Ecological Analysis of the Association Between 2,4-Dichlorophenoxyacetic Acid (2,4-D) Exposure and Thyroid Cancer Incidence in the United States

Abstract

Background/Objectives: The increasing incidence of thyroid cancer in the United States suggests a possible role for environmental exposures that potentially alter thyroid hormone regulation through cellular toxicity. 2,4-Dichlorophenoxyacetic acid (2,4-D) is a commonly used herbicide that can act as an endocrine disruptor and may be toxic to the thyroid gland. This ecological study examined county-level associations of 2,4-D use and age-adjusted incidence of thyroid cancer in the United States. Methods: The study utilized age-adjusted incidence rates for thyroid cancer for 2017–2021 through the CDC database. County-level 2,4-D use data was obtained from the U.S. Geological Survey and averaged over two time periods: 2003–2007 (10-year lag) and 2008–2012 (5-year lag). The association between 2,4-D use and thyroid cancer incidence was tested using linear mixed effects models. We further stratified the data analyses by sex and rurality (using the Rural–Urban Continuum Codes, RUCC 6–9 codes). For this study, statistical significance was set at p < 0.05. Results: In nationwide analyses, no significant associations were found in combined (5-year lag: p = 0.218; 10-year lag: p = 0.276) and sex-stratified models (males: p = 0.540 and p = 0.497; females: p = 0.222 and p = 0.353, for the 5- and 10-year lags, respectively). In analyses limited to rural counties, 2,4-D exposure was significantly associated with greater thyroid cancer incidence for both the 5-year (p = 0.004) and 10-year (p = 0.006) lag periods for the total population. Conclusions: The strong associations observed in rural counties suggest that agricultural context may influence the observed relationship between county-level 2,4-D use and thyroid cancer incidence. Additional mechanistic and longitudinal studies are required to establish causality and to elucidate underlying endocrine pathways.

1. Introduction

Thyroid cancer incidence rates have tripled in the United States, from around 4.6 to 14.4 cases per 100,000 people per year between 1975 and 2016 [1]. An increase in advanced-stage disease and thyroid cancer mortality suggests that factors beyond improved diagnosis are contributing to this rising trend [1,2]. Ionizing radiation is still the most established environmental risk factor, but there is growing evidence that exposure to environmental pollutants may be involved in the epidemiology of thyroid cancer [3].

2,4-Dichlorophenoxyacetic acid (2,4-D) was one of the first commercialized synthetic herbicides and is still one of the most widely used herbicides in U.S. agriculture, with an estimated 46 million pounds applied annually [4]. This compound is an affordable weed killer that, while effective for maintaining large-scale crop yields, poses risks to environmental and human health [5]. The use of 2,4-D is linked with contamination of surface and ground water and toxic exposure to beneficial insects and aquatic organisms [5,6]. Due to its high solubility and poor binding to soil, 2,4-D can leach into drinking water or drift onto neighboring homes and gardens, contaminating residential areas near treated fields [5,7]. Humans may be exposed to 2,4-D by dermal contact during application, the ingestion of residues on food, and inhalation of airborne drift [8,9]. The U.S. EPA has stated that 2,4-D is not very persistent in soil (typical half-life of days to weeks) but both aerial and ground spraying can cause drift and deposition in nearby areas, leading to transient increases in air and surface-water concentrations in areas of high use [6,10]. Consequently, exposure of the general population is mainly dependent on the proximity of the agricultural or residential areas that are being treated.

The International Agency for Research on Cancer (IARC) has classified 2,4-D as “possibly carcinogenic to humans” (Group 2B) [11]. It has also been shown to impact thyroid and reproductive hormones in animal and mechanistic studies, resulting in the EPA placing it in its Endocrine Disruptor Screening Program for further evaluation of its endocrine-disrupting potential [12,13]. Studies in humans on 2,4-D and thyroid health have, to date, been largely inconclusive [9,14]. Correlations have been reported between herbicide exposure and hypothyroidism [14,15], although not specifically for 2,4-D. Co-exposure bias and misclassification are limitations of these studies. A case–control study found significant associations between higher urinary concentrations of 2,4-dichlorophenol, a metabolite related to 2,4-D, and risk of thyroid cancer in women [16]. In a rodent model, Tayeb et al. reported thyroid effects in rats after exposure to high doses of 2,4-D [16]. Mechanistically, when 2,4-D levels exceed a toxicokinetic threshold, the compound can interfere with hormone pathways in the thyroid axis by displacing thyroxine in rats [16]. However, these ROS-related effects were only observed at doses above environmentally relevant concentrations [16].

Overall, the epidemiology of 2,4-D and thyroid dysfunction in human populations remains underexplored. 2,4-D was selected for this analysis because it is among the most widely used pesticides in the United States and has been classified by the International Agency for Research on Cancer as a potential human carcinogen (Group 2B) [17]. There are also county-level application estimates for 2,4-D, which allow for an ecological investigation of its potential adverse effects [4]. Therefore, the current ecological study explores whether there is an association between thyroid cancer incidence and 2,4-D exposure by integrating US county-level data on herbicide use with age-adjusted thyroid cancer incidence rates and lagged exposure models that account for agricultural intensity and rurality.

2. Materials and Methods

2.1. Data Sources and Measures

County-level data for 2,4-D was collected using the U.S. Geological Survey Pesticide National Synthesis Project, a program that collects county-level and crop-specific estimates of pesticide/herbicide use since 1992. We obtained age-adjusted thyroid cancer incidence rates per 100,000 persons through the CDC database. The information from these sources provides official federal cancer statistics, including data from the National Cancer Institute Surveillance, Epidemiology, and End Results (SEER) Program as well as the CDC National Program of Cancer registries. The CDC provides statistics on newly diagnosed cancer cases for the US. County-level data on thyroid cancer incidence rates were only available over a 5-year period [4]. The CDC suppresses incidence rates when fewer than 16 cases are reported in a given category or time frame to ensure statistical reliability and protect patient confidentiality. This threshold is used because small case counts can lead to unstable estimates with large relative standard errors and increased risk of re-identification. The unit of analysis was the county; counties were included if they had complete data on thyroid cancer incidence, 2,4-D exposure, and rurality classification. The data was cleaned using R to harmonize county-level identifiers across datasets and exclude counties with missing thyroid cancer incidence rates or incomplete exposure information. The final cleaned dataset included counties with available age-adjusted incidence, corresponding 2,4-D usage estimates, and complete rurality classification. Age-adjusted thyroid cancer incidence rates were analyzed both in aggregate and stratified by sex (male and female). In order to account for agricultural intensity, the total cropland acreage was used as a proxy for agricultural activity, with data drawn from the USDA Farm Service Agency (FSA). For this study, categorization of rurality was defined through Rural–Urban Continuum Codes (RUCC). This spectrum classifies populations along an urban-to-rural gradient. Codes 6–9 were used as a marker for rural counties (population under 20,000). Rurality was also outlined according to the U.S. Census Bureau, which defined any population, housing, or territory outside urban areas or urban clusters (greater than 50,000 = urban; greater than 2500 = urban cluster) as rural. Outcomes were assessed across all counties with available data as well as within rural counties specifically.

2.2. 2,4-D Exposure Indicators

County-level herbicide usage was reported using EPest-low and EPest-high estimates. These estimates were used to approximate 2,4-D application in areas without survey data, drawing on 2,4-D by crop usage rates from nearby or regionally similar crop-reported districts [4]. EPest-low assumes no herbicide use for missing data, whereas EPest-high fills gaps by extrapolating data for similar crops in neighboring counties. When aiming to better capture 2,4-D exposure patterns, EPest-high was a more reliable measure because it accounts for missing data. The annual 2,4-D usage in kilograms per county was divided by total cropland acreage and then averaged over two separate 5-year intervals, 2008–2012 and 2003–2007, corresponding to a 5-year and 10-year lag, respectively. Chronic exposure to agricultural herbicides can produce different trends in cancer incidence, so two 5-year periods were extracted to more accurately capture the potential impact of herbicide exposure.

2.3. Statistical Methods

We first consolidated the data to highlight key variables that were most useful in identifying any correlations between herbicide exposure and thyroid cancer incidence. County-level datasets were merged using standardized state and county identifiers, and RUCC classifications were assigned based on county FIPS codes to define rurality for stratified analyses. Linear mixed-effects models with random intercept were fitted to evaluate the relationship between herbicide exposure per acreage and thyroid cancer incidence while accounting for fixed effects of exposure and random effects related to differences across counties. A variance components covariance structure was used for the linear mixed-effects models. Analyses were conducted separately for each lag interval and performed both a combined and a stratified (by sex) analysis. Age-adjusted thyroid cancer incidence was modeled as the outcome, with county-level 2,4-D usage standardized by cropland acreage (kg/acre) specified as the exposure of interest. No additional fixed-effect covariates were used in the models. Standardizing 2,4-D usage by cropland acreage accounted for agricultural intensity, while potential sex differences were evaluated through separate analyses instead of including sex as a model covariate. Linear mixed-effects models with county-level random intercepts were then used to evaluate the association between county-level 2,4-D use and thyroid cancer incidence.

Analyses were conducted using both untransformed and log-transformed outcomes across all models (combined and stratified), with log transformation applied to normalize skewed distributions. In subgroup analyses, models were restricted to rural counties, defined as RUCCs 6–9. Results were considered statistically significant at p < 0.05 for both untransformed and log-transformed models. All analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC, USA) and R version 4.4.2 (R Foundation for Statistical Computing, Vienna, Austria).

3. Results

3.1. All-County Analysis

We observed geographic variability in thyroid cancer incidence across the counties included in the mixed-effects analyses. For the 5-year lag dataset, the median incidence rate of thyroid cancer was 13.0 cases per 100,000 population (interquartile range [IQR], 10.7–15.6; range, 5.1–36.7). The median county-level 2,4-D exposure was 0.0463 kg per planted acre (IQR, 0.0215–0.1389; range, 0.000016–61,276.8). Similarly, for the 10-year lag dataset, the median incidence rate of thyroid cancer was 13.0 cases per 100,000 (IQR, 10.7–15.6; range, 5.1–36.7), and the median 2,4-D exposure was 0.0431 kg per planted acre (IQR, 0.0208–0.1590; range, 0.000007–717,906.2). The exposure distributions were right-skewed, suggesting substantial heterogeneity in the intensity of pesticide application across counties (

Table 1. County-level data demonstrating 2,4-D exposure and thyroid cancer incidence across the U.S., combined and stratified by sex.

	5-Year Lag		10-Year Lag
	95% Cl	p-value	95% Cl	p-value
Total	−6.0 × 10−5 (−1.6 × 10−4, 3.6 × 10−5)	0.218	−5.5 × 10−6 (−2.0 × 10−5, 4.4 × 10−6)	0.276
Female	−1.0 × 10−4 (−2.5 × 10−4, 5.9 × 10−5)	0.222	−7.4 × 10−6 (−2.0 × 10−5, 8.21 × 10−6)	0.353
Male	−3.0 × 10−5 (−1.4 × 10−4, 7.4 × 10−5)	0.540	−7.0 × 10−5 (−2.8 × 10−4, 1.4 × 10−4)	0.497

). Median age-adjusted thyroid cancer incidence rate in the 5-year lag dataset was 13.0 per 100,000 population (interquartile range [IQR], 10.7–15.6; range, 5.1–36.7). Median county-level 2,4-D exposure was 0.0463 kg per planted acre (IQR, 0.0215–0.1389; range, 0.000016–61,276.8). Median thyroid cancer incidence rate was 13.0 per 100,000 (IQR, 10.7–15.6; range, 5.1–36.7) in the 10-year lag dataset and median 2,4-D exposure was 0.0431 kg per planted acre (IQR, 0.0208–0.1590; range, 0.000007–717,906.2). Exposure distributions were right-skewed, indicating significant heterogeneity in the intensity of pesticide use across counties (Table 1).

The 5-year lag dataset included 1468 counties, while the 10-year lag dataset included 1466 counties. No statistically significant associations between 2,4-D exposure and thyroid cancer incidence in either the combined population or the sex-stratified groups were found for the 5- and 10-year lag (Table 1). For the 5-year lag, the estimated coefficients were negative and non-significant across all groups (p = 0.218 for the combined population, p = 0.222 for females, and p = 0.540 for males). Similarly, in the 10-year lag analysis, results remained non-significant, with p = 0.276 for the combined population, p = 0.353 for females, and p = 0.497 for males. Although the primary models showed no significant relationships, applying a log transformation to thyroid cancer incidence data revealed statistically significant associations with 2,4-D exposure in the 10-year lag analysis with p < 0.0001; however, these values are negligible as they likely reflect numerical instability rather than a true epidemiologic effect (Supplementary Table S1).

3.2. Rural County-Specific Analysis

In a subset analysis limited to rural U.S. counties, 385 counties contributed data to the 5-year lag models and 386 counties to the 10-year lag models. There were significant positive associations between 2,4-D exposure and thyroid cancer incidence in the combined population for the 5-year lag (p = 0.004) and the 10-year lag (p = 0.006). However, the sex-stratified analyses did not indicate significant associations. For females, p = 0.108 in the 5-year lag and p = 0.173 in the 10-year lag. For males, p = 0.453 and p = 0.520 for the 5-year lag and 10-year lag, respectively (

Table 2. County-level data demonstrating 2,4-D and thyroid cancer incidence in rural U.S. counties, combined and stratified by sex.

	5-Year Lag		10-Year Lag
	95% Cl	p-value	95% Cl	p-value
Total	0.017 (0.006, 0.029)	0.004 *	0.034 (0.010, 0.057)	0.006 *
Female	0.025 (−0.005, 0.054)	0.108	0.038 (−0.016, 0.093)	0.173
Male	−0.334 (−1.277, 0.610)	0.453	0.520 (−0.766, 1.805)	0.393

* Statistically significant if p < 0.05.

). In the log-transformed models, 2,4-D exposure was significantly associated with higher thyroid cancer incidence in the combined population for both the 5-year lag (p = 0.021) and 10-year lag (p = 0.027) (Supplementary Table S2).

4. Discussion

In this county-level ecological analysis, while no statistically significant associations were found between 2,4-D exposure and thyroid cancer incidence across all counties for either the 5-year or 10-year lag models, 2,4-D use was significantly associated with thyroid cancer incidence for both the 5-year and 10-year lag intervals when restricted to rural counties. The findings suggest that higher county-level 2,4-D use in rural areas may be associated with higher thyroid cancer incidence at the ecological level.

The significance noted in rural counties may reflect higher environmental and residential exposure levels in those areas. In rural areas, cropland density is generally higher and homes are closer to treated fields, both of which are associated with greater potential for non-occupational exposure [18,19,20]. Multiple studies have shown that residential proximity to agricultural land, amount of cropland surrounding homes, and amount of pesticide applied in the local area are important factors that could increase community-level exposure [18,19,20,21]. Additionally, house-dust pesticide concentrations were 64% lower in homes located 250 m from treated fields versus those located 23 m from treated fields and were 2.3 times higher in the homes of farmers who had applied pesticides more recently or frequently [22]. These proximity-dependent exposure gradients are particularly relevant in rural communities, where residential properties are often adjacent to or surrounded by treated cropland [22].

In rural communities, the main routes of exposure to 2,4-D are spray drift, deposition, and contamination of food and water [19,21]. Measurable concentrations of 2,4-D in air and soil can result from spray drift during and after the application season, with higher levels found in areas of greater agricultural intensity [18,20,21]. Biomonitoring studies have also found significantly higher urinary 2,4-D concentrations in people living near treated fields, particularly during application periods and for women and children [9,23,24]. A study of private wells in agriculturally intensive northeast Iowa found pesticides and related transformation products, including 35 herbicide-related compounds, in tap water, often exceeding health-based hazard index screening levels [25]. Environmental transport mechanisms of this nature are especially important in rural areas, where an estimated 23 million U.S. households use private drinking water wells that are not subject to federal monitoring or treatment requirements [26].

In addition to environmental contamination, rural populations are exposed through para-occupational pathways that are not typical of non-agricultural settings [27]. The take-home pathway, where residues of pesticides are transported from the workplace to the home on clothing, footwear, skin and in vehicles, has been documented as an important contributor to household exposure in farming families [27,28]. Rural agricultural communities often have work and living spaces that are interconnected, which may increase the likelihood of unintentional pesticide exposure among all household members [28].

Numerous experimental studies have examined the mechanistic effect of 2,4-D exposure on thyroid-related outcomes. Experimental studies demonstrate that 2,4-D disrupts thyroid hormone synthesis, metabolism, and signaling [13,16]. Significant (p < 0.01) decreases in serum triiodothyronine (T3) and thyroxine (T4) concentrations were observed in Wistar rats after administration of sublethal oral doses of 2,4-D, indicating a disturbance in thyroid hormone balance within days of exposure [29]. At the molecular level, 2,4-D changes the expression of thyroid-related genes like Mct10 and Dio2, which regulate hormone transport and local activation [30]. Downregulation of these genes implies disruption of thyroid hormone signaling in reproductive tissues and is mechanistic evidence of the endocrine disrupting effects of 2,4-D [30]. Long-term changes in thyroid hormone balance may cause sustained stimulation by TSH and activation of signaling pathways that favor proliferation, which may be linked to papillary thyroid carcinogenesis [31]. The Agricultural Health Study identified an increased risk of hypothyroidism with 2,4-D exposure after analyzing a large cohort of pesticide applications [32]. The data indicated an exposure–response relationship that was dependent on intensity-weighted lifetime use. The National Report on Human Exposure to Environmental Chemical (NHANES) biomonitoring studies have also demonstrated associations between urinary 2,4-D metabolites and altered thyroid hormone profiles, consistent with experimental evidence showing disruption of thyroid hormone homeostasis through the displacement of thyroxine from plasma-binding proteins [9,23]. Endocrine-disrupting chemicals have also been implicated in autoimmune thyroid diseases, including Hashimoto thyroiditis and Graves disease [28,33]. Taken together, these findings suggest that 2,4-D can affect thyroid physiology at the hormonal, genetic and systemic levels, possibly contributing to thyroid carcinogenesis [13,29,30].

Beyond thyroid effects, experimental studies suggest that 2,4-D may disrupt additional endocrine pathways [13,30]. Exposure to environmentally relevant concentrations of 2,4-D drastically impacted the expression of genes involved in endocrine regulation, such as those associated with 20-hydroxyecdysone and juvenile hormone signaling in aquatic invertebrates [30]. Vertebrate models have demonstrated similar effects. Freshwater fish chronically exposed to 2,4-D at 0.5–4.0 mg/L also show decreased plasma testosterone in males, increased ovarian atresia in females, and altered expression of genes along the hypothalamic–pituitary–gonadal axis [34,35,36]. Downregulation of estrogen and androgen receptors was also involved in the mechanism [36]. In mammals, chronic oral exposure to 2,4-D has been associated with a 36% reduction in testosterone, a dramatic increase in insulin levels, and increases in body weight and body fat [13]. These results suggest that 2,4-D may interfere with the regulation of reproduction and metabolism. The results in human populations are less consistent, but some epidemiologic studies have reported increased odds of hypothyroidism and other endocrine disorders associated with occupational or environmental 2,4-D exposure [32].

Although the CDC database did not specify the histologic subtype of thyroid cancer, papillary thyroid carcinoma (PTC) is most likely the primary form of cancer observed in the data, as it has been identified to greatly contribute to the drastic increase in thyroid cancer incidence [3]. Based on previous studies, PTC has possible associations with environmental pollutants and endocrine-disrupting chemicals [36]. The disruptive factors rooted in papillary thyroid carcinogenesis involve changes in thyroid hormone regulation through modifications in thyroid-stimulating signaling pathways [37]. More specifically, the binding of TSH to the TSH receptor in thyroid follicular cells triggers multiple downstream signaling cascades, such as cAMP/protein kinase A (PKA), phospholipase C (PLC), and PI3K-Akt pathways, which regulate cell proliferation, survival, and thyroid gland growth. Endocrine-disrupting chemicals that disrupt thyroid hormone homeostasis can cause chronic compensatory TSH signaling [38]. This has been shown to cooperate with the BRAF V600E oncogene, the most frequent driver of mutation in PTC. Overall, the importance of subtype stratification was demonstrated in the work of Rusiecki et al., where associations with the same pesticide produced opposite results in different variants of PTC [39]. Thus, identifying the histologic subtype of thyroid cancer is influential in precisely identifying the differential effects among papillary, follicular, medullary, and anaplastic thyroid cancers due to significant variation.

This ecological study found exposure to 2,4-D in rural counties to be significantly associated with thyroid cancer incidence at both 5- and 10-year lag times. Part of the explanation for this is the relatively short half-life of 2,4-D in the environment, which varies from a few days to several weeks, depending on soil and water conditions [5,6,10]. Despite its relatively short environmental half-life, repeated seasonal application may result in chronic exposure in agricultural communities. The significance of shorter and longer lag periods might be reflected in the fact that 2,4-D can elicit both acute and cumulative biological effects, especially in rural areas where environmental and occupational exposures are more common [13,29,30]. A large case–control study in California showed a link between residential exposure to multiple agricultural pesticides, including 2,4-D, and a higher risk of thyroid cancer, with risk increasing with the number of pesticides used in the 20 years before diagnosis [40]. These findings suggest that 2,4-D may be associated with thyroid carcinogenesis through mechanisms that operate during both short- and extended-term exposure periods.

This overall study adds to the understanding of the association between 2,4-D exposure and thyroid cancer incidence, but there are limitations to this study. As an ecological study, these findings are subject to ecological fallacy and associations at the county level may not translate to individual-level risk. The analysis is also subject to possible residual confounding by unmeasured factors such as disparities in healthcare access, socioeconomic status, thyroid-screening practices, and co-exposure to other pesticides. Differences in healthcare access between rural and urban populations can impact thyroid cancer diagnosis as well as the stage at presentation. Rural residence and greater distance from treatment centers have been reported to be independently associated with later-stage thyroid cancer at diagnosis, emphasizing the role of geographic differences in access to care [41]. A Surveillance Epidemiology and End Results (SEER)-based analysis confirmed lower incidence but worse survival in rural areas, pointing to decreased diagnostic intensity in these regions [42]. For this analysis, information on healthcare access and diagnostic practices was not available in our dataset and thus could not be included. The county-level exposure measures employed in this study are useful proxies, but also obscure finer distinctions, especially in mixed-use or suburban areas where true exposure may vary from agricultural usage averages. In ecological analyses, nondifferential misclassification of exposure can bias true exposure–disease associations towards the null, especially when populations with little or no environmental exposure are combined with highly exposed agricultural populations [18]. Another limitation is the suppression of county-level data by the CDC when case counts fall below the reporting threshold. Due to the fact that suppressed counties are not identifiable, the exact number of counties excluded from analyses cannot be determined. Many low-count counties are rural areas, so some of the areas that are most relevant to this question are underrepresented in the data. This may bias results toward the null, and reduce power for sex-stratified analyses. The counties with incidence data available do not perfectly overlap with all counties that use pesticides, and this difference should be kept in mind when thinking about generalizability. Furthermore, 2,4-D is not often used alone, and co-application with other pesticides raises the possibility that the observed associations could reflect the synergistic effects of other chemicals. In U.S agriculture, 2,4-D is commonly tank-mixed or applied with glyphosate, dicamba, and atrazine, particularly in corn and soybean production systems [24,43]. Many of these co-applied pesticides have established or suspected thyroid-disrupting properties. In the Agricultural Health Study, use of glyphosate, dicamba, and 2,4-D was each independently associated with increased risk of incident hypothyroidism among pesticide applicators [15]. Future studies could employ mixture analysis approaches to disentangle the independent and joint effects of co-applied herbicides on thyroid cancer incidence. A final limitation to this ecological analysis is the lack of publicly available county-level datasets, which in turn prevents reliable identification of counties practicing exclusively pesticide-free agriculture. Future studies looking at comparisons between conventional and organic agricultural regions could be insightful in further understanding the contribution of pesticide exposure to thyroid cancer incidence.

5. Conclusions

This analysis of the association between 2,4-D exposure and thyroid cancer incidence suggests a positive ecological association between county-level 2,4-D use and thyroid cancer incidence, particularly in rural regions where agricultural use is most concentrated. As a prominent agricultural chemical used across the United States, further analysis to understand not only the environmental persistence but also the potential to interfere with endocrine pathways relevant to the thyroid function of 2,4-D is important. Future studies should capture the impact of 2,4-D using an individual-level assessment, using 2,4-D and thyroid function biomarkers to test dose–response relationships in a more direct manner. Studies incorporating multi-pesticide exposure frameworks are also necessary to disentangle single-agent and combined effects relevant to thyroid cancer risk. Since the associations observed in this study indicate rural specificity in terms of impact, follow-up studies should focus on rural populations to better understand susceptibility in vulnerable communities, especially those with higher occupational or environmental exposure.