Gray et al raise several important issues and we are grateful for the opportunity to continue the discussion regarding pesticides and their potential for carcinogenicity to humans.
Gray et al posit that a 7% to 18% deficit of cancer among pesticide applicators and their spouses in the Agricultural Health Study cohort is compelling evidence that pesticides are not associated with cancer. Of course, Gray et al are aware of the vast epidemiological literature that establishes that farmers in the United States are at a reduced risk of cancer overall, in large part because farmers smoke much less tobacco than the general population. Conversely, many epidemiologic studies, using a more rigorous methodology, as cited in our Table 5, show significant exposure-response patterns linking a particular pesticide to a particular cancer after carefully controlling for smoking and other potential confounders. We think most scientists would agree that observing a statistically significant exposure-response relationship between a pesticide and a cancer has more etiologic meaning than the indirect and uncontrolled observation made by Gray et al.
In our review, we make no conclusion regarding the relationship between 2,4-dichlorophenoxyacetic acid (2,4-D) and non-Hodgkin lymphoma (NHL) because NHL is a complex disease and relatively few epidemiologic studies with cell type-specific data relevant to NHL etiology are available. We do highlight other widely used pesticides that show etiologic evidence of human cancer (eg, terbufos and malathion with aggressive prostate cancer, diazinon with lung cancer) and state that the implications of this new evidence in public health need serious consideration.
In a field as dynamic as cancer etiology, we find it curious that Gray et al rely on an International Agency of Research on Cancer (IARC) document from 1987 that makes a conclusion about the adequacy of the carcinogenicity data on any compound. There have been major leaps in the understanding of the biology and etiology of cancer within the past 26 years and we believe a comprehensive scientific review of the extant literature on the association between specific pesticides and specific cancers by the IARC is greatly needed.
Gray et al also argue that animal toxicity testing of pesticides performed by private chemical companies and the Environmental Protection Agency have adequately demonstrated that these compounds do not pose a significant health risk to humans. Although animal models are absolutely necessary to predict toxicities in humans that might occur after exposure to environmental chemicals, they are not sufficient for public health surveillance and protection. This issue has been amply demonstrated in the pharmaceutical industry during the development of drugs. Indeed, human drug development is often stymied by unpredictable adverse health effects that arise in clinical trials, but were not predicted by either preclinical animal models or human cell culture experimentation. Furthermore, environmental toxicants such as arsenic, which have been shown to be highly toxic in humans in epidemiologic studies, do not concord with animal models of toxicity. New strategies using quantitative high-throughput screening technology with human cells to identify toxicity pathways and molecular mechanisms leading to the prediction of an in vivo response will become valuable tools in the future.
Gray et al state that: “Hypothesizing alternative mechanisms of cancer causation for these well-studied compounds relying on a minimal number of exploratory molecular studies appears blatantly speculative in comparison with the rich animal toxicology database,” and that “Reaching for new pathways to toxicity needs to seriously consider the substantial mass of current understanding and justify departure from well-established principles.” Unfortunately, it is not so clear that the mechanisms by which many environmental chemicals cause cancer actually do follow the “well-established principles.” Indeed, a chemical found to be negative in simple genotoxicity assays or lifetime animal bioassays could still contribute to human cancer formation by a variety of potential mechanisms, including epigenetic, endocrine disrupting, and proinflammatory pathways, none of which rely on genotoxicity. These mechanisms are important aspects of the current understanding of the carcinogenic process; they are certainly not “blatantly speculative.” Evaluating the health effects of pesticides in humans via epidemiologic studies, and molecular epidemiologic and toxicologic studies to evaluate mechanisms important to humans, will be necessary to safeguard health.