County-Level Pesticide Exposure and Cancer Incidence in the United States

An ecological study of 3,248 US counties examining the association between agricultural pesticide application and cancer incidence rates, using 20+ analytical methods from bivariate correlations through Bayesian spatial models and four risk-factor gauntlets.

3,248 Counties
160+ Variables
20+ Methods
12 Compounds
4 Gauntlets

Key Findings

After controlling for 13 sociodemographic, health-behavior, and environmental confounders—and accounting for spatial autocorrelation via Bayesian BYM2 models—agricultural pesticide exposure shows a statistically credible positive association with two cancer types:

Kidney & Renal Pelvis
RR = 1.034
94% HDI: [1.009, 1.044]
P(RR > 1) = 0.997 · BYM2 v2B (13 covariates)
Colorectal
RR = 1.018
94% HDI: [1.003, 1.027]
P(RR > 1) = 0.991 · BYM2 v2B (13 covariates)
All Other Types
NS
6 of 8 cancer types null
NHL, Leukemia, Bladder, Prostate, Lung, All-Site

Compound Specificity

The signal is driven by herbicides (glyphosate, atrazine, 2,4-D, dicamba, metolachlor-S): 5 of 6 herbicides significantly associated with colorectal cancer, while 0 of 3 insecticides and 0 of 3 fungicides show colorectal associations. This chemical-class specificity argues against residual confounding by general agricultural intensity.

Six Lines of Evidence

Evidence 1

Bayesian Spatial Models

BYM2 models with ICAR spatial random effects and 9–13 covariates. Kidney and colorectal survive all model specifications.

Evidence 2

Instrumental Variables

IV/2SLS using crop acreage as instrument (F=377.8). IV estimate exceeds OLS (0.035 vs 0.019), suggesting OLS is biased toward zero.

Evidence 3

Long-Difference

Within-county changes in pesticide use (1997→2012) predict changes in kidney (β=0.068, p=0.003) and colorectal (β=0.079, p<0.001) rates.

Evidence 4

Compound Specificity

12 individual compounds tested: herbicides consistently significant, insecticides and fungicides (for colorectal) null.

Evidence 5

Negative Controls

Livestock density (6/6 NS), diabetes (NS), fungicides→colorectal (0/3 NS) confirm signal is not general agricultural or metabolic confounding.

Evidence 6

Risk Factor Gauntlets

Pesticide associations survive as covariates across all 4 gauntlets (smoking, obesity, alcohol, inactivity). Independent of established risk pathways.

Cross-Gauntlet Scorecard

Gauntlet IARC Score Best Hit Pest → Kidney Pest → Colorectal
Smoking 7/8 PASS Larynx RR=1.205* Survives —
Obesity 4/8 MIXED Myeloma RR=1.081* Survives Survives
Alcohol 2/7 MIXED Oral Cavity RR=1.044* Survives Survives
Inactivity 2/7 MIXED Liver RR=1.085* Survives Survives

See the full Gauntlets page for detailed results per risk factor.

Effect Estimates Across Methods

Forest plot showing pesticide effect estimates across analytical methods, from bivariate correlation through BYM2
Figure 1. Pesticide effect estimates on all-site cancer rates across analytical methods, ordered by increasing methodological rigor. Effects attenuate as spatial controls strengthen, with kidney and colorectal cancer surviving the most stringent models.
Pesticide rate ratio stability across risk factor gauntlets
Figure 4. Pesticide rate ratio stability across risk factor gauntlets. Kidney and colorectal associations remain significant regardless of which established risk factor is modeled as primary exposure.

Positive Control Validation

To validate the analytical pipeline, we tested the well-established PM2.5–lung cancer relationship. Our IV/2SLS estimate yields a 17.9% relative increase in lung cancer incidence per 10 μg/m³ PM2.5, consistent with published meta-analytic estimates of 10–15%. This confirms the pipeline can detect known environmental carcinogens at realistic effect sizes.

This is an ecological (county-level) study and cannot establish individual-level causation. See Limitations for a full discussion of the ecological fallacy, exposure misclassification, and spatial confounding.