Exploring the Joint Association Between Agrichemical Mixtures and Pediatric Cancer

Abstract

Nebraska's age-adjusted incidence rates for childhood cancers are among the highest in the US. Previous studies indicated associations between agrichemical exposures (atrazine and nitrates) and pediatric cancer rate, assuming single pollutant exposure. We evaluated the joint association between the agricultural mixture and pediatric cancer. Agrichemical exposures at a county scale were quantified using the USGS Pesticide National Synthesis Project for frequently applied pesticides from 1992 to 2014 in 93 Nebraska counties. Outcomes were quantified using pediatric cancer diagnosed among children <20 years of age (1992-2014) from the Nebraska cancer registry. We adjusted for social vulnerability factors such as race, income, employment, and access to care. The associations between 32 agrichemicals and cancer subtypes were assessed using the Generalized Weighted Quantile Sum Regression (gWQS) model. The model was fit assuming a Poisson distribution and using the pediatric population as an offset-term and social vulnerability factors as covariates. We observed a statistically significant positive association between the 32 agrichemicals and overall pediatric cancer and subtypes. The strength of associations was slightly stronger among brain and CNS cancers (β = 0.36, CI = 0.14, 0.57) compared to overall cancer (β = 0.30, CI = 0.16, 0.44) and leukemia (β = 0.23, CI = 0.09, 0.38). Dicamba, glyphosate, paraquat, quizalofop, triasulfuron, and tefluthrin largely contributed to the joint association. These findings may explain the joint associations of the agrichemical mixture on childhood cancer. Alternative biomarker-based approaches to measuring human exposure are worth investigating for chemicals of concern, particularly in counties with high agrichemical and cancer rates.

Keywords: environmental mixtures; pediatric cancer; pesticides; rural health.

Figure 1

(a) Spatial distribution of most‐applied pesticides in Nebraska Counties (1992–2014) using a latent variable (PC1) and (b) the loadings plot showing individual pesticide contribution to PC1.

Figure 2

Bivariate maps using agrichemical index and cancer rate by types. Cancer rates represent the annual rate per 100,000 children at a county scale. (a) Overall cancer. (b) Brain and other CNS cancer. (c) Leukemia. (d) Lymphoma.

Figure 3

Association between agrichemical mixture and pediatric cancer outcomes by cancer types. The y axis represents the beta coefficient (effect estimate) of agrichemical mixture. The x axis represents the overall pediatric cancer and its sub‐types. Adjusted model includes agrichemicals as a mixture and SDoH as covariates. Effect estimates with statistically significant association were represented using blue and non‐significant using orange. The lines with a dot represent constrained model and with a triangle represent un‐constrained model. The effect estimates and 95% CIs were generated using gQWS regression with 5 repeated holdouts. Overall cancer (β1 = 0.30, 95% CI 0.16, 0.44), brain and other CNS (β1 = 0.36, 95% CI 0.14, 0.57), and leukemia (β1 = 0.23, 95% CI 0.09, 0.38) showed positive associations with the agrichemical mixture. Whereas malignant bone (β1 = 0.33, 95% CI −0.19, 0.86) and germ cell (β1 = 0.19, 95% CI −0.33, 0.42) showed negative associations with the agrichemical mixture. Refer to Table 2. For chemical contributions to the joint associations.