Association Between Pesticide Use and Incidence of Diffuse Large B-Cell Lymphoma

Results and Discussion

A total of 2,258 patients with diagnosis of DLBCL in 2012 were extracted from SEER Lymphoma-Leukemia dataset using ICDO-3 code 9680. These patients were distributed over 167 counties that also had pesticide application data. USGS data included 290 compounds reported as being used in agriculture. Of these compounds, 65 were excluded due to no reported use in the studied counties. The remaining 225 were included in the analysis. To investigate the correlation between compounds' exposure and the incidence of DLBCL, the quantity used of each compound (kg/km²) was matched to the 167 counties that had DLBCL cases reported. Pearson correlation analysis was then applied and showed a statistically significant correlation (p<0.05) between the incidence of DLBCL and area density of 40 compounds from the 225. Only one (Pendimethalin) of the 40 compounds was listed as a probable carcinogen by the International Agency for Research on Cancer (IARC) (15), and none are listed in the latest Report on Carcinogens from the United States National Toxicology Program (7). This number decreased to 14 compounds after applying Benjamini-Hocheberg method with alpha set to 0.05 to account for false discovery (Table I). None of the 14 compounds were on the WHO list of obsolete pesticides (16). Only one (Methyl-Parathion) had a WHO hazard classification >3 for acute toxicity based on animal-model LD50. To visually show the correlation between area density of pesticide application and incidence of DLBCL, we used Epi Info software to generate maps for incidence as well as pesticide. These maps show that areas with higher application of the 14 pesticides correspond to areas with increased DLBCL incidence (Figure 1).

To analyze how these pesticides are being used in agriculture, we used University of Hertfordshire Pesticides Properties Database to extract the use as well as the chemical class of the compounds with significant correlation. Of the 14 compounds, 13 are used as herbicides; Methyl Parathion is an organophosphate used as an insecticide. Other organophosphate chemicals have been associated with cancer development (7, 8). However, other compounds belong to different chemical classes with dinitroaniline being the most frequent (Table I). Analysis of the underlying mechanism of action of these compounds is needed to uncover any possible common pathway, such as their effects on hematopoietic cell mutagenesis, B cell hyperstimulation, alteration of intestinal microbiome, and other potential mechanisms for carcinogenicity. Glyphosate, is a herbicide that has been extensively studied due to its association with NHL (17); it was used in 160 (95.8%) of the studied counties in 1992 with a total of 0.2 million pounds applied, and was not significantly associated with diagnoses of DLBCL in 2012. In 1992, 0.8 million pounds were applied to all US crops compared to 78.2 million pounds in 2001 (18).

This analysis advances our understanding of a dose response relationship between the quantity of pesticides used in a particular county and the incidence of DLBCL. The identified compounds warrant study as possible carcinogens. A significant effect size was recently reported in a cohort of 244 French patients with DLBCL, with a 2-year event-free survival hazard ratio adjusted for confounders of 3.5 for patients with occupational exposure to pesticides (9). Future investigations using genetic data can identify whether exposure to pesticides is associated with specific mutational patterns.

This study makes use of high-quality data sets designed to reliably reflect geographic and temporal trends. Population-level data offer a broader perspective than prior case-control studies utilizing recall of chemical exposure in patients with lymphoma (19). Agricultural chemicals are used in the global food chain to improve productivity and reduce crop loss due to pests (20). The long-term effects of specific occupational chemicals should be investigated. With an increasing number of high-quality epidemiologic studies showing an association between pesticides exposure and incidence of cancer, the next step is to find specific associations that warrant multi-pronged analyses for carcinogenesis (21).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Heat maps showing incidence of diffuse large B-cell lymphoma (DLBCL) (A) and use of Mecoprop (B), one of the pesticides reported in Table I. Areas with higher pesticides use corresponds to increase incidence of DLBCL. Enlarged areas shown under each map for clearer view. Heat maps for other pesticides reported in Table I are available upon request.

This study is limited by the SEER*Stat database only having data for 167 counties of the United States that also had pesticide application data, representing an estimated 15% of DLBCL diagnoses in 2012 (22). This lessens the power to detect associations between additional chemicals and the incidence of DLBCL. It is not unusual that multiple pesticides are used in the same area. This adds to the complexity of the analysis as these are not only confounding variables, but also can have important interactions with each other. The time lag between exposure and diagnosis biases our investigation toward the null hypothesis, since relocation and variable latency would be expected to dilute the association (23). Finally, it is important to note that while this study highlights an association to generate hypotheses, it does not establish causality. Further studies designed to mitigate the effect of confounding variables, discover putative biologic mechanisms, and evaluate dose-response relationships are needed to further characterize the apparent link between these pesticides and DLBCL.