||Hayes, TB. 2004. There is no denying this: defusing the confusion about atrazine. BioScience 54:1138-1149.
In 2002 and again in 2003, Univ. California Berkeley endocrinologist Tyrone Hayes published results documenting hermaphroditism in frogs caused by exposure to extremely low levels (0.1 ppb) of atrazine, the most widely used herbicide in the world. These results were immediately and repeatedly challenged by scientists linked through funding and a private consulting firm, EcoRisk, to the manufacturer of atrazine, Syngenta.
Recent press coverage of
Hayes's work on atrazine
The scientists funded by EcoRisk announced they were unable to replicate Hayes's work. Initially, at a time when the EPA was investigating whether atrazine should be allowed to remain on the market, all that was available to evaluate these conclusions were press releases. Over the next year, the original studies were made public either by EPA or through publication in the scientific literature.
Now in this paper in BioScience, Hayes analyzes the research of his EcoRisk critics.
His analysis reveals a series of flaws in the studies conducted by EcoRisk, the interpretation of data and their public descriptions of the results in press releases. These flaws include control groups contaminated with enough atrazine to cause effects, high mortality in laboratory animals, and misrepresentation of data. Indeed, although the published and press release versions of their research conclude that Hayes's results cannot be repeated, the data published in one paper show a series of strong effects consistent with Hayes's observations, with probabilities exceeding 99% that the effects are due to atrazine. One of the authors is an editor of the journal in which the paper was published, raising questions about editorial policies at this journal, Environmental Toxicology and Chemistry. Is it captive, or legitimate?
Hayes also presents a statistical assessment of the factors associated with negative vs. positive results. Compared to studies funded by other sources, studies funded by Syngenta and carried out by EcoRisk are highly unlikely to find an effect of atrazine.
Key points reported by Hayes :
1. Control groups in studies were contaminated by atrazine. In two different studies, EcoRisk recorded hermaphroditism in their control animals, and used those observations to argue that hermaphroditism occurs naturally in the species in question. If it occurs naturally, then atrazine isn't at fault.
One of these studies was based upon lab work conducted at Michigan State University under John Giesy's direction.
The graph (to right) shows the actual levels of atrazine measured in control and 0.1 ppb treatment group in these experiments. In an evaluation of this work, the US EPA concluded that "the negative controls were contaminated with atrazine at levels comparable to those in the ... atrazine treatment."
The red line shows the level of atrazine sufficient to cause hermaphroditism, according to Hayes et al. 2002, 2003.
As Hayes notes in BioScience, atrazine levels experienced by some of the controls exceeded the level at which he had detected an effect by more than four-fold.
The second contaminated study involved fieldwork in South Africa. This study examined the frequency of hermaphroditism in corn-growing and non-corn-growing areas in South Africa. Atrazine is commonly used on corn fields. They found hermaphrodites in both regions, and therefore concluded that hermaphroditism was not caused by atrazine exposure. But according to the EPA's assessment of this research: "reference sites [i.e., the control sites] all contained measureable residues of atrazines that were, in some cases, higher than sites considered representative of atrazine exposure." Atrazine exceeded the level sufficient to cause an effect by fourfold.
2. Data published by EcoRisk scientists contradict their own conclusions and public statements. Even though the data published in one study, led by James Carr, were consistent with Hayes's results, public statements by EcoRisk scientists and the published version of the study itself concluded that Hayes's findings could not be replicated. These statements were issued during a period when the US EPA was actively reviewing its policies on atrazine.
Carr et al. found hermaphrodites and males with multiple testes, which were morphologically identical to what Hayes found, and their statistical analysis revealed that several adverse effects of atrazine were highly unlikely to be due to chance (p values ranging from 0.0003 to 0.02). Specifically for hermaphroditism, p = 0.0042. Yet the authors of this study referred to these as "weak trends" and concluded that Hayes's results were not repeatable.
In a Syngenta press release listing Ronald Kendall, John Geisy and other authors as available for further comment, James Carr stated "We have been unable to reproduce the low-concentration effects of atrazine on amphibians reported elsewhere in the scientific literature. Kendall is quoted: "No conclusions can be drawn at this time on atrazine and its purported effect on frogs. We must get the science done in order to have the facts to make sound conclusions... these latest results simply do not support Hayes' initial work." These comments were issued on 22 June 2002, at a time when the data purportedly supporting them were not available for independent assessment.
3. Animal husbandry problems led to high mortality rates in experiments. In work in John Giesy's laboratory at Michigan State University, up to 86% of animals died in some treatments, including 80% of controls. This compares with mortality rates typically under 10% for Hayes's laboratory. EPA standards require that 70% to 90% survive. With as few as 14% of test animals surviving, any conclusions about the effects of atrazine from these experiments is suspect. Surviving animals were in poor condition, also.
4. Probability of negative conclusions is highly associated with industry funding. Hayes performed a statistical analysis of 16 experiments examining the effect of atrazine on amphibian gonads. In this study he considered the finding negative if the conclusions of the study were described as negative. This meant, for example, that the publication by Carr et al., above, was treated as negative because they concluded that Hayes's data were not repeatable, even though their published data demonstrated impacts. Hayes considered five possible factors that might contribute to positive or negative findings:
- study type (lab or field)
- study design (appropriate or inappropriate)
- principal authors
- financial sponsorship
Of these, only financial sponsorship by itself was a strong predictor (p = 0.009). 100% of negative studies were funded by Syngenta. However, "principal authors" and "study design" were highly correlated with "funding source." Examined simultaneously in a path analysis, all three variables contribute significantly to determining the outcome of the work.
The only "negative" study in this analysis that did not suffer from high mortality or contaminated controls was carried out by Hayes's lab. This work was funded by EcoRisk/Syngenta. Hayes had begun this line of research under contract to EcoRisk, and had submitted to them data showing effects of atrazine. After they did not accept his interpretation, he ended his work for them and obtained funding from other sources. All of his published data in Proceedings of the National Academy of Sciences, Nature and Environmental Health perspectives are from research carried out after ending his work for EcoRisk. EcoRisk reinterpreted his data and submitted to EPA an analysis concluding atrazine had no significant effect. When this study is removed from the analysis, authorship and study design become highly significant.
What does this mean? Hayes's analysis reveals a pattern of behavior by manufacturer of atrazine, its intermediary EcoRisk and scientists receiving funding from them, that strikingly resembles the practices of tobacco companies and their hired scientists, who for decades mounted highly sophisticated campaigns to undermine the ability and willingness of government regulators to take legal and regulatory steps against tobacco companies. These practices have been extensively documented by public health scientists based on analysis of internal memos from Philip Morris made public through legal discovery. Similar documentation is not available on this case, as all relevant internal memos from EcoRisk and Syngenta have not been made public. Were they public, it is possible that the case Hayes's constructs would fall apart. Alternatively, it is possible that the details would corroborate Hayes's analysis.
Hayes is not an impartial observer of these events [see note below], as his studies have been subject to repeated attacks by EcoRisk associated scientists. A truly independent assessment of his interpretation and conclusions would be helpful. This should include a review of the integrity of editorial decisions for the journal Environmental Toxicology and Chemistry, where two of the flawed studies were published after rejection by other scientific journals, and where one of the EcoRisk authors (Kendall) sits as an editor.
As publisher of this website, I should note that from 1990 to 2001 I directed the W. Alton Jones Foundation. In 2001, the foundation made a grant of $5,000 to U.C. Berkeley to support Hayes's research. His total funding for this work exceeded $500,000. Other funders include the National Science Foundation and the Howard Hughes Medical Institute..
John Peterson Myers, Ph.D.
Environmental Health Sciences
Charlottesville, VA 22902