Associations Between Gestational Polybrominated Diphenyl Ether (PBDE) Serum Concentrations and Child Sleep Outcomes from Ages 2 to 8 Years

Abstract

Gestational polybrominated diphenyl ethers (PBDEs) exposures have been associated with thyroid disruption in pregnant women and adverse neurobehavioral outcomes in their children, but it is unknown if they interfere with children’s sleep patterns. We assessed gestational PBDE exposure (16 weeks) and child sleep patterns from ages 2 to 8 years using 410 mother-child dyads in the Health Outcomes and Measures of the Environment (HOME) Study. Gestational biomarkers of serum PBDEs include PBDE-153 (GM±GSD: 5.2±2.8 ng/g lipid), PBDE-100 (4±2.6), PBDE-99 (4.6±2.7), PBDE-47 (20.2±2.6), PBDE-28 (1.3±2.2), and ΣPBDEs (37.03±2.52). We measured child sleep patterns using the adapted Child Sleep Health Questionnaire, which includes sleep irregularity (mean±SD: 2.5±0.8), hypersomnolence (4.7±1.5), sleep disruption (6.7±1.6), and sleep duration. We assessed longitudinal associations between gestational PBDEs and sleep patterns using generalized estimating equations, adjusting for covariates. For PBDE-visit interactions (p < 0.1), visit-specific estimates with 95% CIs were calculated; otherwise, the overall estimate was reported. Every 10-fold increase in PBDE-99 (β=0.18, 95% CI: −0.04, 0.23), PBDE-47 (0.15, 95% CI: 0.001, 0.3) and ΣPBDEs (0.13, 95% CI: −0.21, 0.27) was associated with increased sleep irregularity for all years, and PBDE-28 was associated with this outcome at age 5 and 8 years (0.43, 95% CI: 0.09, 0.77). PBDE-153 (−0.5, 95% CI: −1.06, 0.05) was associated with decreased hypersomnolence at age 4 years. PBDE-47 (0.3, 95% CI: 0.004, 0.61), PBDE-99 (0.38, 95% CI: 0.1, 0.67 and 0.62), and ΣPBDEs (0.27, 95% CI: −0.02, 0.56) were associated with increased sleep disruption for all ages. We observed no significant associations between PBDEs and sleep duration. We found that gestational PBDEs were associated with sleep irregularity and sleep disruption in children, highlighting the need to explore sleep as a mediator of PBDE-associated neurobehavioral problems.

1. INTRODUCTION

Polybrominated diphenyl ethers (PBDEs) have been widely used in the U.S. for decades as flame retardants. Their bio-accumulative properties cause them to accumulate in fatty tissues in humans and wildlife.^1,2 Due to concerns over their developmental neurotoxicity, PBDEs were phased out of use in the United States in the mid-2000s^3,4, and globally, the commercial penta- and octa-BDE mixtures were banned under the Stockholm Convention in 2009.⁵ Unfortunately, humans will continue to be exposed to these persistent organic pollutants (POPs) for decades due to their widespread presence in the home, school, and office environment and dietary fat sources.⁶ PBDEs (BDE-28, −47, −99, −100, and −153) have a relatively long half-life ranging from 1.4–7.4 years resulting in relatively high Intraclass Correlation Coefficients (ICC: 0.96–0.99).^7–9 Pregnant women and young children can encounter PBDEs from everyday consumer products containing flame retardants and through dust, food, and water. PBDEs readily cross the placental barrier by passive diffusion due to their lipophilic nature and are taken up by the fetus.¹⁰ PBDEs are mostly found in dust, resulting in ingestion and dermal absorption as the main routes of exposure, but airborne PBDEs are also inhaled.^11–13 Biomarkers of exposure (serum PBDEs) provide biologically relevant doses from all exposure routes (inhalation, ingestion, and dermal absorption).

PBDEs are endocrine disrupting chemicals (EDCs) that interfere with maternal and neonatal thyroid hormones, which are essential for optimal neurodevelopment and sleep.^14–20 Thyroid function directly impacts sleep health^14,17, and disruptions can contribute to sleep disturbances. Sleep is critical for healthy neurodevelopment, and inadequate sleep or disruptions can lead to lasting cognitive deficits, psychosocial challenges, and increased risk of mental health disorders such as depression and anxiety.^21–24 Gestational PBDE exposure has been linked with impaired child cognitive function, attention deficits, executive dysfunction, and behavioral problems, including externalizing and internalizing symptoms.^25–28 No prior studies have systematically examined the role of gestational PBDEs on child sleep irregularity, insufficiency, disruption, and duration prior to school age.^29–32 Exploring the comprehensive implications of PBDEs on children’s sleep health is paramount, especially considering the ways poor sleep can impact development and behavior in children. This study investigates whether gestational PBDE concentrations affects clinically significant sleep patterns in children at ages 2 to 8 years.

2. METHODS

2.1. Study Participants and Design

This study includes participants from the Health Outcomes and Measures of the Environment (HOME) Study, a longitudinal pregnancy and birth cohort of 410 children born to 400 pregnant women enrolled between 2003 and 2006 in the Greater Cincinnati area, Ohio. The cohort was designed to examine potential adverse effects of known and suspected environmental toxicants on child health and neurobehavioral outcomes.^33,34 Detailed information on enrollment and inclusion criteria are described in Braun et al., 2017.³⁵ The Institutional Review Board at Cincinnati Children’s Hospital Medical Center (CCHMC) approved the study protocol. The Centers for Disease Control and Prevention (CDC) deferred to the CCHMC IRB as the IRB of record. The analysis of de-identified serum specimens at the CDC laboratory was determined not to constitute human subjects research.

2.2. Data collection and measures

2.2.1. Gestational PBDE biomarker

Serum PBDE concentrations were measured using gas chromatography high-resolution mass spectrometry after automated liquid/liquid extraction and removal of coextracted biogenic material on a silica/acid-silica column.³⁶ The methodology for analyzing serum samples for PBDEs has been previously described by Sjödin et al., 2004.³⁷ The primary exposure of interest was gestational (16 weeks) serum PBDEs. PBDEs were expressed using lipid-standardized concentrations (nanograms per gram lipid). The serum lipid concentration was calculated from the total cholesterol and triglyceride concentrations according to Phillips et al.³⁸ The PBDEs of interest are BDE-28, −47, −99, −100, −153, all with a detection frequency >80%, as these were the most detectable serum PBDEs in the study period.^36,39 PBDE concentrations were examined on a continuous scale, and concentrations below the limit of detection (LOD) were replaced with the LOD divided by the square root of 2.⁴⁰ PBDEs concentrations were right-skewed, so log₁₀ transformation was used. Spearman correlation coefficients were calculated to understand patterns of PBDE congeners concentration correlations (Table 2).

Table 2.

Univariate Statistics of Sleep Outcomes and Maternal Serum Gestational PBDE Concentrations

	2 Years	3 Years	4 Years	5 Years	8 Years
Sleep Irregularity (Mean ± SD)	2.55 ± 0.79 n = 280	2.56 ± 0.84 n = 258	2.54 ± 0.88 n = 187	2.49 ± 0.77 n = 210	2.52 ± 0.81 n = 221
Hypersomnolence (Mean ± SD)	4.79 ± 1.52 n = 279	4.76 ± 1.52 n = 258	4.54 ± 1.47 n = 186	4.50 ± 1.50 n = 210	3.97 ± 1.36 n = 220
Sleep Disruption (Mean ± SD)	6.75 ± 1.62 n = 279	6.78 ± 1.66 n = 258	6.72 ± 1.60 n = 187	6.48 ± 1.58 n = 210	6.23 ± 1.43 n = 220
Sleep Duration [min.] (Mean ± SD, minutes)	751.43 ± 90.7 n = 280	704.16 ± 75.61 n = 258	662.57 ± 77.53 n = 187	631.00 ± 72.55 n = 210	580.72 ± 72.1 n = 221

	PBDE-153 (n = 258)	PBDE-100 (n =252)	PBDE-99 (n = 259)	PBDE-47 (n = 266)	PBDE-28 (n = 184)	ΣPBDEs (n = 266)
PBDE Levels (GM ± GSD, ng/g lipid)	5.2 ± 2.8	4 ± 2.6	4.6 ± 2.7	20.2 ± 2.6	1.3 ± 2.2	37.03 ± 2.52
Percent Detection	96.9	96.5	98.3	99.7	86.6	100

^*Sleep outcome sample sizes reflect participants who completed each outcome; PBDE sample sizes reflect participants with PBDE data and at least one outcome

^*Biospecimens collected from 2003–2006; Some samples experienced laboratory processing issues, resulting in a lower sample size

In primary analyses with sleep patterns, we focused on individual PBDE congeners with reported concentrations in Table 2. In secondary analyses, we examined the association of the sum of five PBDE congeners (ΣPBDEs: PBDE-28, −47, −99, −100, −153), creating a regression coefficient for the sum of PBDEs.⁴¹

2.2.2. Child Sleep Health Outcome

Sleep assessments in the HOME Study were completed by caregivers using an adapted version of the Children’s Sleep Habits Questionnaire (CSHQ) at ages 2, 3, 4, 5, and 8 years. The CSHQ has been validated for children 2–10 years.^42,43 The CSHQ is a retrospective caregiver questionnaire that informs about crucial aspects of sleep behaviors in children, such as bedtime behavior, sleep onset and duration, sleep-related anxiety, nighttime behaviors and awakenings, parasomnias, morning waking/daytime sleepiness, and sleep-disordered breathing.⁴² For this analysis, we derived the following sleep outcomes using 11 clinically relevant questions from the CSHQ: sleep irregularity, hypersomnolence, sleep disruption, and sleep duration (minutes) using the information derived from answering the CSHQ (Supplemental Table S1). Caregivers rated each questionnaire item on a 3-point scale, indicating its frequency as either “usually” (i.e., 5–7 times within the past week), “sometimes” (i.e., 2–4 times within the past week), or “rarely” (i.e., never or 1 time within the past week)⁴⁴. We scored the three domains (irregularity, hypersomnolence, and disruption) based on CSHQ scoring metrics. Sleep irregularity included two questions and was measured on a scale from 0 to 6 with reverse scoring, such that higher scores reflect more disturbed sleep behavior. Hypersomnolence included three questions and was assessed on a 9-point scale. Sleep disruption included five questions and was assessed on a 15-point scale. Sleep duration was used as a continuous variable of total minutes.

To ensure that our outcomes were measured accurately, we assessed item variability, McDonald’s omega or the Greatest Lower Bound (GLB)⁴⁵, test-retest reliability, intercorrelation matrix, and conducted confirmatory factor analysis on the study’s sleep domains (Supplemental Tables S2–6 & Figure S1). Test-retest reliability was assessed by evaluating correlations over time among participants with multiple surveys completed. We chose GLB instead of Cronbach’s alpha because of the strict assumptions of unidimensionality, uncorrelated errors, and essential tau-equivalence of all items in Cronbach’s alpha.^46,47 Based on covariance alone, the tau-equivalence assumption is violated (Supplemental Table S3).

2.2.3. Covariates

We used a directed acyclic graph (DAG) (Supplemental Figure S2) to outline the causal pathway and adjusted for covariates based on the DAG and actual changes in regression coefficient by covariate inclusion.⁴⁸

For all models, we included the following covariates: 16 week maternal serum cotinine for smoking (log₁₀-transformed), 16 week blood lead concentrations (log₁₀-transformed), child sex, child race, household income, marital status, maternal IQ, the Home Observation for Measurement of the Environment Inventory score at 1 year (designed to measure the quality and quantity of stimulation and support available to a child in the home environment)⁴⁹, maternal age at delivery, parity, breastfeeding status, baseline Beck Depression Inventory (BDI-II) (designed to measure maternal depression), baseline region (urban & suburban/rural), and neighborhood deprivation index (NDI) at birth. Maternal IQ can influence a mother’s understanding of child sleep health and serves as a key pathway through which maternal education shapes sleep-related knowledge, as well as socioeconomic status (SES). The linkage between IQ and SES is complex, but maternal IQ may influence exposure to PBDEs indirectly through behavioral, environmental, and socioeconomic pathways.^50–52 Lower maternal IQ is often associated with lower SES, which can affect housing conditions, such as the presence of older furniture, carpet paddings, mattresses, insulation, and electronics that are more likely to contain PBDEs. Families with lower SES may also have limited resources to renovate homes or replace broken or contaminated household items. In contrast, mothers with higher IQs may have greater awareness of environmental toxicants and adopt behaviors that reduce exposure, such as wet-dusting and vacuuming with HEPA filters. Finally, maternal IQ may be related to parenting practices around sleep in children.^53,54 The NDI is a census tract-level deprivation index derived from the American Community Survey that consists of six census tract-level variables related to material deprivation. We used principal component analysis, then rescaled and normalized to range from 0 to 1, with higher values indicative of higher deprivation.⁵⁵ NDI plays a crucial role in sleep health, as neighborhood factors like noise, light pollution, and safety directly impact sleep quality and duration.^56,57

2.3. Statistical analyses

We investigated associations between log10-transformed 16-week serum PBDE concentrations and sleep outcomes. To inform model selection, we evaluated potential non-linear associations using tertile-based models. As these exploratory analyses did not suggest meaningful non-linearity, we proceeded with linear regression for the main analysis. We used Generalized Estimating Equations (GEE) to evaluate the longitudinal associations between gestational PBDE concentrations and child sleep outcomes at 2, 3, 4, 5, and 8 years, adjusting for covariates. We used an exchangeable covariance structure in the GEE models based on it providing the lowest Quasi-likelihood Information Criterion (QIC) after comparing with an autoregressive and unstructured covariance structure. We determined final model specifications through our DAG, AIC-based variable selection, and variance inflation factor (VIF) for multicollinearity.⁵⁸ The model was a better fit without mother’s education due to collinearity with mother’s race (AIC with: 3447.37 vs AIC without: 3441.28). This brought the VIF values to acceptable ranges⁵⁹, with the final model covariates listed above. This strategy enabled us to adhere to our theory-driven DAG and uphold statistical rigor by using maternal IQ as a proxy for maternal education, thereby confirming the stability of our models. We used maternal serum cotinine and lead measured at 16 weeks as a proxy for in utero exposure, recognizing that it reflects different pharmacokinetics than postnatal child exposure; however, we selected this time point to align with the gestational PBDE exposure used in our analysis^60–63 We examined the interaction between serum PBDE concentrations and study visit by using interaction terms to determine if the association between PBDEs and child sleep health persisted as children aged. If the interaction term had a p-value < 0.1, we calculated visit-specific estimates and 95% confidence intervals (CIs) using linear combination.^64–66 If the interaction term had a p-value ≥ 0.1, the interaction term was dropped, and the regression estimate was reported as the overall average association across ages 2–8 years (across all ages when outcome was measured).^67,68 We estimated βs and 95% CIs for PBDE-28, −47, −99, −100, −153, and ΣPBDEs with separate models. This was a complete case analysis to address missing data in the longitudinal study. All statistical analyses were performed using R version 4.3.1.

Sensitivity analysis

We derived three sleep domains that align closely with those most relevant to sleep medicine clinicians. We began our analysis by examining the original CSHQ domains and performing Pearson correlation analyses between these original domains and our study’s sleep outcomes. Results showed that sleep irregularity correlated most strongly with CSHQ subscales for bedtime resistance (0.38) and sleep duration (0.41, note this is a composite of caregiver ratings, not the actual duration in min), hypersomnolence with daytime sleepiness (0.5), sleep disruption with night wakings (0.39) and parasomnias (0.33). Based on these correlations, we conducted further GEE analyses using the original CSHQ domains, repeating the methods described previously.

Child sex is a biological variable with particular value in neurodevelopment, especially with regard to PBDE exposure.^69,70 Males are at higher risk of developing ADHD, and females are more likely to have depression and anxiety.^71,72 In addition, some studies have reported effect modification based on the child’s sex relating to environmental exposures and neurodevelopmental outcomes.⁷³ We adjusted for child sex in the overall analysis and tested whether child sex modifies PBDEs and overall 2–8 year sleep associations in the sensitivity analysis.

Lastly, we conducted an analysis incorporating cotinine as a time-varying variable to account for its potential impact on sleep health, as suggested by previous studies.^74,75 Cotinine, a biomarker of tobacco smoke exposure, has been linked to various sleep disturbances, including sleep-disordered breathing, parasomnias, daytime sleepiness (hypersomnolence), overall sleep disruption, and difficulty sleeping.^74,75 In this sensitivity analysis, child cotinine was measured from ages 2 to 4; for ages 5 and 8, we substituted the 4-year value since those time points were unavailable in the cohort. While there are differences in metabolism and exposure patterns across these time points, this approach offered the best available estimate of tobacco smoke exposure. There were strong positive correlations between cotinine concentrations across all time points, with the highest correlations observed between postnatal visits (ranging from r = 0.81 to r = 0.83) and slightly lower but still strong correlations between gestational and postnatal measurements (ranging from r = 0.65 to r = 0.71), suggesting that while cotinine concentrations before birth were still strongly related to those after birth, they showed more variation compared to postnatal concentrations. The same visit for cotinine concentration and the corresponding sleep outcome were matched in the model (e.g., 2-year sleep outcome was paired with 2-year cotinine concentration, etc.).

3. RESULTS

3.1. Sample characteristics, PBDE serum concentrations, and Sleep outcomes

The study sample included 398 pregnant women and their offspring at child follow-up visits from ages 2–8 years. The maternal and child characteristics were similar at baseline (n=398) and at the 8-year (n=223) follow-up (Table 1). The study sample included 59.3% White and 30.7% Black participants, and about 60% were urban with 36.9% being suburban and rural.

Table 1.

Descriptive Characteristics of HOME Study Cohort (n = 398)

Characteristic	% (n)
Sex assigned at birth
Female	53.8 (214)
Male	46.2 (184)
Child’s Race
Black or African American	30.7 (122)
Other/Multiracial	7.5 (30)
White or Caucasian	59.3 (236)
Unknown	2.5 (10)
Annual Household Income (baseline)
< $5k - $40k	38.9 (155)
$40k - $80k	35.2 (140)
$80k - $150k	20.6 (82)
>$150k	5.3 (21)
Marital Status (baseline)
Married, living apart	1.0 (4)
Married, living together	61.6 (245)
Not married, but living with someone	14.1 (56)
Not married, living alone	20.4 (81)
Unknown	3.0 (12)
Parity Category
0	43.0 (171)
1	31.2 (124)
>1	23.1 (92)
Unknown	2.8 (11)
Breastfed
No	17.8 (71)
Yes	75.9 (302)
Unknown	6.3 (25)
Region (baseline)
Suburban & Rural	36.9 (147)
Urban	59.5 (237)
Unknown	3.5 (14)
Characteristic	Mean ± SD
WASI (maternal): Full Scale IQ	105.62 ± 14.70
HOME Inventory: Total score (1yr)	39.36 ± 4.83
Maternal age at delivery (years)	29.34 ± 5.84
Breastfeeding duration (weeks)	23.43 ± 24.57
Maternal BDI-II Total Score - Baseline	10.01 ± 7.02
Neighborhood Deprivation Index (Birth)	0.36 ± 0.15

^*WASI= Wechsler Adult Intelligence Scale; HOME Inventory = Home Observation for Measurement of the Environment; BDI-II= Beck Depression Inventory-Second Edition

3.2. Associations of gestational serum PBDEs with sleep outcomes

Gestational biomarkers of serum PBDEs include PBDE-153 (GM±GSD: 5.2±2.8 ng/g lipid), PBDE-100 (4±2.6), PBDE-99 (4.6±2.7), PBDE-47 (20.2±2.6), PBDE-28 (1.3±2.2), and ΣPBDEs (37.03±2.52) (Table 2). Sleep irregularity remained consistently low and showed minimal change over time (Table 2). Hypersomnolence was moderate and similarly stable across ages. Sleep disruption also remained moderate and largely unchanged over time. In contrast, sleep duration, as expected, decreased with age.

3.2. Gestational serum PBDEs and sleep outcomes

A 10-fold increase in PBDE-47 (p = 0.05), PBDE-99 (p= 0.01), and ΣPBDEs (p= 0.09) were associated with more sleep irregularity reflecting an average effect from ages 2 to 8 years by about the same amount (β: 0.13–0.18), and PBDE-28 was associated with this outcome at age 5 (β=0.35, 95% CI: −0.03, 0.73, p= 0.09) and 8 years (β=0.43, 95% CI: 0.09, 0.77, p= 0.01) (Figure 1, Figure 2 & Supplemental Table S7). A 10-fold increase in PBDE-47 (p= 0.05), PBDE-99 (p= 0.01), and ΣPBDEs (p= 0.07) were associated with more sleep disruption reflecting an average effect from ages 2 to 8 years by about the same amount (β: 0.27–0.38). A 10-fold increase in PBDE-153 was associated with 0.5 point (95% CI: −1.06, 0.05, p= 0.09) lower hypersomnolence at age 4 years. We found null associations between PBDEs and sleep duration.

Figure 1.

Associations Between Gestational PBDE Serum Concentrations and Sleep Outcomes: Overall Estimates

Adjusted differences and 95% CI of child sleep outcomes at 2–8 years per 10-fold increase in gestational PBDE serum concentrations

*Adjusted covariates: Gestational Serum Cotinine, Gestational Blood Lead, Child Sex, Child Race, household Income, Marital Status, Maternal IQ, HOME Inventory Score, Age at Delivery, Parity, Breastfeeding Status, Baseline Beck Depression Inventory (BDI-II), Baseline region, Neighborhood Deprivation Index (NDI) at Birth

*If interaction term had p<0.1, we performed linear combination to calculate visit-specific estimates and 95% CI. If interaction term has p ≥0.1, dropped interaction term, and reported regression estimate as the overall effect

-Estimates are provided for significant (p <0.05) and marginally significant (p < 0.1) p-values

Figure 2.

Associations Between Gestational PBDE Serum Concentrations and Sleep Outcomes: Age-Specific Estimates

Adjusted differences and 95% CI of child sleep outcomes at 2–8 years per 10-fold increase in gestational PBDE concentrations

*Adjusted covariates: Gestational Serum Cotinine, Gestational Blood Lead, Child Sex, Child Race, household Income, Marital Status, Maternal IQ, HOME Inventory Score, Age at Delivery, Parity, Breastfeeding Status, Baseline Beck Depression Inventory (BDI-II), Baseline region, Neighborhood Deprivation Index (NDI) at Birth

*If interaction term had p<0.1, we performed linear combination to calculate visit-specific estimates and 95% CI. If interaction term has p ≥0.1, dropped interaction term, and reported regression estimate as the overall effect

-Estimates are provided for significant (p <0.05) and marginally significant (p < 0.1) p-values

Sensitivity analysis

Sex interaction

A 10-fold increase in PBDE-99 concentration was associated with more sleep irregularity in females by 0.19-point (95% CI: 0.01, 0.38, p= 0.04), while in males (β = −0.02, 95% CI: −0.32, 0.27, p= 0.89) no significant association was observed (Supplemental Table S8). However, the overlapping confidence intervals indicate that the difference between female and male estimates was not statistically significant.

Original CSHQ domains

When using the original CSHQ domains, bedtime resistance and sleep duration positively correlated with sleep irregularity. At age 3 years, a 10-fold increase in PBDE-99 concentration was associated with longer sleep duration (β = 0.5, 95% CI: −0.003, 1). For the original CSHQ domains, daytime sleepiness positively correlated with hypersomnolence. The original CSHQ domains, night wakings and parasomnias correlated with sleep disruption. A 10-fold increase PBDE-99 concentration was associated with more night wakings across all time points by 0.27-points (95% CI: 0.02, 0.52) (Supplemental Table S9).

Cotinine sensitivity results

Serum cotinine concentrations were most correlated with sleep irregularity (0.2, p-value < 0.01) and sleep duration (−0.48, p-value < 0.01) (Supplemental Table S10). While the estimate might have slightly increased or decreased for each PBDE and sleep outcome, the trend was similar (Table S11).

4. DISCUSSION

Our findings suggest that higher concentrations of PBDEs during gestation were adversely associated with various aspects of child sleep, with associations varying by PBDE congener, concentration (i.e., high, moderate, low), and child age and sex. Higher concentrations of several PBDE congeners, including PBDE-28, PBDE-47, PBDE-99, and ΣPBDEs, were associated with higher sleep irregularity. Notably, PBDE-28 showed age-specific associations, with significant associations observed at ages 5 and 8 years. These findings may reflect a lasting effect of prenatal PBDE exposure on child sleep irregularity or, alternatively, the influence of accumulating PBDE exposure during childhood, though the current analysis focused solely on gestational exposures. Higher continuous concentrations of PBDE (e.g., PBDE-47, PBDE-99, and ΣPBDEs) were associated with more sleep disruption, representing an average effect observed from ages 2 to 8 years. This suggests a robust and persistent relationship between PBDE exposure and disruptions in sleep patterns, possibly reflecting disruptions in neurodevelopmental processes that regulate sleep architecture. The contradictory association was lower hypersomnolence at age 4 years with higher PBDE-153 concentrations, suggesting specific developmental windows or vulnerabilities for this outcome, where higher PBDE-153 concentrations could have a differential effect on sleep outcomes, potentially reflecting age-specific or developmental factors that modify the impact of these chemicals. While PBDE exposure showed null associations with sleep duration, suggesting the complexity of PBDE associations with sleep duration. Although sleep duration decreased with age, it remained within the total sleep recommendations set by the American Academy of Sleep Medicine and the CDC.^76–78 As a result, our findings may be influenced by developmental factors not accounted for in our analysis. We found consistent evidence suggesting gestational PBDE concentrations were associated with sleep irregularity, sleep disruption, and sleep duration at age 8 years, and it is possible that sleep may be a mediator of PBDE-associated neurobehavioral problems such as cognitive function and behavior since sleep is crucial for healthy neurodevelopment. In addition, exploring window-specific susceptibility analysis would clarify the postnatal effects demonstrated (e.g. age 5 and 8 years). While previous studies have examined associations between gestational and delivery PBDE concentrations and child sleep outcomes^29–31, our study systematically extends this work by assessing gestational PBDE exposure measured at 16 weeks of pregnancy in relation to clinically-relevant child sleep outcomes spanning early to middle childhood (ages 2 to 8 years).

Although we found no significant evidence of sex differences in our main findings, examining sex-specific associations was important given prior research showing that male infants tend to have higher placental PBDE concentrations than female infants.⁷⁹ PBDE serum concentrations in infants are attributable to placental and lactation transfer from the mother during gestation and post-delivery^80,81, though childhood exposures were not analyzed in this analysis. Additionally, previous studies have demonstrated that PBDE exposure can disrupt thyroid hormone function differently in males and females, further underscoring the importance of considering sex in our analysis.⁸⁰ We did not find a difference in the original CSHQ domains and this study’s sleep health domains, suggesting that the results were not influenced by our choice to create specific sleep health categories. We did not find difference in our findings when including cotinine as time-varying, other than PBDE-28 at age 5 years. Our findings provide evidence to further support PBDEs’ neurotoxicity during childhood, particularly when exposure assessment is conducted during gestation.

We measured PBDE congeners −28, −47, −99, −100, and −153, the most prevalent congeners in humans, with PBDE-47 being the most predominant.^82–84 We used sleep questionnaires for the CSHQ to create clinically relevant sleep domains, which will provide better insight for pediatric clinicians. Sleep irregularity has been linked to heart disease and hypertension^85,86, while hypersomnolence is linked to sleep disorders⁸⁷, and sleep disruption is linked to mental health disorders, chronic health conditions, and overall quality of life.⁸⁸ The mechanisms underlying PBDE neurotoxicity remain unclear, but several biological pathways provide plausible explanations for the observed effects. PBDEs are known to cross the blood-brain barrier and accumulate in the central nervous system of wildlife and humans^12,89–91,, which could account for the repeated associations with sleep outcomes observed at ages 5 and 8 years. Furthermore, PBDEs have been shown to alter neurotransmitter uptake and release, directly influencing sleep regulation by promoting or disrupting sleep pathways.⁹² As previously discussed, PBDEs disrupt thyroid hormone homeostasis, which may indirectly affect sleep. Maternal thyroid hormones play a critical role in fetal development, particularly as fetal thyroid hormone production does not commence until approximately 18–22 weeks of gestation. Even beyond this point, the fetus remains partially reliant on maternal thyroid hormones to maintain homeostasis, especially in cases of insufficient fetal production.^4,93,94 PBDEs may interfere with thyroid hormone function through multiple mechanisms, including competitive binding to thyroid transport proteins such as thyroxin-binding globulin (TBG) and transthyretin (TTR), increased hormone metabolism and excretion, disruption of receptor activity, and altered deiodinase activity, which can affect circadian rhythms.^4,95,96 This interference suggests that thyroid disruption could serve as a mediating pathway for PBDE-induced neurotoxicity.⁴

The few age-specific sleep outcomes observed in this study may reflect developmental differences in vulnerability to PBDE exposure, with congener-specific impacts on critical windows of thyroid-regulated brain maturation. Previous research demonstrated that higher maternal PBDE-28 and PBDE-47 serum concentrations were associated with increased maternal serum T4 levels at 16 weeks of gestation, while PBDE-47 also elevated maternal T3 levels, suggesting that PBDEs can disrupt fetal thyroid homeostasis, which is a critical period for brain development.⁹⁷ Additionally, PBDE-99 has been reported to downregulate thyroid receptor alpha-1 and alpha-2 expression in rat cerebellar granular neurons, further implicating thyroid disruption in sleep and neurodevelopmental effects.⁹⁷ Thyroid hormone sulfotransferase (SULT), an enzyme involved in thyroid hormone metabolism, has been shown to alter T4 conversion rates and exhibits sex- and age-dependent activity in response to PBDE exposure. For example, PBDE exposure was positively associated with T3 SULT activity in males but inversely associated in females.^{4,80,91,98,99} Since altered T3 and T4 levels have been linked to deviations in sleep patterns, these findings may explain the observed sleep changes identified in this study. Given that thyroid hormones play a key role in neuronal differentiation, myelination, and circadian regulation, the interaction between specific PBDE congeners and thyroid pathways may affect sleep-related brain structures and process differently depending on timing of exposure. These mechanisms collectively highlight how PBDE exposure, particularly during sensitive developmental windows, may affect sleep health and neurodevelopment through complex interactions with thyroid hormones and other neurobiological pathways. Thus, developmental timing may be a key modifier of PBDE toxicity, helping to explain the age-specific patterns in sleep outcomes observed in our findings.

This study has several methodological strengths, including its use of a prospective study design and longitudinal data. We had sleep health outcomes at multiple time points, and examined several clinically-relevant sleep domains, while also accounting for a number of important confounders. This study has limitations due to its moderate sample size from a single site, potentially limiting the generalizability of the findings to other populations. However, the inclusion of repeated measures in the longitudinal dataset provided an in-depth analysis of temporal changes within individual participants, offering valuable insights into the progression of outcome variables under investigation. The PBDE serum concentrations in the HOME Study also reflect exposures at that time in the USA.¹⁸ This approach enhanced the scope of the research outcomes and underscored the need for future studies with larger sample sizes. The reliance on self-reported data from sleep health questionnaires posed another limitation, as it may have been influenced by recall bias or subjective interpretation. To mitigate these challenges, the study emphasized the use of clear, precise, and well-validated questionnaire items to reduce potential sources of bias and error. We performed several sensitivity analyses to better validate and understand our results. Despite these limitations, the prospective cohort design allowed for a comprehensive exploration of the relationships between repeated PBDEs serum concentrations and sleep health over time.

In addition, PBDEs have relatively long half-lives of several years, and we only assessed exposure at a single time point during pregnancy, so future research would benefit from examining cumulative gestational and postnatal exposure to better assess window-specific effects. While this study used summed PBDE concentrations (ΣPBDE) to align with prior literature and support interpretability, future research should leverage advanced mixture modeling techniques to better account for more complex mixtures and assess potential joint or interactive effects; however PBDEs are highly correlated¹⁰⁰ so mixture methods may or may not work well in these circumstances. Still, these approaches might enhance our understanding of how complex chemical mixtures influence child sleep health and may uncover patterns not captured through traditional summation methods.

5. CONCLUSION

This study provides evidence that higher gestational PBDE serum concentrations are associated with altered sleep patterns in 2–8 year old children, including increased sleep irregularity and sleep disruption. These associations varied across specific PBDE congeners and child ages, highlighting the potential for developmental windows of vulnerability. The findings contribute to the body of literature indicating that gestational exposure to endocrine disrupting chemicals, such as PBDEs, may have lasting impacts on sleep health in childhood, with potential implications for broader neurodevelopmental and psychosocial outcomes.

Understanding the effects of gestational PBDE exposure on postnatal outcomes could inform improved sleep health screening and exposure mitigation strategies. Given the critical role of sleep in cognitive, emotional, and physical development, reducing PBDE exposure during key developmental windows, particularly gestation, is important. Future research should investigate the mechanisms driving these associations and explore interventions to reduce PBDE-related sleep disturbances, assess population-level effects and individual variability, and interrogate the impact of chemical mixtures (i.e. other POPs and flame retardants).

Highlights.

• Increased gestational PBDE-99, PBDE-47, and ΣPBDEs were associated with higher sleep irregularity from ages 2 to 8 years.

• Increased gestational PBDE-28 was associated with higher sleep irregularity, specifically at ages 5 and 8 years.

• Increased gestational PBDE-47, PBDE-99, and ΣPBDEs were associated with greater sleep disruption from ages 2 to 8 years.

• Gestational PBDEs may impact multiple aspects of sleep health in children, underscoring sleep as a potential mediator in PBDE-related neurobehavioral outcomes.