How Does EPA Decide Whether a Substance is a Carcinogen?
by Allan Felsot, Environmental Toxicologist

In January, EPA released a revised list of pesticides classified as carcinogens. I counted more than 100 active ingredients. In 1987, there were 53 pesticide carcinogens, according to a National Academy of Science report. The difference seems largely due to increased testing and/or evaluation as part of the reregistration process rather than new compounds registered after 1987. Considering that there are about 300 registered active ingredients for food uses, the current number of classified carcinogens represents about 33% of these pesticides. This is curious, considering that Bruce Ames and Lois Gold’s analysis of the Cancer Potency Database, funded by the National Institute of Health at the University of California, Berkeley, shows that about 50% of all synthetic and natural chemicals tested for carcinogenicity are associated with tumors. Does this mean that pesticides are less likely to be carcinogenic on average than all other chemicals? Such a question shows how easy it is to put a positive (or negative) spin on available data. This brings me to the central question of this essay: How does the EPA decide if a chemical is carcinogenic? What spin does EPA place on animal testing data?

The Federal Insecticide, Fungicide, and Rodenticide Act requires testing to determine the carcinogenic potential of pesticides. Actually, the tests determine oncogenic potential, which is the ability to cause benign and/or malignant tumors; a substance causing malignant tumors would be considered carcinogenic. One of the first tests for oncogenic potential is the determi-nation of mutagenic potential, the ability to directly damage or alter DNA. Various types of cellular studies have been developed to determine mutagenicity. EPA views with suspicion any compound testing positive for mutagenicity, because the longstanding but now outdated hypothesis of cancer induction assumes that all carcinogens damaged DNA directly. New research has shown that many nonmutagenic chemicals can cause tumor formation at very high doses but not at low or moderate doses.

EPA’s theoretical assumptions about cancer etiology dictate the design of oncogenic potential studies and decisions to classify a compound as carcinogenic. EPA assumes that thresholds for cancer do not exist; a single molecule can eventually lead to tumor production. This rule seems to be a holdover of the idea that carcinogens work by directly mutating DNA, and a single mutation can lead to cancer. Nevertheless, the Food Quality Protection Act (FQPA) speaks of threshold and nonthreshold pesticides to distinguish those that are noncarcinogenic from those classified as carcinogens.

A second EPA assumption is that exposure to the maximum tolerated dose (MTD) over a rat's lifetime is a valid predictor of human risk. Thus, all studies must include the maximum tolerated dose. This dose is determined by feeding a range of doses to test animals for about 90 days (a subchronic test). The maximum tolerated dose is an estimate of the highest dose over a lifetime that will not alter longevity through non-cancer effects. It should cause no more than a 10% weight decrease in treated animals compared to the undosed group. Other benchmarks for the MTD besides no mortality include absence of clinical signs of toxicity and pathologic lesions that would be predicted to shorten an animal’s life span.

Usually two to three doses including the MTD are fed to rats during their normal two-year lifespan. Pesticide is mixed with the rats’ food, and daily consumption is measured so that doses can be expressed as milligrams pesticide per kilogram body weight per day (mg/kg/day). Researchers dose about 50 animals of each sex and monitor them daily for signs of distress. At the end of the test period, the animals are killed and tissues from multiple organs are examined for a variety of cellular changes in addition to the presence of abnormal growths indicative of tumors. One such chronic feeding study can cost several million dollars.

The EPA reviews studies and classifies compounds as carcinogenic or noncarcinogenic. In addition to animal testing, the EPA may use epidemiological evidence from humans. Epidemiological studies are not conducted on new active ingredients, but older compounds undergoing reregistration may be examined. For example, 2,4-D has received much epidemiological study as well as notoriety. During the mid-1980s, National Cancer Institute (NCI) epidemiologists claimed that Midwest farmers had a risk of non-Hodgkin’s lymphoma that increased in proportion to how long they used the product. A study by Dow Chemical scientists published in Fundamental and Applied Toxicology reviewed 2,4-D epidemi-ological studies and concluded that a majority of studies supported no association with cancer. I noted that the current carcinogenic pesticide listing excludes 2,4-D.

Once mutagenicity, oncogenicity, and epidemiological studies are completed, the EPA considers the results of the mutagenicity and oncogenicity tests, the types and diversity of tumors, the structural similarity of the pesticide to known carcinogens, and whether positive results have been replicated. A study showing elevated incidence of tumors in dosed animals will weigh heavily in the classification decision, especially if other studies corroborate increased tumor incidence in the same tissue and/or the compound is a mutagen. To characterize carcinogenic hazard to humans, the EPA has adapted a multi-tiered classification scheme developed by the International Agency for Research on Cancer. A substance may be classified within one of six possible groups as: human carcinogen (Group A), probable human carcinogen (Groups B1 and B2), possible human carcinogen (Group C), not classifiable (Group D), or noncarcinogenic (Group E). All pesticides EPA considers carcinogenic fall into Groups B2 and C. (See Table 1)

TABLE 1

  Group B1 & B2 Carcinogens  
Acetaldehyde ETO Orthophenylphenol & Na Salt
Acetochlor (Harness, Surpass) Etridiazole (Terrazole) Oxythioquinox (Morestan)
Aciflourfen, Sodium (Blazer) Fenoxycarb (Comply) Pentachlorophenol
Alachlor (Lasso) Folpet Procymidone
Amitrole Formaldehyde Pronamide (Kerb)
Cacodylic Acid (Montar) Heptachlor Propargite (Omite)
Captan Iprodione (Rovral) Propoxur (Baygon)
Chlordimeform Lactofen (Cobra) Propylene oxide
Chloroform Lindane Sufallate
Chlorothalonil (Bravo) Mancozeb (Dithane) Telone (Dichloropropene, 1,3-)
Creosote Maneb Thiodicarb (Larvin)
Cyproconazole (Sentinel) Metiram (Polyram) Toxaphene
Daminozide (Alar) Metam Sodium (Vapam) TPTH
Dichloromethane MGK Repellent 326 Vinclozolin (Ronilan)
     
  Group C Carcinogens  
Quantified Group C    
Acephate (Orthene) Ethalfluralin (Sonalan) Oxadixyl
Amitraz (Mitac) Fenbuconazole (Indar) Oxyfluorfen (Goal)
Atrazine Fomesafen (Reflex) Permethrin (Ambush, Pounce)
Benomyl (Benlate) Hexythiazox (Savey) Pyrithiobac-sodium
Calcium cyanamide Hydrogen cyanamide Simazine (Princep)
Carbaryl (Sevin) Imazalil Tetrachoroethane, 1,1,2,2-
Cyanazine (Bladex) Metolachlor (Dual) Tetrachlorvinphos
Dacthal (DCPA) Molinate Triallate
Dichlorvos (DDVP) Oryzalin (Surflan) Trifluralin (Treflan)
Diclofop-methyl (Hoelon) Oxadiazon (Chipco, Ronstar)  
     
Unquantified Group C    
2-Benzyl-4-chlorophenol Fipronile PCNB (Terraclor)
Acrolein Fosetyl-Al (Aliette) Piperonyl Butoxide (Butacide)
Asulam (Asulox) Hexaconazole (Anvil) Prodiamine (Barricade)
Bifenthrin (Brigade) Hydramethylnon (Amdro) Propazine
Bromacil (Hyvar) Isoxaben (Gallery) Propiconazole (Tilt, Orbit)
Bromoxynil (Buctril) Linuron Tebuconazole (Folicur)
Clofentizine Mercaptobenzothiazole, 2- Terbutryn
Cypermethrin (Cymbush) Methidathion (Supracide) Tetramethrin
Dichlobenil (Casoron) Methylphenol, 3- Triadimefon (Bayleton)
Dichlorethylene, 1,1- MGK-264 Triadimenol (Baytan)
Dicofol (Kelthane) Norflurazon (Evital, Predict) Tribenuron methyl (Express)
Difenconazole (Dividend) Paradichlorobenzene Tribuphos (DEF)
Dimethenamid (Frontier) Pendimethalin (Prowl) Tridiphane (Tandem)
Dimethipin (Harvade) Phosphamidon (Dimecron) Uniconazole
Dimethoate Parathion  

The process seems straight forward, but details can be messy. First consider that, well-trained pathologists must examine thousands of tissue slices, and they may not all agree that a damaged cell has changed to a tumor cell. Manufacturers will argue with the EPA about the interpretation of feeding studies. Second, the criteria that EPA uses to classify compounds into Group B2 and C seem mysterious. B2 represents sufficient evidence of carcinogenicity from animal studies, with inadequate or no epidemiologic data. Group C represents limited evidence of carcinogenicity in the absence of human data. Classification of carcinogenicity seems to depend on evidence ranging from limited to sufficient.

I was curious how EPA judged whether evidence was limited or sufficient. I was particularly interested in how negative data are handled and whether thresholds exist. With readily available data in the book Drinking Water Health Advisory: Pesticides, compiled by the EPA Office of Drinking Water, I used chlorothalonil (Bravo, Daconil) as a test case to see how EPA makes a decision. Chlorothalonil has been classified as a B2 carcinogen. The fungicide is important for many agricultural and nonagricultural uses; its registration could be jeopardized on some crops when its exposure risk is reassessed under the expanded requirements of the FQPA.

The chapter on chlorothalonil lists 13 mutagenicity studies; some were based on cell culture studies, while others involved whole animal feeding studies. Only one test showed evidence of a mutagenic effect. Thus, chlorothalonil is not considered mutagenic.

Fourteen subchronic and chronic feeding studies used various dosing regimens ranging from 0.2 mg/kg/day to 900 mg/kg/day. In 90-day feeding studies, doses did not exceed 3 mg/kg/ day; at this dose there were no changes indicative of a tumor. A two-year feeding study with dogs at a dose of 3 mg/kg/day revealed no conclusive compound-related trends. A 12-week feeding study reported by the NCI revealed no effect in mice dosed with 800 mg/kg/day.

Two-year high dose studies gave evidence of tumor formation that EPA evidently weighed very heavily. An NCI study observed kidney tumors at doses of 253 and 506 mg/kg/day. What I find curious is the percentage of animals with tumors. At the high dose, 4 of 49 males and 5 of 50 females had tumors. EPA apparently chose to consider 10% incidence of tumors after feeding rats amounts of chlorothalonil equivalent to nearly a billion times the estimated human exposure (<0.0000001 mg/kg/day, according to the Food and Drug Administration Total Diet Study) representative of sufficient evidence of ability to cause tumors in animals.

Is there corroboration of the NCI rat study? A study in mice at doses of 119, 251, and 517 mg/kg/day showed significant tumor incidence only in males fed 251 mg/kg/day. There was no linear relationship to dose. A different two-year study with rats concluded that dose-dependent increases in kidney tumors occurred at feeding levels greater than 40 mg/kg/day. A study filed in 1989 reported a 13% and 20% increase in tumors in male and female rats fed 175 mg/kg/day for nearly 2.5 years. No tumors formed at doses of 1.5 and 3.3 mg/kg/day.

EPA’s classification of chloro-thalonil as a B2 carcinogen appears to be based on evidence from the two-year rat and mice feeding studies. It does not seem to matter that the incidence of tumors is incredibly low given the high dose rates. It does not seem to matter that the low doses even in the two-year studies were negative. It does not matter that the 90-day low dose feeding studies showed no toxic or cellular effects at doses of 1.5 mg/kg/day, a dose still a million times greater than the estimated human dietary intake. EPA’s decision, however, does demonstrate one final assumption the agency makes about pesticides: negative epidemiologic or animal results cannot prove safety.

(Allan Felsot is the Environmental Toxicologist at the Food and Environmental Quality Laboratory, Washington State University.)

REF: Agrichemical and Environmental News, 132, February 1997.