Urinary biomonitoring of glyphosate exposure among male farmers and nonfarmers in the Biomarkers of Exposure and Effect in Agriculture (BEEA) study

Abstract

Glyphosate is the most widely applied herbicide worldwide. Glyphosate biomonitoring data are limited for agricultural settings. We measured urinary glyphosate concentrations and assessed exposure determinants in the Biomarkers of Exposure and Effect in Agriculture (BEEA) study. We selected four groups of BEEA participants based on self-reported pesticide exposure: recently exposed farmers with occupational glyphosate use in the last 7 days (n = 98), farmers with high lifetime glyphosate use (>80th percentile) but no use in the last 7 days (n = 70), farming controls with minimal lifetime use (n = 100), and nonfarming controls with no occupational pesticide exposures and no recent home/garden glyphosate use (n = 100). Glyphosate was quantified in first morning void urine using ion chromatography isotope-dilution tandem mass spectrometry. We estimated associations between urinary glyphosate concentrations and potential determinants using multivariable linear regression. Glyphosate was detected (≥0.2 μg/L) in urine of most farmers with recent (91 %) and high lifetime (93 %) use, as well as farming (88 %) and nonfarming (81 %) controls; geometric mean concentrations were 0.89, 0.59, 0.46, and 0.39 μg/L (0.79, 0.51, 0.42, and 0.37 μg/g creatinine), respectively. Compared with both control groups, urinary glyphosate concentrations were significantly elevated among recently exposed farmers (P < 0.0001), particularly those who used glyphosate in the previous day [vs. nonfarming controls; geometric mean ratio (GMR) = 5.46; 95 % confidence interval (CI): 3.75, 7.93]. Concentrations among high lifetime exposed farmers were also elevated (P < 0.01 vs. nonfarming controls). Among recently exposed farmers, glyphosate concentrations were higher among those not wearing gloves when applying glyphosate (GMR = 1.91; 95 % CI: 1.17, 3.11), not wearing long-sleeved shirts when mixing/loading glyphosate (GMR = 2.00; 95 % CI: 1.04, 3.86), applying glyphosate exclusively using broadcast/boom sprayers (vs. hand sprayer only; GMR = 1.70; 95 % CI: 1.00, 2.92), and applying glyphosate to crops (vs. non-crop; GMR = 1.72; 95 % CI: 1.04, 2.84). Both farmers and nonfarmers are exposed to glyphosate, with recency of occupational glyphosate use being the strongest determinant of urinary glyphosate concentrations. Continued biomonitoring of glyphosate in various settings is warranted.

1. Introduction

Glyphosate, the active ingredient of a broad-spectrum herbicide formulation commercialized in 1974, is widely used to control weeds in agricultural and other settings (Benbrook, 2016). Following the introduction of genetically engineered glyphosate-resistant crops in 1996, glyphosate use rose sharply, and it has since become the most heavily applied agricultural pesticide in the United States and globally (Benbrook, 2016). Glyphosate is also the second most commonly used pesticide in the U.S. home and garden sector (Atwood and Paisley-Jones, 2017). Given its broad application, glyphosate and its potential health effects, including cancer, have received increasing public health and scientific attention (Myers et al., 2016); however, evidence regarding glyphosate toxicity and carcinogenicity remains inconclusive and controversial (Mesnage and Antoniou, 2017; Kogevinas, 2019). Notably, the International Agency for Research on Cancer classified glyphosate as a probable (group 2A) human carcinogen in 2015, based in part on limited epidemiologic evidence of an association with non-Hodgkin lymphoma (Guyton et al., 2015), whereas the U.S. Environmental Protection Agency and several other regulatory agencies concluded that glyphosate is unlikely to be carcinogenic to humans (U.S. Environmental Protection Agency (EPA), 2017; Portier et al., 2016).

Glyphosate can enter the human body through dermal contact, inhalation, and oral ingestion of contaminated foods and water (Agency for Toxic Substances and Disease Registry (ATSDR), 2020). Among occupationally exposed individuals directly involved in handling pesticides, including agricultural workers, research suggests that dermal absorption is the predominant route of exposure (Agency for Toxic Substances and Disease Registry (ATSDR), 2020; Connolly et al., 2019). Although there is limited knowledge regarding the toxicokinetics of glyphosate in humans, current data suggest that glyphosate does not undergo significant metabolism upon absorption and is mainly excreted as the unchanged parent compound in urine (Connolly et al., 2019), with an elimination half-life of approximately 5–10 h (Connolly et al., 2019; Zoller et al., 2020; Kohsuwan et al., 2022). As such, quantification of glyphosate concentrations in urine represents a noninvasive method of assessing the internal dose of recent exposure to glyphosate.

Despite widespread use of glyphosate and ongoing debate surrounding its potential impacts on human health, there is a relative paucity of biomonitoring data on glyphosate exposure in both occupational and non-occupational settings (Gillezeau et al., 2019; Gillezeau et al., 2020; Connolly et al., 2020; Connolly and Koch, 2023; Muñoz et al., 2023; Ospina et al., 2022). This may be partly due to analytical challenges associated with laboratory measurements of glyphosate, a highly polar compound with metal-chelating properties (Valle et al., 2019). In a review that was further updated in 2020, Gillezeau et al. (Gillezeau et al., 2019; Gillezeau et al., 2020) identified a total of 24 studies measuring glyphosate concentrations in human biofluids (mainly urine), of which only 11 studies focused on occupational exposures, including eight assessing urinary glyphosate among farmers (Acquavella et al., 2004; Curwin et al., 2007; Mesnage et al., 2012; Jayasumana et al., 2015; Rendón-von Osten and Dzul-Caamal, 2017; Wongta et al., 2018; Perry et al., 2019; Balderrama-Carmona et al., 2020). However, many of the prior investigations among farmers, including additional studies published after the review (Bootsikeaw et al., 2021; Connolly et al., 2022; Campbell et al., 2022; Mueller et al., 2024), were limited by the small number of participants (mostly n < 50 farmers), unavailability of data on glyphosate usage and other factors potentially influencing exposure, lack of adjustment for urinary dilution, and/or absence of a nonfarming comparison group. Furthermore, different methods were employed to quantify urinary glyphosate across studies, with variable detection limits and generally low frequencies of detection, especially among individuals who may not have been exposed within a short time period prior to urine sampling (Gillezeau et al., 2019; Gillezeau et al., 2020; Connolly et al., 2020).

In 2021, the Centers for Disease Control and Prevention (CDC) published an ion chromatography–mass spectrometry approach suitable for quantifying urinary glyphosate concentrations across various settings, including investigations of background exposures (Schütze et al., 2021). Applying this method to urine samples from 2,310 participants in the 2013–2014 National Health and Nutrition Examination Survey (NHANES), urinary glyphosate concentrations were assessed at the national level for the first time in the United States, with 81 % of the population ≥ 6 years of age found to have detectable concentrations (Ospina et al., 2022). However, no information on occupational use of glyphosate was available in this general population study.

Utilizing the recently developed CDC approach to quantify urinary glyphosate, we investigated the extent of glyphosate exposure and its potential determinants in a well-characterized population of male farmers, as well as a group of demographically similar nonfarmers, in the Biomarkers of Exposure and Effect in Agriculture (BEEA) study. Specifically, we evaluated reported use of glyphosate, including recency of use and application practices, in relation to urinary glyphosate concentrations.

2. Materials and methods

2.1. Study design and population

The BEEA study is a molecular epidemiologic subcohort of the Agricultural Health Study (AHS), a large prospective cohort of pesticide applicators in Iowa and North Carolina (Alavanja et al., 1996). Details of the BEEA study design and data collection procedures have been described elsewhere (Hofmann et al., 2015). Briefly, male farmers in the AHS were eligible for participating in BEEA if they were ≥ 50 years of age, still resided in Iowa or North Carolina, completed the AHS enrollment (1993–1997) and follow-up [Phase 2 (1999–2003) and Phase 3 (2005–2010)] questionnaires, and had never been diagnosed with cancer other than nonmelanoma skin cancer. A total of 1,681 eligible farmers from the AHS were enrolled between 2010 and 2017. As a nonfarming comparison group, BEEA additionally enrolled 211 males from Iowa and North Carolina who were ≥ 50 years of age, had no history of cancer, and had not lived or worked on a farm or held a job that involved mixing, loading, or applying pesticides in the last 10 years or for > 12 months since 18 years of age. The nonfarmers were identified using voter registration lists and recruited from 20 counties (10 from each state) located outside major metropolitan areas from which many of the participating BEEA farmers were recruited (Chang et al., 2023). The BEEA study protocol was approved by institutional review boards at the National Cancer Institute and other participating institutions. All participants provided written informed consent. The involvement of the CDC laboratory did not constitute engagement in human subjects research.

2.2. Assessment of glyphosate use and selection of study groups

Occupational use of glyphosate was assessed based on self-report from the AHS and BEEA questionnaires. At AHS enrollment, farmers reported whether they had ever personally mixed or applied specific pesticides, including glyphosate, and if so, the frequency and duration of use and application practices; this information was further updated from follow-up interviews in 1999–2003 and 2005–2010. At the BEEA enrollment interview (2010–2018) that took place during the home visit for biospecimen collection, farmers provided information on specific pesticides used on crops or animals (or for any other farming-related activities) in the last 12 months, including details on the most recent dates of use, frequency of use, application method and type, and whether they normally used gloves or other personal protective equipment (PPE) while mixing/loading or applying each pesticide. The BEEA questionnaire also collected information on home and garden pesticide use in the last 12 months from both farmers and nonfarmers. To estimate cumulative lifetime occupational use of glyphosate among farmers, two metrics were created by combining responses across questionnaires: a) lifetime days of use (number of days of use per year × number of years used); and b) intensity-weighted lifetime days of use, calculated by multiplying lifetime days of use by an exposure intensity score that accounts for factors known to influence exposure, such as mixing/loading of pesticides, application method, and PPE use (Coble et al., 2011).

In the current investigation, we selected four groups of participants (total n = 368) with respect to their use of glyphosate (Chang et al., 2023). First, a group of recently exposed farmers (n = 98) included those with occupational glyphosate use within 7 days before urine collection. Second, high lifetime exposed farmers (n = 70) included those who had no occupational glyphosate use in the last 7 days but whose lifetime use was in the top quintile among all BEEA farmers for both lifetime days (≥197) and intensity-weighted lifetime days (≥11,334). Third, as a minimally exposed farming control group (n = 100), we selected farmers who had never used glyphosate occupationally during their lifetime or had not used glyphosate since after the first AHS follow-up interview (1999–2003) and were in the lowest tertiles of both lifetime days (<14.5) and intensity-weighted lifetime days (<677) of use among all BEEA farmers. Finally, a nonfarming control group (n = 100) was selected from nonfarmers who reported no home or garden use of glyphosate during the 7 days prior to urine collection. Participants from each of the two control groups (farming and nonfarming controls) were selected such that they were frequency matched to the exposed groups (recently exposed and high lifetime exposed combined) on age (50–60, 61–70, and > 70 years), state of residence (Iowa and North Carolina), and season of interview/urine collection (April–September and October–March).

2.3. Urinary glyphosate measurements

Each participant provided a first morning void urine sample on the day of the BEEA enrollment interview (2010–2018). Participants were asked to store the samples in the refrigerator before the arrival of the study staff. Samples were shipped on ice to a central processing laboratory, where they were aliquoted and transferred for long-term storage at −80°C until shipment for analysis.

Urinary glyphosate concentrations were quantified at the CDC’s National Center for Environmental Health (Atlanta, GA, USA) using an ion chromatography isotope dilution tandem mass spectrometry method described previously (Ospina et al., 2022; Schütze et al., 2021). Briefly, an internal standard of isotope-labeled (¹⁵N, 2–¹³C) glyphosate (Cambridge Isotope Laboratories, Andover, MA, USA) and ultrapure water were added to each urine sample. Glyphosate was extracted and separated from other matrix components using a Dionex ICS-5000+ ion chromatography system (Thermo Fisher Scientific Inc., Sunnyvale, CA, USA) equipped with a dual-column and switching valve assembly. Quantitation by isotope dilution-tandem mass spectrometry was performed with an AB Sciex 5500 triple quadrupole mass spectrometer (Applied Biosystems, Foster City, CA, USA) equipped with a TurboIonSpray^® source. Samples from each of the four study groups were distributed evenly across all batches. The limit of detection (LOD) was 0.2 μg/L. Consistent with the NHANES analyses (Ospina et al., 2022; National Center for Environmental Health, 2022), concentrations below the LOD [n = 45 (12 %) of all 368 samples] were assigned a value of LOD/√2, which has been recommended for lognormally distributed data that are not highly skewed (geometric standard deviation < 3) (Hornung and Reed, 1990). To assess measurement reproducibility, we included 68 blinded quality control (QC) samples from 34 participants (selected from all four study groups), with duplicate urine samples from each participant distributed either within or across batches. The within-batch, between-batch, and overall coefficients of variation were 2.9 %, 5.3 %, and 6.0 %, respectively, and the intraclass correlation coefficient was 0.997. We also analyzed pooled QC samples across batches, and no evidence of laboratory drift was observed. In addition, to adjust glyphosate concentrations for urinary dilution, urinary creatinine was measured using an enzymatic method at the University of Minnesota Advanced Research and Diagnostic Laboratory (Minneapolis, MN, USA).

2.4. Statistical analysis

We summarized participant characteristics among each of the four study groups using means and standard deviations for continuous variables and frequencies and proportions for categorical variables. Differences across study groups were assessed using analysis of variance for continuous variables and chi-squared or Fisher’s exact tests for categorical variables. Urinary glyphosate concentrations were presented as both unadjusted (μg/L) and creatinine-corrected (μg/g creatinine) concentrations, summarized using percentiles and geometric means (GMs) with corresponding 95 % confidence intervals (CIs), and visualized using bar graphs and box-and-whisker plots. Differences in GM glyphosate concentrations between study groups were assessed by pairwise comparisons of means of natural log (ln)-transformed concentrations using linear regression. Within each study group, we further computed GM glyphosate concentrations across categories of participant characteristics [age, state, season of urine collection, time of urine collection, body mass index (BMI), and home/garden glyphosate use in the last 12 months] and used analysis of variance to assess differences. Among nonfarming controls with home/garden glyphosate use, we also examined glyphosate concentrations according to number of days since last use.

Our main analyses evaluated urinary glyphosate concentrations in relation to recent (last 7 days) occupational glyphosate use and related characteristics, including days since last use (≤1, 2–4, 5–7 days before urine collection), use of gloves and/or other PPE (e.g., long-sleeved shirt, goggles, rubber boots, respirators) while mixing/loading or applying glyphosate, application method (hand spray only, broadcast/boom spray only, both), and application type [non-crop only (e.g., around buildings, driveways, and fence rows), crops only, both]. For glove and/or other PPE use, we created a composite variable (gloves only, other PPE only, both, neither); we also separately examined use (yes vs. no) of gloves and other specific PPE worn by ≥ 5 farmers. We performed linear regression analyses with ln-transformed urinary glyphosate concentrations (unadjusted for creatinine) as the dependent variable and each glyphosate use-related characteristic as the independent variable. Comparisons were made separately with either the farming or nonfarming controls as the referent category [e.g., farming or nonfarming controls (referent), recently exposed farmers who used gloves, recently exposed farmers who did not use gloves] such that when one of the control groups was designated as the referent, the other control group was excluded from the analysis; the high lifetime exposed group (i.e., no occupational use in the last 7 days) was not included in these analyses. We reported associations as geometric mean ratios (GMRs), which were calculated by exponentiating the regression coefficients and interpreted as the ratio of the GM urinary glyphosate concentrations comparing subsets of recently exposed farmers (e.g., with or without glove use) to participants in the respective control groups. We also performed analyses restricted to the recently exposed group [e.g., glove use (referent), no glove use]. All models adjusted for age (continuous, years), age squared, urinary creatinine concentration (ln-transformed; continuous), state (Iowa, North Carolina), season of urine collection (April–September, October–March), time of urine collection (before 4:00 a.m., 4:00–5:59 a.m., 6:00 a.m. or later), and BMI (continuous, kg/m²). For analyses restricted to the recently exposed group, we also included a model that additionally adjusted for days since last use of glyphosate, the strongest predictor of urinary glyphosate concentrations.

To examine exposure associated with longer-term occupational glyphosate use, we also evaluated past-year use (tertiles of number of days or intensity-weighted days in the last 12 months) and cumulative lifetime use (tertiles of lifetime days or intensity-weighted lifetime days) in relation to urinary glyphosate concentrations among the recently exposed and high lifetime exposed groups (separate or combined) using multivariable linear regression as described above.

All statistical analyses were performed using SAS, version 9.4 (SAS Institute Inc., Cary, NC, USA). Tests were two-sided, with statistical significance evaluated at P < 0.05.

3. Results

Selected demographic and other participant characteristics by study group are shown in Table 1. The vast majority of participants were White (98 %), residents of Iowa (76 %), and with overweight or obesity (83 %). Distributions of most characteristics were similar across the four study groups, except, as expected, recently exposed farmers were more likely to have their urine collected during April–September (i.e., in the growing season) than other groups. Approximately 40 % of all participants used glyphosate in the home or garden in the last 12 months, with no differences observed by study group (P = 0.83).

Table 1.

Selected characteristics of glyphosate-exposed and control groups in the Biomarkers of Exposure and Effect in Agriculture study, 2010–2018.

Characteristica	Recently exposed (n = 98)	High lifetime exposed (n = 70)	Farming controls (n = 100)	Nonfarming controls (n = 100)	P-valueb
Age (years)
Mean ± SD	63.2 ± 8.3	62.6 ± 9.4	64.4 ± 10.1	63.5 ± 8.7	0.62
50–59	35 (36)	31 (44)	34 (34)	40 (40)	0.85
60–69	37 (38)	24 (34)	39 (39)	33 (33)
≥ 70	26 (27)	15 (21)	27 (27)	27 (27)
State
Iowa	77 (79)	49 (70)	76 (76)	76 (76)	0.64
North Carolina	21 (21)	21 (30)	24 (24)	24 (24)
Race
White	97 (99)	70 (100)	96 (96)	98 (98)	0.37
Otherc	1 (1)	0 (0)	4 (4)	2 (2)
Season of urine collection
April–September	92 (94)	29 (41)	71 (71)	63 (63)	< 0.0001
October–March	6 (6)	41 (59)	29 (29)	37 (37)
Time of urine collection
Before 4:00 a.m.	22 (22)	8 (11)	14 (14)	16 (16)	0.61
4:00–5:59 a.m.	33 (34)	29 (41)	40 (40)	38 (38)
6:00 a.m. or later	43 (44)	33 (47)	46 (46)	46 (46)
Body mass index (kg/m2)
Mean ± SD	29.2 ± 5.3	28.6 ± 4.4	30.1 ± 5.9	30.0 ± 6.1	0.28
< 25	16 (16)	11 (16)	20 (20)	15 (15)	0.31
25– < 30	46 (47)	38 (54)	35 (35)	44 (44)
≥ 30	36 (37)	21 (30)	45 (45)	41 (41)
Home/garden glyphosate used
No	61 (62)	42 (60)	58 (58)	56 (56)	0.83
Yes	37 (38)	28 (40)	42 (42)	44 (44)

Abbreviation: SD, standard deviation.

^aPresented as frequencies and percentages [n (%)] unless otherwise specified.

^bP-value for differences across the four groups, calculated using analysis of variance for continuous variables and chi-squared or Fisher’s exact (for race) tests for categorical variables.

^cBlack (n = 3), American Indian or Alaska Native (n = 1), or not reported (n = 3).

^dIn the last 12 months.

Glyphosate was detected (LOD = 0.2 μg/L) in the urine of > 80 % of participants across all study groups, with detection frequencies ranging from 81 % among nonfarming controls to 93 % among high lifetime exposed farmers (Table 2). We observed a clear pattern of differences in urinary glyphosate concentrations across the four study groups defined by glyphosate use, with GMs of 0.89 μg/L (0.79 μg/g creatinine), 0.59 μg/L (0.51 μg/g creatinine), 0.46 μg/L (0.42 μg/g creatinine), and 0.39 μg/L (0.37 μg/g creatinine) among recently exposed farmers, high lifetime exposed farmers, farming controls, and nonfarming controls, respectively (Table 2, Supplementary Figure S1, and Fig. 1). In pairwise comparisons, we observed statistically significantly elevated GM concentrations in recently exposed farmers relative to each of the other three groups (for creatinine-corrected concentrations: P < 0.0001 vs. each control group and P = 0.0002 vs. high lifetime exposed farmers), and in high lifetime exposed farmers relative to nonfarming controls (P = 0.009). The upper percentile concentrations were also considerably higher for recently exposed farmers (e.g., 75th, 90th, and 95th percentiles: 1.52, 3.05, and 7.72 μg/g creatinine, respectively, compared with 0.55, 0.92, and 1.08 μg/g creatinine for nonfarming controls; Table 2 and Supplementary Figures S2 and S3).

Table 2.

Selected percentiles and geometric means of urinary glyphosate concentrations (unadjusted and creatinine-corrected) among glyphosate-exposed and control groups in the Biomarkers of Exposure and Effect in Agriculture study, 2010–2018.

Study group	n	Detection frequency (%)	Urinary glyphosate concentration
			Selected percentiles						Maximum	Geometric mean (95 % CI)
			10th	25th	50th	75th	90th	95th
Recently exposed
μg/L	98	90.8	0.24	0.40	0.77	1.44	3.84	15.2	35.2	0.89 (0.70, 1.13)
μg/g creatinine	98		0.25	0.39	0.72	1.52	3.05	7.72	20.0	0.79 (0.64, 0.98)
High lifetime exposed
μg/L	70	92.9	0.21	0.29	0.55	1.12	1.72	2.89	6.01	0.59 (0.48, 0.73)
μg/g creatinine	70		0.20	0.30	0.47	0.79	1.41	1.64	2.67	0.51 (0.43, 0.60)
Farming controls
μg/L	100	88.0	<LOD	0.27	0.49	0.76	1.09	1.31	1.81	0.46 (0.40, 0.52)
μg/g creatinine	100		—a	0.28	0.41	0.65	0.88	1.10	2.17	0.42 (0.37, 0.47)
Nonfarming controls
μg/L	100	81.0	<LOD	0.24	0.38	0.62	1.02	1.56	2.69	0.39 (0.34, 0.46)
μg/g creatinine	100		—a	0.24	0.37	0.55	0.92	1.08	1.74	0.37 (0.33, 0.42)

Abbreviations: CI, confidence interval; LOD, limit of detection (0.2 μg/L).

^aNot calculated because the uncorrected concentration is < LOD.

Fig. 1.

Geometric mean concentrations of creatinine-corrected urinary glyphosate (μg/g creatinine) among control and glyphosate-exposed study groups in the Biomarkers of Exposure and Effect in Agriculture study, 2010–2018. Error bars represent 95% confidence intervals. P-values were obtained from pairwise comparisons of means of natural log-transformed creatinine-corrected urinary glyphosate concentrations using linear regression.

Urinary glyphosate concentrations across participant characteristics are presented by study group in Supplementary Table S1. In general, glyphosate concentrations increased slightly with age, especially among recently exposed farmers and nonfarming controls, and levelled off at older ages for both unadjusted and creatinine-corrected concentrations among recently exposed farmers. We also observed statistically significant differences according to time of urine collection among recently exposed farmers, with higher concentrations for samples collected at 6 a.m. or later compared to those collected before 6 a.m., although the pattern was not as clear for creatinine-corrected concentrations. No notable differences were observed across other characteristics, including home/garden glyphosate use in the last 12 months. However, among a subset of nonfarmers with information on days since last home/garden use of glyphosate (n = 20), we observed marginally elevated concentrations in those with more recent use [GM (95 % CI) = 0.61 (0.34, 1.11) μg/L for use in the last 30 days vs. 0.36 (0.23, 0.57) μg/L for use > 30 days ago, P = 0.15]; the difference was more pronounced after creatinine correction [GM = 0.54 (0.38, 0.77) vs. 0.29 (0.21, 0.42) μg/g creatinine, P = 0.04] (Supplementary Table S2).

Table 3 presents urinary glyphosate concentrations in relation to recent (last 7 days) occupational glyphosate use and related characteristics, including PPE use and application method and type. We observed statistically significantly elevated glyphosate concentrations among farmers who reported use of glyphosate on the day before urine collection (GMR = 5.46; 95 % CI: 3.75, 7.93) and those with use 2–4 days before urine collection (GMR = 1.86; 95 % CI: 1.38, 2.51) compared to nonfarming controls. Similar associations were observed relative to farming controls. We also found a strong association among recently exposed farmers (≤1 vs. 5–7 days since last use of glyphosate; GMR = 3.94; 95 % CI: 2.40, 6.46; P_trend < 0.0001). The majority (55 %) of recently exposed farmers wore gloves and/or other PPE when mixing or loading glyphosate, whereas only 31 % reported glove/PPE use when applying glyphosate. Urinary glyphosate concentrations were highest among farmers who did not use any PPE and those who only used gloves (with no other PPE) while mixing/loading glyphosate, as well as among farmers who did not use any PPE while applying glyphosate, although comparisons within the recently exposed group were not statistically significant. When we examined specific PPE used within the recently exposed farmers, elevated glyphosate concentrations were observed among those who did not wear gloves while applying glyphosate (GMR = 1.91; 95 % CI: 1.17, 3.11) and those who did not wear a long-sleeved shirt while mixing/loading glyphosate (GMR = 2.00; 95 % CI: 1.04, 3.86); however, these associations were attenuated after additional adjustment for number of days since last use of glyphosate [GMRs (95 % CIs) = 1.49 (0.95, 2.34) and 1.61 (0.90, 2.88), respectively]. Furthermore, farmers who used broadcast/boom sprayers as the only glyphosate application method (31 %) had higher glyphosate concentrations than those who used hand sprayers (alone or with broadcast/boom sprayers), with an association that persisted after adjusting for days since last use (vs. hand spray only; GMR = 2.06; 95 % CI: 1.29, 3.27). Similarly, farmers who applied glyphosate to crops only (31 %) had higher glyphosate concentrations than those who used glyphosate for non-crop applications only (GMR adjusting for days since last use = 1.83; 95 % CI: 1.18, 2.82).

Table 3.

Associations of recent occupational glyphosate use (last 7 days) and related characteristics with urinary glyphosate concentrations in the Biomarkers of Exposure and Effect in Agriculture study, 2010–2018.

Characteristic	n	Geometric mean concentration		Geometric mean ratio (95 % CI)
		μg/L (95 % CI)	μg/g creatinine (95 % CI)	Model 1a				Model 2b
				Compared to nonfarming controls	Compared to farming controls	Recently exposed only		Recently exposed only
Nonfarming controls	100	0.39 (0.34, 0.46)	0.37 (0.33, 0.42)	1.00 (Ref)	—	—		—
Farming controls	100	0.46 (0.40, 0.52)	0.42 (0.37, 0.47)	—	1.00 (Ref)	—		—
Recently exposed	98	0.89 (0.70, 1.13)	0.79 (0.64, 0.98)	1.97 (1.53, 2.54)	1.82 (1.41, 2.33)	—		—
Number of days since last use
5–7 days	42	0.52 (0.41, 0.66)	0.52 (0.42, 0.66)	1.30 (0.98, 1.72)	1.18 (0.89, 1.56)	1.00 (Ref)		1.00 (Ref)
2–4 days	35	0.81 (0.57, 1.13)	0.71 (0.51, 0.98)	1.86 (1.38, 2.51)	1.70 (1.26, 2.28)	1.42 (0.95, 2.11)		1.42 (0.95, 2.11)
≤ 1 day	21	2.99 (1.61, 5.56)	2.20 (1.32, 3.69)	5.46 (3.75, 7.93)	5.29 (3.65, 7.66)	3.94 (2.40, 6.46)		3.94 (2.40, 6.46)
Glove and other PPEc use while mixing/loadingd
Both gloves and other PPE	17	0.49 (0.30, 0.81)	0.41 (0.26, 0.64)	1.25 (0.80, 1.95)	1.06 (0.68, 1.64)	1.00 (Ref)		1.00 (Ref)
Gloves only	31	1.09 (0.76, 1.57)	0.94 (0.67, 1.31)	2.28 (1.61, 3.23)	2.19 (1.54, 3.10)	1.69 (0.89, 3.21)		1.70 (0.98, 2.95)
Other PPE only	4	0.60 (0.12, 2.99)	0.76 (0.25, 2.31)	1.55 (0.67, 3.61)	1.49 (0.64, 3.50)	1.19 (0.36, 3.95)		0.97 (0.34, 2.75)
Neither gloves nor other PPE	43	0.99 (0.65, 1.52)	0.93 (0.65, 1.32)	2.23 (1.63, 3.05)	2.07 (1.52, 2.82)	1.66 (0.89, 3.08)		1.17 (0.68, 2.03)
Glove use while mixing/loadingd
Yes	48	0.82 (0.61, 1.11)	0.70 (0.53, 0.93)	1.84 (1.36, 2.50)	1.68 (1.24, 2.28)	1.00 (Ref)		1.00 (Ref)
No	47	0.95 (0.64, 1.41)	0.91 (0.66, 1.26)	2.15 (1.58, 2.92)	1.99 (1.46, 2.71)	1.14 (0.73, 1.75)		0.81 (0.54, 1.20)
Wearing long-sleeved shirt while mixing/loadingd
Yes	11	0.57 (0.34, 0.95)	0.44 (0.27, 0.71)	1.15 (0.68, 1.95)	1.00 (0.59, 1.71)	1.00 (Ref)		1.00 (Ref)
No	84	0.94 (0.72, 1.22)	0.86 (0.68, 1.09)	2.14 (1.65, 2.76)	1.98 (1.53, 2.56)	2.00 (1.04, 3.86)		1.61 (0.90, 2.88)
Wearing goggles while mixing/loadingd
Yes	11	0.38 (0.19, 0.75)	0.41 (0.23, 0.73)	1.30 (0.76, 2.25)	1.09 (0.63, 1.87)	1.00 (Ref)		1.00 (Ref)
No	84	0.99 (0.76, 1.28)	0.87 (0.69, 1.09)	2.10 (1.62, 2.73)	1.97 (1.52, 2.55)	1.30 (0.64, 2.66)		1.11 (0.60, 2.07)
Glove and other PPEc use while applyinge
Both gloves and other PPE	8	0.50 (0.22, 1.13)	0.44 (0.19, 1.04)	1.21 (0.67, 2.19)	1.06 (0.58, 1.92)	1.00 (Ref)		1.00 (Ref)
Gloves only	15	0.61 (0.43, 0.87)	0.49 (0.33, 0.74)	1.22 (0.77, 1.92)	1.15 (0.73, 1.82)	0.97 (0.41, 2.31)		0.95 (0.44, 2.06)
Other PPE only	7	0.47 (0.21, 1.08)	0.49 (0.19, 1.24)	1.31 (0.69, 2.49)	1.11 (0.58, 2.11)	1.09 (0.40, 3.00)		0.90 (0.36, 2.22)
Neither gloves nor other PPE	66	1.12 (0.81, 1.54)	1.01 (0.78, 1.32)	2.50 (1.90, 3.29)	2.34 (1.79, 3.07)	1.99 (0.96, 4.13)		1.52 (0.79, 2.95)
Glove use while applyinge
Yes	23	0.57 (0.41, 0.79)	0.48 (0.33, 0.68)	1.21 (0.83, 1.78)	1.11 (0.76, 1.63)	1.00 (Ref)		1.00 (Ref)
No	73	1.03 (0.76, 1.39)	0.94 (0.73, 1.21)	2.35 (1.80, 3.07)	2.18 (1.67, 2.84)	1.91 (1.17, 3.11)		1.49 (0.95, 2.34)
Wearing long-sleeved shirt while applyinge
Yes	9	0.54 (0.28, 1.03)	0.50 (0.25, 1.00)	1.24 (0.68, 2.27)	1.06 (0.58, 1.94)	1.00 (Ref)		1.00 (Ref)
No	87	0.94 (0.73, 1.22)	0.84 (0.67, 1.05)	2.08 (1.60, 2.69)	1.95 (1.51, 2.52)	1.87 (0.86, 4.03)		1.29 (0.64, 2.58)
Wearing goggles while applyinge
Yes	5	0.34 (0.17, 0.66)	0.42 (0.23, 0.77)	1.03 (0.48, 2.22)	0.91 (0.42, 1.98)	1.00 (Ref)		1.00 (Ref)
No	91	0.94 (0.73, 1.21)	0.83 (0.66, 1.04)	2.06 (1.59, 2.66)	1.91 (1.48, 2.47)	1.93 (0.75, 4.99)		1.62 (0.70, 3.78)
Application methode,f
Hand spray only	32	0.90 (0.58, 1.39)	0.79 (0.53, 1.19)	1.83 (1.30, 2.59)	1.73 (1.22, 2.46)	1.00 (Ref)		1.00 (Ref)
Broadcast/boom spray only	30	1.17 (0.70, 1.98)	1.07 (0.70, 1.66)	2.91 (2.02, 4.17)	2.60 (1.81, 3.72)	1.70 (1.00, 2.92)		2.06 (1.29, 3.27)
Both hand spray and broadcast/boom spray	32	0.74 (0.53, 1.04)	0.64 (0.47, 0.87)	1.57 (1.12, 2.22)	1.48 (1.04, 2.09)	0.86 (0.52, 1.41)		1.08 (0.70, 1.66)
Application typee
Non-crop onlyg	36	0.76 (0.50, 1.15)	0.72 (0.49, 1.04)	1.73 (1.25, 2.40)	1.58 (1.13, 2.19)	1.00 (Ref)		1.00 (Ref)
Crops only	30	1.25 (0.74, 2.11)	1.08 (0.70, 1.69)	2.97 (2.07, 4.26)	2.69 (1.88, 3.85)	1.72 (1.04, 2.84)		1.83 (1.18, 2.82)
Both non-crop and crops	30	0.78 (0.56, 1.08)	0.68 (0.50, 0.91)	1.61 (1.13, 2.30)	1.53 (1.07, 2.19)	0.89 (0.53, 1.47)		1.01 (0.64, 1.57)

Abbreviations: CI, confidence interval; PPE, personal protective equipment; Ref, reference.

Bold values indicate statistical significance at P < 0.05.

^aAdjusted for age (continuous, years), age squared, urinary creatinine concentration (natural log-transformed; continuous), state (Iowa, North Carolina), season of urine collection (April–September, October–March), time of urine collection (before 4:00 a.m., 4:00–5:59 a.m., 6:00 a.m. or later), and body mass index (continuous, kg/m²).

^bAdjusted for Model 1 covariates (listed in footnote a) and additionally for number of days since last occupational use of glyphosate (5–7, 2–4, ≤1 days).

^cOther PPE includes long-sleeved shirt, goggles, rubber boots, and/or respirator.

^dAmong recently exposed farmers who personally mixed or loaded glyphosate (n = 95).

^eAmong recently exposed farmers who personally applied glyphosate (n = 96).

^fAnalysis excludes one farmer who used airblast only and one farmer who used backpack sprayers only.

^gExamples of non-crop applications include use around buildings, driveways, and fence rows.

In additional analyses examining occupational glyphosate use in the last 12 months (Table 4), we observed higher urinary glyphosate concentrations among recently exposed farmers in the highest tertiles of glyphosate use, including a statistically significant trend for intensity-weighted days of use (tertile 3 vs. 1; GMR = 1.62; 95 % CI: 0.98, 2.68; P_trend = 0.02); however, the association was attenuated and lost statistical significance after adjusting for days since last use. Although glyphosate concentrations were also generally elevated for all tertiles of lifetime glyphosate use relative to the control groups, we did not observe any statistically significant trend across tertiles of lifetime days or intensity-weighted lifetime days of use among exposed farmers (Supplementary Table S3).

Table 4.

Associations between occupational glyphosate use in the last 12 months and urinary glyphosate concentrations in the Biomarkers of Exposure and Effect in Agriculture study, 2010–2018.

Glyphosate use in the last 12 months	n	Geometric mean concentration		Geometric mean ratio (95 % CI)
		μg/L (95 % CI)	μg/g creatinine (95 % CI)	Model 1a				Model 2b
				Compared to nonfarming controls	Compared to farming controls	Exposed farmers only		Exposed farmers only
Nonfarming controls	100	0.39 (0.34, 0.46)	0.37 (0.33, 0.42)	1.00 (Ref)	—	—		—
Farming controls	100	0.46 (0.40, 0.52)	0.42 (0.37, 0.47)	—	1.00 (Ref)	—		—
Recently exposed
Number of days of use in the last 12 months
Tertile 1 (1–8)	35	0.56 (0.41, 0.76)	0.57 (0.41, 0.78)	1.40 (1.01, 1.94)	1.27 (0.91, 1.76)	1.00 (Ref)		1.00 (Ref)
Tertile 2 (9–15)	38	1.09 (0.69, 1.74)	0.96 (0.65, 1.40)	2.41 (1.76, 3.32)	2.21 (1.60, 3.05)	1.69 (1.07, 2.09)		1.40 (0.92, 2.12)
Tertile 3 (16–260)	25	1.22 (0.80, 1.86)	0.96 (0.65, 1.42)	2.40 (1.63, 3.52)	2.31 (1.57, 3.39)	1.52 (0.88, 2.62)		1.36 (0.84, 2.20)
				Ptrendc < 0.0001	Ptrendc < 0.0001	Ptrendc = 0.13		Ptrendc = 0.21
Intensity-weighted days of use in the last 12 months
Tertile 1 (88–528)	33	0.59 (0.41, 0.87)	0.63 (0.44, 0.91)	1.65 (1.18, 2.32)	1.46 (1.04, 2.06)	1.00 (Ref)		1.00 (Ref)
Tertile 2 (>528–1,080)	34	0.83 (0.58, 1.19)	0.66 (0.48, 0.90)	1.62 (1.16, 2.27)	1.54 (1.10, 2.14)	0.90 (0.54, 1.50)		0.93 (0.59, 1.48)
Tertile 3 (>1,080–11,375)	31	1.45 (0.89, 2.38)	1.24 (0.83, 1.87)	3.03 (2.14, 4.29)	2.82 (1.99, 3.99)	1.62 (0.98, 2.68)		1.21 (0.76, 1.93)
				Ptrendc < 0.0001	Ptrendc < 0.0001	Ptrendc = 0.02		Ptrendc = 0.32
High lifetime exposed
Number of days of use in the last 12 months
Tertile 1 (0–5)	24	0.72 (0.51, 1.02)	0.57 (0.42, 0.78)	1.58 (1.17, 2.14)	1.31 (0.97, 1.76)	1.00 (Ref)		1.00 (Ref)
Tertile 2 (6–20)	28	0.58 (0.41, 0.83)	0.50 (0.39, 0.66)	1.40 (1.06, 1.84)	1.24 (0.95, 1.62)	0.92 (0.61, 1.38)		0.93 (0.60, 1.43)
Tertile 3 (21–120)	18	0.47 (0.31, 0.71)	0.45 (0.33, 0.59)	1.29 (0.93, 1.80)	1.17 (0.84, 1.63)	0.92 (0.55, 1.53)		0.93 (0.54, 1.61)
				Ptrendc = 0.12	Ptrendc = 0.30	Ptrendc = 0.74		Ptrendc = 0.81
Intensity-weighted days of use in the last 12 months
Tertile 1 (0–440)	24	0.69 (0.48, 0.98)	0.58 (0.43, 0.77)	1.60 (1.18, 2.16)	1.31 (0.97, 1.76)	1.00 (Ref)		1.00 (Ref)
Tertile 2 (>440–1,485)	23	0.52 (0.34, 0.79)	0.42 (0.31, 0.56)	1.21 (0.89, 1.65)	1.08 (0.80, 1.46)	0.73 (0.46, 1.15)		0.73 (0.45, 1.19)
Tertile 3 (>1,485–4,800)	23	0.58 (0.41, 0.83)	0.54 (0.41, 0.72)	1.48 (1.10, 1.98)	1.34 (1.01, 1.78)	1.00 (0.65, 1.54)		1.00 (0.63, 1.59)
				Ptrendc = 0.04	Ptrendc = 0.10	Ptrendc = 0.69		Ptrendc = 0.63
Recently exposed and high lifetime exposed
Number of days of use in the last 12 months
Tertile 1 (0–6)	56	0.59 (0.46, 0.74)	0.55 (0.44, 0.69)	1.44 (1.11, 1.88)	1.26 (0.97, 1.65)	1.00 (Ref)		1.00 (Ref)
Tertile 2 (7–15)	60	0.93 (0.67, 1.29)	0.81 (0.62, 1.06)	2.07 (1.60, 2.68)	1.87 (1.44, 2.43)	1.41 (1.01, 1.97)		1.21 (0.89, 1.65)
Tertile 3 (16–260)	52	0.76 (0.58, 1.01)	0.63 (0.50, 0.81)	1.72 (1.31, 2.26)	1.58 (1.20, 2.07)	1.18 (0.82, 1.69)		1.09 (0.79, 1.52)
				Ptrendc < 0.0001	Ptrendc = 0.0002	Ptrendc = 0.50		Ptrendc = 0.69
Intensity-weighted days of use in the last 12 months
Tertile 1 (0–512)	56	0.68 (0.52, 0.89)	0.63 (0.50, 0.81)	1.69 (1.30, 2.20)	1.48 (1.13, 1.94)	1.00 (Ref)		1.00 (Ref)
Tertile 2 (>512–1,320)	56	0.67 (0.51, 0.87)	0.54 (0.44, 0.68)	1.46 (1.12, 1.90)	1.32 (1.01, 1.72)	0.84 (0.59 1.19)		0.81 (0.59, 1.12)
Tertile 3 (>1,320–11,375)	56	0.92 (0.66, 1.29)	0.83 (0.63, 1.10)	2.09 (1.61, 2.72)	1.92 (1.47, 2.50)	1.20 (0.85, 1.69)		1.01 (0.74, 1.40)
				Ptrendc < 0.0001	Ptrendc < 0.0001	Ptrendc = 0.11		Ptrendc = 0.57

Abbreviations: CI, confidence interval; Ref, reference.

Bold values indicate statistical significance at P < 0.05.

^aAdjusted for age (continuous, years), age squared, urinary creatinine concentration (natural log-transformed; continuous), state (Iowa, North Carolina), season of urine collection (April–September, October–March), time of urine collection (before 4:00 a.m., 4:00–5:59 a.m., 6:00 a.m. or later), and body mass index (continuous, kg/m²).

^bAdjusted for Model 1 covariates (listed in footnote a) and additionally for number of days since last occupational use of glyphosate (>180, 31–180, 8–30, 5–7, 2–4, ≤1 days).

^cCalculated by modeling within-category median values (setting as “0″ for the nonfarming and farming control groups) as continuous variables.

4. Discussion

This urinary biomonitoring study among farmers and nonfarmers is one of the largest and most detailed investigations to date of occupational glyphosate exposure and its potential determinants. We found that high proportions of both farmers (88–93 %, depending on the study group) and nonfarmers (81 %) from Iowa and North Carolina had detectable concentrations of glyphosate (LOD = 0.2 μg/L) in their urine. The strongest determinant of urinary glyphosate concentrations was recency of occupational glyphosate use, with substantial exposure contrast observed between those with and without self-reported use within 7 days prior to urine sampling, as well as a clear trend of increasing concentrations with shorter time elapsed since last use of glyphosate. We also observed elevated glyphosate concentrations among recently exposed farmers who did not wear gloves when applying glyphosate, those who did not wear a long-sleeved shirt when mixing/loading glyphosate, and those who applied glyphosate exclusively using broadcast/boom sprayers or to crops only.

Glyphosate was detected in the urine of approximately 90 % of farmers in our study. This is higher than those reported by most previous occupational biomonitoring studies of farmers (0–75 %; LODs ranging from 0.05 to 5 μg/L) in the United States (Acquavella et al., 2004; Curwin et al., 2007; Perry et al., 2019) and abroad (Rendón-von Osten and Dzul-Caamal, 2017; Wongta et al., 2018; Balderrama-Carmona et al., 2020; Connolly et al., 2022; Mueller et al., 2024), although several studies in Sri Lanka (Jayasumana et al., 2015), Thailand (Bootsikeaw et al., 2021), and New Zealand (Campbell et al., 2022) had detection frequencies of > 95 %. Urinary glyphosate concentrations also varied widely across studies, with central tendency measures (median or GMs) ranging from < LOD to 73.5 μg/L (mostly ~ 0.5–3 μg/L) among farmers (Gillezeau et al., 2019; Gillezeau et al., 2020; Connolly et al., 2020); concentrations in our study are in the lower end of this range. However, comparisons across studies should be interpreted with caution in light of differences in analytical methods (e.g., enzyme-linked immunosorbent assay, mass spectrometry with different chromatography methods), method sensitivities, and urine sample type (e.g., spot, 24-hour, first morning void) (Gillezeau et al., 2019; Gillezeau et al., 2020; Connolly et al., 2020). The large variability in urinary glyphosate concentrations may also reflect differences in participant characteristics (e.g., age), timing of glyphosate use with respect to urine sampling, regulations surrounding glyphosate use in agricultural and other settings, application practices, and PPE use across farming populations from different parts of the world.

Urinary glyphosate concentrations among nonfarmers in our study [GM = 0.39 μg/L (0.37 μg/g creatinine)] are similar to those observed in a nationally representative sample of the general U.S. population in NHANES [GMs = 0.41 and 0.35 μg/L (0.44 and 0.37 μg/g creatinine) in 2013–2014 and 2015–2016, respectively] (Ospina et al., 2022; National Center for Environmental Health, 2022). Of note, glyphosate concentrations in our study were quantified by the same laboratory that measured glyphosate in NHANES using the same analytical approach with the same LOD (Ospina et al., 2022; Schütze et al., 2021). Similar concentrations were also observed among NHANES subgroups with demographics comparable to our study (e.g., GM = 0.42 μg/L for both males and non-Hispanic White persons) (Ospina et al., 2022). Although our study consisted of an older population (age ≥ 50 years), the association we observed between increasing age and increasing glyphosate concentrations among nonfarmers (and recently exposed farmers) is consistent with that among adults ≥ 20 years of age in NHANES, (Ospina et al., 2022). In contrast, a large study of the French general population observed a continuous decrease in glyphosate concentrations with increasing age (Grau et al., 2022), while the few occupational studies examining age as a potential determinant in farmers found no associations (Perry et al., 2019; Bootsikeaw et al., 2021; Mueller et al., 2024). Beyond possible age-related differences in exposure sources and pathways, detailed toxicokinetics of glyphosate remain to be fully elucidated (Connolly et al., 2020). We also noted a potential difference in urinary glyphosate concentrations by time of urine collection among recently exposed farmers, with a less clear pattern after correction for creatinine; this may be related to factors such as time since last urination, although this information was not available in our study. Furthermore, while our study lacked information on fasting status and dietary consumption, data from NHANES suggested higher urinary glyphosate concentrations among individuals who fasted for ≤ 8 (vs. > 8) hours prior to urine collection and those who consumed cereal in the last 24 h, highlighting diet as a primary source of exposure in the general population (Ospina et al., 2022) and the importance of evaluating glyphosate residues in contaminated foods (European Food Safety Authority (EFSA), 2019; Kolakowski et al., 2020; Xu et al., 2019). Additionally, an organic diet was found to reduce urinary glyphosate concentrations in two intervention trials (Fagan et al., 2020; Hyland et al., 2023), including one reporting a stronger effect among individuals living further away from agricultural fields (Hyland et al., 2023). Given the relatively large proportion of nonfarmers with detectable glyphosate, future biomonitoring studies, especially those in the general population, should incorporate detailed dietary assessment to provide a fuller picture of potential exposure sources.

Similar to our findings, higher detection frequencies and/or concentrations of urinary glyphosate were observed among farmers compared to nonfarmers in most other studies that quantified glyphosate in both subgroups (Curwin et al., 2007; Connolly et al., 2022; Campbell et al., 2022; Grau et al., 2022; Jayasumana et al., 2015; Rendón-von Osten and Dzul-Caamal, 2017; Wongta et al., 2018), suggesting that agricultural use of glyphosate is an important source of exposure. With detailed information on the timing of occupational glyphosate use, our study further confirmed that recency of use, particularly across a range of ≤ 1–7 days since last use, is an important predictor of urinary glyphosate concentrations among farmers. Consistent with the short urine elimination half-life of glyphosate (~5–10 h) (Connolly et al., 2019; Zoller et al., 2020; Kohsuwan et al., 2022), several studies among farmers (Acquavella et al., 2004; Mesnage et al., 2012; Bootsikeaw et al., 2021; Mueller et al., 2024) and amenity horticulturists (Connolly et al., 2017; Connolly et al., 2018) with serial urine samples also reported higher concentrations shortly after (vs. before) glyphosate application and declining concentrations over time following application. Notably, in addition to the much higher concentrations among farmers who used glyphosate in the last day, we also found significantly elevated concentrations among those with use in the last 2–4 days (and elevated, albeit not statistically significant, concentrations for use in the last 5–7 days) compared to each of the control groups, suggesting that despite the short half-life, glyphosate applicators may have exposures even several days after reported use (e.g., through contaminated clothing and equipment). Furthermore, although we found no associations with home/garden glyphosate use in the last 12 months (yes vs. no) among both farmers and nonfarmers and had no information on timing of home/garden use for farmers, the observation of higher glyphosate concentrations among a small number of nonfarmers with more recent use (8–30 vs. > 30 days before urine collection) warrants additional investigation. There is a paucity of data on exposure specifically related to residential use of glyphosate. In NHANES, use of any chemical products in the lawn or garden to kill weeds in the last 7 days was not associated with urinary glyphosate concencentrations (Ospina et al., 2022).

Our study was one of the very few to assess urinary glyphosate concentrations in relation to PPE use among farmers. Our results suggest that PPE use, particularly gloves when applying and long-sleeved shirts when mixing/loading glyphosate, may help reduce exposure. In an investigation of 48 farmers from the Farm Family Exposure Study in South Carolina and Minnesota, farmers who did not wear rubber gloves when mixing/loading glyphosate had statistically significantly higher urinary glyphosate concentrations compared to those who wore gloves; however, use of gloves or other PPE during glyphosate application was not evaluated (Acquavella et al., 2004). Another study of 43 Thai farmers found that wearing long (vs. short) sleeved shirts when mixing and spraying glyphosate was associated with lower glyphosate concentrations in both urine and dermal patch samples (Bootsikeaw et al., 2021). Together, these findings highlight skin contact as a potential major route of exposure among farmers who use glyphosate. However, it should be noted that while approximately two-thirds of farmers in our study were hand spray applicators, all farmers in the Farm Family Exposure Study applied glyphosate using boom sprayers and tractors (Acquavella et al., 2004), and most farmers in the Thai study used backpack sprayers (Bootsikeaw et al., 2021). Interestingly, we found the highest urinary glyphosate concentrations among the one-third of farmers who applied glyphosate exclusively using broadcast/boom sprayers and lower concentrations among those who used hand sprayers. Conversely, the AHS pesticide exposure intensity algorithm assigned greater weight to hand spray than broadcast and boom spray (Coble et al., 2011). Our findings may be explained by potential differences in glyphosate-based formulations used, work practices, frequency and duration of applications, and exposure pathways and routes across different application methods (Connolly et al., 2017; Arbuckle et al., 2002). Notably, we found that application method was highly correlated with application type (i.e., broadcast/boom spray and hand spray for crop and non-crop applications, respectively); it is likely that glyphosate applications to crops involved more frequent and longer periods of spraying and products with higher active ingredient concentrations compared to non-crop applications. Farmers applying glyphosate to crops may also be exposed to residues on crops during re-entry activities (Toumi et al., 2019), which warrant further exploration in relation to urinary glyphosate concentrations in other studies conducted in different agricultural settings.

Given the short half-life of glyphosate in urine, it is not surprising that cumulative measures of long-term (past 12-month or lifetime) occupational glyphosate use showed no exposure–response associations with urinary glyphosate concentrations among farmers after adjusting for days since last use. However, when compared with nonfarmers, we found statistically significantly elevated concentrations among farmers who were high lifetime (but not recent) users of glyphosate, as well as non-statistically significantly higher concentrations among farmers with minimal lifetime use. Because urinary glyphosate is a marker of recent exposure, this novel finding is interesting and suggests that these farmers, though not having personally used glyphosate recently, may be indirectly exposed in their work environment, especially given the relatively longer half-life of glyphosate in soil (~2–200 days, but typically < 60 days) (Giesy et al., 2000; Agricultural Research Service (ARS), 1995). Exposure may have also occurred through household dust, which can serve as a long-term reservoir of pesticide residues entering the farmers’ homes via contaminated clothes and shoes and from nearby agricultural applications (Simcox et al., 1995; Curl et al., 2023). Notably, a small study of five farm and six nonfarm households in Iowa observed higher dust concentrations of glyphosate in farm households (Curwin et al., 2005), with positive but nonsignificant correlations between dust and urine concentrations (Curwin et al., 2007). Another study in the general population of California reported significantly higher glyphosate dust concentrations in households that had members with occupational pesticide exposure, households with pesticide use around the home to treat lawn/garden weeds, and those with nearby (within 1 km of the home) agricultural use of glyphosate (Ward et al., 2023). Additional studies quantifying glyphosate in both urine and dust can help better characterize recent and longer-term glyphosate exposure and potential determinants among farmers as well as their spouses and children.

The main strengths of our study include the larger sample size compared to other occupational biomonitoring studies of glyphosate among farmers, highly reproducible measurements of urinary glyphosate quantified by the same method used in NHANES, inclusion of both farming and nonfarming control groups, and availability of detailed information on occupational glyphosate use, including timing of use and other related characteristics (e.g., PPE use, application method/type), as well as home/garden glyphosate use. Our study also had several limitations. First, because glyphosate was quantified in a single first morning void urine sample, we could not evaluate within-individual changes in concentrations over time, and given the short urine half-life of glyphosate and the timing of urine collection with respect to glyphosate use, we may have missed peak exposures and underestimated exposure even among farmers who used glyphosate on the day prior to the urine sampling. As such, exposure differences between groups may be even greater than those reflected by measured urine concentrations. Second, unlike several occupational biomonitoring studies in farmers (Perry et al., 2019; Balderrama-Carmona et al., 2020; Connolly et al., 2022; Campbell et al., 2022), we did not measure urinary concentrations of aminomethylphosphonic acid (AMPA), the main metabolite of glyphosate. However, because of limited glyphosate metabolism in humans, AMPA is generally detected not as frequently and at lower concentrations than glyphosate in urine (Agency for Toxic Substances and Disease Registry (ATSDR), 2020; Gillezeau et al., 2019; Connolly et al., 2020). AMPA is also known to form as a breakdown product of other phosphonate-containing compounds, such as detergents, and thus may not specifically reflect glyphosate exposure (Grandcoin et al., 2017). Nevertheless, future biomonitoring studies incorporating AMPA, as well as surfactant co-formulants (e.g., polyethoxylated amine) present in glyphosate-based products (Mesnage et al., 2019), may help provide a more comprehensive assessment of exposures related to glyphosate use in both occupational and non-occupational settings (Connolly and Koch, 2023). Third, information on PPE use among recently exposed farmers was based on their usual practices when mixing/loading or applying glyphosate in the last 12 months, which may not necessarily apply to the most recent instance of use. Similarly, for farmers who reported using both hand and broadcast/boom sprayers, it is unclear which method (or both) was used during the most recent application. Fourth, information on recency of home/garden pesticide use was only available for nonfarmers. Finally, our study included a predominantly non-Hispanic White population, and as such, results may not be generalizable to other populations of farmers or farmworkers. Future investigations of urinary glyphosate concentrations and their determinants would benefit from including farmers and farmworkers from diverse racial and ethnic backgrounds in various agricultural settings across different geographic areas.

5. Conclusions

Findings from this large biomonitoring study of male farmers and nonfarmers from Iowa and North Carolina suggest that while a high proportion of both farmers and nonfarmers may be exposed to glyphosate, substantial exposure contrast was detectable in urine between those with and without recent occupational glyphosate use. In addition, factors such as PPE use and application method and type were associated with urinary glyphosate concentrations among farmers, which may have implications for strategies to reduce occupational exposure. Given the widespread use of glyphosate, continued biomonitoring of this herbicide in various settings is warranted.

Supplementary Material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.envint.2024.108644.

Abbreviations:

AHS

Agricultural Health Study

AMPA

aminomethylphosphonic acid

BEEA

Biomarkers of Exposure and Effect in Agriculture

BMI

body mass index

CDC

Centers for Disease Control and Prevention

CI

confidence interval

GM

geometric mean

GMR

geometric mean ratio

LOD

limit of detection

NHANES

National Health and Nutrition Examination Survey

PPE

personal protective equipment

QC

quality control

Data availability

The data underlying this investigation will be provided upon request as described for the BEEA subcohort on the AHS website: aghealth.nih.gov/collaboration/studies.html.