The Effect of Pesticide Spray Season and Residential Proximity to Agriculture on Glyphosate Exposure among Pregnant People in Southern Idaho, 2021 Authors: Cynthia L Curl, Carly Hyland, Meredith Spivak, Lianne Sheppard, Bruce Lanphear, Michael N Antoniou, Maria Ospina, Antonia M Calafat Published: 2023 ABSTRACT AbstractBackground:Glyphosate is one of the most heavily used pesticides in the world, but little is known about sources of glyphosate exposure in pregnant people living in agricultural regions.Objective:Our objective was to evaluate glyphosate exposure during pregnancy in relation to residential proximity to agriculture as well as agricultural spray season.Methods:We quantified glyphosate concentrations in 453 urine samples collected biweekly from a cohort of 40 pregnant people in southern Idaho from February through December 2021. We estimated each participant’s glyphosate exposure as the geometric mean (GM) of glyphosate concentrations measured in all samples (average n=11 samples/participant), as well as the GM of samples collected during the pesticide “spray season” (defined as those collected 1 May–15 August; average n=5 samples/participant) and the “nonspray season” (defined as those collected before 1 May or after 15 August; average n=6 samples/participant). We defined participants who resided <0.5km from an actively cultivated agriculture field to live “near fields” and those residing ≥0.5km from an agricultural field to live “far from fields” (n=22 and 18, respectively).Results:Among participants living near fields, urinary glyphosate was detected more frequently and at significantly increased GM concentrations during the spray season in comparison with the nonspray season (81% vs. 55%; 0.228μg/L vs. 0.150μg/L, p<0.001). In contrast, among participants who lived far from fields, neither glyphosate detection frequency nor GMs differed in the spray vs nonspray season (66% vs. 64%; 0.154μg/L vs. 0.165μg/L, p=0.45). Concentrations did not differ by residential proximity to fields during the nonspray season (0.154μg/L vs. 0.165μg/L, for near vs. far, p=0.53).Discussion:Pregnant people living near agriculture fields had significantly increased urinary glyphosate concentrations during the agricultural spray season than during the nonspray season. They also had significantly higher urinary glyphosate concentrations during the spray season than those who lived far from agricultural fields at any time of year, but concentrations did not differ during the nonspray season. These findings suggest that agricultural glyphosate spray is a source of exposure for people living near fields. https://doi.org/10.1289/EHP12768 Introduction In 2019, Gillezeau et al.1 published a review of the strikingly sparse scientific evidence documenting human exposure to glyphosate. Glyphosate, which entered the agrochemical market over 50 y ago, is the most commonly and intensively used herbicide in the world due to both the introduction of glyphosate-tolerant genetically modified crops starting in 1996 and its more recent adoption as a desiccant prior to harvest of crops such as cereals and pulses.2 Yet by 2019, only 19 studies of glyphosate concentrations in humans had been published. Moreover, these studies usually included small numbers of participants, evaluated glyphosate concentrations in different tissues with varying detection limits, and provided little information on temporal trends or potential sources of exposure. Important advances have been made in our understanding of glyphosate exposure in the 3 y since Gillezeau et al. published their review. In 2023, the U.S. Centers for Disease Control and Prevention (CDC) reported, for the first time, urinary glyphosate concentrations using samples from the 2013–2014 and 2015–2016 National Health and Nutrition Examination Survey (NHANES).3 The population-representative 2013–2014 NHANES data showed that an estimated 81% of the U.S. population ≥6 y of age had recent exposure to glyphosate.4 In a population-based survey in France, glyphosate was detected in 99% of urine samples collected from nearly 7,000 participants recruited between 2018 and 2020.5 Several other studies published in the past year alone have measured glyphosate concentrations in both occupationally and nonoccupationally exposed populations, with most reporting glyphosate concentrations in a single spot urine sample.6–12 Human studies have reported adverse reproductive effects of glyphosate exposure at environmentally relevant levels. Four recent investigations found that gestational glyphosate exposure was associated with shortened gestational length,13 preterm birth,14,15 and reduced fetal growth.16 If the developing fetus is especially vulnerable to glyphosate, it is critical to understand the magnitude and sources of exposure during this critical developmental period. Still, few studies have examined glyphosate exposure during pregnancy. Beyond the four studies of glyphosate exposure and reproductive outcomes, we are aware of only two other studies of glyphosate exposure during pregnancy, neither of which occurred within the past decade.17,18 Given the increasing use of glyphosate and glyphosate-based herbicides (GBH; e.g., Roundup), the results of earlier studies may not be indicative of current exposures. Another important data gap relates to sources of glyphosate exposure. Diet is one potentially important source19,20; agricultural use may be another. More than 250 million pounds of glyphosate are applied each year to cropland in the United States alone,2 and exposure to many types of pesticides is higher among individuals living near agricultural fields.21,22 Pesticide exposure among residents of agricultural regions can occur through inhalation or dermal absorption via spray drift from active applications or through ingestion of pesticide residues in soil or water.23 A few studies have evaluated the effect of residential location on glyphosate exposure and found higher exposure among those in rural/agricultural locations than those living in urban regions.13,18,24 The purpose of this study was to evaluate glyphosate exposure in pregnant women in relation to residential proximity to agriculture as well as agricultural spray season. Methods Study Setting and Design In 2021, we recruited 40 participants living in southern Idaho during the first trimester of pregnancy and collected weekly urine samples from enrollment through delivery. Participants lived either close to (<0.5km) or far from (≥0.5km) actively cultivated agricultural fields. We recruited participants from Women, Infants and Children (WIC) clinics in three health districts of Idaho: Southwest District Health (serving the towns of Nampa and Caldwell), South Central District Health (serving Twin Falls, Shoshone, Gooding, Jerome, and Heyburn), and Central District Health (serving Boise, Meridian, and Garden City). Administered through local agencies, WIC is a federal program that provides nutritional support to low-income pregnant, postpartum, and breastfeeding women and low-income infants and children up to 5 y of age. We calculated a sample size to provide a power of 0.80 at a 0.05 significance level in a two-sided test of the difference in exposure between participants living near and far from agricultural fields, based on the geometric mean (GM) glyphosate concentrations measured in urine samples collected from mothers in farm and nonfarm households25 and an assumed standard deviation (SD) of 0.5μg/L. We estimated this SD from the ratio of mean to variance in organophosphate pesticide (OP) concentrations from a repeated measures study,26 which we used because no comparable glyphosate data exist and half-lives are similar (<12–16 h).27,28 We began our study prior to the annual initiation of agricultural pesticide applications and continued sample collection throughout the year, thus capturing exposure before, during, and after the agricultural pesticide spray season. We defined the agricultural pesticide spray season a priori as 1 May through 15 August, based on conversations with local agricultural experts, including land grant extension specialists and educators regarding weather and conditions for 2021. Participant Recruitment and Eligibility Because of the COVID-19 pandemic, we developed contactless approaches to communicate with and engage our community partners. Specifically, we developed a video (https://www.youtube.com/watch?v=pFH3MIELXNU), which we shared with WIC clinic staff in lieu of visiting the clinics for in-person presentations. We worked remotely with clinic directors to identify on-site “project champions” at each Health District who connected with us regularly and directly, advocated for our research project, and helped answer any of their colleagues’ questions. In addition to the videos we prepared for WIC clinic staff, our research team created brief videos to advertise the study to potential participants (https://www.youtube.com/watch?v=sOG0lppKzUw and https://www.youtube.com/watch?v=p75mIj0hwrc); all materials intended for participants were prepared in both English and Spanish. We asked the WIC staff to share these videos with any clients they thought might be interested in participating in this study. We also created electronic versions of full-page recruitment flyers and shorter index card–sized recruitment flyers that could be emailed to potential participants. When a WIC staff member spoke with a potential participant who wanted to learn more about the study, the staff member asked them for permission to share their contact information with our research team. Our research staff then contacted interested individuals to assess eligibility. Eligible participants were in their first trimester of pregnancy; over the age of 18 y; had not been told by a medical professional that their pregnancy was high risk; did not work with pesticides or have a household member who worked with pesticides; did not regularly eat organic food; had access to a smartphone, tablet, or computer that could connect to the Internet; and spoke either English or Spanish. We scheduled initial meetings with interested and eligible participants for the end of the participant’s first trimester at 13–14 wk of pregnancy. The Boise State University institutional review board reviewed and approved this study. Informed Consent, Demographic Questionnaire, and Pesticide Exposure Interview All contact with participants occurred in English or Spanish, as preferred by the participant, and all were conducted in a contactless manner. For the initial meetings, research staff traveled to each participant’s home and phoned the participant on arrival. Staff then placed paper consent forms outside the participant’s home, returned to their vehicles, and texted the participant the link to a study-specific video that provided a detailed explanation of all components of the informed consent process. They then asked the participant to either call the research staff member with any questions or to sign the consent form and leave it outside. Each participant was assigned a unique Study ID. Following the consent process, research staff phoned the participant and verbally administered a basic demographic questionnaire to collect self-reported information on age, household income, education, race/ethnicity (African American or Black; Asian; Caucasian or White; Hispanic or Latino; Native Hawaiian or other Pacific Islander; American Indian or Alaskan Native), occupation, household size, and whether or not they were a current smoker or vape user. Research staff also verbally administered a pesticide exposure interview about residential use of pesticides to control for indoor and outdoor insects, weeds, fleas, or ticks on domestic pets or for other pests, including snails, slugs, birds, rabbits, squirrels, rats, mice, gophers, raccoons, or deer (see Table S1). Participants were asked about both their personal use of residential pesticides, use by any other member of their household, and use by a pest control company. For any pesticides used by a participant or a member of their household, participants were asked whether they knew what product was used and/or had the containers available. For any containers they had available, participants were asked to locate and photograph those products and to text the pictures to our research staff. Urine Sample Collection During the initial meeting, participants were asked to provide a spot urine sample. Following completion of the surveys, research staff texted the participant with a link to a study-specific video that described how to collect that sample. Research staff then placed a 4-oz polypropylene specimen cup with a unique bar-coded label in a resealable plastic bag outside the participant’s door. The participant then collected and returned the sample. Research staff then left the participant with a USD $10 gift card, a small plastic cooler and reusable ice pack, and a prelabeled specimen cup for the next week’s sample. We established a regular day of the week for urine sample collection, and for each subsequent week, we asked participants to collect their first urine of the day [first morning void (FMV)], to write the time and date of collection on the specimen cup, and to leave it outside their home in the study cooler with ice packs for collection by research staff that same day. We provided participants with posters to hang in their bathrooms as reminders and sent text messages the night before each participant’s scheduled sample collection to remind them to collect their sample the next morning. Research staff then traveled to each participant’s home on the preestablished day to collect that week’s sample and to leave another $10 gift card and a labeled specimen cup for the following week. We collected weekly urine samples from enrollment through delivery except for a 2-wk period from 14 June to 30 June, during which the cohort took part in a randomized crossover trial of an organic diet. During the crossover trial, participants provided daily urine samples; the details of the dietary intervention are described elsewhere.29 None of the results from daily samples collected during this dietary intervention are included in this analysis. We collected an average of 22 weekly urine samples from each of the 40 participants and grouped these samples by calendar week. Because of funding limitations, we selected for analysis all samples collected during every other week. Urine Sample Processing and Laboratory Analysis Research staff completed a chain-of-custody form to record the date and time of each urine collection and make any notes about the collection process or sample condition. Samples were transported on ice to our laboratory at Boise State University, where they were refrigerated for no more than 24 h before processing. All samples were analyzed for color, clarity, and specific gravity (Atago Urine Specific Gravity Refractometer, PAL 10-S) at 5°C. Research staff aliquoted four identical 4-mL subsamples into 5-mL cryovials for storage at −80°C. In addition to participant samples selected for glyphosate quantification, for quality assurance, we also randomly selected a set of duplicates from 10% of the sample set sent for analysis, which were submitted without identifying characteristics. We shipped samples and duplicates to the CDC overnight on dry ice in January 2022 for analysis. Details of the analytical method have been described previously.30 Briefly, the method employs ion chromatography-isotope dilution–tandem mass spectrometry using a Dionex ICS-5000+ion chromatography system (using polyether ether ketone materials to prevent carryover and interaction with metal surfaces) and an AB Sciex 5500 triple quadrupole mass spectrometer. The limit of detection (LOD) was 0.1μg/L. Method accuracy was established by spiking two different urine samples at zero, low, mid, and high glyphosate concentrations. Samples were prepared in triplicates and the results averaged. Recovery of the spiked analytes was calculated as [(final concentration− initial concentration)/spiked concentration], and the mean relative recovery for this method was 99% (97%–103%). Along with study samples and analytical standards, each analytical batch included high- and low-concentration quality control materials (QCs) and reagent blanks to assure the accuracy and reproducibility of the data. The concentrations of the QCs were evaluated using standard statistical probability rules.31 As described above, we also submitted 55 duplicate samples collected during the weekly urine sample processing. The relative percent difference (RPD) between these paired sets was 5.3%, indicating excellent reproducibility and precision. The analysis of de-identified specimens at the CDC laboratory was determined not to constitute human subject research. Assessment of Residential Proximity to Agriculture We collected participants’ residential addresses at baseline and recorded whether they moved or lived in additional locations during the study. We geocoded all addresses and physically verified the existence and location of all actively cultivated agricultural fields within a 0.5-km radius of each address during the growing season concurrent with sample collection (August 2021); if no potential fields were present within this radius, we expanded to a 1-km buffer. As we described in detail elsewhere,32 we entered each address into Google Earth and identified any area within the 0.5-km buffer that could potentially be an agricultural field (defined as areas of green or brown that did not contain homes or other structures). We drove to each of the locations and visually inspected it to identify whether it was an agricultural field in current cultivation. If the location was not accessible (e.g., if it was surrounded by private property), we used a combination of National Land Cover Database (NLCD) and Landsat imagery paired with Normalized Difference Vegetative Index (NDVI) thresholds to determine whether it was an agricultural field likely to be in current cultivation, using the method described by Hyland et al.32 We then created polygons identifying each actively cultivated agricultural field in ArcGIS and calculated the distance from each residence to the nearest such field (up to 1km). We also calculated total agricultural acreage within a 0.5-km buffer. Data Analysis We imputed glyphosate concentrations below the LOD as the LOD/√2.33 We corrected glyphosate concentrations for hydration by adjusting for specific gravity according to: CSG=C×1.016−1SG−1, where CSG is the adjusted result (μg/L), C is the reported glyphosate concentration (μg/L), 1.016 is the mean specific gravity measured within the study population, and SG is the specific gravity of the individual sample.34 All results presented in this manuscript represent these SG-corrected concentrations. We evaluated the normality of the data set with the Kolmogorov-Smirnov test and found that the glyphosate concentrations were not normally distributed but were right-skewed with long tails both within and between participants. Therefore, for each participant, we estimated the central tendency of her individual exposure as the GM of all her samples (“overall exposure”; average per participant=11 samples), and the GMs of samples collected during the pesticide spray season (“spray season exposure”; average per participant=5 samples) and the nonspray season (“nonspray season exposure”; average per participant=6 samples). These individual GMs represent the central tendencies of the within-participant distributions of glyphosate concentrations; thus, we consider them to be estimates of each participant’s glyphosate exposure. These overall exposure estimates were also not normally distributed (n=40, p=0.012). Therefore, to characterize the distribution of estimated glyphosate exposures in the cohort (i.e., the between-participant distribution), we used these exposure estimates to calculate the cohort GM and percentiles. These were also calculated stratified by spray season and residential proximity to agriculture. We evaluated the effect of spray season on glyphosate exposure with Wilcoxon signed rank tests for paired data; we evaluated the effect of residential proximity using Wilcoxon rank sum tests for independent data. As a sensitivity analysis, we evaluated the potential for confounding by age, income, education, race/ethnicity, occupation, and household size. These variables were selected because they are commonly collected demographics and because we hypothesized that they could be indicators of, or associated with, socioeconomic status or community identification, which could influence location of residence. First, we ran independent t-tests to evaluate the association between residential proximity and each of these variables. We also conducted a multivariable regression analysis to evaluate the relationship between glyphosate concentration during the spray season and residential proximity to agriculture adjusting for these variables. Finally, we evaluated the relationship between glyphosate concentration and spray season among the near field participants and adjusted for these variables using generalized estimating equations (GEE) to account for the repeated measures within individuals. We conducted all analyses in SPSS Statistics [version 28.0.0.0; IBM]. Results Study Participants We enrolled our first participant on 23 February 2021 and completed enrollment on 3 June 2021. Weekly urine sample collection continued for each participant until she delivered her baby, with delivery dates occurring from 20 August through 28 December 2021. We collected the last weekly sample on 20 December (see Figure 1). Timeline of the study, including community engagement, recruitment, weekly urine sample collection, and dates of delivery. The spray season is also indicated. Table 1 shows the demographic characteristics of the study participants collectively and separately for those who live near and far from fields. Participants ranged in age from 18 to 37 y, with approximately half of the cohort (n=21) age 18–27 y. Gestational length at enrollment averaged 14.4 wk (SD=2.2 wk). One participant identified as Asian and one as American Indian or Alaskan native; the rest of the cohort identified as either White (52.5%, n=21) or Hispanic/Latina (47.5%, n=19), with two participants identifying as both White and Hispanic/Latina. We ran these analyses categorizing these two participants first in one category and then in the other; findings were unchanged. For the results presented here, we included them in the Hispanic/Latina category. Most study participants reported graduating high school or earning their general equivalency diploma (35%, n=14) or completing some college (47.5%, n=19) as their highest level of education. Two-thirds of the participants had a combined household income of USD $39,999 or less. Approximately half the cohort (52.5%, n=21) reported working outside the home, 16 (40%) described themselves as “stay-at-home moms,” and 3 (7.5%) were unemployed. We did not observe any differences between the near and far from field participants in any of these variables (independent samples t-test: age, p=0.83; income, p=0.35; education, p=0.59; race/ethnicity, p=0.10; occupation, p=0.57; household size, p=0.78). No participants reported applying herbicides during their pregnancies (Table S1). Table 1Demographic characteristics of 40 pregnant people participating in a study of the association between residential proximity to agricultural and glyphosate concentrations. Twenty-two (22) participants lived <0.5km of an agricultural field (“near field”), and 18 participants lived ≥0.5km of an agricultural field (“far from field”) in southern Idaho, 2021.Full sample (n=40) mean±SD or n (%)Near field participants (n=22) mean±SD or n (%)Far field participants (n=18) mean±SD or n (%)Age (y) 18–2211 (28%)5 (23%)6 (34%) 23–2710 (25%)7 (32%)3 (17%) 28–3214 (35%)7 (32%)7 (39%) 33–375 (13%)3 (14%)2 (11%) Gestational age at enrollment (wk)14.4 (2.3)14.2 (2.2)14.6 (2.4)Race and ethnicitya Caucasian or White21 (53%)11 (50%)10 (56%) Hispanic or Latina19 (48%)11 (50%)8 (45%) Asian1 (2.5%)1 (4.5%)0 (0%) American Indian or Alaskan Native1 (2.5%)1 (4.5%)0 (0%) African American or Black0 (0%)0 (0%)0 (0%)Highest level of education Less than high school5 (13%)3 (14%)2 (11%) Graduated high school or earned GED14 (35%)9 (41%)5 (28%) Some college19 (48%)9 (41%)10 (56%) Bachelor’s degree1 (2.5%)0 (0%)1 (5.5%) Graduate degree or other advanced degree1 (2.5%)1 (4.5%)0 (0%)Income [last year, household (USD)] <$10,000–$19,9999 (24%)4 (20%)5 (28%) $20,000–$29,99910 (26%)6 (30%)4 (22%) $30,000–$39,9998 (21%)3 (15%)5 (28%) $40,000–$59,9996 (16%)3 (15%)3 (17%) ≥$60,0005 (13%)4 (20%)1 (5.5%) Missing or declined to answer220Employment “Stay-at-home mom”16 (40%)10 (46%)6 (34%) Working outside the home21 (53%)11 (50%)10 (56%) Unemployed3 (7.5%)1 (4.5%)2 (11%)Number living in household 1–315 (38%)9 (41%)6 (34%) 4–623 (58%)11 (50%)12 (67%) 7–92 (5%)2 (9%)0 (0%)Current tobacco smoker or vape user Yes2 (5%)2 (9%)0 (0%) No38 (95%)20 (91%)18 (100%)Note: GED, general equivalency diploma; SD, standard deviation; USD, U.S. dollars. a Participants could choose more than one option. Demographic characteristics of 40 pregnant people participating in a study of the association between residential proximity to agricultural and glyphosate concentrations. Twenty-two (22) participants lived <0.5km of an agricultural field (“near field”), and 18 participants lived ≥0.5km of an agricultural field (“far from field”) in southern Idaho, 2021. Note: GED, general equivalency diploma; SD, standard deviation; USD, U.S. dollars. Participants could choose more than one option. Residential Proximity to Agriculture Thirty-one participants (78%) had only one residence throughout the study period. No participant lived in more than two locations during the study. For seven of the nine participants who had two residences during the study, both locations were either near fields or far from fields. The other two participants were assigned to the location where they spent most of their time (one near field and one far from field). As shown in Table 2, 22 (55%) of the study participants lived near fields, whereas 18 (45%) lived far from fields. Of the 22 participants who lived near fields, 3 lived within 100m of a field, 9 lived 100 to<300m from a field, and 10 lived 300 to<500m from the nearest field. Of the 18 participants who lived far from fields, 4 lived 500–1,000m from a field and 14 lived more than 1,000m from the nearest field. By definition, none of the participants who lived far from a field had any cultivated acreage within a 0.5-km radius of their homes. Of the participants who lived near fields, approximately one-third had more than 25 acres, 10–25 acres, and <10 acres each within a 0.5-km radius of their homes. Alfalfa was the most common crop nearest to participants’ homes (35%), followed by corn (15%). Table 2This table describes residential proximity to agriculture for the 40 participants in this study, including the number living “near” vs. “far from” fields (dichotomized at 0.5km) in southern Idaho, 2021. n %Cultivated field <0.5km Yes (“Near field”)2255% No (“Far from field”)1845%Distance from nearest field <100m38% 100 to <300m923% 300 to <500m1025% 500 to <1,000m410% ≥1,000m1435%Total cultivated acreage <0.5km from residence 25 + acres718% 10 to <25 acres820% >1 to <10 acres718% 0 acres1845%Crop type closest to residence Alfalfa1435% Corn615% Dry beans25% Wheat13% Mint13% Onions13% Sugar beet13% None1435% This table describes residential proximity to agriculture for the 40 participants in this study, including the number living “near” vs. “far from” fields (dichotomized at 0.5km) in southern Idaho, 2021. Sample Collection We collected a total of 864 weekly urine samples, averaging 21.6 samples per participant and representing 98% of our study goal. Thirty-eight (95%) of the study participants were enrolled in the study until they delivered their babies. Of the remaining two, one participant developed pregnancy complications and relocated out of state, where she remained until she delivered her baby, and one participant experienced housing instability and was unable to continue in the study; both were lost to follow-up in mid-July. Samples collected from these participants were included in data analysis. As described in the “Methods” section, we submitted biweekly samples from each participant for analysis; this process ultimately resulted in the selection of 453 samples (52% of all collected). Figure 2 illustrates the timing and density of sample collections, indicates when samples were missed, and shows which samples were analyzed for glyphosate concentrations. Weekly urine sample collections by participant among 40 pregnant people in southern Idaho. Darker circles indicate that a sample was collected and analyzed for glyphosate concentration; lighter circles indicate a sample that was collected but not analyzed; and circles containing an “x” indicate a missed sample. The y-axis shows Participant IDs, ordered chronologically by date of first sample. For the most part, these IDs were assigned in numerical order, but some were assigned out of order because of logistical or scheduling issues. Glyphosate Concentrations, Residential Proximity to Agriculture, and Spray Season Two-thirds (66%) of all individual samples contained detectable concentrations (≥0.1μg/L) of glyphosate (Table 3). The detection frequency across all samples was similar among participants living near and far from fields (67% and 65%). Estimated individual-level glyphosate exposure, calculated as the GM glyphosate concentrations for each participant, ranged from