Comments on EPA's Evaluation of the Atrazine
Mechanism of Carcinogenic Actionby
James D. Wilson, Ph.D.
Resources for the Future, Washington, DCJames D. Wilson has nearly 30 years of experience in evaluating human and environmental safety of chemicals and in developing methods and procedures for this purpose, with particular expertise in risk characterization. He has served on the National Academy of Sciences Committee on Risk Characterization (1993-1996) and as president of the Society for Risk Analysis (1993) and was named an SRA fellow in 1993. He currently holds the position of senior fellow at Resources for the Future, an independent, not-for-profit research institute headquartered in Washington, D.C. However, the comments appearing here represent only his views and should not be attributed to Resources for the Future. (Note: The author holds the copyright to this paper. He can be contacted at wilson@rff.org.)
This short essay pertains to comments I sent to the U.S. Environmental Protection Agency in January, 2000, to help a Scientific Advisory Panel committee evaluate an EPA report, issued in December, that describes results of research on the herbicide atrazine and presents some conclusions drawn from that research (Report: Atrazine: Carcinogenicity Hazard Assessment and Characterization, Docket Number OPP-00637). The submitted comments outline how mechanistic data should be used when the regulatory decision in question is the acceptability of a tolerance for a food-use pesticide that may pose a carcinogenic risk. This decision is almost unique in the way risk assessment is to be used. The law applying here is a section of the Food, Drug and Cosmetic Act that governs approval of food additives and pesticide residues and demands a decision based solely on comparison between an estimated exposure and an exposure judged safe. (I know of only two other laws structured in this way; the others are the “no action” decision embedded in Superfund remedy selection and California’s “Proposition 65.”)
The SAP committee’s scheduled January meeting was snowed out but has been rescheduled for late June. The review is of more than passing importance to the health risk assessment profession, especially to those of us interested in improving the use of science in risk assessment, and to EPA.
The ultimate question is whether solid science can overcome politics. The research on atrazine is of high quality, and the results imply a departure from standard risk assessment practices that would tend to increase the amount of atrazine that would be tolerated as a residue in food. However, atrazine has attracted notoriety, and certain people will oppose any regulatory actions that might increase its use, or even maintain its use at the present level.
In an unusual action, EPA’s instructions to the SAP committee forbid the committee to offer comments on the applicability of this mode-of-action data to the risk assessment. As noted below, it bears on two different judgments reached in any risk assessment, whether the study observations are valid and “relevant” to human risk, and, for carcinogens, if the carcinogenic response should be treated with a “threshold” or a “linear, no-threshold” algorithm in setting the tolerance. EPA states that it has already concluded the mammary tumors “relevant” to human risk; the evidence for this conclusion is not stated. The agency’s instructions to the committee forbid them to comment on the applicability of the data to the question of how the data should be treated. It leads one to ask, “Why hold an expensive review of the data if it is not going to be used in risk assessment?”
Atrazine is a very cost-effective pre-emergent herbicide for grassy weeds in corn, and thus is used in the United States practically everywhere that corn is grown. Traces of it show up in surface waters, particularly in the spring after heavy rains, and in some wells. Its notoriety has been increased by an observation that it induces an increased rate of mammary tumors in one strain of lab rats, those called “Sprague-Dawley,” although the same effect is not induced in standard test rats, those called “Fisher 344.” Further, early research results suggested that the increased mammary tumor rates in Sprague-Dawley rats was somehow related to changes in hormone levels. So atrazine’s use is opposed because: (1) some of it finds its way into drinking water sources, (2) it may act on humans to cause breast cancer, and (3) it may be a “hormone disrupter.”
EPA has the authority to review the use of any pesticide at any time new information becomes available. In addition, under laws passed in 1988 and in 1996, it is required to perform periodic reviews even in the absence of new data, especially for pesticides used on food crops. Under the 1996 statute, known as the “Food Quality Protection Act,” this review focuses on the safety of pesticide residues left in food. These residues are regulated through a tool called a “tolerance,” that being an amount that is permitted to be found in foodstuffs. The December report slated for Scientific Advisory Panel review would constitute a part of that tolerance review process for atrazine.
Residue tolerances are to be approved if found to be “safe”-defined in 1996 as an amount producing a “reasonable certainty of no harm.” That term may sound vague, but, in fact, it refers to conclusions drawn using specific test data interpreted according to specified procedures. These procedures date to the mid-1950s but were modified in the late 1970s to deal with carcinogens. (Their original application was to food additives. Since the Delaney Clause prohibits use of carcinogenic food additives, there was no need for a procedure to deal with carcinogens. But this clause does not apply to pesticide residues, so a modification to the then-standard procedures had to be introduced.)
The modified procedure calls for the “safe” level to be identified by application of a computer program known as “linearized multistage” to a single set of data from a cancer bioassay test, on either rats or mice. This program computes a “plausible upper bound risk”; the exposure corresponding to a plausible upper bound risk of 1 in a million (10-6) is defined as “safe” for these purposes. (EPA usually refers to this exposure as the “Risk-specific Dose” or “RsD.”)
This procedure was adopted when it was widely assumed that all carcinogens act by damaging DNA, and that one “hit” could suffice to cause cancer. Since then that assumption has been found to be wrong. A large number of cancer-causing substances do their damage by inducing cells to divide, without harming the cells’ DNA. Because of the controls organisms exert on cell division, for all such carcinogens there exists some exposure that causes no increase in cancer risk. That is, for these “non-mutagenic” carcinogens, some “threshold” exists in the dose-response curve. So exposures larger than the standard “safe” level often will also be safe.
Starting about 1995, EPA began to evaluate certain non-mutagenic carcinogens using a new procedure. This new procedure is, in fact, very similar to that used since 1955 for substances not found to cause cancer (non-carcinogens). Generally speaking, in this procedure, the largest “no effect” dose in the cancer bioassay test is divided by 100 to identify the exposure judged to pose “a reasonable certainty of no harm.”
The substances to which this new procedure has been applied, so far, are all members of a class that are not observed to cause mutations and which reduce thyroid hormone levels in the body.
The obvious difference between this new procedure and that used for non-carcinogens is that for the latter substances, some kind of adverse effect other than tumor development serves to identify the “critical dose.” This “critical dose” is then divided by 100 to identify the safe exposure. (EPA usually refers to this number as the “Reference Dose” or “RfD.”)
In these standard procedures, there are just two places that allow exercise of expert scientific judgment. The first is when toxicologists review the available toxicology tests to judge them as to validity and “relevance.” By “relevance” they mean whether or not the response observed to occur in animals is judged to predict how humans will respond, when similarly exposed. Test data sets judged not reliable or not relevant are set aside and not used in the safety-assessment procedure.
The second place allowing the exercise of expert scientific judgement is in the selection of an algorithm used to act on the test data to produce the “safe” exposure. For instance, in the case of non-carcinogens, the default algorithm is “divide by 100 the largest ‘no effect’ dose from the data set exhibiting the smallest ‘Lowest Observed Adverse Effect Level.’” But in cases where certain standard tests have not been done, or other defects occur in the data base, a divisor of 200 or larger may be used. In the case of carcinogens, the default is the “linearized multistage” procedure noted above. EPA’s recent change introduces a third choice into this second exercise of professional judgment: simplifying, now the algorithm for “carcinogens” may be either of the two broad kinds of algorithm described.
Generally speaking, the expected effect of this change in procedures is to increase the exposure to pesticide residues that is judged “safe.” (The increase may be quite small or it may be substantial.)
The atrazine mode-of-action research is unusual in that the results inform both of these two places where exercise of scientific judgment is permitted. The results bear on the issue of relevance of the finding that exposure to atrazine induces mammary tumors in female Sprague-Dawley rats but not in Fisher rats. If this response is judged “not relevant,” then the standard “non-carcinogen” procedure is to be used. If it is judged “relevant,” then the research results bear on the choice of algorithm. If atrazine is mutagenic or if the mode of action cannot be determined, then the old standard “linearized multistage” (or an equivalent) procedure is to be used. However, if atrazine is not mutagenic and the tumors are induced because of some effect on a hormone or endocrine organ, then the new procedure should be used.
In this case, researchers from EPA and Novartis collectively have concluded that atrazine is not mutagenic. Further, they conclude that in the Sprague-Dawley rats, treatment with atrazine affects a key organ in the endocrine system, the hypothalamus. In these rats, one or more hypothalamic hormones are affected, which leads, through a complex pathway, to an increase in the rate at which mammary tumors develop. This mechanism appears to be unique to this strain of rat and to atrazine (and atrazine-like chemicals).
If the scientific community accepts the research results and their interpretation as valid, EPA should decide either that the observed response is not relevant to humans, or, if it is “relevant,” that the hormonally mediated pathway makes the traditional “Linearized Multistage” procedure not appropriate for standard setting. In either case, it would appear that the appropriate RfD value would come by dividing an appropriate No-Observed Adverse Effect Level by 100.
We shall see what happens.