Glyphosate exposure and preterm birth: a nested case-control pilot study

Abstract

Preterm birth (PTB) is associated with a high risk of infant mortality and long-term adverse health effects. Glyphosate is a broad-spectrum herbicide applied in agricultural and non-agricultural settings. Studies suggested an association between maternal exposure to glyphosate and PTB among mostly racially homogenous populations, though results were inconsistent. The objective of this pilot study was to inform the design of a larger and more definitive study of glyphosate exposure and adverse birth outcomes in a racially-diverse population. Urine was obtained from 26 women with a PTB as cases and 26 women with a term birth as controls, from participants enrolled in a birth cohort in Charleston, South Carolina. We used binomial logistic regression to estimate associations between urinary glyphosate and the odds of PTB, and multinomial regression to estimate associations between maternal racial identity and urinary glyphosate among controls. Glyphosate was unrelated to PTB (odds ratio (OR) = 1.06, 95% CI: 0.61, 1.86). Women who identified as Black had greater odds (OR = 3.83, 95% CI: 0.13, 111.33) of having categorical “high” glyphosate (> 0.28 ng/mL) and lesser odds (OR = 0.79, 95% CI: 0.05, 12.21) of “low” glyphosate (< 0.03 ng/mL) relative to women who identified as white, suggesting a potential racial disparity, although the effect estimates were imprecise and included the null. Given concerns of potential reproductive toxicity of glyphosate, the results merit confirmation in a larger investigation to determine specific sources of glyphosate exposure, incorporating longitudinal urinary glyphosate measures during pregnancy and a comprehensive measure of diet.

1. Introduction

Preterm birth is associated with an increased risk of infant mortality and morbidity [1]. In a multi-country analysis of nearly 2 million live births, infants born at less than 32 weeks gestation were approximately 29 times more likely to die in the first 28 postnatal days than infants born at term [2]. A U.S. study of almost 3.5 million births found that infants born at 34–36 weeks gestation experienced significantly greater neonatal (3.7 to 9.5-fold) and infant (2.6 to 5.4-fold) mortality risks compared to term births [3]. Infants born preterm also experience greater risks for infections [4], sepsis, respiratory distress, intraventricular hemorrhage, hypoglycemia, feeding difficulties, necrotizing enterocolitis, hypothermia, and hyperbilirubinemia than their term counterparts [5]. Furthermore, preterm infants have greater long-term risks for neurodevelopmental delays, hypertension, cardiovascular disease, renal disease, and pulmonary impairment [6].

In addition to socioeconomic status and racial/ethnic identity [7], exposure to chemical pollutants, including glyphosate, may be associated with an increased risk of preterm birth [8]. Glyphosate is used to control broadleaf weeds and grasses and is the active ingredient in several commercial and agricultural herbicides including Roundup^® [9]. It is applied in agricultural and non-agricultural settings and can be detected in the soil, air, and surface and ground waters [10]. Glyphosate use grew dramatically over the past 40 years in the U.S., from 635,000 kg used in 1974 to 125,384,000 kg used in 2014 [11], increasing the likelihood of human exposure. Based primarily on in vitro studies, glyphosate has been proposed to possess anti-estrogenic activity [12], raising concerns for potential human reproductive toxicity [13]. Glyphosate inhibited cytochrome P450 aromatase activity in human placental JEG3 and embryonic cell lines at levels lower than those recommended for use in agriculture, and the effect was potentiated in the presence of adjuvants found in the commercial Roundup^® formulation [14,15].

In women with a pregnancy, glyphosate was detected in maternal and umbilical cord serum [16] and urine [17], and glyphosate exposure may differ among women with different racial identities. A 2003–2005 study conducted in an agricultural area of North Carolina found that among women with a pregnancy, 59% of women who identified as white were exposed to glyphosate applied to nearby crops compared to only 22% of women who identified as Black [18]. However, median urinary glyphosate concentrations were modestly greater among U.S. women 18–44 years of age who identified as non-Hispanic Black (0.47 ng/mL) in 2013–2014, compared to women who identified as non-Hispanic white (0.36 ng/mL) [19]. Diet is a common source of glyphosate exposure in the general population, via ingestion of residues on fruits and vegetables [20] and in processed food obtained from genetically modified organism (GMO) crops such as corn, wheat, and soybean [21]. Agricultural use is an important source of occupational glyphosate exposure, as farmers had higher urinary glyphosate concentrations than their spouses, who had glyphosate levels ranging from <1 part per billion (ppb) to 1 ppb [22]. Women residing in closer proximity to glyphosate crop applications had greater urinary glyphosate concentrations than women residing farther away [16], demonstrating that levels of glyphosate exposure may also vary based on proximity to its agricultural use.

Observational studies have suggested associations between glyphosate exposure and preterm birth among human populations, although the results have been inconsistent, and were mostly limited to racially homogenous populations. Greater glyphosate exposure was associated with an increased odds of preterm birth among women in Puerto Rico [23]. Greater prenatal glyphosate exposure was also correlated with shorter gestational length among women in urban and rural areas of the U.S. [17,24]. However, studies of women from agricultural regions of Missouri and California showed no association between glyphosate exposure and preterm birth [25,26]. Still, U.S. study populations overwhelmingly comprised women who identified racially as white, potentially limiting generalizability of the results to a more diverse population including women of color, who experience an increased risk of preterm birth [27,28]

To the best of our knowledge no studies published to date have investigated an association between glyphosate exposure and preterm birth in a racially-diverse southeastern U.S. study population. Therefore, the objectives of this hypothesis-generating pilot study were to estimate the association between glyphosate exposure and preterm birth, and to estimate a potential disparity in glyphosate exposure between women who identified as Black or white. These results will help to inform the design of a larger and a more definitive future study of race disparities of glyphosate exposure and adverse birth outcomes.

2. Materials and Methods

2.1. Study population

The women in this study were selected from participants in the larger Reproductive Development Study, a prospective birth cohort study of n = 319 women with a pregnancy and residing in the Charleston, South Carolina metropolitan area between 2011–2014 [29,30]. The current analysis includes women with a preterm birth (< 37 weeks gestation) selected as cases (n = 26) and randomly selected women with a term birth (> 37 weeks gestation) as controls (n = 26), frequency matched to cases on racial identity (self-identified). Eligible women were at least 18 years of age, had a fetal ultrasound between 18–22 weeks gestation, had a singleton fetus, and planned to deliver at the Medical University of South Carolina (MUSC). Women were excluded if they did not have a first trimester ultrasound, had a fetus with anomalies or aneuploidy, used progesterone or any other steroids, and had a known endocrine disorder. At the time of enrollment (18–22 weeks gestation), women provided a urine specimen for analysis of environmental chemicals, completed a study questionnaire to collect sociodemographic information, and had height and weight measured to calculate body mass index (BMI, kg/m²). Women provided information about their daily food and beverage consumption habits as part of the study questionnaire, both “in general” and “since learning of pregnancy.” Clinical data were extracted from electronic medical records. All participants provided informed consent prior to enrollment and the study protocol was approved by the Institutional Review Board of the MUSC.

2.2. Glyphosate analysis

Urine samples were processed and stored at −70 °C until shipment to the University of California at San Francisco Clinical Toxicology and Environmental Biomonitoring Laboratory for glyphosate analysis early in 2020. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) with an Agilent LC 1260-AB Sciex 5500 Triple Quadrupole mass spectrometer (Agilent Technologies, Inc., St. Clara, CA USA; AB Sciex LLC, Framingham, MA USA) was employed to quantify urinary glyphosate using the standard addition method [17]. The limits of detection (LOD) and quantification were 0.03 ng/mL. In this analysis, we imputed values measured below the detection limit as $\text{LOD}/\sqrt{2}$ [31]. Urinary specific gravity was measured using a handheld digital refractometer (Atago USA, Inc., Bellevue, WA USA) prior to freezing. We corrected urinary glyphosate concentrations for dilution using the formula, G_c = G((1.016–1)/(SG-1)), where G is the individual glyphosate concentration (ng/mL), 1.016 is the average urine specific gravity for all women, and SG is the individual sample specific gravity [32].

2.3. Confounding variables

Potential confounding factors, identified as previously reported predictors of glyphosate exposure and preterm birth, were determined a priori based on the literature [33,34]. Potential confounding factors included maternal age at enrollment (years), maternal BMI at enrollment (kg/m²) [35], self-identified maternal race (Black including African-American or white) [28], maternal education as a dichotomous variable (“less than college” or “college degree or more”) [36,37], and season of specimen collection as a categorical variable (fall, spring, summer, or winter). We incorporated the potential confounders into a directed acyclic graph using DAGitty 3.0 software [38], which describes our proposed causal framework as shown in Figure 1. We retained maternal race, education, age, BMI, and season of urine collection for adjustment in regression models, based on the “back door test” for confounding pathways [39].

Figure 1.

Directed acyclic graph showing hypothesized causal pathways between urinary glyphosate and preterm birth.

2.4. Statistical analysis

First, we analyzed the distributions of all covariates, and conducted Student’s t-tests, chi-square tests, Fisher’s exact tests, or Mann-Whitney U tests to examine differences in covariate values between case women with preterm births and control women with term births. Among control women with a term birth, we estimated pairwise spearman correlations between the specific gravity corrected urinary glyphosate concentration and each self-reported food and beverage consumption habit from the study questionnaire data. We categorized urinary glyphosate concentration as “low” values below the detection limit (< 0.03 ng/mL), and as “medium” (0.03 < × < 0.28 ng/mL) and “high” (≥ 0.28 ng/mL) values dichotomized at the median concentration of the detected values.

We used binomial logistic regression to estimate the association between uncorrected urinary glyphosate concentration and the odds of preterm birth, adjusted for urinary specific gravity as a covariate to accommodate urinary dilution [40], and self-identified maternal race to accommodate the matching factor [41]. We repeated the binomial logistic regression model, with additional adjustment for maternal education, age, BMI, and season of urine specimen collection as confounding variables, using the Multiple Imputation Chained Equation (MICE) package in R statistical software [42] to account for n = 4 women with missing covariates. We exponentiated the regression coefficients to express the associations as odds ratios (OR) scaled to an interquartile range (IQR) difference in measured urinary glyphosate, to accommodate the low exposure levels in our study population. We also conducted an exploratory analysis of urinary glyphosate concentration categories as the outcome predicted by maternal racial identity, adjusted for maternal age and BMI using multinomial logistic regression with “medium” glyphosate exposure as the reference category. We used Dfbetas to identify and examine influential observations. Consistent with the hypothesis-generating nature of our study, statistical significance was defined as p < 0.10 for a two-tailed hypothesis test. Data were analyzed using R software (version 4.1.2, R Foundation for Statistical Computing, Vienna, Austria).

3. Results

3.1. Demographic characteristics and urinary glyphosate concentrations

There were 26 case women with preterm births and 26 control women with term births as shown in Table 1. The women with preterm births tended to be younger and have a greater BMI at the time of enrollment than the women with term births, however the differences were not statistically significant. For women with preterm births, 38.5% had a college education level or higher compared to 57.7% of women with term births. During the winter season 34.6% of urinary specimens were collected from preterm births compared to 19.2% among term births. Most preterm infants were assigned male sex at birth (73.1%), whereas most term infants were assigned female sex at birth (57.7%). There were 42.3% (n=22) of urinary glyphosate values measured above the detection limit, of which 11 experienced preterm birth and 11 delivered at term.

Table 1.

Demographic characteristics of women with a pregnancy from Charleston, South Carolina, among case women with preterm births and control women with term births.

Covariate	Preterm births (n = 26)	Term births (n = 26)	p-value
Age, years (mean ± SD)	27.4 ± 4.9	28.9 ± 5.8	0.333a
BMI, kg/m2, (mean ± SD)	30.1 ± 7.9	27.0 ± 6.1	0.129a
Race, n (%)			1.000b
White	13 (50.0)	13 (50.0)	-
Black	13 (50.0)	13 (50.0)	-
Education, n (%)			0.145b
< College	15 (57.7)	8 (30.8)	-
> College	10 (38.5)	15 (57.7)	-
Missing	1 (3.8)	3 (11.5)	-
Season of collection, n (%)			0.613b
Spring	5 (19.2)	7 (26.9)	-
Summer	7 (26.9)	7 (26.9)	-
Fall	5 (19.2)	7 (26.9)	-
Winter	9 (34.6)	5 (19.2)	-
Smoking status, n (%)			1.000c
Never smoked	25 (96.2)	24 (92.3)	-
Current smokerd	1 (3.8)	2 (7.7)	-
Infant sex, n (%)			0.049b
Male	19 (73.1)	11 (42.3)	-
Female	7 (26.9)	15 (57.7)	-
Specific gravity corrected urinary glyphosate concentration, ng/mL (median ± IQR)	0.06 ± 0.35	0.05 ± 0.15	0.634e

NOTE:

^aStudent’s t-test;

^bChi-Square test of independence;

^cFisher’s exact test;

^dcurrent smoker or quit since learning of pregnancy;

^eMann Whitney U test

Abbreviations: BMI, body mass index; IQR, interquartile range; SD, standard deviation

Figure 2 shows the distribution of specific gravity-corrected urinary glyphosate concentrations among control women with term births and case women with preterm births. There was a median difference of 0.01 ng/mL (p = 0.634) urinary glyphosate concentration between women with a preterm birth and women with a term birth. Among women with term births, those who identified as Black and white had similar median urinary glyphosate concentrations (median difference = 0.01 ng/mL; p = 0.858). However, a disproportionate number of women who identified as Black were categorized as “high” (19.2%) and “low” (23.1%) urinary glyphosate concentrations relative to women who identified as white (3.8% and 34.6%, respectively), using categorical glyphosate exposure (Figure 3).

Figure 2.

Boxplots of the distributions of specific gravity-corrected urinary glyphosate concentration between case women with a preterm birth (n = 26) and control women with a term birth (n = 26) from Charleston, South Carolina.

Figure 3.

Categories of urinary glyphosate concentrations among control women with a term birth, by self-identified maternal race (n = 26) from Charleston, South Carolina.

3.2. Correlations between urinary glyphosate concentration and food and beverage consumption habits

Table 2 shows the associations between sociodemographic factors and self-reported food and beverage consumption habits and specific gravity-corrected urinary glyphosate concentrations, among control women with a term birth. Women who smoked cigarettes (median difference = −0.29, p = 0.962) and women with female fetuses (median difference = −0.06, p = 0.011) tended to have lower urinary glyphosate concentrations than non-smokers and women with male fetuses, respectively. There were also positive correlations, albeit non-statistically significant, between urinary glyphosate concentrations and consuming organic, eco-friendly, chemical-free, or environmentally-friendly food in general (r = 0.32, p = 0.135), consuming foods and beverages in plastics they thought were safe in general (r = 0.29, p = 0.176), consuming foods stored in a clear, unbreakable, plastic container in general (r = 0.30, p = 0.161), microwaving food in a plastic container in general (r = 0.25, p = 0.256), and consuming fresh fruits or vegetables since learning of pregnancy (r = 0.33, p = 0.131).

Table 2.

Spearman correlation coefficients and differences of medians for sociodemographic factors and food and beverage consumption habits with specific gravity corrected urinary glyphosate concentrations, among control women with a term birth from Charleston, South Carolina (n = 26).

Covariate/Exposure	Correlation Coefficient or Difference of Medians	p-value
Age	−0.05	0.800
BMI	−0.11	0.579
Educationa	0.02b	0.539
Season of urine specimen collection c	−0.02b	0.569
Smoking statusd	−0.29b	0.962
Infant sexe	−0.06b	0.011
Food and Beverage Consumption Habits
“In general, how often did you …”:
Consume organic, eco-friendly, chemical-free, or environmentally-friendly food	0.32	0.135
Food and beverages in safe plastics	0.29	0.176
Drink water from a soft, crushable plastic bottle	−0.06	0.788
Drink water from a hard, reusable plastic bottle	0.11	0.630
Consume foods stored in a clear, unbreakable, plastic	0.30	0161
container
Microwave food in a plastic container	0.25	0.256
Caffeinate coffee in the past week?	−0.02b	0.911
Caffeinate tea in the past week?	−0.02b	0.423
Caffeinate soft drinks or sodas in the past week?	−0.01b	0.598
“Since learning of pregnancy, how often did you …”:
Consume organic, pesticide-free, chemical-free food	0.13	0.555
Consume food grown by family/friends	0.10	0.649
Consume unprocessed food	0.11	0.608
Consume canned fruits or vegetables	0.23	0.281
Consume frozen fruits or vegetables	0.03	0.909
Consume fresh fruits or vegetables	0.33	0.131
In a typical week consume food from a can	0.18	0.420
In a typical week drink from a can	0.12	0.600
In a typical week drink from a plastic bottle	0.24	0.270

NOTE:

^aEducation (< college / > college);

^bdifference of medians;

^cdifference of medians between winter and spring;

^dsmoking status (never smoked / current smoker or quit since learning of pregnancy);

^einfant sex assigned at birth (male / female)

Abbreviation: BMI, body mass index

Supplemental Table 1 shows the frequency of self-reported food and beverage consumption habit questionnaire responses according to case status. Among case women with preterm births, 12 (46.2%) “never” consumed organic, eco-friendly, chemical-free, or environmentally-friendly food, whereas 15 (57.7%) control women with term births “always,” “usually,” or “sometimes” did. On a daily basis, 11 (42.3%) women with preterm births drank water from a soft, crushable plastic bottle compared to six (23.1%) women with term births. Similarly, 11 (42.3%) women with preterm births “always,” “usually,” or “sometimes” consumed organic, pesticide-free, and chemical-free food since learning of pregnancy, whereas 15 (57.7%) women with term births did. Since learning of pregnancy, 20 (76.9%) women with preterm births “always” or “usually” consumed fresh fruits or vegetables, compared to 18 (69.2%) women with term births who did.

3.3. Association between urinary glyphosate and preterm birth

Table 3 shows the estimated associations between urinary glyphosate concentrations and preterm birth from binomial logistic regression models. Model 1 shows that an IQR difference in urinary glyphosate concentration was associated with minimally greater odds of preterm birth, (OR = 1.13, 95% confidence interval (CI): 0.72, 1.83), adjusted for urinary specific gravity and maternal race, though the confidence interval included the null hypothesis. However, the association was close to the null hypothesis (OR = 1.06; 95% CI: 0.61, 1.86) when additionally adjusted for confounding by maternal race, education, age, BMI, and season of urine specimen collection in Model 2.

Table 3.

Associations between urinary glyphosate concentration and preterm birth among women with a pregnancy from Charleston, South Carolina (n = 52).

	Model 1		Model 2
	OR	95% CI	OR	95% CI
Urinary glyphosate concentration (ng/mL)a	1.13	0.72, 1.83	1.06	0.61, 1.86
Race
White	REF	-	REF	-
Black	1.02	0.30, 3.52	0.40	0.08, 2.04
Education
≥ College	-	-	REF	-
> College	-	-	0.31	0.06, 1.61
Age (years)	-	-	0.94	0.82, 1.08
BMI (kg/m2)	-	-	1.06	0.96, 1.18
Season of collection
Fall	-	-	REF	-
Winter	-	-	3.55	0.53, 23.76
Spring	-	-	1.16	0.18, 7.61
Summer	-	-	1.42	0.23, 8.81

NOTE: Binomial logistic regression models adjusted for urinary specific gravity;

^aGlyphosate expressed per interquartile range difference

Abbreviation: BMI, body mass index; CI, confidence interval; OR, odds ratio; REF, reference group

3.4. Associations between urinary glyphosate and maternal racial identity

Table 4 shows the estimated associations between maternal racial identity and categorized urinary glyphosate concentration among 26 control women with a term birth, adjusted for urinary specific gravity. Both the crude (Model 1) and adjusted (Model 2) results show that women who identified as Black had a greater odds of “high” categorical urinary glyphosate. Model 2 shows that women who identified as Black had a 3.83-fold greater odds (95% CI: 0.13, 111.33) of “high” categorical urinary glyphosate and a 0.79-fold lesser odds (95% CI: 0.05, 12.21) of “low” categorical urinary glyphosate compared to women who identified as white, adjusted for maternal age and BMI. Still, the effect estimates were imprecise as suggested by the large 95% confidence intervals, which included the null hypothesis of no association. Therefore, this preliminary result requires confirmation in a larger investigation.

Table 4.

Associations between self-identified maternal race and categorical urinary glyphosate, among women with a term birth from Charleston, South Carolina (n = 26).

	Model 1				Model 2
	High		Low		High		Low
	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
Race
White	REF	-	REF	-	REF	-	REF	-
Black	3.91	0.12, 132.21	0.67	0.04, 10.49	3.83	0.13, 111.33	0.79	0.05, 12.21
Age (years)	-	-	-	-	0.89	0.70, 1.14	0.93	0.76, 1.14
BMI (kg/m2)	-	-	-	-	0.93	0.74, 1.16	0.94	0.76, 1.15

NOTE: Multinomial logistic regression models adjusted for urinary specific gravity; “Medium” urinary glyphosate category was the reference level

Abbreviation: BMI, body mass index; CI, confidence interval; OR, odds ratio; REF, reference group

4. Discussion

4.1. Key findings

In this hypothesis-generating nested case-control study, we did not find evidence to suggest an association between urinary glyphosate concentration and the odds of preterm birth. However, urinary glyphosate levels were low among our study population relative to previous studies and women in the general U.S. population. Among control women with term births, there were positive correlations between fruit and vegetable consumption habits and urinary glyphosate concentration. Thus, our results suggest a potential racial disparity in glyphosate exposure among women who identified as Black compared to women who identified as white, although the results are imprecise and require confirmation.

4.2. Glyphosate exposure

Urinary glyphosate levels in our study were lower than reported for the general U.S. population [19]. U.S. women, 18–44 years of age, who self-identified as Black or white race, had a median urinary glyphosate concentration of 0.40 ng/mL (95% CI: 0.36, 0.44) in 2013–2014 [19], compared to a median urinary glyphosate of 0.02 ng/mL (95% CI: 0.02, 0.07) among control women with a term birth in our study population. This difference may in part reflect our primarily urban study population recruited from the Charleston, South Carolina Metropolitan Area, with a large proportion of values measured below the detection limit, whereas women from rural areas, who tended to have greater glyphosate exposure than women from urban areas [17], may be included in the U.S. data. Still, The Infant Development and Environment Study (TIDES) also reported a greater median second trimester specific gravity corrected urinary glyphosate concentration of 0.23 ng/mL among n = 94 women with term births from San Francisco, California, Rochester, New York, Minneapolis, Minnesota, and Seattle, Washington [24], than in our study. The Puerto Rico Testsite for Exploring Contamination Threats (PROTECT) study reported a median of 0.52 ng/mL second trimester urinary glyphosate among n = 142 women with a term birth and residing in the Northern Karst aquifer region of Puerto Rico [23]. The reasons for the discrepancy in urinary glyphosate levels across studies are unclear, but may be related to our limited sample comprised only of women with low risk pregnancies using prenatal care in the Charleston, SC Metropolitan Area. Median urinary glyphosate levels were much higher, however, in a study of n = 71 women with a pregnancy from Central Indiana, with greater concentrations among women from rural areas (4.05 ng/mL) than among women from urban (3.41 ng/mL) and suburban (3.05 ng/mL) areas [17].

4.3. Association between urinary glyphosate and preterm birth

Similar to previous epidemiologic studies in general U.S. populations, we found little evidence of an association between urinary glyphosate concentration and preterm birth. Silver et al. [23] reported an OR of 1.11 (95% CI: 0.71, 1.74) per IQR greater urinary glyphosate concentration and preterm birth among 177 women (142 women with term births and 35 women with preterm births) in the PROTECT study, adjusted for maternal age, education, prepregnancy BMI, and smoking. Similarly, Lesseur et al. [24] found little association between increased glyphosate exposure and gestational age at delivery (hazard ratio = 1.08, 95% CI: 0.91, 1.29) for 163 women with a pregnancy as part of TIDES, adjusted for maternal age, race/ethnicity, and education level. In contrast, at substantially greater urinary glyphosate levels, Parvez et al. [17] reported a correlation between greater urinary glyphosate and shorter gestational length at delivery (correlation = −0.30, p = 0.01) among 71 women from Central Indiana, adjusted for maternal age, BMI, tobacco use, alcohol use, and trimester of pregnancy, although the effect size was very weak and within the range of full-term delivery [43].

The results of epidemiologic studies of glyphosate exposure and preterm birth among women residing in agricultural areas, where populations tend to experience greater levels of glyphosate exposure than in non-agricultural areas, were also mixed. For example, the South Carolina and Minnesota Farm Family Exposure Study reported urinary glyphosate concentrations ranges from <1 ppb to 15 ppb pre-application to <1 ppb to 126 ppb post-application among farmers using glyphosate to treat crops, and <1 ppb to 3 ppb pre-application to <1 ppb to 1 ppb post-application among their spouses [22], demonstrating greater glyphosate exposure among applicators compared to their spouses. There was no association between the risk of preterm birth and county-level density of some glyphosate-intensive crops, including corn (relative risk (RR) = 0.97, 95% CI: 0.94, 0.99), soybean (RR = 1.00, 95% CI: 0.98, 1.01), and wheat (RR = 1.01, 95%CI: 0.96, 1.06), although an increased risk for cotton (RR = 1.06, 95% CI: 1.03, 1.09) among 140,329 women with live births in rural areas of Missouri, adjusted for individual and county-level factors [25]. However, among 444,135 women with live births in California between 1998 and 2010, the odds of preterm birth were 1.04 (95%CI: 1.01, 1.07) times greater among women who lived within 2 km of glyphosate compounds being applied to agricultural areas compared to women residing farther away, adjusted for individual and neighborhood-level confounding factors [26].

4.4. Correlations between individual maternal food and beverage consumption habits and urinary glyphosate concentration.

We found positive, albeit imprecise associations between specific food and beverage consumption habits in general and since learning of pregnancy and urinary glyphosate concentrations. Maximum glyphosate residue limits in the U.S. vary for fruits (0.1–0.5 mg/kg), vegetables (0.1–5 mg/kg), legumes and pulses (5–8 mg/kg), and grains, including wheat (5–100 mg/kg) [44]. Although our study questionnaire did not refer to specific foods consumed, we found that consumption of fresh and canned fruits and vegetables since learning of pregnancy were correlated with greater urinary glyphosate concentrations, thus glyphosate exposure may be related to dietary habits in this non-agricultural study population. Given the recognized health benefits of fruit and vegetable consumption to mother and fetus during pregnancy [45], and ongoing concerns about the potential adverse health effects of exposure to glyphosate and other pesticides [46], these results merit a more comprehensive investigation to identify specific fruits, vegetables, legumes, pulses (i.e., beans, lentils, or peas inside the pods), and grains that may contribute to greater glyphosate exposure during pregnancy.

4.5. Associations between maternal racial identity and food and beverage consumption habits

In the current study, we found that a greater number of women who identified as Black were in the “high” urinary glyphosate category than women who identified as white, and fewer women who identified as Black were in the “low” urinary glyphosate category than women who identified as white. The categorical differences in urinary glyphosate concentrations were small and imprecise, so this result requires confirmation. Although our study did not look at specific dietary practices by racial identity, previous studies have shown differences in dietary habits among women with a pregnancy and different racial identities [47]. In our previous work, women who identified as Black reported greater consumption of canned fruits and vegetables since learning of pregnancy than women who identified as white [30]. In a study of 125 women with a pregnancy in North Carolina, women who identified as African-American consumed fewer whole grains but more fruit than women who identified as white, adjusted for age, education, planned or unplanned pregnancy, and marital status [48]. A study of 1151 women with a pregnancy from Tennessee found that 1% of those who identified as African-American had a diet consisting primarily of fruits and vegetables, compared to 30.2% of women who identified as European-American [49]. A study looking at dietary glyphosate exposure among different U.S. diet styles (i.e., healthy-U.S. style, healthy Mediterranean style, and healthy vegetarian style) found higher levels of glyphosate residues associated with specific foods, specifically legumes and grains, but not associated with a specific U.S. dietary style [50]. Differences in glyphosate residue on specific foods may be due to differences in the treatment of the product prior to consumption (e.g., washing and drying) [51]. These studies suggest that differences in consumption of glyphosate-intensive produce may contribute in part to the racial disparity in urinary glyphosate that we found, although we were unable to investigate the hypothesis in this pilot study. A future study with a comprehensive dietary assessment, including differentiating foods with and without GMOs, in a racially-diverse population of women with a pregnancy is necessary.

4.6. Study strengths and limitations

There were several important strengths to this hypothesis-generating study. We used a nested case-control design to estimate associations between urinary glyphosate in preterm birth cases and term birth controls selected from our larger birth cohort, ensuring that the controls were representative of the source population for cases and minimizing the likelihood of a selection bias. Our study population was also racially diverse, comprised equally of women who identified as Black and women who identified as white, allowing us to generate a preliminary estimate of the potential race disparity in glyphosate exposure. We employed clinically validated birth outcomes collected from electronic medical records, with dating of gestational age using first trimester ultrasound, minimizing outcome misclassification. Also, we employed a urinary biomarker to estimate glyphosate exposure during pregnancy, which incorporates toxicokinetic processes that are not captured by indirect exposure assessment strategies, such as using exposure questionnaires or residential proximity measures.

However, the pilot nature of our study also presents limitations. First, we had a small sample size. Therefore, we had limited power to detect modest associations and we were unable to conduct a stratified analysis according to maternal-racial identity. Also, we only had a single urinary biomarker of exposure to measure glyphosate, which has a half-life measured in hours [52], and thus may have misclassified exposure among some women. The reported intraclass correlation coefficient, which describes the expected correlation between two urinary glyphosate measures from the same woman, was only 0.24 (95% CI: 0.10, 0.46) in two urine specimens collected over eight weeks from women with a pregnancy in the PROTECT study [23]. These results demonstrate the low reliability of using a single urinary glyphosate sample to estimate gestational glyphosate exposure. However, urine specimens were collected from all women between 18–22 weeks gestation in the current study and so we anticipate a bias towards the null hypothesis (i.e., non-differential exposure measurement error between preterm birth cases and term birth controls). A future study with longitudinal urine specimen collections during pregnancy in a racially-diverse population will be necessary for more definitive results. Our study also did not measure urinary aminomethylphosphonic acid (AMPA), the primary environmental degradation product of glyphosate, which has also been shown to be associated with gestational age at delivery [23,24], and may have underestimated the overall glyphosate exposure. A future study should measure both urinary glyphosate and AMPA. There were a large proportion of glyphosate values measured below the detection limit in our study, although our detection limit was lower than in previous reports [23,24]. Therefore, these results may not be generalizable to populations with greater glyphosate exposures. Due to the small sample size, we also did not distinguish late (34–36 weeks) from early (< 34 weeks gestation) preterm birth, which may have different etiologies [53] and undermined our ability to detect an association with urinary glyphosate. Finally, these are preliminary data and suffer from limitations of a pilot study, however we are hopeful the results might warrant further investigation.

5. Conclusions

Overall, the results of this hypothesis-generating pilot study suggest that maternal urinary glyphosate concentration was not associated with preterm birth in a racially-diverse southeastern U.S. study population with low levels of urinary glyphosate. Our results also suggest that consumption of fruits and vegetable may contribute to glyphosate exposure among women with a pregnancy, and importantly that women who identified as Black may have experienced greater glyphosate exposure than women who identified as white. However, given our small sample size, low urinary glyphosate levels, and limited data collection, these results are preliminary and require confirmation. These results can be used to inform a larger study, incorporating longitudinal urinary glyphosate measures during pregnancy, a comprehensive dietary measure, and assessing sources of glyphosate other than dietary routes among racially-diverse populations.

Highlights.

• Preterm birth is associated with infant mortality and adverse health effects

• Glyphosate exposure has been suggested to be associated with preterm birth

• Black women had greater urinary glyphosate levels than white women during pregnancy

• No association between urinary glyphosate concentration and preterm birth