Prenatal maternal pesticide exposure in relation to sleep health of offspring during adolescence

Abstract

Study Objectives:

The neurobiological processes involved in establishing sleep regulation are vulnerable to environmental exposures as early as seven weeks of gestation. Studies have linked in utero pesticide exposure to childhood sleep-disordered breathing. However, the impact of in utero pesticide exposure on the sleep health of adolescents remains unexplored.

Materials and Methods:

Data from 137 mother-adolescent pairs from a Mexico City cohort were analyzed. We used maternal urinary 3-phenoxybenzoic acid (3-PBA, pyrethroid metabolite) and 3, 5, 6-trichloro-2-pyridinol (TCPy, chlorpyrifos metabolite) from trimester three to estimate in utero pesticide exposure. Among adolescents, we obtained repeated measures of objectively assessed sleep duration, midpoint, and fragmentation using wrist-actigraphy devices for 7 consecutive days in 2015 and 2017. Unstratified and sex-stratified associations between maternal urinary 3-PBA and TCPy and adolescent sleep measures were examined using linear mixed models. We also examined the interactive effects of maternal pesticide exposure and offspring sex on sleep outcomes.

Results:

3-PBA and TCPy were detected in 44.4% and 93% of urine samples, respectively. In adjusted models, we observed monotonic associations between TCPy with longer sleep duration and later sleep midpoint. Adjusted findings demonstrated that higher exposure to maternal TCPy was associated with longer sleep duration (p, trend = 0.01), and later sleep timing (p = 0.07). Findings from interaction tests between exposure and offspring sex were not statistically significant. Adjusted sex-stratified findings showed that the association between TCPy with duration and midpoint was evident only among female offspring; those in the highest tertile of exposure had a 59 minute (95% CI: 12.2, 104.8) (p, trend = 0.004) longer sleep duration and a 0.6 h (95% CI: −0.01, 1.3) (p, trend = 0.01) later sleep midpoint. We found no significant associations between 3-PBA and sleep outcomes.

Conclusion:

Although findings from interaction tests were not statistically significant, effect estimates demonstrated that associations between maternal pesticide exposure and longer sleep duration and later sleep timing were stronger in female offspring.

Introduction

The development of the neurobiological circuitry necessary for healthy sleep occurs throughout gestation. The pineal gland, which produces melatonin and is important in establishing sleep/wake patterns, develops in the first trimester (Seron- Ferre et al., 2007). In the second trimester, fetal circadian rhythms begin (Li et al., 2018; Seron- Ferre et al., 2007). By the third trimester, there are clear distinctions between rapid eye movement (REM) and non-rapid eye movement (NREM) (Okai et al., 1992). Thus, it is conceivable that prenatal maternal exposure to various environmental (chemical and non-chemical) and behavioral factors throughout pregnancy could impact offspring sleep in the infancy period and beyond. Indeed, maternal exposures, including sleep (Lyu et al., 2020), stress (Mrdalj et al., 2013), depression (Toffol et al., 2019), smoking (O’Callaghan et al., 2019), and dietary intake (Crossland et al., 2017) have all been linked to poor offspring sleep health.

Prenatal toxicant exposures have been widely studied for their potential developmental effects, particularly on neurological outcomes. However, only a few studies have examined the effects of prenatal toxicant exposure and offspring sleep (Bose et al., 2019; Kim et al., 2018). These studies have found that prenatal mercury and manganese were linked to sleep problems among offspring during childhood (Kim et al., 2018), and prenatal exposure to PM_2.5 was linked to altered sleep in preschool-aged offspring (Bose et al., 2019). Pesticides are a particular class of environmental toxicants widespread in the environment and especially pervasive in Latin America that can cause chronic and acute pesticide poisoning (Laborde Amalia et al., 2015). Among Latin American children, acute pesticide poisoning can occur from non-intentional ingestion of pesticides found in medicine or soft drink bottles or from pesticide containers reused for storing food or drinks (Laborde Amalia et al., 2015). Pyrethroids and chlorpyrifos (CPF) are two specific chemical classes of pesticides that are widespread (Klimowska et al., 2020). The exposure route to these pesticides is likely to occur via dietary intake, with pyrethroids residuals found in various foods, including grains, salmon, and fresh produce (Morgan, 2020), and CPF often found in fruits, vegetables, and nuts (Morgan et al., 2005).

Pesticides have been linked to poor sleep in animals (Crossland et al., 2017; Darwiche et al., 2018) and human studies (Serrano-Medina et al., 2019; Zhao et al., 2010). The negative impact of pesticide exposure among farmworkers has been well documented in the literature (Amr et al., 1997; Baumert et al., 2018; R. Kori et al., 2018; Serrano-Medina et al., 2019; Sultana et al., 2014). For example, a study among Chinese farmworkers showed that organophosphorus pesticide exposure was related to poor sleep quality (Zhao et al., 2010). Cross-sectional studies have also demonstrated links between pesticide exposure and trouble sleeping (i.e., insomnia) among farmworkers in Mexico (Serrano-Medina et al., 2019) and Bangladesh (Sultana et al., 2014); and sleep and memory disorders among North Carolina (Elmore and Arcury, 2001) and Indian farmworkers (R. K. Kori et al., 2018). Adding support to the relationship’s biological plausibility, one specific class of pesticides, carbamates, has been shown to interact with melatonin (sleep/wake hormone) in experimental studies (Popovska-Gorevski et al., 2017).

Prenatal pesticide exposure has also been associated with sleep-disordered breathing in offspring. Moreover, a longitudinal birth cohort study found that maternal prenatal exposure to organophosphate pesticides (i.e., chlorpyrifos, parathion, and diazinon, etc.) was associated with offspring respiratory symptoms, which encompassed trouble going to sleep or being awakened from sleep because of wheezing, whistling, shortness of breath, or coughing that was not associated with a cold (i.e., sleep-disordered breathing) (Raanan et al., 2015). Animal studies have also underlined the impacts of prenatal maternal pesticide exposure and sleep-disordered breathing symptoms (Darwiche et al., 2018; el Khayat el Sabbouri et al., 2019).

Given the correlation between sleep-related outcomes and neurodevelopment and cognitive outcomes (Engleman et al., 2000; Vaughn et al., 2015), studies on the association between pesticides and neurodevelopmental or cognitive outcomes may be of relevance. For example, research among non-occupationally pesticide-exposed mothers has provided evidence for motor and cognitive effects among human fetuses and infants exposed to chlorpyrifos (CPF) (Silver et al., 2017; Timchalk et al., 2007) and pyrethroids (Watkins et al., 2016). These studies also shed light on differential sex impacts of prenatal exposure to pesticides (Silver et al., 2017; Watkins et al., 2016). For example, one study found that maternal exposure to CPF was linked to motor function deficits among 9-month old Chinese infants, with the adverse associations being stronger among girls compared to boys (Silver et al., 2017). Another study found that maternal exposure to pyrethroids was linked to cognitive and behavioral deficits in their infants, with findings similarly demonstrating stronger negative associations among girls than boys (Watkins et al., 2016). Although unclear, it is possible that these sex differences are potentially driven by the interaction between sex-steroids and pesticides in relation to the outcomes under study. Several currently used pesticides with known endocrine-disrupting properties impair male reproductive functioning (Han et al., 2008). For example, a study conducted among a sample of non-occupationally exposed males found that urinary 3-PBA levels were associated with an increased level of luteinizing hormone and reduced estradiol, suggesting that pyrethroids may disrupt the male endocrine function (Han et al., 2008). Moreover, multiple studies have demonstrated the estrogenic (Jin et al., 2010) and anti-estrogenic effects (Du et al., 276 2010; Sun et al., 2014) of 3-PBA. Notably, sex steroids have also been shown to impact sleep in both males and females (Manber and Armitage, 1999). Further, there is growing evidence that males are more susceptible than females to pesticides’ adverse effects (Torres-Rojas and Jones, 2018).

Although animal and human studies have examined the links between pesticides and sleep-related health outcomes, as well as correlated cognitive/neurodevelopment outcomes, no studies have examined the impacts of prenatal pesticide exposure on other measures of human offspring sleep health, including sleep duration, timing, or fragmentation (a measure of sleep quality). To address the current gaps in the literature, we utilized the Early Life Exposures in Mexico to ENvironmental Toxicants (ELEMENT) cohort. We aimed to evaluate the associations between prenatal urinary concentrations of 3-phenoxybenzoic acid (3-PBA), a metabolite of pyrethroids exposure, and 3, 5, 6-trichloro-2-pyridinol (TCPy), a metabolite of CPF with subsequent repeated measures of sleep duration, timing, and fragmentation among adolescent offspring. Based on a priori knowledge, which demonstrated sex differences in the impacts of prenatal pesticide exposure and neurological outcomes (Silver et al., 2017; Watkins et al., 2016) and other toxicants with sleep measures (Jansen et al., 2020b), we decided to stratify by sex. Based on a priori knowledge, we hypothesized that higher exposure to CPF and 3-PBA would be associated with shorter sleep duration, later sleep timing, and more fragmented sleep, with findings being stronger among female offspring. A secondary aim was to evaluate associations between prenatal urinary concentrations of 3-PBA and TCPy with snoring and difficulty breathing while sleeping among adolescent offspring at a single time point. We hypothesized that higher exposure to maternal pesticides would be associated with difficulty breathing while sleeping and snoring among adolescent offspring and that both of these associations would be more robust among female offspring.

Materials and Methods

Study Sample

The analytic sample included mother-adolescent pairs (N = 137) from two sequentially enrolled birth cohorts of the ELEMENT Study (Perng et al., 2019) conducted in Mexico City, Mexico, between 1997 to 2005. Both cohorts enrolled low-to-moderate-income women from the Mexican Social Security Institute’s public maternity clinics. For the entire cohort, mothers provided second-morning void urine samples at each trimester of pregnancy. However, for the analytical sample (N=137) we only used urine samples collected during trimester three.

In 2015, a subset of 554 adolescent offspring (ages 9 to 17) participated in a follow-up visit called time 1 that included collecting sociodemographic, behavioral, and objective-sleep assessment (wrist-actigraphy) over seven consecutive days. In 2017, 519 (94%) of the same adolescents participated in an additional follow-up visit identical to the 2015 visit (time 2). Of the 519 adolescent participants who completed both follow-up visits, 137 had complete information on maternal measures (urinary pesticide exposure during trimester three and relevant maternal characteristics) and adolescent measures (objectively assessed sleep outcomes and relevant adolescent characteristics). Urinary data from trimesters 1 and 2 were not analyzed due to small sample sizes (pilot funding was available for only a limited analysis of pesticides). Compared to the overall sample of 519 adolescents, the analytic sample was older but did not differ on other demographic characteristics (See Supplemental Table 1). The Mexico National Institute of Public Health (INSP) Research, Ethics, and Biosafety Committees and the University of Michigan Human Subjects Committee approved all research protocols and procedures, and all participants provided informed consent.

Exposure Assessment of Maternal Pesticides

Maternal urine samples (2 mL) were separated into aliquots, frozen at −80 °C, and transported on dry ice to Emory University for processing using standard protocols. Samples were spiked with stable isotopically labeled 3-PBA and subjected to enzyme hydrolysis. Hydrolysates were extracted using mixed-polarity solid-phase extraction cartridges, and eluates were concentrated and analyzed using high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS) with both quantification and confirmation ions monitored. A matrix-based isotope dilution calibration was used for quantification (Ao et al., 2004). The limit of detection (LOD) for urinary 3-PBA was 0.25 ng/mL and 0.10 ng/mL for TCPy. Values below the LOD were assigned a value of LOD/√2. Urinary specific gravity (SG) was determined using a handheld digital refractometer (ATAGO Company Ltd., Tokyo, Japan). SG-corrected urinary 3-PBA and TCPy concentrations were calculated for use in specific statistical analyses.

Outcome Assessment of Sleep Measures

The sleep outcomes of interest were sleep duration, midpoint of sleep, and sleep fragmentation obtained from wrist-actigraphy devices (ActiGraph GT3X+; ActiGraph LLC, Pensacola, FL) that were worn by adolescent participants on the non-dominant wrist for seven consecutive days. Actigraphy devices were placed on the adolescent’s wrist by trained personnel at the end of time 1 and time 2 visits. Nightly sleep measures were estimated from the actigraphy data using a fused LASSO (least absolute shrinkage and selection operator)-based calculator package developed in R (R Foundation for Statistical Computing, Vienna, Austria) as previously described (Zamora et al., 2021). The fused LASSO approach incorporated the self-reported bedtime and wake time as part of its algorithm. For analysis, we obtained weekday (Sunday through Thursday) sleep duration (minutes), weekday midpoint of sleep (the median of sleep onset and wake time; reported in decimal hours), and average sleep fragmentation index during the weekdays. The sleep fragmentation index was calculated as the percentage of one-minute periods of sleep out of the total number of sleep bouts of any length, with higher values representing more fragmented sleep (Chung et al., 2016). The present study did not have issues with missing data, particularly because among adolescents participating in time 1 and time 2 visits, actigraphy devices were rarely removed since they had to be physically cut off in order to be removed. We found that <1% of all participants (N = 28/519) had removed their actigraphy devices, and these participants were not included in the analytic sample. Based on a priori knowledge, we decided to conduct analyses separately for weekdays versus weekends because sleep habits are known to vary, with adolescents having more freedom to choose their sleep schedule on the weekend(Jansen et al., 2020a). However, in this study, we only present findings for weekday sleep.

From the time 2 visit, we also obtained self-reported information on the presence of snoring (yes or no) and difficulty breathing while sleeping (yes or no). These variables were not collected during the time 1 visit.

Potential Confounders

We considered the following maternal and adolescent characteristics as potential confounders based on a priori knowledge maternal age, marital status, parity, maternal education, adolescent age, and sex. Maternal information reported at the initial enrollment visit included age, marital status, parity, years of education. Age was categorized into tertiles (16–23 years, 24–28 years, and 29–39 years). Marital status was classified as married or in a civil union or other (i.e., single, separated, divorced, or widowed). Maternal education was categorized as < 9 years, 9 to < 12 years, 12 years, and > 12 years. Parity was categorized as ≤ 1, 2, or ≥ 3. Maternal urinary specific gravity was also included as a covariate to adjust for urinary dilution. Adolescent age and sex were abstracted from a general questionnaire. Adolescent age was reported at the time 1 and time 2 visits and operationalized into tertiles (13–14 years, 15 years, and 16–19 years). We operationalized adolescent sex at birth as a binary variable (male or female).

Statistical Analysis

To assess crude associations between objectively assessed adolescent offspring sleep measures according to maternal and adolescent characteristics, we first created three continuous summary variables that were an average of each participant’s (1) sleep duration, (2) midpoint of sleep, and (3) sleep fragmentation index from both study visits. We then examined the mean (SD) for all three continuous sleep measures according to binary or categorical maternal and adolescent characteristics using the Wilcoxon-Mann-Whitney test. We calculated summary statistics (geometric means, standard errors, and percentiles) to measure maternal urinary pesticide exposure distribution. All pesticides were evaluated for non-normality and were log-transformed before subsequent regression analysis. In addition, we assessed the distribution of each sleep outcome at both time visits prior to regression analysis to assess for the presence of outliers. Next, we used chi-square (χ²) tests to determine differences in categorical maternal and adolescent characteristics between mothers who had detectable (detected group) levels of urinary 3-PBA. Because most of the sample had detectable levels of TCPy, we calculated median (Q1, Q3) levels of TCPy according to binary or categorical maternal and adolescent characteristics.

Generalized linear mixed models (LMMs) were used to evaluate the unadjusted and adjusted associations between tertiles of maternal urinary pesticide concentrations and adolescent sleep outcomes using a random intercept for subject ID to account for intra-individual correlation of repeated measurements over time. Model fit was assessed through examination of residual plots and residual diagnostics. TCPy was categorized into tertiles based on our previous work, which demonstrated that classifying TCPy concentrations into broad categories may be more robust to temporal within-person variability and measurement error (Meeker et al., 2005). Further, 3-PBA was categorized into tertiles with the lowest category of the three-level variable comprising all samples below the LOD (N =72); the middle and highest categories were formed by dividing the samples with detectable 3-PBA into two reasonably equal-sized groups (N = 32 for the middle category and N = 33 for the highest category). The rationale for this categorization has been described elsewhere (Watkins et al., 2016). We used LMMs separately for each maternal urinary pesticide concentration as the exposure (tertiles) and one sleep outcome (continuous) per model to explore the longitudinal relationships between maternal urinary pesticides and adolescent sleep outcome and adjusted for maternal education, marital status, parity, and urinary specific gravity and adolescent age and sex. Linear trend estimates for the categorical 3-PBA and TCPy measure were calculated by entering the 3-level ordinal exposure variables as continuous variables in the LMMs. In post-hoc analyses we created separate models to evaluate the association between maternal exposures and sleep measures at each time point. We also we created separate models to test for interaction between each maternal urinary pesticide and offspring sex. To examine potential effect modification by sex, we conducted additional unadjusted and adjusted LMMs separately for adolescent males and females using the same approach.

Finally, using adolescent data from only the T2 visit, we generated multivariable-adjusted logistic regression models in which the presence of snoring (yes or no) served as the outcome, and maternal urinary pesticide concentration based on level of detection (not detected vs. detected) were modeled as the exposures. We created similar separate logistic regression models to model the association between difficulty breathing while sleeping (yes or no) and continuous maternal urinary pesticide concentration of TCPy. Because the urinary pesticide concentration of TCPy was non-normally distributed, we log-transformed values to reduce the influence of outliers and presented estimates per interquartile range (IQR) concentrations in all regression models. We conducted overall and sex-stratified analysis for both models and adjusted for maternal urinary specific gravity and adolescent age group. Findings were considered statistically significant at p < 0.05 and marginally significant at p < 0.10. All statistical analyses were performed using SAS 9.4 software (Cary, NC, USA).

Results

Among adolescents, the mean (SD) age at time 1 was 13.9 (2.1) years, and 53.3% of the sample was female. The mean (SD) sleep duration was 8.4 (1.2) hours, the mean (SD) midpoint sleep and sleep fragmentation were 3.6 (1.3) decimal hours and 11.9 (3.6) %, respectively. The mean (SD) follow-up time between time 1 and time 2 was 1.3 (0.5) years. The mean for sleep duration and sleep fragmentation was not significantly different from time 1 and time 2 (p = 0.35 and 0.88, respectively). However, we found that at time 2, the mean (SD) for sleep midpoint was later than time 1 (4.2 [1.4] decimal hours in time 2 versus 3.6 [1.3] in time 1; p, difference <0.0001).

Maternal and adolescent characteristics according to adolescent sleep outcomes are presented in Table 1. Maternal age was the only characteristic significantly associated with sleep fragmentation (p = 0.04), with children of mothers in the middle age category (24–28 years) having the highest sleep fragmentation levels. Findings demonstrated a significant difference in sleep duration and midpoint sleep according to adolescent age. To illustrate, adolescents aged 15 years had a longer sleep duration than younger and older adolescent counterparts (p = 0.03). The 15 year-olds also had a later sleep midpoint compared to their counterparts (p = 0.05).

Table 1.

Maternal characteristics according to adolescent sleep measures (n = 137)

	N	Sleep Duration, minutesa Mean (SD)	Midpoint Sleep, decimal hourb Mean (SD)	Sleep Fragmentation, percentagec Mean (SD)
Maternal Characteristics
Maternal age (years)
16–23	46	502.7 (72.0)	3.9 (1.3)	11.8 (3.6)
24–28	46	503.2 (86.2)	4.1 (1.5)	12.7 (3.7)
29–39	45	512.4 (81.5)	4.0 (1.4)	11.4 (3.9)
p		0.47	0.84	0.04*
Marital Status
Married/Civil Union	122	508.1 (80.6)	4.0 (1.4)	12.0 (3.7)
Single, separated, divorced, or widowed	15	488.2 (72.5	3.8 (1.1)	11.6 (4.2)
p		0.16	0.80	0.69
Parity
≤ 1	48	496.8 (78.2)	3.7 (1.2)	11.4 (3.5)
2	49	501.5 (86.1)	4.1 (1.5)	11.9 (3.7)
≥ 3	40	521.9 (72.8)	4.1 (1.5)	12.7 (4.0)
p		0.13	0.14	0.13
Maternal Education (years)
< 9	16	530.6 (66.4)	4.2 (1.3)	12.4 (4.2)
9–11.9	21	497.1 (80.8)	4.1 (1.4)	12.2 (4.1)
12	43	496.0 (85.0)	3.7 (1.3)	11.2 (3.4)
>12	57	511.0 (78.0)	4.0 (1.5)	12.3 (3.7)
p		0.30	0.25	0.21
Adolescent Characteristics
Age (years)
13–14	47	511.7 (82.3)	3.8 (1.3)	11.5 (3.9)
15	45	520.9 (71.1)	4.2 (1.6)	12.2 (3.5)
16–19	45	487.2 (94.6)	4.1 (1.3)	12.2 (3.8)
p		0.03	0.05	0.18
Sex
Male	64	503.8 (82.5)	4.1 (1.5)	12.1 (4.0)
Female	73	509.9 (85.4)	4.0 (1.3)	11.8 (3.5)
p		0.42	0.66	0.52

^aSleep duration averaged over the weekday nights (Sunday through Thursday)

^bSleep midpoint (median of bedtime and wake time) averaged over the weekdays

^cCalculated as the percentage of 1-min periods of sleep out of the total number of sleep bouts of any length, averaged over the weekdays

¹Spearmans correlation coefficient; ABR: SD: Standard Deviation; P-values from Wilcoxon-Mann-Whitney test

^**p <0.05 (2-tailed)

^*p < 0.10 (2-tailed)

The distributions of 3-PBA and TCPy concentrations are presented in Table 2. We detected 3-PBA in 44% of the sample and TCPy in 93% of the sample. The geometric mean (SE) for 3-PBA and TCPy was 0.3 (0.0) ng/mL and 1.7 (0.2) ng/mL, respectively (Table 2).

Table 2.

Distribution of prenatal urinary pesticides measured from maternal urine samples collected in the third trimester of pregnancy (N = 137)

	LOD	% < LOD	GM ± SE (ng/mL)	Min	25th	75th	Max
3-PBA	0.25	55.6	0.3 ± 0.0	< LOD	< LOD	0.3	11.1
TCPY	0.1	7.0	1.7 ± 0.2	< LOD	0.9	3.5	44.8

ABR: LOD: Limit of Detection; GM: Geometric Mean; SE: Standard Error; ng/mL: nanograms per milliliter; 3-PBA: 3-phenoxybenzoic acid; TCPy: 3, 5, 6-trichloro-2-pyridinol; DCCA: 2,2-dimethylcyclopropane carboxylic acid

Associations between maternal and adolescent characteristics with 3-PBA and TCPy are presented in Table 3. Mothers with detectable levels of 3-PBA in their urine were of older age (p = 0.05). To illustrate, 43.1% of mothers with detectable levels of 3-PBA were between the ages of 29–39, compared to only 23.6% of mothers without detectable levels (Table 3). In addition, women who were married or in a civil union had higher levels of TCPy than those who were not.

Table 3.

Maternal and adolescent characteristics according to 3-PBA and TCPy detection from maternal urine samples collected in the third trimester of pregnancy (N = 137)

	3-PBA			TCPy
Maternal Characteristics	Not Detected (N = 72)		Detected (N = 65)	N	Median (Q1, Q3)
Maternal age (years)
16–23	37.5		29.2	46	1.7 (0.91, 3.2)
24–28	38.9		27.7	46	2.0 (0.97, 3.8)
29–39	23.6		43.1	45	1.9 (0.92, 3.9)
p		0.05*			0.85
Marital Status
Married/Civil Union	87.5		90.8	122	2.0 (1.1, 3.9)
Single, separated, divorced, or widowed	12.5		9.2	15	0.85 (0.91, 1.7)
p		0.54			0.0023
Parity
≤ 1	41.7		27.7	48	1.7 (0.90, 2.9)
2	36.1		35.4	49	1.5 (0.85, 3.3)
≥ 3	22.2		36.9	40	2.2 (1.2, 4.7)
p		0.10			0.26
Maternal Education (years)
< 9	11.1		12.3	16	2.0 (0.96, 4.5)
9–11.9	19.4		10.8	21	1.5 (0.69, 3.3)
12	30.6		32.3	43	1.7 (0.87, 3.1)
>12	38.9		44.6	57	2.0 (0.97, 3.3)
p		0.56			0.76
Adolescent Characteristics Age (years)
13–14	33.3		35.4	47	1.7 (0.74, 3.3)
15	34.7		30.8	45	2.0 (1.1, 2.8)
16–19	31.9		33.9	45	1.8 (0.93, 3.9)
p		0.88			0.59
Sex
Male	52.8		40.0	64	1.6 (0.92, 3.8)
Female	47.2		60.0	73	2.0 (1.1, 3.3)
p		0.13			0.88

ABR: 3-PBA: 3-phenoxybenzoic acid; TCPy: 3, 5, 6-trichloro-2-pyridinol; DCCA: 2,2-dimethylcyclopropane carboxylic acid; LOD: Limit of Detection < LOD: Value below limit of detection; Q1: Lower quartile; Q3: Upper quartile; P-values from chi-square (χ²) test

^**p <0.05 (2-tailed)

^*p < 0.10 (2-tailed)

Results from urinary specific gravity-adjusted generalized LMMs analysis among the overall population are presented in Table 4. Findings from crude models demonstrated that compared to the reference tertile (T1), adolescent offspring in the highest tertile of TCPy had approximately 39 minutes longer sleep duration (β = 38.5 [95% CI: 9.8, 67.2]) (p, trend = 0.0087) and 0.52 h later midpoint of sleep (β = 0.5 [95% CI: 0.0, 1.1]) (p, trend = 0.0395). No significant associations were observed between tertiles of 3-PBA and sleep measures from crude or adjusted models. When we adjusted for confounders, the association between maternal TCPy with offspring sleep duration and midpoint of sleep persisted (Table 4). To illustrate, adjusted findings demonstrated that adolescents in the highest tertile of maternal TCPy exposure had 36.2 minutes longer sleep duration (β = 36.2 [95% CI: 5.2, 67.3]) (p, trend = 0.0113) and nearly 0.50 h later midpoint of sleep (β = 0.5 [95% CI: −0.1, 1.0]) (p, trend = 0.07) compared to the reference tertile. In a post-hoc analysis, we tested for interactions between maternal exposures and sex. Results from the adjusted models demonstrated no statistical significant for all of the p for interaction tests (Table 4). Findings from post-hoc analyses also demonstrated that associations between maternal exposures and sleep measures among offspring at time point 1 (Supplemental Table 4) and time point 2 (Supplemental Table 5) were not significantly different.

Table 4.

Association between tertiles of maternal urinary pesticide and sleep duration, midpoint and fragmentation among mother-adolescent pairs (N = 137)

	N	Sleep Duration, Minutes p (95% CI)a	Mid-point, Decimal hours p (95% CI)b	Sleep Fragmentation, Percentage p (95% CI)c
3-PBA
Unadjusted
T1	72	Ref	Ref	Ref
T2	32	−5.4 (−33.0, 22.2)	0.2 (−0.7, 0.2)	−0.7 (−2.0, 0.5)
T3	33	0.5 (−29.7, 30.8)	0.3 (−0.2, 0.8)	1.0 (−0.4, 2.4)
P for trend		0.95	0.42	0.33
Adjusted
T1	72	Ref	Ref	Ref
T2	32	−2.0 (−30.2, 26.3)	−0.2 (−0.7, 0.2)	−0.6 (−1.9, 0.7)
T3	33	−4.0 (−35.5, 27.5)	0.4 (−0.2, 0.9)	1.0 (−0.4, 2.5)
P for trend		0.78	0.53	0.35
P for int		0.18	0.57	0.53
TCPy
Unadjusted
T1	45	Ref	Ref	Ref
T2	46	10.3 (−17.3, 37.9)	0.4 (−0.1, 0.8)	−0.8 (−2.1, 0.5)
T3	46	38.5 (9.8, 67.2)**	0.5 (0.0, 1.1)**	0.5 (−0.9, 1.9)
P for trend		0.009**	0.04**	0.46
Adjusted
T1	45	Ref	Ref	Ref
T2	46	7.2 (−22.4, 36.9)	0.4 (−0.1, 0.9)	−0.8 (−2.2, 0.7)
T3	46	36.2 (5.2, 67.3)*	0.5 (−0.1, 1.0)*	0.4 (−1.1, 1.9)
P for trend		0.01**	0.07*	0.37
P for int		0.49	0.37	0.55

^aSleep duration averaged over the weekday nights (Sunday through Thursday)

^bSleep midpoint (median of bedtime and wake time) weekday nights

^cCalculated as the percentage of 1-min periods of sleep out of the total number of sleep bouts of any length, weekday nights ABR: β: Beta Coefficient; CI: Confidence Interval; T: Tertile; 3-PBA: 3-phenoxybenzoic acid; TCPy: 3, 5, 6-trichloro-2-pyridinol Unadjusted models (only adjusted for maternal urinary specific gravity); Adjusted models for maternal (education, parity, marital status, specific gravity) and adolescent characteristics (sex, age)

^**p <0.05 (2-tailed)

^*p < 0.10 (2-tailed)

^*Int p (interaction p value) <0.1 considered significant for interaction between maternal urinary pesticide*sex

In sex-stratified models, we found that the associations between TCPy and sleep measures were evident among females, whereas there were no associations among males. To illustrate, the adjusted model showed that female offspring in the middle tertile had a longer sleep duration of approximately 27 minutes (β = 27.4 [95% CI: −13.5, 68.2]) and those in the highest tertile had a longer sleep duration of approximately 59 minutes (β = 58.5 [95% CI: 12.2, 104.8]) (p, trend =0.0042) compared to the reference group. The association with later midpoint of sleep similarly persisted among female offspring in the middle tertile (β = 0.5 [95% CI: 0.0, 1.1]) and highest tertile (β = 0.6 [95% CI: 0.0, 1.3]) (p, trend = 0.01) of maternal TCPy exposure (Table 5), after adjusting for confounders.

Table 5.

Sex-stratified association between tertiles of maternal urinary pesticide and sleep duration, midpoint and fragmentation among mother-adolescent pairs (N = 137)

	3-PBA				TCPy
	N	Sleep Duration, Minutes	Mid-point, Decimal hours	Sleep Fragmentation, Percentage	N	Sleep Duration, Minutes	Mid-point, Decimal hours	Sleep Fragmentation, Percentage
		β (95% CI)a	β (95% CI)b	β (95% CI)c		β (95% CI)a	β (95% CI)b	β (95% CI)c
Males
Unadjusted
T1	38	Ref	Ref	Ref	23	Ref	Ref	Ref
T2	12	−7.9 (−54.1, 38.3)	−0.3 (−1.2, 0.6)	−1.6 (−4.0, 0.7)	19	−6.6 (−47.7, 34.6)	0.2 (−0.6, 1.0)	−1.2 (−3.3, 1.0)
T3	14	24.5 (−20.1, 69.1)	0.4 (−0.5, 1.3)	1.1 (−1.1, 3.4)	22	30.6 (−11.6, 72.8)	0.2 (−0.6, 1.1)	0.6 (−1.6, 2.8)
P for trend		0.33	0.48	0.50		0.17	0.59	0.66
Adjusted
T1	38	Ref	Ref	Ref	23	Ref	Ref	Ref
T2	12	−8.4 (−57.3, 40.5)	−0.4 (−1.3, 0.6)	−1.4 (−3.9, 1.2)	19	−24.5 (−70.2, 21.2)	0.01 (−0.9, 0.9)	−1.7 (−4.1, 0.7)
T3	14	4.4 (−45.2, 54.2)	0.2 (−0.7, 1.2)	1.2 (−1.4, 3.8)	22	20.1 (−23.9, 64.0)	0.2 (−0.7, 1.1)	0.5 (−1.8, 2.9)
P for trend		0.84	0.80	0.51		0.34	0.75	0.83
Females
Unadjusted
T1	34	Ref	Ref	Ref	22	Ref	Ref	Ref
T2	20	−6.0 (−43.0, 30.9)	0.1 (−0.5, 0.6)	−0.2 (−1.7, 1.3)	27	23.8 (−14.9, 62.4)	0.5 (−0.0, 1.1)*	0.0 (−1.7, 1.7)
T3	19	−20.8 (−62.9, 21.4)	0.3 (−0.3, 0.9)	0.6 (−1.2, 2.4)	24	48.6 (8.3, 88.9)**	0.8 (0.2, 1.4)**	0.6 (−1.1, 2.3)
P for trend		0.34	0.29	0.59		0.02**	0.01**	0.46
Adjusted
T1	34	Ref	Ref	Ref	22	Ref	Ref	Ref
T2	20	−5.7 (−45.2, 33.7)	−0.0 (−0.6, 0.5)	−0.4 (−2.0, 1.3)	27	27.4 (−13.5, 68.2)	0.5 (0.0, 1.1)*	0.3 (−1.5, 2.1)
T3	19	−15.8 (−61.1,29.5)	0.5 (−0.2, 1.1)	0.6 (−1.4, 2.5)	24	58.5 (12.2, 104.8)*	0.6 (0.0, 1.3)*	0.8 (−1.2, 2.9)
P for trend		0.34	0.41	0.73		0.004**	0.01**	0.26

^aSleep duration averaged over the weekday nights (Sunday through Thursday)

^bSleep midpoint (median of bedtime and wake time) weekday nights

^cCalculated as the percentage of 1-min periods of sleep out of the total number of sleep bouts of any length, weekday nights ABR: β: Beta Coefficient; CI: Confidence Interval; T: Tertile; 3-PBA: 3-phenoxybenzoic acid; TCPy: 3, 5, 6-trichloro-2-pyridinol; Unadjusted models (only adjusted for maternal urinary specific gravity); Adjusted models for maternal (education, parity, marital status, specific gravity) and adolescent characteristics (age)

^**p <0.05 (2-tailed)

^*p < 0.10 (2-tailed)

Finally, we performed a secondary analysis of children with information on snoring (Supplemental Table 2) and difficulty breathing while sleeping at T2 (Supplemental Table 3). 30.7% of the sample reported snoring, and 7.3% reported difficulty breathing. There was no evidence of an association between prenatal pesticide exposure and sleep-disordered breathing-related outcomes after adjusting for maternal urinary specific gravity and adolescent age.

Discussion

In this prospective study, maternal in utero exposure to chlorpyrifos, but not pyrethroids, was associated with longer sleep duration and later midpoint of sleep among adolescent offspring. Sex-stratified findings demonstrated that the association between maternal TCPy with longer sleep duration and later midpoint of sleep was only evident among female adolescent offspring.

The present study findings add to the literature on pregnancy as a potentially vulnerable period to pesticide exposure. As in the present sample, other studies have identified 3-PBA, TCPy, and other pesticide metabolites in the urine sample of pregnant women (Barkoski et al., 2021; Dalsager et al., 2019; Jaacks et al., 2019; Zhang et al., 2020), at levels similar to general populations. Regarding 3-PBA, the detection levels in this Mexican cohort (44%) were lower than those reported in a Danish pregnancy cohort (94%) (Dalsager et al., 2019) but higher than those in a Bangladeshi pregnancy cohort (20%) (Jaacks et al., 2019). Further, compared to general adult populations, urinary 3-PBA concentrations among pregnant mothers in the present cohort were similar to adult populations in Germany (Heudorf et al., 2006) and metropolitan Canada (Fortin et al., 2008). The levels of TCPy were comparable across all three pregnancy cohorts (Mexican, Danish, and Bangladeshi) and similar to the general adult population of US adults aged 18–40 from NHANES 1999–2002 (Fortenberry et al., 2012) (2.0 ng/mL compared to 1.7 ng/mL in the present sample).

Previous studies have demonstrated that the neural connectivity required to process signals involved in sleep regulation begins to form as early as seven weeks of gestation (Bennet et al., 2018; Scher and Loparo, 2009), otherwise providing biological evidence that maternal exposures during the first trimester of gestation may alter fetal processes related to offspring’s sleep regulation. In the present study, we did not assess the impact that first trimester maternal exposure to pesticides had on the sleep health of adolescent offspring due to low data availability. However, the decision to focus on trimester three was related to both availability of exposure data, as well as the fact that the components of sleep regulation, including rapid eye movement (REM) and non-rapid eye movement (NREM), are established during the start of the third trimester of gestation (Okai et al., 1992). The noted components of sleep regulation are of importance to fetal development, given that a host of sensory systems, including the hippocampal, limbic, and auditory systems all require REM sleep for normal fetal development (Graven and Browne, 2008). In addition, one human study found that maternal exposure to chlorpyrifos during the third trimester was associated with a higher odds of sleep-disordered breathing among offspring during childhood (Raanan et al., 2015). To our knowledge, no other studies among humans have assessed the impact of maternal pesticide exposure during the first or second trimester of pregnancy on offspring sleep outcomes later in life.

Findings from the present study may be difficult to compare given a lack of literature on the associations of interest. However, previous studies have provided evidence of links between pesticides and sleep, primarily among homogeneous samples of male farmworkers (17–21). To illustrate, a longitudinal study found that Chinese farmworkers with long-term exposure to organophosphorus pesticides had poor sleep quality, resulting in sleepiness during the day (Zhao et al., 2010). A multicenter case-controlled study among European adults found that occupational pesticide exposure and farming were potential risk factors for rapid eye movement (REM) sleep behavior disorder. However, comparing our findings may be inappropriate given differences in life-stage for the exposed populations, with participants in the prospective study being prenatally exposed, while most existing literature is among adult farmworkers with occupational exposure to pesticides.

Although existing studies are non-existent, several potential molecular pathways exist to explain links between prenatal pesticide exposure and offspring sleep health. There is evidence that exposure to chlorpyrifos (CPF) causes dysfunction in the nervous system by interrupting neurotransmitters, including acetylcholine, due to inhibition of acetylcholinesterase (AChE) (Colovic et al., 2013). Acetylcholine can affect sleep as it has been shown to play a critical role in REM sleep regulation (Yamada and Ueda, 2020). Further, a study conducted among pregnant rats found that males exposed to lower prenatal CPF had higher serotonin turnover in different regions of the brain, whereas females showed a similar effect at higher CPF concentrations than males (Slotkin and Seidler, 2007). Though we found that females’ sleep measures rather than males’ were associated with maternal CPF exposure, our study and the study conducted by Slotkin and Seidler reinforce the idea that exposure to CPF during pregnancy is a critical period for sex-dependent effects of CPF on the offspring during adolescence. Moreover, Slotkin and Seidler’s findings are imperative to understanding the present study findings. They found that maternal CPF was associated with a higher turnover of serotonin in offspring’s brain during adolescence (Slotkin and Seidler, 2007). Serotonin is a critical neurotransmitter important for regulating mood and emotions, and it also serves as a direct precursor for melatonin (Thor et al., 2007). Melatonin is vital for regulating sleep-wake cycles and circadian rhythm (Brown, 1994); therefore, serotonin turnover may directly result in a lower level of melatonin, which ultimately would affect offspring sleep outcomes.

A few prior animal and human studies have focused on the role of prenatal pesticide exposure and sleep-disordered breathing and respiratory-related offspring outcomes. One animal study demonstrated that pregnant rats exposed to chlorpyrifos (CPF) had adolescent offspring mice with higher sleep apnea index and diaphragm contractility and modified respiratory patterns (Darwiche et al., 2018). Another study found that peri-gestational exposure to a low dose of CPF combined with a high-fat diet induced an increase in the sleep apnea index in offspring during adulthood, which was associated with a significant decrease in the acetylcholinesterase activity compared to controls (el Khayat el Sabbouri et al., 2019). Results from a longitudinal birth cohort of Mexican/Mexican-American mother-child pairs found that early-life exposure to organophosphate pesticides, including chlorpyrifos, during the third trimester was associated with a higher odds of respiratory symptoms during childhood, including a measure of sleep-disordered breathing (Raanan et al., 2015). Although we did not measure obstructive sleep apnea or sleep-disordered breathing among the present sample, we obtained information on the presence of difficulty breathing while sleeping and snoring among adolescent offspring during the time 2 visit. Unlike the previous study, we did not find evidence of an association between prenatal pesticide exposure and snoring or difficulty breathing. However, these secondary analysis results were underpowered, and ultimately, a larger sample size is needed to assess these relationships further.

Of note, although we found that over half of the present analytic sample had non-detectable levels of maternal 3-PBA, a recent report from Mexico highlighted 140 pesticide ingredients that are authorized for use in thousands of commercial products that are banned in other countries. Among these, over 40 pesticides authorized for agricultural, domestic, urban, and livestock use in Mexico are classified as dangerous by international pesticide organizations. Although the present analysis did not investigate any of the aforementioned pesticides, our findings highlight the need for additional examinations of a wide range of pesticides.

Of clinical relevance, we found that adolescents born to mothers in the highest tertiles of TCPy had a longer sleep duration of >30 minutes. For sleep deprived populations, 30 minute longer sleep duration has been linked to improvements in well-being (Cepeda et al., 2016). However, among populations who tend to have longer sleep duration, 30 minute longer sleep has been associated with higher depressive symptoms (Liu et al., 2020). Although we do not have other information regarding the mental health status of our population, it is possible that the adolescents in our analytic sample were are more aligned with the latter scenario, given that overall they demonstrated longer sleep duration than other adolescent cohorts (Chattu et al., 2018; Kerkhof, 2017).

Strengths and Limitations

This study includes several strengths, including the use of urinary 3-PBA and TCPy as biomarkers of exposure, rather than relying on self-reported pesticide exposure. Another strength in using a urinary biomarker of 3-PBA was that pyrethroids urinary measures are likely to be more reliable over time compared to parent compounds in blood, given the rapid metabolism and low detection rates of non-persistent pesticides in the blood (Watkins et al., 2016). Furthermore, the study was prospective with repeated time-points for sleep outcomes. Sleep data were recorded objectively using an actigraph device, limiting errors in self-reports which may be particularly common among adolescents. However, although we used an objective method to assess sleep measure outcomes, research among adolescents has shown that actigraph technology begins to record the onset of sleep as when the adolescent stops moving, which may not reflect actual sleep (Fekedulegn et al., 2020). Nonetheless, this measurement error is unlikely to be differential and thus may only decrease the precision of the effect estimates. In addition, because there is no validated actigraph sleep fragmentation algorithm for adolescents, sleep fragmentation may not be the best measure to assess sleep quality among adolescent populations. This limitation may partially explain why we did not see observe the associations we hypothesized between the exposure of interest and sleep fragmentation among adolescent offspring. Other limitations include that our study was observational; thus, causality cannot be assumed. We also used a modest sample size to explore relationships between urinary 3-PBA and TCPy and adolescent sleep outcomes and a single urinary measure to estimate maternal exposure. Further, nearly 56% of the 3-PBA measurements were below the limit of detection, limiting our ability to assess linear relationships between continuous exposure measures and outcome and variability within individuals over time. Although biomonitoring has several strengths, multiple urine samples are required to accurately classify long-term exposure to chemicals with short half-lives, such as those in the present study (Sexton and Ryan, 2012).

Similarly, given that urinary pesticide metabolite concentrations generally have a half-life of < 24 h, they tend to reflect recent exposure (Glorennec et al., 2017). Further, previous studies have demonstrated that both pyrethroids and chlorpyrifos are metabolized quickly and using a single specimen to assess pesticide concentrations may result in exposure misclassification of study participants (Barkoski et al., 2018). Our study is also limited in that it relied on one urine sample obtained during the third trimester to determine maternal exposure to both pyrethroids and chlorpyrifos. However, as previously mentioned, a strength to having used urinary measures of chlorpyrifos and pyrethroids is that compared to blood, urinary measures of these pesticides may be more reliable over time since blood has a more rapid metabolism and lower detection rates of non-persistent pesticides (Fortenberry et al., 2012; Watkins et al., 2016). In addition, although this study sought to examine the impact early-life exposure to pesticides had on adolescent offspring sleep, assessing concurrent exposure would allow us to control for variation due to exposure in adolescence. However, we did not have access to information on offspring exposure to pesticides at the time that we collected information on sleep measures. Further, we adjusted for maternal specific gravity to adjust spot urinary concentrations for hydration status. Although there are some limitations to adjusting for specific gravity, including varying by BMI, there are strengths to adjusting for specific gravity among pregnant women(MacPherson et al., 2018). Specifically, compared to creatinine, specific gravity has been shown to be a more appropriate correction approach for pregnant women because of better within-person reproducibility and because it is less affected by participant demographics (MacPherson et al., 2018).

Previous studies among the present cohort of adolescents have demonstrated significant variation in sleep outcomes by pubertal status (Jansen et al., 2018; Zamora et al., 2021). Furthermore, adolescence is a unique period marked by physiological and behavioral changes that may impact sleep health. Future studies should assess whether the relationships under study hold in pre-puberty versus post-puberty. Study findings may also be subject to confounding by season and geographic location. Because our analysis did not account for the season or geographic location in which participants were questioned or sampled, we could not consider the seasonal and geographic variation that may have attributed to study findings. In addition, we did not ask about depression symptoms and thus could not explore the potential pathways related to depression. Finally, generalizability may be limited to mothers and adolescents living in urban Latin American settings.

Conclusion

In summary, we found that adolescent females with the highest prenatal exposure levels to TCPy had longer sleep duration and later sleep timing than those with lower prenatal exposure levels. Overall, these results are of public health importance considering the continued widespread agricultural and possibly residential use of pyrethroids and chlorpyrifos in Mexico. Moreover, the Mexican population is likely exposed to hundreds of unregulated pesticides. Thus, our results underline the importance of additional research studies that include both larger samples and assessment of unregulated pesticides, as well as studies that consider the underlying mechanisms explaining sex differences.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author, EJ, upon reasonable request.