Our study is the largest study to date of cancer risk associated with exposure to the herbicide dicamba. We observed a suggestion of increased risk for lung and colon cancer when the referent group comprised low-exposed applicators but not when the referent group comprised unexposed applicators. With further follow-up and accumulation of exposed cases, we believe we will better understand this phenomenon.
The association with lung cancer we observed in our study is similar to that reported in an earlier nested case–control study in the AHS cohort. Alavanja et al. (2004)
observed a positive trend in risk for lung cancer with lifetime exposure days of dicamba, for the highest exposure tertile (highest tertile divided at the median) relative to the lowest exposure tertile [odds ratio (OR) = 1.0, 1.3, 1.7, 3.1; p
-value for trend = 0.04). A significant trend was not observed when the investigators used unexposed participants as the referent group (OR = 1.0, 0.7, 0.9, 1.1, 1.6; p
-value for trend = 0.15). When we modeled risk of lung cancer associated with lifetime dicamba exposure days using the low exposed as the referent group and the same tertile cut points used in the analysis conducted by Alavanja et al. (2004)
, not surprisingly, our results were very similar (OR = 1.0, 1.1, 1.6, 2.1; p
-value for trend = 0.02). There is no prior evidence that suggests an association between dicamba exposure and colon cancer.
Exposure to dicamba has been associated with increased risk for NHL in a few previous case–control studies. In a case–control study of NHL and pesticide exposure conducted in Canada, information on pesticide exposure was collected through a combination of mailed questionnaires and telephone interviews (McDuffie et al. 2001
). After adjusting for demographic characteristics, family history of cancer in a first-degree relative, and history of selected medical conditions, increasing days per year of dicamba application was associated with an increased risk of NHL (OR = 1.7; 95% CI, 1.0–2.8). When exposure to dicamba as a general class was evaluated, which included dicamba-only products as well as mixtures of dicamba and glyphosate and mixtures of dicamba, 2,4-D and mecoprop, NHL risk increased slightly with increasing days per year of application (OR = 1.9; 95 %CI, 1.3–2.7). Conversely, results from a case–control study of NHL and farming in the United States suggested no association between risk of NHL and ever handling either benzoic acids as a class (OR = 1.3; 95% CI, 0.9–1.9) or dicamba in particular (OR = 1.2; 95% CI, 0.7–2.0) (Cantor et al. 1992
). After restricting the analyses to pesticides handled prior to 1965, risk for NHL was elevated among dicamba users (OR = 2.8; 95% CI, 0.96–8.1). Our prospective data do not provide evidence of an association between dicamba and NHL. Our findings, however, may be influenced by the small number of cases and relatively short follow-up time.
Results from other studies provide no evidence for an association between dicamba and risk of leukemia (OR = 0.7; 95% CI, 0.4–1.4) (Brown et al. 1990
) or multiple myeloma (OR = 1.3; 95% CI, 0.6–2.8) (Brown et al. 1993
). However, Burmeister (1990)
reported a nonsignificant, marginal association between exposure to benzoic acids as a class and risk of multiple myeloma (OR = 1.22; confidence interval/p
-value not reported). In our study, exposure to dicamba is likely to be a combination of exposure to dicamba-only products as well as to dicamba mixtures, making it difficult to disentangle the effect of dicamba from other pesticides included in dicamba mixtures. After stratifying dicamba models for lung and colon cancers by never/ever use of 2,4-D, atrazine, and glyphosate (three herbicides commonly mixed with dicamba), there was no evidence for either increased risk among dicamba-only users, or for increased risk among participants who also used 2,4-D, atrazine, or glyphosate. Because this analysis is based on information pertaining to ever/never use of individual pesticide active ingredients, however, we were unable to unambiguously differentiate between use of dicamba-only products and dicamba mixtures at this time.
There is little experimental evidence to suggest that dicamba is carcinogenic or mutagenic (U.S. EPA 1999
). Feeding studies in rats, mice, dogs, and rabbits have shown no increased incidence of tumors (Extension Toxicology Network 1999). There is evidence that dicamba acts as a peroxisome proliferator (PP) by increasing fatty acyl–coenzyme A oxidase activity in the livers of rats and activating the peroxisome proliferator receptor in a dose-dependent fashion (Espandiari et al. 1998
). It is thought that PPs may induce liver tumors in rats through mechanisms related to oxidative stress, inducing replicative DNA synthesis, or by promoting growth of preneoplastic lesions (Espandiari et al. 1998
). Dicamba induced DNA damage in one study of rats (Perocco et al. 1990
). In another study of mice, dicamba caused mortality in two of four mice injected with dicamba but did not increase xenobiotic-metabolizing activities in the two surviving mice (Moody et al. 1991
). Few epidemiologic studies on the effect of PPs on humans have been conducted (Nakajima et al. 2002), but there are marked species differences in response to PPs (Lai 2004
). Humans seem to exhibit a weak response to PP chemicals (including certain pesticides, industrial solvents, and hypolipidemic drugs), which may be due to low levels of peroxisome proliferator-activated receptor alpha in human liver (Lai 2004
; Maloney and Waxman 1999
The Agricultural Health Study is the largest study to date of pesticide applicators exposed to dicamba. The potential for recall bias is minimal, as exposure information was collected prior to cancer diagnosis. AHS applicators have been shown to provide reliable information about their histories of pesticide use (Blair et al. 2002
; Hoppin et al. 2002
), although misclassification can occur. Misclassification in a prospective study is likely to be nondifferential with regard to cancer occurrence and, although it could diminish estimates of relative risk, it is unlikely to create false positives (Checkoway et. al. 2004
Our results may be affected by simultaneous exposure to other pesticides of varying intensity that has changed over time, although we attempted to account for this by adjusting for total lifetime days of any pesticide use. In addition, we could not differentiate use of dicamba-only products from dicamba mixtures. This is one of the biggest challenges in conducting epidemiologic research on pesticides, as many pesticides are most often used in combination with others in complex mixtures and not as individual pesticides. The existing toxicologic data, however, pertain to dicamba as an individual chemical. Our findings may also be limited because of a relatively short period of follow-up and small numbers of cases for some cancer sites. Because of the lack of consistency among results from the four evaluations of exposure metric and referent type for lung and colon cancer, these findings should be interpreted with caution.
Despite these limitations, our prospective study of cancer incidence among dicamba-exposed pesticide applicators provided an opportunity afforded in few other studies to evaluate cancer risks associated with exposure to dicamba while adjusting for lifetime use of other pesticides and lifestyle factors. We did not detect much evidence for an association between dicamba exposure and any of the cancer sites investigated, but the patterns of associations observed for lung and colon cancers warrant further attention. We will re-examine dicamba in the future when larger numbers will allow for a more comprehensive evaluation of lung and colon cancer, as well as additional cancer sites.