This study evaluated the relationship between crop application of 2,4-D and urine biomarkers of exposure in a large, nationally representative sample of 14,395 participants. There were significant associations between crop application of 2,4-D and the percent of NHANES participants with high 2,4-D urinary concentrations (adjusted odds ratio 2.27 (95% CI: 1.71, 3.01, p < .0001)). This study demonstrates that as average annual use of 2,4-D increased, individuals had increased odds of having higher urinary 2,4-D concentrations.
Overall, the amount of 2,4-D applied in agriculture has risen nearly 67% between 2012 and 2020, and over 240% between 1991 and 2020 (see Additional file 1), a trend that is unlikely to change due to unabated weed resistance. 2,4-D use began in 1945 and dramatically increased over the next two decades. In the late 1960’s, reliance on 2,4-D declined due to the emergence of other selective herbicides such as atrazine, although it remained the herbicide of choice for most wheat growers [2, 10, 11]. Commercial sale of genetically engineered glyphosate-tolerant “Roundup™ Ready” soybeans and cotton began in 1996. Cost-effectiveness and simplicity of use led to rapid and widespread adoption of the Roundup™ system (see Additional files 3 and 4), causing sales of 2,4-D and all other herbicides to decline sharply in these three crops (see Additional files 5 and 6). By the mid 2000s, three hard-to-control weeds had developed resistance to glyphosate in the southeastern US and were spreading rapidly north and west from farm to farm and across state lines. The problem worsened from a few resistant weeds in scattered fields to economically damaging resistant weed populations in many fields by the mid 2010s [25, 26]. The growing diversity and spread of resistant weeds have forced farmers to add additional herbicides, often including 2,4-D, to their weed control programs .
In response to increasing herbicide resistance in weeds, the pesticide industry developed DuoEnlist™ technology, which EPA approved in 2017. This technology couples seeds genetically engineered to tolerate 2,4-D, glyphosate, and 4-Hydroxyphenylpyruvate dioxygenase inhibitors (HPPD inhibitors, aka “fop” herbicides) with new formulations of glyphosate and 2,4-D designed to reduce volatility, drift, and off-target crop damage. Reliance on the DuoEnlist™ system has caused the use of 2,4-D to rise sharply in both soybean and cotton production, with further increases highly likely through around 2025 (see Additional files 7 and 8).
The expected trend of increased use of 2,4-D raises concerns about changes in population exposure, particularly for sensitive populations who may be more vulnerable to harmful effects of exposure. The U.S. EPA provides a “chronic reference dose” (chronic RfD) for certain pesticide chemical exposures. This is defined as “an estimate (with uncertainty spanning perhaps an order of magnitude) of a daily exposure to the human population (including sensitive subgroups) that is likely to be without an appreciable risk of deleterious effects during a lifetime” . For 2,4-D, the chronic RfD based on oral consumption is 0.005 mg/kg/day . A biomonitoring equivalent (BE) provides information about how to compare urinary biomonitoring data to the RfD . Toxicokinetic research indicates that the RfD of 0.005 mg/kg/day of 2,4-D is roughly equivalent to a creatinine adjusted urinary concentration of 300 μg/g in the general population . Prior CDC evaluation of NHANES data indicates urinary concentrations of 2,4-D in the general population continued to be orders of magnitude below the BE of the RfD through 2014 . This is the latest survey in which NHANES includes 2,4-D data, and other biomonitoring studies have confirmed these results .
The results of this study deserve careful consideration in the context of herbicide exposure given all that is already known about the human health effects of herbicides from animal and epidemiologic studies. While the carcinogenicity of 2,4-D has been intensively studied and long debated [31, 32], new studies have heightened concern. A 2020 longitudinal biomarker study linked 2,4-D with increased systemic markers of oxidative stress . Association between 2,4-D exposures and non-Hodgkin lymphoma (NHL) have been reported, with a recent meta-analysis showing that highly exposed groups experience an elevated relative risk of NHL (RR = 1.73, CI: 1.10–2.72) . The risk of pediatric leukemia is increased in children residing near areas sprayed heavily with herbicides, including 2,4-D and dicamba , raising concerns about the impact of exposure to this herbicide on pediatric populations.
Non-cancer outcomes such as birth defects and pediatric anatomical abnormalities have also been linked to 2,4-D. A 1996 study linked licensed pesticide applicators to birth defects from the state birth registry and found an increased rate of birth defects in children of applicators who applied chlorophenoxy herbicides such as 2,4-D . More recently, a case control study assessing birth defects in infants found an association between 2,4-D exposure and hypertrophic pyloric stenosis, patent ductus arteriosus, and hypospadias in male infants . In adults, health outcomes as diverse as allergic wheeze , hypothyroidism [39, 40] and olfactory deficits  have all been linked to 2,4-D exposure.
In this analysis, several vulnerable population subgroups demonstrated an increased odds of higher urinary levels of 2,4-D associated with increased magnitude of 2,4-D use in agriculture. These subgroups include the youngest age group evaluated in NHANES surveys (children aged 6–11), as well as the oldest age group (adults aged 60+). Likewise, women of reproductive age also demonstrated increased odds of high exposure with rising agricultural use of 2,4-D.
Children may be at risk of higher exposure levels due to children’s play behaviors that include more time outdoors and on the floor, where there are higher quantities of dirt or dust particles with herbicide residue [41,42,43]. NHANES did not evaluate children under the age of 6 during the survey cycles included in this study. While the results here indicate a trend toward increasing exposure as age decreases, the data may not generalize to children under age 6 because of differences in types and amounts of food eaten, the ratio between amount of food eaten and unit of body weight, and the amount of accidental herbicide ingestion through play and mouthing behaviors . This gap in knowledge is of critical concern because current knowledge of 2,4-D exposure dynamics  suggest that these differences may place younger children at even greater risk for higher exposure than their older counterparts. Further, young children are generally at higher risk for adverse health and developmental outcomes due to physiologic and developmental differences from adults . Biomonitoring studies of 2,4-D exposure in pregnant women and young children are needed because of their unique vulnerabilities to both exposure and adverse health outcomes.
Only a small number of participants classified as agricultural workers were included in this study (n = 139). As expected, they were significantly more likely to have higher urinary 2,4-D concentrations compared to participants not classified as agricultural workers. This correlates with previous research on exposure to 2,4-D and other pesticides in agricultural workers [44, 45].
Non-Hispanic White participants had increased odds of higher exposure in both unadjusted and adjusted models compared to participants of other races/ethnicities. Because the models controlled for other variables related to socioeconomic status (SES), the differences probably cannot be explained entirely by factors related to SES. Although Non-Hispanic White participants had the highest odds of exposure compared to all other racial/ethnic groups, the difference was greatest between Non-Hispanic White and Non-Hispanic Black participants. It is possible that racial differences in geographic distribution, in particular legacy effects of racism resulting from formal and informal real estate redlining and segregation [46,47,48] may play a role in the differences in exposure between Black and White survey participants.
Race was included in this analysis recognizing it may serve as a proxy variable for geographic location and proximity to agricultural land and non-agricultural managed greenspace, including residential lawns [49,50,51]. In 2003, 2,4-D was applied at higher rates to American lawns and greenspaces than any other household herbicide on market . The increase of southern Black migrants into northern cities through the 1960’s and 70’s drove many White families into suburban areas, where each home was often accompanied by a lawn in the front or back, and sometimes both . Even in cities, urban greenspace is likely to be disproportionately distributed to White residents . While these forms of structural racism often explain higher environmental exposure burdens among Black Americans [55,56,57], in this case, because non-agricultural uses relate primarily to lawns and green space, they may explain the apparent lower exposure burden among Black and other minority NHANES participants compared to Whites.
There was not a significant difference in biomarker levels by time of year of testing. It is expected that exposures are likely higher during the spray season for some people because of increased inhalation and dermal exposures due to proximity to land in agricultural use. However, NHANES reports testing season data in 6-month intervals that do not directly align with herbicide spraying season, which may explain why differences by time of year of testing were not statistically significant. The lack of control for geographic location in the analysis may have attenuated actual differences, or exposure may occur primarily through sources not dependent on proximity to agricultural land, such as through dietary sources or nearby lawn care applications.
There are some important limitations of this study, including that it did not evaluate health endpoints. The purpose of this study focused on determining whether changes in agricultural uses of 2,4-D are affecting human exposures. Variable limits of detection of 2,4-D biomarkers across survey cycles and a high percentage of biomarker values below the LOD affect the quantification of population exposure; it was not possible to evaluate associations with the mean population exposure level or its changes over time. Because 2,4-D has an expected half-life in humans of 10.2–28.5 h and is nearly completely cleared within 3 days , one-time spot urine samples provide information on a window of exposure immediately prior to sample collection but reflect neither the expected high variability in exposure levels from individual to individual, nor seasonal changes.
With these recognized limitations, studies tracking associations between agricultural pesticide application and human pesticide exposures are scant. As far as we are aware, this is the first study to evaluate 2,4-D biomonitoring levels and agricultural use of 2,4-D in a large nationally representative survey. In 2020, agricultural use of 2,4-D reached 33.3 million pounds nationally, reflecting a nearly 200% increase over the 2002 level (see Additional file 1). This rate of growth in the last two decades, however, will likely be dwarfed by the rate of growth and absolute annual increases in total pounds of 2,4-D applied in the next decade. Particularly sharp increases are expected in the next 3–5 years on soybean and cotton crop acres as the supply of 2,4-D-tolerant DuoEnlist™ seeds expands.
2,4-D serves as a sentinel for anticipated changes in other herbicide exposure levels, the application of which are changing in concert with changes in 2,4-D use (see Additional Files 5 and 6). Many of these herbicides have never been included in NHANES or other national biomonitoring studies. Continued monitoring of urinary 2,4-D levels by NHANES is strongly recommended, along with other herbicides that are increasing in use and potentially in exposure (e.g., glufosinate, dicamba, and 4-hydroxyphenylpyruvate dioxygenase (HPPD) inhibitors), as these will be used in progressively higher amounts over the coming years [58,59,60]. If these application trends unfold as predicted, the analyses reported here signal higher exposures to humans in the near future.