Herbicide NA Rice Paraquat 1961 Herbicide 2,800,000 Rice, soybeans ( G ram ox o ne ) Parathionb 1955 Insecticide 7,000,000 Citrus, cotton PCNB 1955 Fungicide 2,500,000 Cotton, peanuts, vegetables Permethrinb 1978 Insecticide 500,000 Vegetables (Ambush, Pounce) Pronamideb (Kerb) 1972 Herbicide 100,000 Lettuce Sodium arsenate NA Insecticide NA Pears Sodium arsenite NA Fungicide, NA Grapes herbicide, insecticide Terbutrynb 1959 Herbicide 600,000 Barley, wheat Tetrachlorvinphos 1969 Insecticide NA Cattle, poultry Thiodicarb (Larvin) 1985 Insecticide NA Cotton, soybeans Thiophanate-methyl 1972 Fungicide 30,000 Fruits, nuts, vegetables Toxaphene 1955 Insecticide NA Cattle Trifluralin (Treflan) 1963 Herbicide 39,000,000 Soybeans Zinebb 1955 Fungicide 3,500,000 Fruits, small gra~ns, vegetables aWebb, S.E.H., 1981, Preliminary Data: Pesticide Use on Selected Deciduous Fruits in the United States, 1978, Economic Research Service Staff Report No. AGES810626, Washington, D.C.: U.S. Department of Agriculture; Parks, J. R., 1983, Pesticide Use on Fall Potatoes in the United States, 1979, Economic Research Service Staff Report No. AGES830113, Washington, D.C.: U.S. Department of Agriculture; Ferguson, W. L., 1984, 1979 Pesticide Use on Vegetables in Five Regions, Springfield, Va.: National Technical Information Service; Gianessi, L. P., 1986, A National Pesticide Usage Data Base, Resources for the Future, Washington, D.C., photocopy; and unpublished data from the EPA for the years 1981 through 1985, excluding 1983, for crops affected by PIK. bThese are compounds for which risk estimates were performed.

54 REGULATING PESTICIDES IN FOOD The "Q Star" A pesticide's oncogenic potency is expressed quantitatively as a "Q star" or Q*. The Q* is the slope of the dose response curve from animal tests yielding a positive oncogenic response. The slope represents the change in tumor incidence (YJ over the change in dose (X). The units of the potency factor are tumors/mg of pesticide/kg of body weight/clay. The Q* represents the estimated tumor incidence expected to occur at the relatively low doses of pesticides in the human diet. It is based on a purely mathematical extrapolation of tumor incidence observed at the high doses used in animal tests. The potency factor does not consider the type, site, or diversity of tumors observed in animal tests. In most cases, however, the potency factors used by the EPA express a combination of malignant and benign tumors. A high Q* indicates a strong oncogenic response (more tumors) to the administered dose; a low number indicates a weak response. Most Q*'s that the committee obtained from the EPA are average Q* calculations derived from several positive oncogenicity stuclies. These Q*'s were calculated by EPA scientists and have not been formally peer reviewed. The Q* is considered a conservative or risk-averse model for quanti- fying oncogenic potency. As such, it represents the 95 percent upper- bounri confidence limit (UCL) of tumor induction likely to occur from a given close. On the other hand, the maximum likelihood estimate (MLE) represents the average probability for tumor induction from a given dose. Oncogenic potency factors derived by the two methods are similar in many cases. In some cases, however, the factors differ by several orders of magnitude, with the Q* calculation generally characterizing a com- pound as more potent. The EPA relies on the Q* at the 95 percent UCL in risk assessment to provide a margin of safety for uncertainties in characterizing the oncogenic response, for the existence of more sensitive individuals in the exposed population, and for possible synergism of pesticides and metabol ites. The committee relied solely on the Q* in estimating oncogenic potential. Therefore, the estimated oncogenic risks for certain pesticides may appear overstated. More sophisticated judgments of the human risk from dietary exposure to oncogenic agents consider qualitative evidence. This evidence includes the type of tumors produced and whether they are malignant or benign, have metastasized, or are evident in more than one sex and animal species. Such a judgment would entail a "weight-of-the- evidence" approach to risk assessment, which the EPA relies upon in regulatory decision making. The EPA's weight-of-the-evidence classifica- tion system for carcinogens is explained in the boxed article "The EPA's

ESTIMATES OF DIETARY ONCOGENIC RISKS 55 Classification System for Carcinogens," on p. 67. In Tables 3-9 and 3-17 through 3-19, risk estimates are presented with the EPA's classification of the qualitative weight of the evidence. 1 Quantitative Oncogenic Potency Factors (Q*) for Each Active Ingredient Designated by the EPA as Oncogenic Active Ingredient (trade name) Q* Chlordimeform (Fundal, Galecron) 9.4 x 10-i Linuron (Lorox) 3.28 x 10-, Oxadiazon (Ronstar) 1.3 x 10-, Ethalfluralin (Sonalan) 8.7 x 10-2 Alachlor (Lasso) 5.95 x 10 2 Oryzalin (Surflan) 3.4 x 10-2 Permethrin (Ambush, Pounce) 3.0 x 10-2 Captafol (Difolatan) 2.50 x 10-2 Chlorothalonil (Bravo) 2.4 x 10-2 Asulam 2.0 x 10 2 Cypermethrin (Ammo, Cymbush) 1.9 x 10-2 Mancozeb(Dithane M-45) 1.76 x 10 2 Maneb 1.76 x 10-2 Metiram 1.76 x 10 2 Zineb 1.76 x 10-2 Pronamide(Kerb) 1.6 x 10-2 Diclofop methyl (Hoelon) 1.1 x 10-2 Acephate (Orthene) 6.9 x 10-3 Fosetyl Al (Aliette) 4.3 x 10 3 Folpet 3.5 x 10-3 Cyromazine (Larvadex) 2.4 x 10-3 Captan 2.30 x 10-3 Metolachlor(Dual) 2.10 x 10-3 Benomyl (Benlate) 2.065 x 10-3 Terbutryn 1.87 x 10-3 Parathion 1.80 x 10-3 O-Phenylphenol 1.57 x 10-3 Glyphosate (Roundup) 5.9 x 10-5 Azinphos-methyl (Guthion) 1.5 x 10-7 fungicides. The committee received Q*'s for only 7 of 19 oncogenic insecticides. Therefore, it initially appears that the committee examined a disproportionately small number of insecticides. Conclusions regarding relative risk distribution would thus appear to be significantly influenced by the unevenness of the sample. When the sample is adjusted to account for compounds with significant use cancellation, however, the results appear more even. The situation with insecticides is unique because many oncogenic

56 REGULATING PESTICIDES IN FOOD TABLE 3-4 Number of Pesticides Identified as Oncogens by the EPA Active Ingredients Herbicides Insecticides Fungicides Other Total Number identified as 17 19 14 3 53 oncogens With Q*'sa 11/64 7/37 11/79 1/33 30/57 (number/%) No Q*'sa; major 0/0 5/26 0/0 0/0 NA uses canceledb (number/%) Total (number/%) 11/64 12/63 11/79 0/0 NA Currently used 6/36 7/37 3/21 NA NA oncogenic ingredients not considered in committee's analysis of risk (number/%) aQuantitative calculation of oncogenic potency. bThese chemicals are toxaphene, lindane, sodium arsenite, copper arsenate, and ethylene oxide. insecticides with no Q*'s have suffered widespread or total use cancella- tions. This is not the case with oncogenic herbicides or fungicides. Of the 12 insecticides with no Q*'s, 4 have had most or all uses canceled; one has tolerances for three crops and no available data on use. The risk from these compounds will not increase in the future. The committee did not consider the remaining seven insecticides in its risk analyses. Those insecticides account for only 37.3 percent of the 19 insecticides identified by the EPA as oncogenic or potentially oncogenic. The committee lacked potency estimates for a similar percentage of the fungicides and herbi- cides. Pesticide use and the impacts of certain regulatory scenarios on current patterns of use were characterized for all 53 oncogenic active ingredients when data were available and relevant to the analysis. However, this information was used primarily in the crop-level analyses in Chapters 4 and 5. Estimating Dietary Exposure to Pesticide Residues The average consumer is exposed to pesticide residues, although in minute quantities, in nearly every food, including meat, dairy products, fruits, vegetables, sugar, coffee, oils, dried goods, and most processed foods. Because pesticide residues are ubiquitous, there is great need for

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 57 a scientifically rigorous method for estimating dietary exposure to and risk from these residues. Many assumptions are necessary to estimate dietary exposure to pesticides. Simply, exposure to pesticide residues in food is a function of the quantity and type of foods one eats and the amount of residues on or in those foods. .., . FOOD CONSUMPTION DATA , . . Estimates of dietary exposure to pesticide residues are based on food consumption estimates. These can vary widely, depending on the meth- ods and data sources used. For example, the recently completed EPA TAS replaced an older EPA data base that rested upon a less sophisti- cated set of food consumption data developed in the 1960s. The TAS made possible the analyses presented in this report, and will substantially advance the EPA's capacity to assess pesticide residues in the diet. Some of the key improvements and limitations in the TAS are described below. Until 1985 the EPA used the Food Factor system, which was based on a 1966 (USDA) food consumption survey. As developed by the EPA, the Food Factor system made no distinction between the differing dietary patterns of various segments of the population. The TAS, on the other hand, is derived from the USDA's 1977-1978 food consumption survey, a nationwide study of food consumption patterns. The study contains dietary consumption estimates for 376 food types, which can be broken down by population subgroups according to sex, race, age, and region. Most significantly as far as this study is concerned, the TAS incorporates specific consumption estimates for raw and processed food forms for most crops consumed in the United States. For example, where the Food Factor system estimated consumption of raw and processed apples, TAS estimates consumption of a greater variety of raw and processed forms, such as whole apples, applesauce, apple juice, and dried apples. Tables 3-5 and 3-6 compare selected Food Factor system and TAS consumption estimates. (The TAS is described in more detail in Appendix B.) Some changes in consumption patterns are evident from these tables. Table 3-5 shows that TAS consumption estimates for milk and dairy products, citrus, tomatoes, peppers, apples, and pears are higher than those in the Food Factor system. Consumption of root vegetables, leafy vegetables, red meat, and grains has declined. The TAS ratios of raw to processed food forms of a given crop are significantly different from Food Factor ratios. Table 3-6 shows great differences between TAS and Food Factor estimates of the percentage of selected crops consumed in processed form. For three of the crops analyzed (apples, grapes, and oranges), the percentage of the crop

58 REGULATING PESTICIDES IN FOOD TABLE 3-5 Comparative Consumption of Selected Crop Groups (grams/day) Crop TAS Food Factor System Citrus a 85 57 Fruits, pomeb 47 41 Grains C 118 206 Meat, redo 132 162 Milk and dairy e 635 i 429 Vegetables, fruitingf 49 44 Vegetables, leafy g 33 41 Vegetables, root h 106 165 aCitrus fruits include grapefruit, lemons, limes, oranges, and other fruits. bPome fruits include apples and pears. CGrains include corn, oats, rice, rye, and wheat. Fred meat includes cattle, hogs, and sheep. eMilk and dairy products include fresh fluid milk, processed milk, cream, frozen milk desserts, cheese, butter, and other products. fFruiting vegetables include peppers and tomatoes. "Leafy vegetables include broccoli, cabbage, cauliflower, celery, collards, kale, lettuce, mustard greens, rhubarb, and spinach. hRoot vegetables include beets, carrots, onions, potatoes, sugar beets, sweet potatoes, and turnips. iThis consumption level was calculated with a TAS conversion factor to estimate consumption of whole fluid milk. This conversion was necessary to make TAS milk consumption, otherwise expressed as milk solids, compatible with the Food Factor consumption figures for whole milk. The committee's risk estimates for milk and dairy products do not use this conversion factor. Because of this, the risk from pesticide residues in whole milk may be underestimated. TABLE 3-6 Comparative Consumption of Selected Raw and Processed Crops (in percent) TAS Food Factor System Crop Fresh Processed Fresh Processed Apples 67 33 79 21 Grapes 29 71 60 40 Oranges 12 88 62 38 Potatoes 99 1 87 13 Tomatoes 61 39 49 51

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 59 assumed to be consumed in processed form in TAS rose 57 to 131 percent over Food Factor estimates. For two others (tomatoes and potatoes), it fell by 30 and 90 percent, respectively. Although the TAS permits analyses of estimated food consumption patterns for specific population subgroups, all analyses in this report are based on U.S. mean consumption estimates. Current food consumption patterns may be different from those during 1977-1978; therefore, the estimates used may not reflect the contemporary diet accurately. METHOD FOR ESTIMATING RESIDUES IN FOOD The committee had to use estimates for all pesticide residues in food because of very limited actual data. The estimates used are based on current food tolerances in the Code of Federal Regulations (CFR). Tolerances are typically expressed as parts per million (ppm) of pesticide X on food Y. Because pesticide residues are generally below the tolerance and only occasionally above it, the estimates used may over- estimate actual exposure. Dietary exposure to residues of a given pesticide on a given food is estimated by converting the tolerance from parts per million into milligrams in the daily diet. The conversion of an individual tolerance into a dietary intake estimate is performed by multiplying the tolerance by the consumption estimate. This number is then divided by 1,000 to attain comparable units because tolerances are expressed in parts per million and consumption estimates are expressed in kilograms. Estimated exposure levels, expressed in milligrams of pesti- cide in the daily diet, generally are aggregated for all foodstuffs in which a pesticide could be found. The TAS greatly facilitates analysis of dietary exposure patterns. It incorporates all tolerances published in the CFR and additional estimates of pesticide residues not covered by a published tolerance in processed and raw foods. (These residues are discussed further in the next subsec- tions and in Appendix B.) Although consumption estimates for drinking water and for water added or used during processing or cooking are included in the TAS, these estimates are not considered here. METHOD FOR ESTIMATING EXPOSURE TO RESIDUES The EPA traditonally has estimated dietary exposure conservatively by incorporating worst-case assumptions. Pesticide residues are assumed to be present in foods at the published tolerance level. The agency also generally assumes that 100 percent of the acreage of a crop that could be treated with a pesticide will be treated. Estimating exposure in this way nearly always produces an overestimate of actual dietary exposure across

60 REGULATING PESTICIDES IN FOOD z O 80 m E To He o 0 60 Cal 50 or: ~ —40 ~ 30 <~20 o 10 ~ o _ ~ A\\\\\ FUNGICIDES INSECTICIDES HERBICIDES FIGURE 3-1 Percentage of theoretical maximum residue contribution for oncogenic pesticides by pesticide type. the whole population, assuming pesticides are applied at the prescribed application rate. This conservative bias is reflected in the term the EPA uses for this calculation, the Theoretical Maximum Residue Contribution (TMRC). The agency has acknowledged for a long time the shortcomings of this method. The EPA contends, however, that the two assumptions together introduce a prudent safety factor in its overall assessment of pesticide risks. Figure 3-1 presents the distribution of TMRC for oncogenic fungicides, insecticides, and herbicides. Fourteen oncogenic fungicides account for 73 percent of the TMRC, 19 insecticides for 24 percent, and 17 herbicides for 3 percent. To estimate current oncogenic risk, the committee considered several ways to develop more realistic calculations of current dietary exposure. One method would be to adjust exposure calculations by taking into account the percentage of each crop treated with a pesticide. To carry out this adjustment, data would be needed on the percentage of acres of all crops treated with all pesticides. It would bias the results if the adjustment were made only for crops and pesticides for which accurate use data are available. The committee was able to compile such data for the pesticides used on some crops a step necessary in carrying out the crop analyses in the next chapter. For the risk-estimate tables in this chapter, however, lack of

ESTIMATES OF DIETARY ONCOGENIC RISKS 6~ data on use precluded this adjustment. Moreover, the validity of adjusting estimated average exposure on the basis of percentage of acres treated rests on the unlikely assumption that the residue from any given pesticide is evenly distributed throughout the population. Although the committee did not study this issue in depth, it is aware that many foods are grown and consumed within one or a few regions of the country. This is particularly true with fresh milk, and vegetables and fruits in season. Consumers in the Southeast are likely to consume a different mix and different levels of pesticide residues than consumers in California. In Chapter 5, the committee discusses some regional pesticide use problems in more detail. Other complications in estimating the level of residues in or on a food following application of a pesticide need to be considered in assessing the TMRC concept and in choosing a method to estimate exposure. First, residues found in most foods are consistently below the legal limit or tolerance, but this is not always the case. Recent General Accounting Office (GAO) and FDA reports concluded that about 3 to 4 percent of all samples checked contained unlawful residues.8 These residues were either above the tolerance level or appeared on a food for which there was no tolerance. Moreover, many foods contain residues of several pesti- cides. There is no way to know how or whether these interact. Pesticides are sometimes misused. Unexpected and undetectable residues can find their way into the food supply. For these reasons, the committee found no defensible basis or method to adjust tolerances or residue estimates systematically in the TAS. METHOD FOR ESTIMATING RESIDUES IN PROCESSED FOODS One of the committee's most important tasks was to compare the oncogenic risks to humans in raw versus processed foods. The distinction is the key to assessing the regulatory impact of the Delaney Clause. Most pesticides registered on food crops before 1978 lack tolerances and data for residues in processed foods, even though the EPA suspects that such residues are often present at concentrated levels. Dietary exposure to these pesticides will be underestimated without considering the expected residue levels in these processed foods separately from the residue levels and consumption estimates for fresh foods. To deal with such cases, the EPA built an operating assumption into the TAS: residues are presumed present in processed foods at the same level authorized on raw commodities, unless a published section 409 tolerance exists. In this case, the level that the section 409 tolerance specifies is incorporated into the TAS. Thus, if the EPA has established a tolerance for pesticide X on apples but not in apple juice, the TAS presumes a

62 REGULATING PESTICIDES IN FOOD residue level in apple juice equal to the published tolerance for raw apples. The TAS also contains conversion factors designed to adjust exposure to allow for residue concentration during processing. When a raw commodity tolerance is applied to a processed food in the TAS, the resulting exposure is multiplied by a conversion factor to incorporate the assumption that residues will concentrate during processing. These conversion factors generally are estimates, however, and are not based on scientific data. Therefore, the committee did not use TAS conversion factors in calculating any exposure or risk estimates in this report. The TAS further elaborates on the published tolerances by disag- gregating section 408 and section 409 tolerances by food type, particularly for processed foods. For the hypothesized pesticide with no section 409 tolerances used on apples, the TAS would assume residues equal to the published raw food tolerance for several forms of processed apples: applesauce, apple juice, and dried apples. The TAS makes a similar assumption for pesticides with section 409 tolerances. Often, section 409 tolerances appear in the CFR as tolerances on, for example, processed apple products. The TAS expands this single section 409 tolerance by duplicating the section 409 tolerance level for each distinct processed apple product shown in the TAS. Therefore, a pesticide with one section 408 and one section 409 tolerance on apples in the CFR may have one section 408 and four section 409 tolerances in the TAS. This adjustment produces many more processed-food tolerances in the TAS than in the CFR. Nonetheless, these expanded or duplicated section 409 tolerances are easily distinguished from the processed-food residue estimates created in the TAS when no section 409 tolerance exists in the CFR (see Appendix B). These methodological refinements yield some significant insights. For all 289 pesticides considered by the committee, the TAS contains toler- ances or residue estimates for some 3,178 distinct processed-food forms. For the same 289 pesticides, section 409 tolerances cover only 239 food forms, which is about 1 approved tolerance for every 13 processed-food forms. Consequently, pesticide residues that may be consumed in 12 out of 13 processed foods are incorporated in TAS-based exposure and risk assessments only through the assumption that residues in processed foods will equal section 408 raw food tolerance levels. A list of all oncogenic pesticides with section 409 food or feed additive tolerances is presented in Table 3-7. For oncogenic pesticides, the TAS contains 809 distinct estimates of residues in processed foods. Yet for the same pesticides, there are only 31 processed foods associated with published section 409 tolerances. Table 3-8 compares the number of TAS residue estimates for processed foods with the number of published section 409 tolerances for these same

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 63 TABLE 3-7 Presumed Oncogenic Pesticides with Section 409 Tolerances Food-Additive Tolerances Feed-Additive Tolerances Benomyl Captan Chlordimeform Daminozide Glyphosate Azinphos-methyl Dicofol Maleic hydrazide Mancozeb Oryzalin Paraquat Toxaphene Trifluralin Acephate Amitraz Benomyl Captan Chlordimeform Cyromazine Daminozide Glyphosate Azinphos-methyl Linuron Mancozeb Methanearsonic acid Paraquat Tetrachlorvinphos Thiodicarb Thiophanate-methyl processed foods. The committee believes this discrepancy deserves attention. Although the TAS residue estimates may overestimate exposure to some residues in processed foods and underestimate exposure to others, TAS is the best available tool for estimating pesticide residues in the diet. The consumption data the TAS incorporates, together with published and presumed residue levels, provide a reliable characterization of the relative magnitude and distribution of dietary exposure to pesticides. The range of error in TAS exposure estimates is well characterized compared to the uncertainties encountered at other steps in the risk assessment process. ESTIMATION OF ONCOGENIC RISK Like the EPA, the committee calculates oncogenic risk to humans by a seemingly simple formula: estimated exposure is multiplied by a pesticide's estimated oncogenic potency, which is a single number representing its tendency to induce tumors. This formula is expressed in the following equation: dietary oncogenic risk = exposure (food consumption x pesticide residues) x oncogenic potency This potency factor, or Q*, is derived from mathematical models that

64 REGULATING PESTICIDES IN FOOD TABLE 3-8 Processed Foods in the Tolerance Assessment System (TAS) Compared with Section 409 Tolerances TAS Processed Section 409 Pesticide Food Forms Tolerancesa Pesticides with food tolerances 3,178 239 All oncogens 809 31 All oncogenic herbicides, 755 insecticides, and fungicides All oncogens with Q*'s 502 All oncogenic herbicides, 490 insecticides, and fungicides with Q*'s 24 25 19 All oncogens on— Corn 16 0 Grapes 37 3 Oranges 142 0 Soybeans 60 0 Tomatoes 100 8 aThe numbers in this column represent tolerances published in the CFR plus additional processed food forms in the TAS to which these section 409 tolerances have been applied. For example, in the CFR the fungicide benomyl has one section 409 tolerance for processed tomato products. In the TAS this section 409 tolerance level is applied to five distinct processed tomato products resulting in four additional section 409 tolerances. extrapolate from data derived from animal experiments to estimate human cancer risk (see the boxed article "The 'Q Star' "I. Despite much effort to develop accurate models, there is a margin of error in the Q*'s, il as well as a significant degree of uncertainty regarding their importance to human cancer risk estimates. For example, although the models used to develop Q*'s make the assumption that the potency of the carcinogen extends to the low dose, the data do not exclude the possibility that a threshold dose exists for certain compounds that exhibit oncogenicity in animal studies. The Q* value is a probabilistic estimate of the upper bound on ncidence of extra instances of tumor formation in humans that can be expected following dietary ingestion, or exposure by other routes, of a given level of a particular chemical over a 70-year human lifetime. The Q*'s used by the committee were calculated by EPA scientists and have not been formally peer reviewed. The methods for calculating risks that the EPA and the committee use estimate extra cases of tumor formation

ESTIMATES OF DIETARY ONCOGENIC RISKS 65 in humans above existing average expectations. While uncertain, these methods are used widely by essentially all governments and international organizations faced with assessing the health consequences of human exposure to animal oncogens. It is important to understand what the risk estimates developed by these extrapolation models represent. First, the reported risk of cancer is typically expressed as a conservative upper-bound estimate of the num- ber of additional cancer cases per 10,000 (1o-4), 1OO,OOO (1o-5), or 1 million (10-6) individuals. Risk refers to incidence or frequency of cancer cases, not cancer deaths. The risk is over and above the 1 in 3.85 to 1 in 4, or about 2.5 x 10-~, lifetime risk of cancer now expected for members of the U.S. population.9 Oncogenic risks associated with pesticide residues in the diet have been calculated in many different ways for many crops and pesticides. The highest risks ever calculated for a pesticide are 1 additional cancer case in every 100 people (10-2) following a lifetime's exposure. Such a risk estimate, if accurate, would mean that the odds that an average individual would contract cancer in a lifetime would rise from about 25 percent to 26 percent.~° Further, potential cancer risk is estimated by identifying a conservative upper bound on potential human risk. The estimated risks reported in this chapter, which incorporate many conservative assump- tions regarding crops consumed and pesticide residues, are no greater than 1 x 10-3. This figure represents an increase in average individual human risk from 25 percent to 25.1 percent. The risk estimates offered here incorporate many assumptions that may overstate actual risk. These include · Conservative assumptions in the models used for extrapolating high-dose tumor incidence data in animal tests to expected low-dose . . Incidence; · Assumptions that all acres of all crops are treated with all pesticides for which they have tolerances; · Assumptions that residues are always present at the tolerance when in fact they are usually at lower levels; and · Assumptions that daily exposure to these residues occurs over the course of a 70-year lifetime. On the other hand, several other factors may work to understate the dietary risks posed by these compounds. They include · A lack of toxicological data for some active ingredients and most inerts, degradation products, and metabolites; · The possibility that the models used for extrapolating the results of animal experiments may be insufficiently conservative in certain respects; · The omission of certain routes of exposure; and

66 REGULATING PESTICIDES IN FOOD · Possible synergy of compounds and metabolites. Further, the committee's rather mechanical calculation of risk, which multiplies total exposure by a pesticide's Qua value, leads to a purely quantitative estimate of the distribution and magnitude of these risks. A more sophisticated weight-of-the-evidence approach would be desirable and would improve confidence in quantitative estimates of risk. The risk estimates for linuron and permethrin are two good examples when a weight-of-the-evidence approach probably would alter determinations of human dietary risk. (For a more detailed discussion of the committee's risk assessment procedures, see Appendix B.) Such an approach gener- ally yields somewhat different empirical results that provide a firmer sense of a compound's actual cancer risk to humans. To indicate the relative hazard of animal oncogens to humans, the EPA has developed a classification system (see the boxed article "The EPA's Classification System for Carcinogens" for further discussion). This system, adapted from the approach of the International Agency for Research on Cancer (IARC), is designed to characterize the qualitative weight of the evidence for a carcinogenic compound. The system contains five basic categories: (A) human carcinogen, (B) probable human carcin- ogen, (C) possible human carcinogen, (D) not classifiable as to human carcinogenicity, and (E) evidence of noncarcinogenicity for humans. The EPA's weight-of-the-evidence classification for the 28 pesticides is presented in certain tables to provide greater perspective of the relative human oncogenic hazard of these compounds. The committee did not use the carcinogen classification system to estimate risks or their distribution or to calculate the impacts of the scenarios described in Chapters 4 and 5. An Analysis of Estimated Oncogenic Risk Table 3-9 presents the estimates of dietary oncogenic risk for the 28 pesticides the committee was able to examine. The estimates are arranged in descending order of risk. As stated earlier, the committee examined these compounds by · Raw versus processed food; · Risk from categories of foodstuffs, including animal products; · Pesticide type (herbicides, insecticides, and fungicides); · Active ingredients; and · The date tolerances were granted. In addition, estimated risks were aggregated for combinations of these factors. For example, residues of old and new pesticides were analyzed by pesticide type in fresh and processed food. The data base compiled by

ESTIMATES OFDIETARYONCOGENIC RISKS 67 The EPA's Classification System for Carcinogens The EPA classification system for carcinogens is adapted from a similar system cleve~opecl by the International Agency for Research on Cancer. It is ,~1 hv the EPA to classify all potential human carcinogens, not just , . _ pesticicles. The purpose of the system is to characterize a compound s , . carcinogenic hazard to humans. Substances are classified based on the evaluation of such factors as the results of mutagenicity tests, consicler- ation of any negative oncogenicity results, the types and diversity of tumors inclucecl, the structural similarity of the compound to other carcinogens, and whether positive results have been replicatecl. GROUP A Human carcinogen Sufficient eviclence from epidemiologic studies to support a causal association between exposure to agents and cancer GROUP B Probable human carcinogen B. Sufficient eviclence of carcinogenicity from animal studies with limited eviclence of carcinogenicity from epidemiologic studies B2_Sufficient eviclence of carcinogenicity from animal stuclies, with inadequate or no epidemiologic data GROUP C Possible human carcinogen Limitecl evidence of carcinogenicity in the absence of human data GROUP D Not classifiable as to human carcinogenicity Inadequate or no human and animal data for carcinogenicity GROUP E—Evidence of noncarcinogenicity for humans No eviclence of carcinogenicity in at least two adequate animal tests in different species in adequate epidemiologic and animal stuclies. This classification is based on available eviclence and does not mean that the agent will not be a carcinogen under any circumstances. the committee is designed to allow analyses of dietary pesticide residue risks according to any combination of these factors. DISTRIBUTION OF RISK BY TOLERANCE TYPE: SECTION 408 VERSUS SECTION 409 The distribution of estimated oncogenic risk by tolerance type (raw versus processed food) is important because the Delaney Clause applies only when residues concentrate in processed foods above the levels allowed in raw foods. As noted in Chapter 2, the EPA has about 2,500 section 408 tolerances for oncogenic pesticides. It has approved only 31

68 REGULATING PESTICIDES IN FOOD TABLE 3-9 Estimated Oncogenic Risk from Dietary Exposure to 28 Pesticides Food Crop Weight-of- Yearof Uses Type of the-Evidence First Active Ingredient (number) Pesticide Risk Classification Tolerance Linuron (Lorox) 20 Herbicide 1.52 x 10-3 C 1966 Zineb 83 Fungicide 7.17 x 10-4 B2a 1955 Captafol (Difolatan) 34 Fungicide 5.94 x 10-4 B2 1959 Captan 83 Fungicide 4.74 x 10 4 B2 1955 Maneb 56 Fungicide 4.42 x 10-4 B2a 1957 Permethrin 43 Insecticide 4.21 x 10-4 C 1978 (Ambush, Pounce) Mancozeb 44 Fungicide 3.38 x 10-4 B2a 1962 (Dithane M-45) Folpet 41 Fungicide 3.24 x 10 4b B2 1955 Chlordimeform 24 Insecticide 3.22 x 10 4 B2 1968 (Fundal, Galecron) Chlorothalonil 47 Fungicide 2.37 x 10-4 NA 1961 (Bravo) Metiram 11 Fungicide 1.15 x 10-4 B2a 1967 Benomyl 101 Fungicide 1.13 x 10-4 C 1972 O-Phenylphenol 22 Fungicide 9.99 x 10-5 NA 1955 Acephate (Orthene) 34 Insecticide 3.73 x 10-s NA 1972 Alachlor (Lasso) 25 Herbicide 2.42 x 10-s B2 1969 Parathion . 98 Insecticide 1.47 x 10-5 C 1955 Oxadiazon (Ronstar) 26 Herbicide 1.21 x 10 s B2 1977 Oryzalin (Surflan) 57 Herbicide 1.14 x 10-5 C 1974 Pronamide (Kerb) 25 Herbicide 7.77 x 10-6 C 1972 Cypermethrin 7 Insecticide 3.73 x 10-6 C 1984 (Ammo, Cymbush) Ethalfluralin 18 Herbicide 3.56 x 10-6 NA 1982 (Sonalan) Diclofop methyl 5 Herbicide 2.04 x 10-6 NA 1980 (Holelon) Metolachlor (Dual) 40 Herbicide 1.44 x 10-6 C (pending) 1976 Cyromazine 4 Insecticide 3.58 x 10-7 NA 1984 (Larvadex) Terbutryn 4 Herbicide 2.86 x 10-7 C 1959 Glyphosate 134 Herbicide 2.73 x 10-7 C 1976 (Roundup) Fosetyl Al (Aliette) 1 Fungicide 3.29 x 10-8 C 1983 Azinphos-methyl 78 Insecticide 1.68 x 10-9 D 1956 (Guthion) NOTE: In this table and in those following, risk estimates are based on EPA data and methods using the TAS U.S. mean consumption estimates. They assume that residues are at the tolerance level, that 100 percent of all acres are treated, and that exposure occurs over a 70-year lifetime. B2 indicates a probable human carcinogen. C indicates a possible human carcinogen. D indicates not classifiable as to human carcinogenicity. NA indicates that the pesticide has not been classified by EPA. aThe classification is for the EBDC metabolite ethylene thiourea. bThis risk estimate assumes residues at the level of detection, or 0.05 ppm, for all meat, poultry, and dairy products with tolerances for chlordimeform. CThis risk estimate does not include the use of alachlor on potatoes because this use was withdrawn by the registrant when these risk estimates were calculated.

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 69 TABLE 3-10 Estimated Oncogenic Risk Distribution by Pesticide Type on Fresh and Processed Foods Fresh Food Processed Food Type of Pesticide Risk Percentagea Risk Percentagea Fungicides 2.53 x 10-3 54.5 9.33 x 10-4 77.8 Herbicides 1.44x 10-3 31.0 1.40x 10-4 11.6 Insecticides 6.73 x 10-4 14.5 1.27 x 10-4 10.6 Total 100.0 100.0 Estimated risk/percent 4.64 x 10-3 79.4b 1.20 x 10-3 20.6b total estimated risk NOTE: These risk estimates are derived using EPA data and methods described on pages 50 66 and in Appendix B. aThese percentages represent fresh or processed food risk by pesticide type. bThese are percentages of total dietary risk. section 409 tolerances for the same compounds. Many of these oncogenic pesticide residues probably will be found to concentrate. If they do, the EPA will have to establish many additional food and feed additive tolerances or consider revocation of the underlying raw commodity tolerances. For the 28 oncogenic compounds included in the committee's assess- ment of current risks, the TAS identifies 490 processed foods. Only 19 section 409 tolerances have been approved for these foods. For the remaining 471 processed foods with no section 409 tolerances, the TAS assigns the section 408 tolerance as the assumed residue level. Based on the assumption that residues in processed foods without approved toler- ances do not concentrate but are present at the same level allowed on the raw commodities, about 80 percent of estimated oncogenic risk is associated with raw foods. Processed foods account for about 20 percent (see Table 3-104. This assumed contribution of risk from raw foods suggests that the Delaney Clause will not affect many pesticide uses or substantially reduce risk. The clause's potential impact could be larger, however, for the following reasons: · In most cases, revoking processed-food tolerances would mean revoking section 408 raw-food tolerances for the crop from which the processed foods are derived. The EPA denies raw-commodity tolerances for new oncogenic active ingredients if it determines that their residues will concentrate in processed foods, which would ban them under the Delaney Clause. When the estimated oncogenic risk derived from resi- dues in processed foods is added to the estimated risk from residues in or

70 REGULATING PESTICIDES IN FOOD on their parent raw commodities, the total accounts for more than one-half of the estimated current oncogenic risk from all residues in all foods. · Under current regulations all meat, milk, and poultry products have no processed food forms. Consequently, residues in them are not subject to the Delaney Clause. Yet, the clause could indirectly reduce the dietary risk associated with these foods if residues concentrate in processed feed such as soybeans and corn and tolerances for the feeds are revoked. Because the food additives law applies only to concentrated residues in TABLE 3-11 Crop Requirements for Processing Studies Under Current EPA Guidelines Not Required Required Almonds Lettuce Apples Apricots Loganberries Barley Avocados Mangoes Beans Beet greens Milk Corn, sweet Beets Mushrooms Cottonseed Blackberries Muskmelons Grapefruit Blueberries Mustard greens Grapes Boysenberries Nectarines Lemons Broccoli Nuts Limes Brussel sprouts Onions, dry bulbs Oats Cabbage, sauerkraut Onions, green Oranges Cantaloupes Papayas Peanuts Carrots Peaches Pineapples Cattle, meat, fat Pears Plums Cauliflower Peas Potatoes Celery Peppers Rice Cherries Pimientos Rye Collards Poultry Soybeans Crab apples Pumpkins, squash Tomatoes Cranberries Quinces Wheat Cucumbers, pickles Raspberries Dewberries Rhubarb Eggplant Rutabagas Eggs Shallots Garlic Strawberries Goats Summer squash Hogs Sweet potatoes Honeydew melons Tangerines Horses Taros Kale Turnips Kohlrabi Winter squash Leeks

ESTIMATES OF DIETARY ONCOGENIC RISKS 71 TABLE 3-12 Worst-Case Impact of the Delaney Clause Estimated Risk Number of Crops Type of Reduction Affected Pesticide (number/5ro)a (number/~o)b Fungicides 2.45 x 10-3/70.7 27/19 Herbicides 5.75 x 10-4/36.4 34/20 Insecticides 2.12 x 10-4/26.5 25/16 Total 3.23 x 10-3/55.4C 38/20a' NOTE: These risk estimates are derived using EPA data and methods described on pages 50 66 and in Appendix B. This scenario assumes that tolerances for all processed foods and the parent raw commodities are revoked. aThese percentages represent reduction in estimated dietary oncogenic risk by type of pesticide. bThese are percentages of all crop registrations for oncogenic herbicides, insecticides, or fungicides by pesticide type. CThis percentage represents the total estimated oncogenic risk elimi- nated. This is the percentage of all crop registrations for pesticides comprising total estimated risk. processed foods, residues on crops with no processed-food forms escape the zero-risk standard of the Delaney Clause. Unprocessed fruits, vege- tables, and animal products account for the majority of section 408 tolerances. From Table 3-1 1 it is apparent that the Delaney Clause will not revoke or deny tolerances for many fruit and vegetable crops, all currently defined as minor crops, unless the EPA changes its definition of a processed food. The maximum impact of the Delaney Clause on the 28 pesticides can be estimated by assuming that residues concentrate in processed-food forms for all crops that have these forms. If section 408 and section 409 tolerances for oncogens were revoked or denied for all these crops pursuant to the Delaney Clause, 55.4 percent of current estimated risk would be eliminated (see Table 3-121. It is noteworthy that such a significant percentage of estimated oncogenic risk could be eliminated by tolerance revocations affecting only 20 percent of all crops. But it is also significant that residues accounting for nearly 45 percent of total esti- mated risk would remain in the diet. ONCOGENIC RISK DERIVED FROM RESIDUES IN ANIMAE FEEDS Although pesticide tolerances for most fruits and vegetables currently escape the Delaney Clause because these crops have no recognized

72 REGULATING PESTICIDES IN FOOD TABLE 3-13 Industry Recommendations of Processed By-products Requiring Tolerances Estimated Waste Used for Livestock Feed a Commodity Percentage Tons, Wet Apricots 47 4,897 Asparagus 44 15,272 Bananas b Beets, garden 20 19,733 Cabbages 9 8,024 Carrots 72 161,758 Cauliflower 80 25,326 Celery b Cherries 16 6,696 Cucumbers 25 7,447 Mung beans b Onions, bulb b Papayas b Passion fruit b Peaches 17 32,725 Pears 64 89,126 Peppers b Pimientos b Plums b Spinach 75 20,844 Sweet potatoes 90 52,052 aThese figures were calculated in J. L. Cooper (1976, The Poten- tial of Food Processing Solid Wastes as a Source of Cellulose for Enzymatic Conversion, pp. 251-271 in Biotechnology and Bioengineer~ng Symposium No. 6) from data collected by A. M. Katsuyama (1973, Solid Waste Management in the Food Processing Industry, NTIS Report No. PB 219 019, Springfield, Va.: National Technical Information Service) by questionnaire and site visitation. bFigures were not available for these commodities. processed-food form, by-products from some of these crops are increas- ingly fed to animals. The EPA currently does not recognize many of these by-products as processed animal feeds. The National Food Processors Association has suggested requiring food-processing studies to verify the need for section 409 tolerances for processed by-products of 21 crops fed to animals (see Table 3-134. The EPA has not required such studies or yet considered the need for corresponding tolerances. If the suggested residue studies confirmed the presence of concentrated residues in by-products fed to animals, section 409 tolerances for these residues would be required. For oncogenic pesticides, the Delaney Clause would

ESTIMATES OF DIETARY ONCOGENIC RISKS 73 make these feed additive tolerances difficult for the EPA to grant or, in the case of old compounds, to continue. (The FDA's sensitivity-of-the- method approach might sustain some uses, however.) Section 408 toler- ances for residues of these pesticides in animal food products could also be required. The Delaney Clause would have still greater impact if feed-additive tolerances were required for residues that concentrate in hays and fodders that (1) do not leave the farm or (2) result from certain food-manufacturing processes. The EPA's current definition of processed feed does not include these feed sources. Only certain by-products that result from specific processes such as canning, milling, or hulling are now subject to the feed-additive tolerances when residues concentrate during these processes. Thus, pesticide residues in nonprocessed feeds such as fod- ders and hays do not require section 409 tolerances, even when residues concentrate during the drying of these feeds. Section 408 regulates residues in these feeds instead. Some fodders and hays not subject to feed-additive regulations are listed in Table 3-14. Table 3-15 describes the oncogenic risk from meat, milk, dairy, and poultry products and major feed sources of residues. Were additional feed additive and food tolerances required, estimated oncogenic risk from animal products would probably increase. Because animal products have no processed-food forms under current EPA regulations, the Delaney Clause has smaller current and potential impacts on animal products than on other foods. TABLE 3-14 Animal Feeds Not Subject to Feed-Additive Regulations Alfalfa hay Almond hulls Bean hay Clover Corn fodder Cotton forage, by-products Cowpea fodder (dry vines) Flax straw Grass straw Hops (dried) Hop vines (dehydrated) Lentil hay Lespedeza hay Millet forage (dry) Milo fodder Mint hay Oat hay, fodder, straw Peanut hay Pea vine hay Peppermint hay Pigeon pea hay Pineapple fodder (if dried) Rape straw Rendered meat (cattle, poultry, swine, etc.) Rice straw Rye straw Safflower fodder Sainfoin hay Sorghum hay (fodder) Soybean hay, straw Spearmint hay Sugarcane fodder Sunflower forage (dry) Trefoil hay Vetch hay Wheat straw NOTE: Feed-additive tolerances are not required even in cases where residues concen- trate during the drying of hays or forage.

74 REGULATING PESTICIDES IN FOOD TABLE 3-15 Estimated Oncogenic Risk from Meat, Milk, Dairy, and Poultry Products Estimated Risk Type of Pesticide Number/Percent a Percentage Total Estimated Risk Major Sources of Residues Herbicides Insecticides Fungicides Total 7.73 x 10-4/69.9 3.31 x 10-4/29.9 1.59 x 10-6/00.1 1.10 x 10-3/99.9 13 6 Negligible 19 Corn, soybeans, various hays Corn, cotton, soybeans NOTE: These risk estimates are derived using EPA data and methods described on pages 50-66 and in Appendix B. There are no processed-food tolerances for meat and animal products. Processed meat (such as salami) and dairy products (such as cheese) are considered by the EPA as unique raw agricultural products. aThese figures represent risk and percent of risk from meat, milk, dairy, and poultry products. DISTRIBUTION OF ESTIMATED RISK BY TYPE OF PESTICIDE Table 3-16 and Figure 3-2 present the distribution of estimated oncogenic risk by fungicides, herbicides, and insecticides. Nearly 60 percent of all estimated risk is from fungicides, 27 percent is from herbicides, and 13 percent is from insecticides. Almost all estimated herbicide risk is from a single compound, linuron, which in large part reflects that compound's relatively high tolerances. Two active ingredi- ents, chlordimeform and permethrin, account for nearly all estimated dietary oncogenic risk from insecticides. Estimated risk from fungicides is TABLE 3-16 Distribution of Estimated Oncogenic Risk by Pesticide Type Type of Pesticide Risk (number/%) Fungicides Herbicides Insecticides Total 3.46 x 10-3/59.2 1.58x10-3/27.1 8.00 x 10-4/13.7 5.84 x 10 3 /100 NOTE: These risk estimates are derived using EPA data and methods described on pages 50~6 and in Appendix B. aThis number is the total of all upper-bound estimates of dietary oncogenic risk for the 28 compounds examined by the committee. It does not represent a total estimated dietary oncogenic risk to individuals in the population, but rather serves as a benchmark from which to measure distribution and reduction of risk.

ESTIMA TES OF DIETAR Y ONCOGENIC RISKS 75 60 50 40 - ~o on - c,, 30 - C: 20 10 O' FUNGICIDES HERBICIDES INSECTICIDES FIGURE 3-2 Percentage of estimated dietary oncogenic risk from fungicides, herbicides, and insecticides. fairly evenly distributed among the nine compounds presenting the greatest estimated risks. The committee found that roughly 20 percent of the current estimated total dietary oncogenic risk is associated with consumption of processed foods; fungicides account for three-fourths of this risk and nearly 60 percent of total oncogenic risk (see Table 3-101. This contribution is extraordinary. Most dietary risk from fungicides is attributable to crops that account for only 15 percent of all planted acres, and fungicides comprise only about 10 percent of all pesticides applied to food crops. DISTRIBUTION OF ESTIMATED RISK BY ACTIVE INGREDIENT When the risks attributable to individual active ingredients are ana- lyzed, the markedly different distribution of risks within the categories of herbicides, insecticides, and fungicides is obvious. One herbicide, linuron, represents more than 98 percent of all estimated oncogenic risk from herbicides (see Table 3-17~. Two insecticides, chlordimeform and permethrin, contribute more than 95 percent of all estimated dietary risk from insecticides (see Table 3-181. For the nine principal oncogenic fungicides, however, no single active ingredient accounts for more than 20 percent of all estimated fungicide risk (see Table 3-191. Reducing dietary risk from insecticides and herbicides is primarily a matter of reducing or eliminating exposure to several presumably high-

76 REGULATING PESTICIDES IN FOOD TABLE 3-17 Estimated Oncogenic Risk from Dietary Exposure to Selected Herbicides Estimated Number Dietary Weight-of- of Food TMRCa Oncogen~c the-Evidence Active Ingredient Uses (ma pesticide) Q* Risk Classification Linuron (Lorox) 20 4.65 x 10-3 3.28 x 10-' 1.52 x 10-3 C Alachlor (Lasso) 25 4.08 x 10-4 5.95 x 10-2 2.42 x 10-5 B2 Oxadiazon 26 9.38 x 10-5 1.3 x 10-' 1.21 x 10-5 B2 (Ronstar) Oryzalin (Surflan) 57 3.37 x 10-4 3.4 x 10-2 1.14 x 10-5 C Pronamide 25 4.86 x 10-4 1.6 x 10-2 7.77 x 10-6 C (Kerb) Ethalfluralin 18 4.09 x 10-5 8.7 x 10-2 3.56 x 10-6 NA (Sonalan) Diclofop methyl 5 1.86 x 10-4 1.l X 1O-~ 2.~ x 10-6 NA (Hoelon) Metolachlor 40 6.88 x 10-4 2.10 x 10-3 1.44 X 10-6 C (pending) (Dual) Terbutryn 4 1.53 x 10-4 1.87 x 10-3 2.86 x 10-7 C Glyphosate 134 4.63 x 10-3 5.9 X 10-5 2.73 x 10-7 C (Roundup) NOTE: These risk estimates are derived using EPA data and methods described on pages 5~66 and in Appendix B. B2 indicates a probable human carcinogen. C indicates a possible human carcinogen. NA indicates that the pesticide has not been classified by the EPA. aThis column expresses the theoretical maximum residue contribution in the diet. See Appendix B for further discussion of TMRC. risk compounds. Risks that might be posed by likely replacement chem- icals must also be considered. If exposure to this one herbicide and these two insecticides were significantly reduced, the estimated dietary oncogenic risk would fall by about 30 percent. If this step were taken, the share of the remaining estimated risk contributed by fungicides would increase to nearly 99 percent. In this circumstance, the 10 agents presenting the greatest estimated oncogenic risk would all be fungicides. Unlike herbicides and insecticides, major reductions in estimated fungi- cide risk cannot be attained by reducing or eliminating use of any single agent. DISTRIBUTION OF ESTIMATED RISK BY CROP Examination of the risk from herbicides, insecticides, and fungicides on crops reveals a similar pattern. In each case the same herbicide (linuron) and the same insecticides (chlordimeform and permethrin) represent more than 99 percent of the total estimated herbicide and insecticide risk

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 77 TABLE 3-18 Estimated Oncogenic Risk from Dietary Exposure to Selected Insecticides Estimated Number Dietary Weight-of- of Food TMRCa Oncogenic the-Evidence Active Ingredient Uses (ma pesticide) Q* Risk Classification Permethrin (Ambush, 50 1.40 x 10-2 3.0 x 10-2 4.21 x 10-4 C Pounce) Chlordimeform 24 7.19 x 10-3 9.4 x 10-' 3.22 x 10-4 B2 (Fundal, Galecron) Acephate (Orthene) 34 5.41 x 10-3 6.9 x 10-3 3.73 x 10-5 C Parathion 98 8.19 x 10-3 1.8 x 10-3 1.47 x 10-5 C Cypermethrin 7 1.97 x 10-4 1.9 x 10-2 3.73 x 10-6 C (Ammo, Cymbush) Cyromazine 4 1.49 x 10-4 2.4 x 10-3 3.58 x 10-7 NA (Larvadex) Azinphos-methyl 78 1.13 x 10-2 1.5 x 10-7 1.68 x 10-9 D (Guthion) NOTE: These risk estimates are derived using EPA data and methods described on pages 50-66 and in Appendix B. B2 indicates a probable human carcinogen. C indicates a possible human carcinogen. D indicates not classifiable as to human carcinogenicity. NA indicates that the pesticide has not been classified by the EPA. aThis column expresses the theoretical maximum residue contribution in the diet. TABLE 3-19 Estimated Oncogenic Risk from Dietary Exposure to Selected Fungicides Estimated Number Dietary Weight-of- of Food TMRCa Oncogenic the-Evidence Active Ingredient Uses (ma pesticide) Q* Risk Classification Zineb 83 4.08 x 10-2 1.76 x 10-2 7.17 x 10-4 B2b Captafol 34 2.38 x 10-2 2.5 x 10-2 5.94 x 10-4 B2 Captan 77 2.06 x 10-' 2.3 x 10-3 4.74 x 10-4 B2 Maneb 56 2.52 x 10-2 1.76 x 10-2 4.42 x 10-4 B2b Mancozeb 44 1.92 x 10-2 1.76 x 10-2 3.38 x 10-4 B2b Folpet 41 9.26 x 10-2 3.5 x 10-3 3.24 x 10-4 B2 Chlorothalonil 47 9.91 x 10-3 2.4 x 10-2 2.37 x 10-4 NA (Bravo) Metiram 11 6.58 x 10-3 1.76 x 10-2 1.15 x 10-4 B2b Benomyl (Benlate) 101 5.49 x 10-2 2.07 x 10-3 1.13 x 10-4 C O-Phenylphenol 22 6.37 x 10-2 1.57 x 10-3 9.99 X 10-5 NA Fosetyl Al (Aliette) 1 7.67 x 10-6 4.3 x 10-3 3.29 x 10-8 C NOTE: These risk estimates are derived using EPA data and methods described on pages 50-66 and in Appendix B. B2 indicates a probable human carcinogen. C indicates a possible human carcinogen. NA indicates that the pesticide has not been classified by the EPA. aThis column expresses the theoretical maximum residue contribution in the diet. See Appendix B for further discussion of TMRC. bThe classification is for the EBDC metabolite ethylene thiourea.

78 REGULATING PESTICIDES IN FOOD 80 70 60 ^ 50 To of - on a: 40 30 20 10 : ~ 15 FOODS 186 OTHER FOODS FIGURE 3-3 Concentration of total estimated dietary oncogenic risk in selected foods. (Also see Table 3-20.) TABLE 3-20 Fifteen Foods with the Greatest Estimated Oncogenic Risk Total Dietary Oncogenic Risk Estimates Food Number Percentage Tomatoes 8.75 x 10-4 14.9 Beef 6.49 x 10-4 1l.l Potatoes 5.21 x 10-4 8.9 Oranges 3.76 x 10-4 6.4 Lettuce 3.44 x 10-4 5.8 Apples 3.23 x 10-4 5.5 Peaches 3.23 x 10-4 5.5 Pork 2.67 x 10-4 4.5 Wheat 1.92 x 10-4 3.3 Soybeans 1.28 x 10-4 2.2 Beans 1.23 x 10-4 2.1 Carrots 1.22 x 10-4 2.1 Chicken 1.12 x 10-4 1.9 Corn (bran, grain) 1.09 x 10-4 1.9 Grapes 1.09 x 10-4 1.9 Total 78.0 NOTE: These worst-case risk estimates are derived using EPA data and methods described on pages 50-66 and in Appendix B. They assume residues are at the tolerance level, although actual residues may be different. These numbers are the totals of the committee's upper-bound estimates of dietary oncogenic risk for oncogenic pesticides with tolerances on these crops. As more accurate data are received by the EPA, crops may move on or off the list.

ES TIMA TES OF DIETAR Y ONCOGENIC RISKS 79 TABLE 3-21 Estimated Oncogenic Risk from Herbicides in Major Foods Estimated Risk Crop Number Percentage Beef Potatoes Pork Soybeans Wheat Carrots Corn Asparagus Celery Milk Percentage of total risk from herbicides, insecticides, and fungicides o-5 5.38 x 10-4 10.0 3.89 x 10-4 6.7 2.17 x 10-4 3.7 1.22x10-4 2.1 1.22x10-4 2.1 5.95x 1 4.98 x 10-5 1.48 x 10-5 1.04 x 10-5 7.87x 10-6 1.0 0.9 0.3 0.2 0.1 27.1 NOTE: These worst-case risk estimates are derived using EPA data and methods described on pages 50 66 and in Appendix B. They assume residues are at the tolerance level, although actual residues may be different. These numbers are the totals of the committee's upper-bound estimates of dietary oncogenic risk for oncogenic herbicides with tolerances on these crops. As more accurate data are received by the EPA, crops may move on or off the list. associated with most crops. With fungicides, estimated risk is again more evenly distributed among active ingredients and among crops; no single fungicide accounts for more than 43 percent of total fungicide risk from any crop. Further, the ranking of estimated risk from fungicides varies from crop to crop. When estimated risks from individual foods are ranked, 15 crops and animal products contribute nearly 80 percent of all estimated dietary oncogenic risk from pesticide residues (see Figure 3-3 and Table 3-20~. Tables 3-21, 3-22, and 3-23 show estimated crop risk broken down among fungicides, herbicides, and insecticides. Fungicide residues on just 10 crops represent 42 percent of total estimated dietary risk. In general, relatively few pesticides account for high percentages of total estimated risk in selected high-risk foods (see Tables 3-24, 3-25, 3-26, and 3-274. About 99 percent of herbicide risk from estimated residues in beef is from one herbicide; 72 percent of estimated tomato risk is from five fungicides (see the boxed article "Concentration of Residues in Processed Foods". Many of these fungicides are close substitutes in controlling a range of diseases.

80 REGULATING PESTICIDES IN FOOD TABLE 3-22 Estimated Oncogen Foods ic Risk from Insecticides in Major Estimated Risk Crop Number Percentage Lettuce Chicken Beef Cottonseed Milk Tomatoes Pork Peaches Spinach Cabbage Percentage of total risk from herbicides, insecticides, and fungicides NOTE: These worst-case risk estimates are derived using EPA data and methods described on pages 50-66 and in Appendix B. They assume residues are at the tolerance level, although actual residues may be different. These numbers are the totals of the committee's upper-bound estimates of dietary oncogenic risk for oncogenic insecticides with tolerances on these crops. As more accurate data are received by the EPA, crops may move on or off the list. 1.59 x 10-4 1.11 X 10-4 1.10 x 10-4 9.95 X 10-5 5.19 x 10-5 5.17 x 10-5 5.02 x 10-5 3.50 x 10-5 2.80 x 10-5 1.82 x 10-5 2.5 1.9 1.9 1.7 0.9 0.9 0.9 0.6 0.5 0.3 12.1 TABLE 3-23 Estimated Oncogenic Risk from Fungicides in Major Foods Estimated Risk Crop Number Percentage Tomatoes 8.23 x 10-4 14.1 Oranges 3.72 x 10-4 6.3 Apples 3.18 x 10-4 5.4 Peaches 2.86 x 10-4 4.9 Lettuce 1.81 x 10-4 0.1 Potatoes 1.29 x 10-4 2.2 Beans 1.17 x 10-4 2.0 Grapes 1.08 x 10-4 1.8 Wheat 6.65 x 10-5 1.1 Celery 6.04 x 10-5 1.l Percentage of total risk from herbicides, 42.0 insecticides, and fungicides NOTE: These worst-case risk estimates are derived using EPA data and methods described on pages 50-66 and in Appendix B. They assume residues are at the tolerance level, although actual residues may be different. These numbers are the totals of the committee's upper-bound estimates of dietary oncogenic risk for oncogenic fungicides with tolerances on these crops. As more accurate data are received by the EPA, crops may move on or off the list.

ESTIMATES OF DIETARY ONCOGENIC RISKS 81 Concentration of Residues in Processed Foods The data in the Tolerance Assessment System (TAS) indicate that under EPA's worst-case assumptions the estimated dietary oncogenic risk from tomato products may be 1 5 percent of the total dietary oncogenic risk from pesticide resiclues. The committee was greatly interested in the finding that one crop may constitute such a significant percentage of total dietary risk. Further, more than 90 percent of this estimated risk is attributable to fungicides and is derived assuming that residues do not concentrate in processed tomato products. Most oncogenic fungicides have no section 409 tolerances for pro- cessed tomato products. The EPA has never considerecl the need for section 409 tolerances because it has not yet received the data required to set such tolerances. The committee expects concentration to occur for many crops whenever food processing involves cirying, removing water, or extracting oils or other fractions of raw agricultural commodities. To illustrate the potential impact that residue concentration could have on the distribution and character of dietary oncogenic risk, the committee undertook several analyses of fungicide use on tomatoes. The method was simple. The committee made assumptions regarding the expected residue level on the raw crop as well as the expected ~eve! of concentration in processed products from this raw residue level. The committee then computed the ensuing worst-case risks. The results are presented below. If one assumes that all acres are treated, that residues on raw tomatoes are at the current tolerance level, and that these residues in processed tomato products undergo a 10-fold concentration, then total estimated oncogenic risk from all fungicide residues in tomatoes would increase more than 300 percent above the committee's risk estimates, which assume no concentration of residues in processed foocis. If residues on raw tomatoes are assumed to be one-tenth the published tolerance and to undergo a 10-fold concentration, then estimated oncogenic risk from fungicide residues would decline about 51 percent. Alternatively, if residues on raw tomatoes are present at one-ha~f the tolerance ~eve! and concentrate by a factor of 10, estimated oncogenic risk from fungicides in tomato products wou~cl increase more than 1 1 8 percent. As the residue chemistry data base and tolerance profile are modern- izecl for all older products, the committee expects the following com- pared to current risk estimates: · The actual residues likely to be found on the vast majority of fresh foods will decreases

82 REGULATING PESTICIDES IN FOOD · The percentage of dietary oncogenic risk associated with processed- food forms will increase; and · The share of dietary risk stemming from relatively few foods and pesticides wi ~ ~ increase. Assumed Levels Concentration in Processed Raw Residue Products (from raw residue) New Riska/% Change CFR None 6.19 x 10-4 CFR tolerance 10x 2.67 x 10-3/ + 331 1/2 CFR tolerance 1 0 x 1 .35 x 1 o-3/ + 1 1 8 1/10 CFR tolerance 1 0 x 3.02 x 1 0-4 / - 51 CFR tolerance 2x 8.47 x 10-4 / + 36 1/2 CFR tolerance 2x 4.43 x 10-4/ - 28 aRisk assumes U.S. mean consumption estimates from the TAS, that residues are at the raw food tolerance, that all acres are treated, and that exposure occurs over a 70-year lifetime. The oncogenic fungicides metiram, folpet, O-phenylphenol, and zineb were not included in this analysis due to limited commercial use. Herbicide residues on three foods account for more than 20 percent of estimated dietary risk. Nearly all of this risk is from residues at the tolerance level of one compound, linuron. Risks from insecticides, on the other hand, are evenly distributed among the top 10 crops. Several conclusions emerge in considering the distribution of risks by crop and active ingredient. Where one chemical dominates risk, estimated dietary risk can be significantly reduced through action on tolerances for that chemical. For example, estimated dietary oncogenic risk from beef (about 11 percent of the total), could be reduced by 90 percent through actions lowering or revoking tolerances for linuron on beef products. Revoking all tolerances for linuron would reduce the committee's total estimate of dietary oncogenic risk by about 30 percent. By contrast, estimated risk from tomatoes is primarily from fungicide residues. No single compound accounts for more than 38 percent of the total. Reducing residues from a single chemical will not significantly reduce estimated risk from tomatoes, because the oncogenic fungicides used in tomato production generally substitute for one another. The same generalization applies when one considers all fungicide active ingredients

ESTIMATES OFDIETARYONCOGENIC RISKS 83 TABLE 3-24 Estimated Oncogenic Risk from All Active Ingredients Used on Selected Foods Estimated Risk Active Raw Processed Total Crop Ingredient Food Food Total Risk (~c) Tomatoes Captafol 1.90 x 10-4 1.23 x 10-4 3.14 x 10-4 36 Chlorothalonil 6.09 x 10-5 3.95 x 10-5 1.00 X 10-4 11 Folpet 4.44 x 10-5 2.88 x 10-5 7.32 x 10-5 8 Metiram 3.58 x 10-5 2.32 x 10-5 5.89 x 10-5 7 Beef Linuron 5.34 x 10-4 0 5.34 x 10-4 82 Chlordimeform 7.77 x 10-5 0 7.77 x 10-5 12 Permethrin 2.98 x 10-5 0 2.98 x 10-5 6 Oxadiazon 2.12 x 10-6 0 2.12 x 10-6 0.3 Alachlor 1.94 x 10-6 0 1.94 x 10-6 0.3 Potatoes Linuron 3.87 x 10-4 5.64 x 10-7 3.88 x 10-4 74 Captan 6.79 x 10-5 9.88 x 10-8 6.80 x 10-5 13 Mancozeb 2.08 x 10-5 3.03 x 10-8 2.08 x 10-5 4 Captafol 1.48 x 10-5 2.15 x 10-8 1.48 x 10-5 3 Metiram 1.04 x 10-5 1.51 x 10-8 1.04 x 10-5 2 NOTE: These risk estimates are derived using EPA data and methods described on pages 50 66 and in Appendix B. across all of their uses. Eliminating residues of a single fungicide compound will not achieve dramatic reductions in estimated risks. EPA'S INTERPRETATION OF THE DELANEY CLAUSE TO DATE The EPA has applied the Delaney Clause unevenly to the 289 pesticides examined here. It has never invoked the clause to repeal an existing tolerance. (Many of these tolerances were established in the absence of oncogenicity data or information indicating residue concentration.) Con- versely, the agency has enforced the Delaney Clause strictly to refuse section 408 and section 409 tolerances on all crops when section 409 tolerances are needed for new oncogenic active ingredients registered since 1978. This has been the policy since the agency required more complete data before granting initial tolerances. This policy helps explain why most estimated dietary oncogenic risk is associated with tolerances

84 REGULATING PESTICIDES IN FOOD TABLE 3-25 Foods with the Greatest Estimated Oncogenic Risk from Herbicides Estimated Risk Risk from Active Raw Processed Herbicides Ingredient Food - Food Total on Crop (%) Beef Linuron Oxadiazon Alachlor Terbutryn Glyphosate . Lauren Oryzalin Glyphosate Linuron Oxadiazon Alachlor Terbutryn Glyphosate 5.34 x 10-4 2.12 x 10-6 1.94 x 10-6 5.92 x 10-7 7.23 x 10-l° 3.87 x 10-4 2.01 x 10-6 1.39 x 10-8 2.16 x 10-4 8.54 x 10-7 7.82 x 10-7 2.27 x 10-7 1.68 x 10-1° Potatoes 5.64 x 10-7 2.72 x 10-9 2.0 x 10- Pork o o o o o 5.34 x 10-4 2.12 x 10-6 1.94 x 10-6 5.92 x 10-7 7.23 x 10-1° 3.88 x 10-4 2.01 x 10-6 1.40 x 10-8 2.16 x 10-4 8.54 x 10-7 7.82 x 10-7 2.27x 10-7 1.68 x 10-1° 99 0.4 0.4 Negligible Negligible 99 0.5 Negligible 99 Negligible Negligible Negligible Negligible NOTE: These risk estimates are derived using EPA data and methods described on pages 50-66 and in Appendix B. TABLE 3-26 Foods with the Greatest Estimated Oncogenic Risk from Insecticides Estimated Risk Risk from Active Processed Insecticides Ingredient Raw Food Food Risk on Crop (Jo) Lettuce Permethrin 1.43 x 10-4 0 1.43 x 10-4 90 Acephate 1.54 x 10-5 0 1.54 x 10-5 10 Parathion 4.30 x 10-7 0 4.30 x 10-7 Negligible Chicken Chlordimeform 1.11 x 10-4 0 1.ll X 10-4 99 Permethrin 7.06 x 10-7 0 7.06 x 10-7 0.1 Acephate 3.25x 10-7 0 3.25x 10-7 Negligible Beef Chlordimeform 7.77 x 10-5 0 7.77 x 10-5 71 Permethrin 2.98 x 10-5 0 2.98 x 10-5 27 Cypermethrin 1.55 x 10-6 0 1.55 x 10-6 1.4 Acephate 1.12 x 10-6 0 1.12 x 10-6 1 Azinphos-methyl 2.04 x 10-" 0 2.04 x 10-" Negligible NOTE: These risk estimates are derived using EPA data and methods described on pages 50~6 and in Appendix B.

ESTIMATES OF DIETARY ONCOGENIC RISKS 85 TABLE 3-27 Foods with the Greatest Estimated Oncogenic Risk from Fungicides Estimated Risk Risk from Active Raw Processed Fungicides Ingredient Food Food Total on Crop (56) Tomatoes Captafol 1.90 x 10-4 1.23 x 10-4 3.14 x 10-4 38 Chlorothalonil 6.09 x 10-5 3.95 x 10-5 1.00 X 10-4 12 Folpet 4.44 x 10-5 2.88 x 10-5 7.32 x 10-5 9 Metiram 3.58 x 10-5 2.32 x 10-5 5.89 x 10-5 7 Captan 2.92 x 10-5 1.89 x 10-5 4.81 x 10-5 6 Oranges Zineb 2.07 x 10-5 1.42 x 10-4 1.62 x 10-4 43 Captan 9.68 x 10-6 6.62 x 10-5 7.58 x 10-5 20 Folpet 8.84 x 10-6 6.04 x 10-5 6.92 x 10-5 9 Benomyl 3.48 x 10-6 2.38 x 10-5 2.72 x 10-5 7 O-Phenylphenol 2.64 x 10-6 1.81 x 10-5 2.07 x 10-5 6 Apples Mancozeb 5.97 x 10-5 3.08 x 10-5 9.05 X 10-5 28 Folpet 4.24 x 10-5 2.19 x 10-5 6.43 x 10-5 20 Captan 2.79 x 10-5 1.44 x 10-5 4.22 x 10-5 13 O-Phenylphenol 1.90 x 10-5 9.81 x 10-6 2.88 x 10-5 9 Metiram 1.71 x 10-5 8.79 x 10-6 2.59 x 10-5 8 NOTE: These risk estimates are derived using EPA data and methods described on pages 50 66 and in Appendix B. for older pesticides. In this report, new chemicals are those registered since 1978, and old chemicals are those registered earlier. In 1978, FIFRA amendments imposed new data demands, as well as other regulatory requirements, on registrants. The distribution of estimated oncogenic risk associated with tolerances granted over time indicates that most risk comes from old chemicals, particularly old fungicides (Table 3-281. For herbicides, one active ingre- dient registered in 1965 contributes most of the risk. For fungicides, old pesticides account for 97.7 percent of the estimated risk. Tolerances for one insecticide, permethrin, granted after 1978 account for slightly more than half of all insecticide risk and less than 5 percent of total estimated dietary oncogenic risk. As shown in Figure 3-4, more than 90 percent of all estimated dietary oncogenic risk is associated with tolerances granted before 1978. When all pesticides are considered (see Table 3-28), the trend is clear: the estimated dietary oncogenic risk associated with tolerances granted after 1978 is very small when compared with the risk associated

86 REGULATING PESTICIDES IN FOOD TABLE 3-28 Estimated Oncogenic Risk from Tolerances over Time Tolerance Type Raw Food Processed Food Total Type 0 Pesticide Number Percenta Number Percenta Number Percenta Fungicides Pre-1970 1.20 x 10-3 76.1 1.14 x 10-4 7.26 1.31 x 10-3 83.4 197~1977 2.04 x 10-4 12.9 2.20 x 10-5 1.39 2.26 x 10-4 14.3 1978-1985 1.54 x 10-5 0.97 2.66 x 10-6 0.16 1.81 x 10-5 1.14 Total 1.55 x 10-3 98.84b Insecticides Pre-1970 2.82x 10-6 0.3 1.17x 10-8 0.001 2.83 x 10-6 0.35 1970~1977 2.57 x 10-4 32.1 1.01 x 10-4 12.6 3.58 x 10-4 44.8 1978-1985 4.04 x 10-4 50.6 2.11 x 10-5 2.6 4.26 x 10-4 53.2 Total 7.86 x 10-4 98.35b Herbicides Pre-1970 4.50x 10-4 13.0 4.94x 10-5 1.42 5.00x 10-4 14.46 197~1977 1.69 x 10-3 49.0 6.57 x 10-4 18.9 2.35 x 10-3 68.0 1978-1985 1.25 x 10-5 0.36 3.89 x 10-5 1.1 5.15 x 10-5 1.4 Total 2.90 x 10-3 83.86b NOTE: These risk estimates are derived using EPA data and methods described on pages 50 66 and in Appendix B. aThese columns express the percent of total risk by pesticide type. bNumbers do not equal 100 percent because petition numbers indicating the year of a tolerance petition were not available for all tolerances. with tolerances granted before 1978. The reasons for this are (1) the fact that the few oncogenic pesticides registered since 1978 generally present less dietary oncogenic risk than those registered before 1978 and (2) the EPA's application of the Delaney Clause. The distribution of estimated risk between raw- and processed-food tolerances by time also is presented in Table 3-28. About one-fifth of all estimated dietary oncogenic risk is associated with residues of pesticides in processed food; nearly 80 percent of this risk is derived from tolerances granted before 1978. EPA Application of the Delaney Clause to New Active Ingredients From 1975 through 1981 the EPA issued a series of standards and requirements for data to support pesticide registrations. In 1982 the EPA published a proposed rule consolidating all testing requirements. The EPA's final rule, which did not differ significantly from the 1982 proposal, became

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 87 effective in April 1985. Although the EPA had required oncogenicity testing for many pesticides before the development of these formal data require- ments, it was clear for the first time which studies would be routinely required, which criteria would govern exemptions from data requirements, and which testing protocols registrants would have to follow. In the late 1970s, pesticide oncogenicity and residue concentration data became available for newly registered pesticides. The EPA began explor- ing the regulatory ramifications of the Delaney Clause in the granting of tolerances for new active ingredients. Simultaneously, the agency began confronting questions about the applicability of the Delaney Clause in the review and reregistration of old pesticides. To explore the history of the EPA's use of the Delaney Clause, the committee wrote to the agency, seeking confirmation of a list of decisions in which the clause had been discussed or relied on. In response, the EPA identified six regulatory actions on petitions for new tolerances in which the Delaney Clause was specifically cited. These cases are presented in Table 3-29. (Detailed case studies of these pesticides are contained in Appendix C.) The EPA also noted 10 additional pesticides that the Delaney Clause may affect in the review of existing registrations. These cases are presented in Table 3-30. 100 90 80 70 60 so on 0~, 50 - ~: 40 30 20 10 O ~ ~ PRE-1978 1978-85 FIGURE 3-4 Risk from tolerances granted before and after 1978.

8 8 REG ULA TING PES TI CIDES IN FO OD TABLE 3-29 Tolerance Actions for Which the Delaney Clause Was Cited Risk Associated Chemical Name/ Concern Related to with Tolerance Regulatory Action Delaney Clause Application Agency Decision Fosetyl Al Application for section Need for section 409 1 x 10-8 Denied: Delaney 408 tolerance in hops tolerance and Clause cited oncogenicity of fosetyl Al Amitraz Application for section Need for section 409 1 x 10-6 Denied: apples, 408 tolerance in tolerance in Delaney Clause apples, pears apples and cited oncogenicity of Allowed: pears, no amitraz processed food form Dicamba Application for section Need for section 409 1 x 10-~° Allowed: FDA 408 tolerance in tolerance and constituents policy sugarcane oncogenicity of cited nitrosamine contaminant Larvadex Application for section Need for section 409 1 x 10-8 Allowed: 408 tolerance in tolerance and SOMa/FDC Act eggs/poultry oncogenicity of cited metabolite acetamide Permethrin Application for section Need for section 409 1 x 10-4 Denied: Delaney 408 tolerance in tolerance and Clause cited b tomatoes oncogenicity of permethrin Thiodicarb Application for section Need for section 409 1 x 10-8 Allowed: SOM/FDC 408 and 409 tolerances tolerance and Act cited in cotton oncogenicity of metabolite acetamide NOTE: These risk estimates are derived using EPA data and methods described on pages 50-66 and in Appendix B. aSOM is the sensitivity-of-the-method procedure. This has been used in cases where the residue of an oncogenic chemical in an animal product is deemed to present an additional oncogenic risk of less than 1 in 1 million (1 x 10-6) and can be reliably detected using modern residue chemistry techniques. In these cases the Delaney Clause has been bypassed. (For further discussion, see Chapter 2 and the case studies in Appendix C.) bSection 408 tolerances for tomatoes grown in Florida and destined only for fresh market were granted.

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 89 TABLE 3-30 Pesticide Active Ingredients Under Review for Which the Delaney Clause Has Been a Concern Estimated Major Uses and Oncogenic Volume of Usea Pesticide Pesticide Type Risk (pounds AI/year) Special Review Alachlor (Lasso) Corn, soybean herbicide 2.42 x 10-5 85,100,000 Dicofol (Kelthane) Citrus, cotton acaricide/ No risk assess- 1,200,000 insecticide ment per- formed Captan Fruit, vegetable fungicide 4.74 x 10-4 1O,OOO,OOO Daminozide (Alar) Select fruit, vegetable 8.30 x 10-3 825,000 growth regulator Benomyl (Benlate) Registration Standards Multiple-use systemic 1.13 x 10-4 2,000,000 fungicide EBDCs(mancozeb, Group of four widely 1.11 x 10-3 28,000,000 maneb, metiram, used fruit and zineb) vegetable fungicides Chlorobenzilate Citrus acaracide No risk assess- 1,600,000 ment per- formed Metolachlor (Dual) Corn, soybean herbicide 1.44 x 10-6 38,000,000 Oryzalin (Surflan) Citrus, field crop herbi- 1.14 x 10-5 1,600,000 cide Thiophanate-methyl Fruit, vegetable fungicide No risk assess- 28,000 (Topsin M) ment per- formed NOTE: These risk estimates are derived using EPA data and methods described on pages 50 66 and in Appendix B. aThe pounds active ingredient/year are averaged from selected years and are derived from Webb, S.E.H., 1981, 'preliminary Data: Pesticide Use on Selected Deciduous Fruits in the United States, 1978," Economic Research Service Stad Report No. AGES810626, Washington, D.C.: U.S. Department of Agriculture; Ferguson, W.L., 1984, "1979 Pesticide Use on Vegetables in Five Regions," Springfield, Va.: National Technical Information Service; Parks, J. R., 1983, "Pesticide Use on Fall Potatoes in the United States, 1979," Economic Research Service Staff Report No. AGES830113, Washington, D.C.: U.S. Department of AgIiculture; Gianessi, L. P., 1986, "A National Pesticide Usage Data Base," Washington, D.C.: Resources for the Future, photocopy; and unpublished data from the EPA for the years 1981 through 1985. The agency's policy in recent years of not approving tolerances for oncogenic active ingredients that concentrate in processed foods is amply demonstrated in Table 3-30. The EPA has denied all applications for section 409 tolerances since 1978 that involve oncogenic active ingredi- ents, including at least one active ingredient with very small estimated risk (10-~.

90 REGULATING PESTICIDES IN FOOD Another important conclusion can be drawn from Table 3-29. In at least three cases involving old and new ingredients, the EPA granted new tolerances when it determined that oncogenic potential came from an impurity, metabolite, or contaminant of the parent active ingredient (dicamba, cyromazine, and thiodicarb). In these cases, the EPA relied on the FDA's interpretations of the Delaney Clause. From these cases, it is clear that the EPA will consider applying, in appropriate cases, the FDA's constituents policy and sensitivity-of-the-method procedures in granting food- and feed-additive tolerances for oncogenic pesticides. (For a more detailed discussion, see Chapter 2.) The estimated additional risk sanctioned by these EPA tolerances for dicamba, cyromazine, and thiodicarb is far less than the estimated risk associated with tolerances that have been denied. These include permethrin tolerances on tomatoes and amitraz tolerances on apples. Risks allowed are on the order of 1 x 10-8 or less. Tolerances denied had risks between 1 x 10-4 and 1 x 10-~. The agency also gave the committee a list of 10 active ingredients for which it suspects that manufacturer or registrant concern about the impact of the Delaney Clause significantly influenced the content of tolerance applications (see Table 3-311. In each case, the agency is aware of section 408 and section 409 tolerance petitions that a registrant withdrew or declined to file because of concerns about the Delaney Clause. The committee believes that the Delaney Clause has been more influential than this table reveals. Once a pesticide is determined to be oncogenic, most registrants withdraw or do not submit petitions for section 409 tolerances. One reason is that tolerance petitions must be accompanied by a fee that must be paid regardless of the agency's decision. Companies will often attempt to obtain registrations, however, when they regard evidence of oncogenicity as equivocal or believe that the oncogenic risks are very low TABLE 3-31 Pesticides with Retracted or Unpursued Tolerance Applications Amitraz (Bamm) Benomyl (Benlate) Captan Cypermethrin (Ammo, Cymbush) EBDCs (mancozeb, maneb, metiram, zineb) Fosetyl Al (Aliette) Metolachlor (Dual) Permethrin (Pounce, Ambush) Vinclozolin (Ronilan) NOTE: In these cases, the EPA believes the petitioners either retracted or failed to pursue applications for tolerances under section 408 or 409 because of potential problems from the Delaney Clause.

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 9 ~ and the use proposed may fall within an exception to the Delaney Clause. Another strategy is to change a pesticide's use pattern in a way that keeps residues below the level detectable on raw agricultural commodities (see Appendix E). CASE STUDIES OF POTENTIAL POLICY PRECEDENTS Tolerances for New Active Ingredients Recent reports have criticized the EPA for not articulating a clear policy for application of the Delaney Clause in the tolerance-setting and reassessment process. ]2 In fact, the need for such a policy was the reason the EPA initated this project. Even in cases when the EPA has applied the constituents policy or the sensitivity-of-the-method procedure, it has stressed that such action does not represent a formal change in policy. The agency has defended its authority to use these options on a case-by- case basis until a more definitive policy is adopted. In this section, recent agency actions are analyzed to determine what patterns emerge from the EPA's application of the Delaney Clause in the tolerance process. First, the application of the Delaney Clause to new pesticides and pesticide uses seems clear-cut. New section 409 tolerances are not approved for clearly oncogenic pesticides. New section 408 tolerances are not approved for crops routinely processed into food forms in which oncogenic residues are expected to concentrate. The agency is willing to approve new tolerances for very low risk oncogens if there is a reasonable basis for doing so within FDA precedents, however. The greatest area of uncertainty is how the EPA will proceed in cases involving currently registered pesticides that have been found to be oncogens and have several existing section 409 tolerances or are shown to need these tolerances as residue chemistry data requirements are satis- fied. The committee finds no convincing legal or scientific basis for the EPA, as it completes the special review and reregistration processes, to avoid applying the standards of section 409, including the Delaney Clause, to currently registered compounds. Prior-Sanctioned Pesticides The prior-sanction exception to the Food Additives Amendment of the FDC Act would arguably render the Delaney Clause inapplicable to any pesticide residue in processed foods approved before 1958. (The FDC Act's definition of a food additive excludes substances regulated as food additives before 1958 from the food additive amendments of 1958, including the Delaney Clause.) Because of this, some pesticide residues

92 REGULATING PESTICIDES IN FOOD could technically escape the current requirements for food additives (including the Delaney Clause) if it could be shown that the FDA or the USDA sanctioned these residues before 1958. The committee briefly attempted to determine the number of pre-1958 tolerances to which the prior-sanction exception might apply. The com- mittee could find no tolerances issued between 1954 and 1958 that could be described as food-additive tolerances by current standards. Neverthe- less, such tolerances may have been issued and there may have been earlier approvals of residue-producing uses of agents still in use. The committee believes, however, that the number of prior-sanctioned resi- dues that might technically escape the strict standards of the food- additive regulation is quite small. Even when the prior-sanction exception might be invoked to preserve a tolerance, the committee can discern no health or scientific basis for treating residues sanctioned before 1958 differently from those sanctioned after 1958. The following review of seven case studies sheds some light on how the agency may resolve the issues surrounding Delaney Clause applications. Tolerance Actions and New Active Ingredients FOSETYL AL Fosetyl Al is a systemic organophosphorous fungicide used to control downy mildew and other diseases. It is currently widely used in Europe. In this country, the only registered use of fosetyl Al is on pineapples. The registrant, Rhone-Poulenc, in 1983 applied for tolerances for fosetyl Al on hops. Fosetyl Al residues were determined to concentrate during the drying of hops, and it has demonstrated weak but positive oncogenic effects in animals. Therefore, the EPA cited the Delaney Clause in denying section 408 and section 409 tolerances for residues in or on hops. Significantly, the risk presented by fosetyl Al residues in hops would have been far less than the risk from fungicides currently used on hops. According to the EPA, the additional risk presented by fosetyl Al residues on hops would have been 1 x 10-8, or 1 in 100 million or less. This risk is several orders of magnitude less than the estimated risk from ethylenebisdithiocarbamate (EBDC) fungicide residues widely used on hops, which is between 1 x 10-4 and 1 x 10-s. PERMETHRIN Permethrin is a widely used synthetic pyrethroid insecticide. In setting tolerances for permethrin, the EPA granted a section 408 tolerance for the fresh-market portion of tomato crops, but denied section 409 tolerances

ESTIMATES OF DIETARY ONCOGENIC RISKS 93 for the processed portion. This was the first time the EPA had set a tolerance for an oncogenic pesticide only on the raw portion of a crop with the knowledge that this pesticide's residues concentrate during processing. The EPA's general policy is to deny a raw agricultural tolerance for an oncogen when a section 409 tolerance is also needed. The agency departed from this policy in approving the use of permethrin on fresh tomatoes grown in Florida, because 98 percent of the Florida tomato crop is produced for the fresh market. In this case the agency was prepared to consider fresh tomatoes from Florida a distinct crop from processed tomatoes grown elsewhere. No tolerances for the use of permethrin on tomatoes grown outside Florida have been granted. Under the terms of the EPA's approval, surplus Florida tomatoes may not be processed. The agency has not drawn similar distinctions for other pesticides or tolerance applications. This may be because providing proof that a crop would be sold exclusively through the fresh market would be very difficult. Section 408 tolerances granted for the use of permethrin on corn and soybeans provide other insights. In these cases, the agency initially denied petitions for section 408 and section 409 tolerances because of residue concentrations in processed soybean and corn products. After further testing, the agency granted section 408 tolerances based on proposed changes in the label directions designed to reduce residues in the raw form of the crop below a level detectable by widely accepted analytical methods. The key change was extension of the time between application and harvest, allowing residues to degrade below detection levels by harvest. With residues theoretically eliminated from the raw commodity, the issue of concentration in the processed foods was moot. THIODICARB Thiodicarb is a newly registered carbamate insecticide, effective on a range of insect pests. Thiodicarb itself is not oncogenic. A metabolite of thiodicarb, acetamide, is oncogenic when administered to test animals at relatively high doses (12,500 to 80,000 ppm). Animals fed treated crops metabolize thiodicarb residues into acetamide. Residues of acetamide are then present in minute amounts in animal products. For example, 1.8 parts per billion (ppb) are present in beef liver, assuming that thiodicarb residues are at the tolerance level and that all feed is treated. In issuing section 409 feed-additive tolerances for thiodicarb, the EPA adopted the FDA's sensitivity-of-the-method procedure. This interpreta- tion requires the applicant for a feed-additive tolerance for an oncogenic substance to prove that the risk to humans from eating animals fed treated

94 REGULATING PESTICIDES IN FOOD feed is less than 1 in 1 million or 1 x 10-6. (See Chapter 2 and the thiodicarb case study in Appendix C for further examination of these issues.) On the basis of EPA calculations, meat and poulty could contain up to 90 ppb of acetamide and the risk would be below 10-6; milk and eggs could contain up to 30 and 90 ppb, respectively. Expected residues in meat and poultry, milk, and eggs were 1.8, 0.3, and 0.07 ppb, respec- tively, resulting in risk far less than 10-6. In every case, even at the highest allowable levels, the risk from acetamide in food as a result of thiodicarb use is well below the 10-6 standard. For purposes of the committee's work, it is noteworthy that the risks involved here were insufficient to trigger a special review of thiodicarb. As stated by the EPA in the final rule: There are no regulatory actions pending against the registration of thiodicarb. On the basis of the available studies on acetamide and the chronic oncogenicity studies for thiodicarb, the agency has concluded that the human risks posed by the use of thiodicarb on cotton and soybeans does [sic] not raise prudent concerns of unreasonable adverse effects and that a special review under 40 CFR 162.11 is not warranted. (Federal Register 50(No. 128~: 27464) In the agency's opinion, the regulatory actions surrounding thiodicarb arise entirely from the Delaney Clause and concern the issuance of tolerances, not the granting of product registration. In the absence of the Delaney Clause, therefore, the risk associated with thiodicarb tolerances would not have warranted agency review. D~cAMsA Dicamba is a broadleaf herbicide widely used in the production of soybeans, corn, and other row and field crops. Studies submitted to the EPA do not show dicamba as oncogenic. However, studies have shown a contaminant of dicamba, dimethylnitrosamine (DMNA), to be an animal oncogen. The EPA relied on the FDA's constituents policy in granting section 409 tolerances for dicamba residues in or on sugarcane molasses. The FDA articulated its constituents policy in the April 2, 1982 Federal Register (bracketed phrases are added to describe how the EPA applied the constituents policy to dicamba): "The constituents policy states that the safety of any undesired tin this case oncogenic] nonfunctional constituents tin non-oncogenic substances] should be judged under the general safety clause of the FDC Act Enot the Delaney Clausel, using risk assessment as one of the decision-making tools." The FDA has interpreted the general safety clause of the FDC Act as allowing an additional risk no greater than 1 x 10-6. The EPA assessed

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 95 the additional risk from exposure to DMNA in sugarcane molasses as no greater than 2.9 x 10-~. Accordingly, the agency approved section 409 tolerances. The EPA explained its policy as follows: EPA does not regard deliberately added active or inert ingredients, or metabolites thereof, as potential candidates for clearance under the constitutents policy rationale. Rather, EPA will only consider applying this rationale to impurities arising from the manufacture of the pesticide (residual reactants, intermediates, and products of side reactions and chemical degradates). Furthermore the Agency will consider using this rationale in issuing a food additive regulation only where the potential risk from the impurity is extremely lowed (The Federal Register notice did not define low potential risk. The FDA criteria, however, is 1 x 10-6.) Tolerance Actions and Old Active Ingredients DICOFOL AND CHLOROBENZILATE Dicofol and chlorobenzilate are insecticides, acaricides, and miticides registered before the creation of the EPA in 1970. They are widely used in citrus production. Dicofol is also extensively used on cotton. Both compounds have demonstrated oncogenic effects in animal experiments. The agency has scrutinized each for several years. Residues of both pesticides at concentrated levels have been found in food products, primarily citrus oil. The EPA has not altered the citrus tolerance for either chemical, however, because it believes that the oncogenic potential of the pesticides is so weak, and citrus oil is consumed in such small quantities, that a quantitative assessment of the oncogenic risk from consumption of citrus oil cannot be supported by the available data. In essence, the agency has chosen to defer action on these tolerances indefinitely. These cases suggest that there is a de minimis risk standard below which the agency will not calculate risks. BENOMYL Benomyl is the most widely used systemic fungicide in the world. It is important because its existing section 409 tolerances will probably be the first to force an EPA decision on retroactive application of the Delaney Clause. Benomyl is one of the most extensively studied pesticides in use. It has been through the EPA's special review process and then through its registration standards procedure. The data supporting its current regis- trations are generally of high quality. The registrant and the EPA agree that benomyl causes an oncogenic response in animal experiments, and that it concentrates in certain processed foods. It appears that existing

96 REGUrATING PESTICIDES IN FOOD TABLE 3-32 Number of Cancer Studies Due for Pesticide Active Ingredients, 198~1990 Chronic Year Oncogenicity Feeding Total 1986 10 5 15 1987 27 16 43 1988 21 17 38 1989 24 28 52 1990 3 3 6 Total 85 69 154 SOURCE: U.S. Environmental Protection Agency. 1986. Data Generation Schedule Status Report. Washington, D.C. section 409 tolerances for benomyl violate the Delaney Clause. The agency has deferred action on these tolerances pending public comment on the benomyl registration standard, recently invited by notice in the Federal Register. The resolution of the benomyl issue could provide a basis for agency actions in the future. The impact of tolerance revocations for benomyl and other oncogenic active ingredients included in the committee's risk estimates is projected in the next section and discussed in further detail in Chapter 5. PROJECTING PAST ACTIONS INTO THE FUTURE Over the next five years, the EPA will receive new data on the oncogenicity of many agriculturally important chemicals through the data call-in, special review, and registration standards programs. The ap- proaches the EPA devises for reassessing tolerances in light of the Delaney Clause will have a tremendous impact on how these new data are evaluated and incorporated into the pesticide reregistration process.~4 Table 3-32 shows an approximate schedule for the submission of new chronic feeding and oncogenicity bioassay results for major food-crop pesticides in response to data call-in letters issued from 1982 to 1986. From 1987 through 1989, the EPA should receive about 40 to 50 new tests each year. In completing the call-in, the EPA gave priority to data on chronic health effects. It has requested relatively few new residue concentration studies. The agency has recently begun to seriously evaluate the com- plexity and cost of modernizing residue chemistry data. It is already clear

ESTIMATES OF DIETAR Y ONCOGENIC RISKS 97 that the costs can be sizable. They may exceed the cost of a complete new chronic toxicology data base for active ingredients used on many foods. THE SHORT-TERM POTENTIAL IMPACT OF THE DELANEY CLAUSE Tables 3-33 and 3-34 show the approximate dates the EPA is expected to have enough information to compel decisions on certain pesticide am. ... ~ r · ~ ~~ = ~ Jo ~ ~ tolerances. the committee s criteria for ~nc~ua~ng spec~nc compounds on these lists are that the pesticides are oncogenic compounds used on foods for which a special review and a registration standard will be complete by the date listed.~5 The committee's analysis supports several important conclusions. First, the EPA will soon be faced with several significant decisions regarding section 409 tolerances for oncogenic pesticides. These deci- ~ions involve commercially important chemicals, which present sizable estimated risks. Second, the estimated dietary risk associated with these pesticides represents approximately 85 percent of all estimated dietary oncogenic pesticide risks. Third, agency actions could have the greatest impact on fungicide use and on associated dietary risk. Over the next three years the EPA is scheduled to make decisions on active ingredients that account for about 85 percent of fungicide use. Fourth, most fungi- -~.~ - ~ ~^~ ^~ a ----or TABLE 3-33 Potential Short-Term Impact of the Delaney Clause on Selected Fungicides Fungicide Possible Market Date for Estimated Risk on Commodities Share Active Tolerance (% acre Ingredient Revocation Action Raw Processed Total treatments) Benomyl 1986 Rsa 3.42 x 10-5 7.91 x 10-5 1.13 x 10-4 15 EBDCs 35 Mancozeb 1987 RS 2.43 x 10-4 9.44 x 10-5 3.38 x 10-4 Maneb 1987 RS 3.90 x 10-4 5.22 x 10-5 4.42 x 10-4 Metiram 1987 RS 7.65 x 10-5 3.91 x 10-5 1.15 x 10-4 Zineb 1987 RS 4.71 x 10-4 2.45 x 10-4 7.17 x 10-4 Captafol 1987 sRb 4.34 x 10-4 1.59 x 10-4 5.94 x 10-4 5 Folpet 1987 RS 1.81 x 10-4 1.43 x 10-4 3.24 x 10-4 5 Captan 1988 SR 2.80 x 10-4 1.93 x 10-4 4.74 x 10-4 15 Chlorothalonil 1988 SR 1.89 x 10-4 4.82 x 10-5 2.37 x 10-4 10 NOTE: These risk estimates are derived using EPA data and methods described on pages 50-66 and in Appendix B. aRS is registration standard. bSR is special review.

9~3 REGULATING PESTICIDES IN FOOD TABLE 3-34 Potential Short-Term Impact of the Delaney Clause on Selected Herbicides Possible Date for Estimated Risk on Commodities Active Tolerance Ingredient Revocation Action Raw Processed Total Herbicide Market Share (pro pounds applied) Pronamide 1986 (Kerb) Terbutryn 1986 Trifluralin 1986 (Treflan) Paraquat 1986 Alachlor (Lasso) Linuron (Lorox) 1986 1987 Rsa 7.14 x 10-6 6.28 x 10-7 7.77 x 10-6 RS RS RS SR <1 2.86 x 10-' No risk assessment conducted 2.86 x 10-7 <1 8 No risk assessment conducted 1.36 x 10-5 1.06 x 10-5 2.42 x 10-5 18 1.12 x 10-3 3.96 x 10-4 1.52 x 10-3 <1 NOTE: These risk estimates are derived using EPA data and methods described on pages 50-66 and in Appendix B. aRS is registration standard. bSR is special review. cides have few section 409 tolerances; some will have to be granted if certain uses on food crops are to continue. The EPA faces an especially difficult challenge with the fungicides. To guarantee that its regulatory actions actually reduce real risks, the agency must carefully assess all fungicides registered for each crop and base its actions on reducing risk after predictable substitutions have been made. One principle should guide the EPA's actions to reduce dietary oncogenic risks. It should focus its efforts on all oncogenic pesticides used on the most widely consumed crops that in turn present the greatest dietary risk. NOTES 1. National Agricultural Chemical Association. 1986. Industry Profile Survey: 1985 Washington, D.C. 2. Gianessi, L. P. 1986. A National Pesticide Usage Data Base. Washington, D.C.: Resources for the Future. 3. Ballard, G., W. Cummings, M. Luther, and N. Pelletier. 1980. Fungicides: An Overview of Their Significance to Agriculture and Their Pesticide Regulatory Implications. Washington, D.C.: U.S. Environmental Protection Agency. 4. U.S. Department of Agriculture. 1985. Economic Indicators of the Farm Sector: Farm Sector Review, 1984. ECIFS 4-2. Washington, D.C.: U.S. Government Printing Office. 5. U.S. Environmental Protection Agency. 1986. Guidelines for Carcinogenic Risk As- sessment. Federal Register 51(185): 33992-34003. 6. Paynter, O. E. 1984. Standard Evaluation Procedures for Oncogenicity Potential:

ESTIMATES OF DIETARY ONCOGENIC RISKS 99 Guidance for Analysis and Evaluation of Long-term Rodent Studies. Washington, D.C.: U.S. Environmental Protection Agency. 7. U.S. Environmental Protection Agency. 1985. Captan Special Review Position Docu- ment 2/3. Washington, D.C. 8. U.S. General Accounting Office. October 1986. Pesticides: Need to Enhance FDA's Ability to Protect the Public From Illegal Residues. GAO/RCED-87-7. Washington, D.C. 9. National Research Council. 1984. Cancer Today: Origins, Prevention, and Treatment. Washington, D.C.: National Academy Press. 10. U.S. Environmental Protection Agency. 1983. Ethylene Dibromide Special Review Position Document 4. Washington, D.C. 11. 40 CFR Part 158 (1986). 12. U.S. General Accounting Office. 1986. Pesticides: EPA's Formidable Task to Assess and Regulate Their Risks. GAO/RCED-86-125. Washington, D.C. 13. U.S. Environmental Protection Agency. 1984. Tolerances for Pesticides in Food Administered by the Environmental Protection Agency; Animal Drugs, Feeds, and Related Products; Tolerances for Pesticides in Animal Feeds; Dicamba; Denial of Stay. Federal Register 49(235): 47481~7483. 14. U.S. Environmental Protection Agency. 1986. Data Generation Schedule Status Report. Washington, D.C. 15. U.S. Environmental Protection Agency. 1986. Report on the Status of the Chemicals in the Special Review Program, Registration Standards Program, and Data Call-In Program. Washington, D.C.