As shown in Table 1, median annual endocrine-disrupting pesticide use was higher in rural vs. urban counties (17,523 kg vs. 14,743 kg; p = 0.02), with some variation by pesticide class. The use of phosphonates (9,638 kg vs. 7,503 kg; p = 0.01) and other miscellaneous pesticides (525 kg vs. 418 kg; p = 0.01) was higher in rural vs. urban U.S. counties. In contrast, the use of organochlorines (21 kg vs. 37 kg; p < 0.001) was lower in rural compared to urban counties. Use of carbamates, neonicotinoids, organophosphates, pyrethroids, and triazines did not differ significantly across rural and urban U.S. counties. Phosphonates (glufosinate and glyphosate marketed as Roundup®) were the most heavily used endocrine-disrupting pesticides with a median 8,483 kg applied per year, followed by triazines (atrazine, propazine and simazine) at a median of 2,625 kg annually.
Trends in endocrine-disrupting pesticides over time are shown in Fig. 1. The use of carbamates (β = − 73.7,), organochlorines (β = − 20.2), organophosphates (β = − 106.1), triazines (β = − 143.5), and other miscellaneous pesticides (β = − 56.7) decreased significantly over time from 2001 to 2015 (all ptrend < 0.001). However, increasing trends were observed for neonicotinoids (β = 65.6, ptrend < 0.001), phosphonates (β = 2248, ptrend < 0.001), and pyrethroids (β = 1.5, ptrend < 0.05) over time. Similar trends were observed in analyses stratified by rurality (Supplementary Fig. 2).
Fig. 1The alternative text for this image may have been generated using AI.
Trends in the amount of pesticides with endocrine-disrupting properties used (Kg) from 2001 to 2015 across U.S. Counties by pesticide class. *Beta estimates and p-values obtained from the linear regression model assessing average annual pesticide use over time; **Significant non-monotonic trend; the Mann–Kendall test was not significant
The geographic distribution of endocrine-disrupting pesticide usage (2001–2015) and age-adjusted breast cancer incidence rates from 2016 to 2020 in the U.S. are shown in Supplementary Fig. 3. Endocrine-disrupting pesticide use was higher in the central northern region, with smaller areas of higher use in California, Washington, Montana, and Florida. The geographic distribution of age-adjusted breast cancer incidence rates from 2016 to 2020 was more dispersed, with some areas of higher rates in the central northern region and lower incidence rates in the southwestern U.S. counties.
County-level characteristics varied across tertiles of endocrine-disrupting pesticide use (Table 2). The average percentages of smoking, below 150% poverty, unemployment, and uninsurance all decreased with increasing pesticide tertile (all ptrend < 0.05). Age-adjusted breast cancer incidence rates were higher in counties with higher pesticide use (123.7 vs. 119.4 per 100,000 in Tertile 3 vs. Tertile 1; ptrend < 0.05).
Table 2 Characteristics of 2,457 U.S. counties (2016 – 2020) by total pesticide use (kg) of 38 pesticides with endocrine disruption properties applied from 2001 to 2015
Age-adjusted and multivariable-adjusted associations of breast cancer incidence per IQR increase in county-level pesticide use in the overall U.S. and by rurality are shown in Table 3. Endocrine-disrupting pesticide use was positively associated with breast cancer incidence rates in rural counties. Results were similar, though slightly attenuated in multivariable models. After adjusting for covariates, endocrine-disrupting pesticide use was positively associated with breast cancer incidence in rural (aRR = 1.02,95% CI 1.01–1.03), but not in urban (aRR = 1.00, 95% CI 1.00–1.00) counties (pint < 0.001). Associations differed by pesticide type and by metropolitan status. In multivariable models, neonicotinoids (aRR = 1.01, 95% CI 1.01–1.02) and phosphonates (aRR = 1.02, 95% CI 1.01–1.03) were associated with higher breast cancer rates in rural counties but associations were not significant in urban regions (both pint < 0.05). Carbamates, organochlorines, organophosphates, pyrethroids, triazines, and miscellaneous endocrine-disrupting pesticides were not significantly associated with breast cancer rates in multivariable models.
Table 3 Age-adjusted and multivariable-adjusted associations between breast cancer incidence and interquartile range (IQR) increase in county-level pesticide use in the U.S., overall and by rurality
Moran’s I values ranged from − 0.008 to 0.025, indicating minimal spatial autocorrelation in the models and suggesting no meaningful residual spatial dependence. Local Indicators of Spatial Association (LISA) analyses (Supplementary Fig. 4) were used to explore local spatial associations between breast cancer incidence and endocrine-disrupting pesticide use. Statistically significant clusters were limited and did not form consistent or contiguous spatial patterns. A small number of high–-high clusters (counties with elevated age-adjusted breast cancer rates surrounded by high pesticide use) and low–high clusters were identified, but these were geographically dispersed. Overall, the observed spatial patterns were weak and heterogeneous, consistent with minimal global spatial autocorrelation.
Most county-level characteristics and pesticide usage for counties included in the analysis (n = 2,457) differed significantly from those excluded due to missing data (n = 686), though differences were small in magnitude (Supplementary Table 1). In a supplemental analysis assessing tertiles of total endocrine-disrupting pesticide use in relation to breast cancer incidence (Supplementary Table 2), results were generally consistent with the main findings. Higher pesticide use was associated with modestly increased age-adjusted breast cancer rates overall, particularly in the highest tertile of exposure (aRRT3 vs. T1 = 1.03, 95% CI 1.01–1.04). Stratified analysis indicated that this association was driven by rural counties, where high pesticide use was associated with higher breast cancer incidence rates (aRRT3 vs. T1 = 1.06, 95% CI 1.03–1.08), while no association was observed in urban counties (aRRT3 vs. T1 = 0.99, 95% CI 0.97–1.01). In supplemental analyses of individual pesticide types (Supplementary Table 3), most associations were null, though interpretation was limited by small sample sizes for several pesticides. In rural counties, thiamethoxam (aRR = 1.02, 95% CI 1.01–1.02), chlorpyrifos (aRR = 1.01, 95% CI 1.01–1.02), glufosinate (aRR = 1.01, 95% CI 1.01–1.02), and glyphosate (aRR = 1.02, 95% CI 1.01–1.03) were modestly associated with higher breast cancer rates.
In supplemental analyses using lower-bound EPA pesticide estimates (Supplementary Table 4), associations were generally consistent with the main findings. Total pesticide use was associated with higher breast cancer rates in rural counties (aRR = 1.02, 95% CI 1.01–1.03). By pesticide class, most associations were null or very small; however, phosphonates were associated with higher breast cancer rates in rural counties (aRR = 1.02, 95% CI 1.01–1.03). While secondary analysis of young-onset breast cancer was limited by smaller sample size (n = 1,405 U.S. counties with estimated young-onset breast cancer rates), we did not observe any significant associations between endocrine-disrupting pesticides and young-onset breast cancer incidence rates (Supplementary Table 5).
