Gestational PBDE Concentrations and Executive Function in Adolescents with Self- and Caregiver-Report: The HOME Study

Abstract

Background:

Polybrominated diphenyl ethers (PBDEs), synthetic chemicals previously used as flame retardants in commercial products, impact human behaviors, mood symptoms and cognitive abilities.

Objective:

We estimated the association of gestational PBDE serum concentrations with early adolescent self- and caregiver-reported ratings of executive function in a prospective pregnancy and birth cohort.

Methods:

We measured gestational serum concentrations of five PBDE congeners and created a summary exposure variable (Σ₅BDE: -28, -47, -99, -100 and -153). At age 12 years, we assessed executive function for 237 adolescents using self- and caregiver-reports with the Behavior Rating Inventory of Executive Functioning (BRIEF-2). We used multivariable linear regression models to estimate covariate-adjusted associations of lipid standardized, log₁₀₋ transformed gestational PBDE concentrations with BRIEF-2 T-scores. We evaluated potential effect measure modification (EMM) of sex by examining sex-stratified regression models.

Results:

As higher scores indicate greater deficits in executive function, gestational PBDE concentrations were positively associated with adolescent-reported BRIEF-2 T-scores for Global Executive Composite (BDE-28: β=6.31 (95%CI: 2.59, 10.03), BDE-47: (β=3.32 (95%CI: 0.1, 6.54), Σ₅BDE: (β=3.70 (95%CI: 0.37, 7.03), Behavior Regulation Index (BDE-28: β=5.36 (95%CI: 1.56, 9.15), BDE-99: β=3.53 (95%CI: 0.33, 6.74), Σ₅BDE: β=3.93 (95%CI: 0.57, 7.3), Emotion Regulation Index (BDE-28: β=4.76 (95%CI: 0.88, 8.64) and the Cognitive Regulation Index (BDE-28: β=6.69 (95%CI: 3.08, 10.31), BDE-47: β=3.45 (95%CI: 0.3, 6.59), Σ₅BDE: β=3.57 (95%CI: 0.32, 6.82) and several other scales. We observed stronger associations with gestational PBDE concentrations for all congeners among males, especially for the caregiver-rated scales (all EMM p-values <0.1).

Discussion:

This study provides evidence that gestational PBDE serum concentrations may adversely influence offspring executive function during adolescence.

Introduction

Chronic and persistent exposures to polybrominated diphenyl ethers (PBDEs) prior to and during pregnancy lead to potential adverse effects in the developing fetus. Gestational exposure to PBDEs are of research interest as the offspring of those ubiquitously exposed to significant concentrations of these flame retardants in the United States are reaching adolescence and young adulthood (EPA, 2010; Frederiksen et al., 2009). Cohort studies examining childhood behavioral outcomes describe adverse adaptive, internalizing, externalizing and social behaviors associated with gestational PBDE exposures (Braun et al., 2014; Braun et al., 2017b; Cecil et al., 2024; Chen et al., 2014; Cowell et al., 2015; Eskenazi et al., 2013; Hartley et al., 2022; Sagiv et al., 2015; Vuong et al., 2017). However, only a few studies have examined the relationship of gestational PBDE concentrations with executive function. Executive function encompasses a set of processes involved in inhibiting behavior, modulating emotions, and switching between activities. One study found higher gestational BDE-153 concentrations were significantly associated with poorer behavior regulation in children ages 5 and 8 years participating in the Health Outcomes and Measures of the Environment (HOME) Study, including inhibitory and emotional control (Vuong et al., 2016). A study from the Center for the Health Assessment of Mothers and Children of Salinas (CHAMACOS) Study (Sagiv et al., 2015) observed poorer executive function in children ages 9 and 12 years with higher gestational Σ₄PBDE (BDE-47, -99, -100, -153) concentrations, specifically for working memory. In contrast, a study cohort from Menorca, Spain assessed with the McCarthy Scales of Children’s Abilities (MSCA) at age 4 years found no associations between gestational or childhood PBDEs and executive function (Gascon et al., 2011).

The objective of this study was to assess fetal exposures during gestation by examining the association of gestational PBDE concentrations and executive function at age 12 years as reported by both adolescents themselves and their caregivers. Environmental epidemiological studies often demonstrate adverse effects of a suspected neurotoxicant by reporting changes in childhood executive function with caregiver reporting. However, few have examined the adolescent self-report of these skills. This measurement at the start of adolescence allowed for self- and caregiver-ratings and provided multiple perspectives of executive function. It is important to measure executive function during adolescence when development of these skills is most rapid (Tervo-Clemmens et al., 2023). The study of this period may provide unique perspective and knowledge as these skills are being integrated with one another. Developmental science has explored the effects of age, puberty and brain maturation along with other genetic and social factors to understand how adolescent perturbations impact adult outcomes associated with executive function. However, from a public health perspective, recognizing adverse effects on executive function from gestational PBDE exposures is essential to reinforce efforts at exposure prevention for these and similar toxicants, understand modifiers, and identify potential interventional strategies that could be applied prior to effects lasting into adulthood.

Methods

Study Participants

We analyzed data from the HOME Study, an on-going prospective pregnancy and birth cohort study conducted in the Greater Cincinnati, Ohio area. The HOME Study examines the associations between gestational and childhood exposures to a variety of environmental toxicants and health and developmental outcomes from infancy through early adulthood. Between 2003 and 2006, 468 pregnant women were enrolled, and 409 remained in the study and delivered live-born infants. Subsequent follow-up, occurring at multiple time points, involved 420 children (11 twin sets included) and their caregivers. Details of enrollment and data collection have been described elsewhere (Braun et al., 2020; Braun et al., 2017a).

The Institutional review boards at Cincinnati Children’s Hospital Medical Center (CCHMC) and the enrolling delivery hospitals approved this study. The Centers for Disease Control and Prevention (CDC) laboratory’s involvement did not constitute engagement in human-subjects research. Caregivers provided written informed consent for their own participation as well as their child. At age 12 years, adolescents provided written informed assent.

Gestational PBDE Concentrations

We measured 10 PBDE congeners (−17, −28, −47, −66, −85, −99, −100, −153, −154, −183) in maternal serum samples collected at 16 weeks gestation. If serum samples were not available for mothers at 16 weeks, we substituted measures from 26 weeks (n=2). The samples were stored at −80°C until measurement of PBDE concentrations at the CDC using gas chromatography/isotope dilution high-resolution mass spectrometry (Jones et al., 2012; Sjodin et al., 2004; Sjodin and Pirkle). Details about PBDE measurements including quality assurance/quality control and limits of detection (LOD) have been previously described (Chen et al., 2014; Vuong et al., 2016). PBDE concentrations were expressed on a serum lipid basis (ng/g lipid). In the current study, we focused on BDE-28, -47, -99, -100, -153, the top five most abundant congeners with over 80% detection frequency among the participants. For concentrations below the LOD, we imputed a value of LOD divided by the square root of 2 (Hornung and Reed, 1990). For a small number of samples with missing results of BDE-28 (n=36), BDE-99 (n=1) and BDE-100 (n=1), we imputed with the median value of the corresponding congener. We calculated the arithmetic sum of the 5 congeners (Σ₅BDE) as an aggregate measure.

The BRIEF Assessment

We assessed adolescent executive function at age 12 years using the Behavior Rating Inventory of Executive Function, second edition (BRIEF-2) (Gioia et al., 2015). The BRIEF-2 is a validated assessment of executive function within the context of everyday functioning. We collected both adolescent self-report (55-items) and caregiver-report (63-items) versions. Both versions provide T-scores for a Global Executive Composite and three indices: Behavior Regulation Index, Emotion Regulation Index, Cognitive Regulation Index, and subscales including: inhibit, shift, emotion control, working memory, plan/organize, and self-monitor. Task completion is an additional scale on the self-report version; additional scales on the caregiver-report version include initiate, organization of materials, and task monitoring. For both self- and caregiver-report versions, the BRIEF-2 T-scores have a mean of 50 and standard deviation of 10, with higher scores indicating greater deficits in executive function. A BRIEF-2 T-score of 65 or above is considered potentially clinically elevated (Gioia et al., 2015).

Covariates

We identified potential covariates for our analysis based on prior studies. We collected demographic and socioeconomic characteristic information through questionnaires administered by trained research assistants. We measured blood lead and serum cotinine in maternal samples collected during pregnancy. We also measured serum concentrations of polychlorinated biphenyls (PCBs), dichlorodiphenyldichloroethylene (DDE) and dichlorodiphenyltrichloroethane (DDT) in maternal samples collected at 16 weeks of gestation. We calculated an aggregate concentration measure of total lipid adjusted PCBs (Σ₆PCB) as the sum of the six major PCB congeners measured (PCB-74, -99, -118, -138/-158, -153, -180). In addition, we measured organophosphate (OP) pesticide metabolites in maternal urine samples collected during pregnancy. We further derived an aggregate measure of total dialkylphosphate (DAP) as the sum of the OP pesticide metabolites measured as (ΣDAP). The caregivers completed the Beck Depression Inventory (BDI-II) (Beck et al., 1996a; Beck et al., 1996b) at the age 12 visit. We used the BDI-II total score as a measure of maternal mental health in our analysis. The caregivers also completed the Behavior Assessment System for Children, third edition, (BASC-3) Parenting Relationship Questionnaire (PRQ) at the age 12 visit (Kamphaus and Reynolds, 2006). We used the PRQ Relational Frustration subscale, which measures the caregivers’ levels of stress or frustration in relating to or controlling the behavior of the adolescent, as an indicator of caregiver-child relationship.

Statistical Analysis

We started with univariate analyses examining the distributions and computed descriptive statistics for the PBDE concentrations, BRIEF-2 T-scores, and the covariates. We reported mean (standard deviation), frequency (percent), median (inter-quartile range, IQR) and geometric mean (95% confidence interval), as appropriate. We compared caregiver and adolescent characteristics for the 237 adolescents included in the study with the 183 who were not included due to loss to follow-up using t-test, Chi-square test or analysis of variance, as appropriate. We defined statistical significance at a threshold of p<0.05. We compared the BRIEF-2 T-scores from adolescent self-report and caregiver-report versions using paired t-test and assessed their correlation using Pearson’s correlation analysis. We log-transformed the PBDE concentrations prior to further analysis to reduce the influence of right-skewed distribution of PBDE concentrations. We chose a base of 10 for the log-transformation, therefore the regression coefficients for each PBDE correspond to each tenfold increase of the PBDE concentration. We fitted general additive models (GAM) to evaluate the non-linear trend in the dose-response relationship between PBDE concentrations and BRIEF-2 T-scores. We did not observe any significant non-linear relationships; therefore, we relied on linear regression analysis. We examined the associations of PBDE concentrations (−28, −47, −99, −100, −153, and Σ₅BDE, separately) with BRIEF-2 T-scores in both unadjusted and adjusted linear regression models. Guided by a directed acyclic graph (DAG, see Supplemental Figure 1), we adjusted for the following covariates in our primary analysis: maternal age at delivery, maternal education, maternal BDI-II total score and PRQ relational frustration T-score at the age 12 visit, and maternal serum cotinine (calculated as the mean of maternal serum cotinine at 16-week and 26-week gestation). In addition to regression estimates and their 95%CI, we also calculated false discovery rate (FDR) adjusted p-values using the Benjamini and Hochberg method to facilitate result interpretation (Benjamini and Hochberg, 1995). We evaluated effect measure modification (EMM) of adolescent sex by running sex-stratified models and estimating EMM p-values (Buckley et al., 2017). We also evaluated EMM of breastfeeding duration, given findings reported in a previous study (Wallenborn et al., 2024). We examined models stratified by median breastfeeding duration (<16 weeks vs. >=16 weeks). The duration was obtained by maternal report and accounted for any breastfeeding. We considered the EMM to be statistically significant if p-value <0.1 (Cecil et al., 2024; Chen et al., 2014; Kuiper et al., 2023; Strawn et al., 2022; Vuong et al., 2017). As a secondary analysis, we additionally adjusted for maternal urinary DAP to account for its potential confounding since it was correlated with both maternal PBDE and the BRIEF scores. We also explored maternal blood lead, maternal serum Σ₆PCB, maternal serum DDT, and maternal serum DDE as potential confounders. Since none of them were significantly associated with maternal PBDE concentrations or BRIEF-2 T-scores (p-values>0.2), we did not pursue further adjustment of these biomarkers in our models. In sensitivity analyses, we examined models excluding participants with extreme values (>1.5 IQR beyond the 75^th percentile value) in PBDEs (n = 9, 6, 6, 6, 6, 4 for analysis of BDE-28, -47, -99, -100, -153, and Σ₅BDE, respectively), as well as models without imputed values for missing results in BDE-28, -99, -100 and Σ₅BDE described earlier. In addition, we examined BRIEF-2 T-scores as a dichotomous outcome, for which we assessed the association between gestational PBDE concentrations and risk of having an at-risk BRIEF-2 T-score (>=65) using modified Poisson regression models (Zou, 2004). We examined BRIEF-2 T-scores for composites and indices in the potentially clinically elevated range (>= 65), given that this is a low-risk and typically developing cohort. Statistical analyses were performed using SAS^® version 9.4 (SAS Institute) and graphs were generated using SAS version 9.4 or R package ggplot2.

Results

Descriptive Statistics - Participants

From the 420 children in the HOME Study cohort, 394 had PBDE concentrations collected from mothers during pregnancy and 244 (62%) completed the BRIEF-2 at the age 12 visit. We randomly excluded one adolescent from each of the 7 twin sets, resulting in a final sample of 237 adolescents for the current study. The adolescents included in this analysis were approximately 12 years (12.4 ± 0.7) at the study visit, and 55.7% were female. The mothers had a mean age of 29.3 years at delivery, were mostly non-Black, married or living with a partner, and completed college or graduate/professional school. The median household income at the age 12 visit was $75 K (IQR: 35K – 145 K). The maternal and adolescent characteristics of the 237 participants were very similar to the 183 not included in the current study (Table 1).

Table 1.

Characteristics of adolescents included in current study and those not included

Characteristic	Included (n=237)	Not included (n=183)
Child sex, female	132 (55.7%)	94 (51.4%)
Maternal age at delivery (Mean ± SD)	29.3 ± 5.7	29.3 ± 5.8
Maternal racea
Black	83 (35%)	48 (27%)
Non-Black	154 (65%)	130 (73%)
Missing		5
Maternal marital status at baseline#a
Married	149 (65.4%)	120 (67.4%)
Not married, living with partner	25 (11%)	31 (17.4%)
Not married, living alone	54 (23.7%)	27 (15.2%)
Missing	9	5
Maternal marital status at age 12 visit
Married	156 (65.8%)
Not married, living with partner	23 (9.7%)
Not married, living alone	58 (24.5%)
Maternal education at baselinea
High school or less	49 (21.5%)	46 (25.8%)
Some college	66 (28.9%)	36 (20.2%)
College graduate	70 (30.7%)	50 (28.1%)
Graduate or professional	43 (18.9%)	46 (25.8%)
Missing	9	5
Maternal education at age 12 visit
High school or less	34 (14.3%)
Some college	75 (31.6%)
College graduate	69 (29.1%)
Graduate or professional	59 (24.9%)
Household income at baseline (Median (IQR))	$55K ($27.5K, $85K)	$55K ($22.5K, $85K)
Household income at age 12 visit (Median (IQR))	$75K ($35K, $145K)
Maternal BDI-II score at baseline (Mean ± SD)	9.9 ± 6.6	10 ± 7.3
Maternal BDI-II score at age 12 visit (Mean ± SD)	6.7 ± 6.6
BASC-3 PRQ relation frustration score at age 12 visit (Mean±SD)	48.2 ± 9.1

^#Comparison between children included and those not included significantly different (p<0.05);

^aMissing not counted in the denominator for percentage;

SD, standard deviation; IQR, interquartile range; Beck Depression Inventory-II, BDI-II; Behavioral Assessment System for Children, Third Edition, Parenting Relationship Questionnaire, BASC-3 PRQ

Gestational PBDE Concentrations

The five PBDE congeners were detected in at least 89% of the maternal serum samples. The geometric means of PBDE serum concentrations were 1.1, 20.7, 4.7, 3.8, 4.7 and 37.4 ng/g lipid for BDE-28, -47, -99, -100, -153 and Σ₅, respectively (Table 2). The individual concentrations of the 5 congeners were moderately to strongly correlated as illustrated in a heat map (Supplemental Figure 2), with Pearson’s correlation coefficients ranging from 0.48 (between BDE-28 and BDE-153) to 0.92 (between BDE-47 and BDE-99). They were also strongly correlated with Σ₅BDE (Pearson’s correlation coefficient ranging from 0.74 to 0.97), as expected.

Table 2.

Gestational serum PBDE concentrations (ng/g lipid) in maternal samples (2003–2006)

Analyte	N	Min	Max	Mean	SD	P25	Median	P75	Geomean (95%CI)	LOD	Percent>LOD
BDE-28	237	0.3	16.7	1.5	1.8	0.7	1.0	1.5	1.1 (1.0 – 1.2)	0.4	89.0
BDE-47	237	2.2	539.0	34.7	56.6	11.0	19.8	34.5	20.7 (18.3 – 23.3)	0.5	99.2
BDE-99	237	0.6	193.0	8.4	15.8	2.5	4.5	7.8	4.7 (4.2 – 5.3)	0.4	99.2
BDE-100	237	0.4	107.0	6.8	11.6	2.0	3.5	6.6	3.8 (3.4 – 4.3)	0.4	95.8
BDE-153	237	0.6	152.0	8.7	15.9	2.4	4.1	8.1	4.7 (4.2 – 5.3)	0.4	95.8
Σ5BDE	237	5.2	905.8	60.1	91.2	19.9	34.2	65.6	37.4 (33.4 – 41.9)	N/A	N/A

Minimum, Min; Maximum, Max; Arithmetic mean, Mean; Standard Deviation, SD; 25th percentile, P25; 75th percentile, P75; Geometric mean, Geomean; 95% confidence interval, 95%CI; Limit of detection, LOD; Σ₅BDE is calculated as arithmetic sum of BDE-28, -47, -99, -100 and -153 gestational serum concentrations, N/A – not applicable.

BRIEF-2 outcomes

Data were available from 235 adolescent self-reported and 237 caregiver-reported BRIEF-2 assessments. In general, T-scores were within a typical range for the scale (e.g., for global executive composite, mean (SD) for adolescent self-reported = 56.5 (10.1) and caregiver-reported = 52.7 (10.6)) forms. On average, the T-scores were higher across all subscales for the adolescent self-reported version (means ranging from 52.7 to 57.3) than the caregiver-reported version (means ranging from 50.8 to 54.6). Using paired t-tests, we consistently found that adolescents reported more problems than their caregivers. For the 10 scales/indices available in both adolescent-reported and caregiver-reported versions, mean differences of T-scores ranged from 1.3 to 5.7 (p-values <0.001 for all, except the Emotional Control scale). T-scores on the adolescent-reported measure were moderately correlated with the caregiver-reported measure (Pearson’s correlation coefficient = 0.24 – 0.48, Supplemental Table 1). Proportions of T-scores at or above 65 (potentially clinically elevated cutoff) were generally higher in the adolescent-reported version (10.6% to 26%) than the caregiver-reported version (10.5% to 20.3%) (Supplemental Tables 1 and 2).

Associations of PBDE concentrations with BRIEF-2 outcomes

Bivariate regression models showed that higher gestational PBDE concentrations were significantly associated with higher adolescent-reported BRIEF-2 T-scores (Table 3). The pattern of the associations was consistent across different BRIEF-2 scales and PBDE congeners. Higher gestational PBDE concentrations were consistently associated with higher caregiver-reported T-scores, but we found fewer significant associations, and the estimates were smaller relative to the observed adolescent associations.

Table 3.

Differences in adolescent self-reported Behavior Rating Inventory of Executive Functioning (BRIEF-2) T-scores at age 12 years per each tenfold increase of gestational serum PBDE concentrations

		Unadjusted	Adjusted*
BRIEF Scales and Indices	Congeners	Estimate (95%CI)	Estimate (95%CI)
Inhibit	BDE-28	6.58 (2.6, 10.56)	5.5 (1.69, 9.31) a
	BDE-47	4.02 (0.75, 7.29)	3.28 (0.002, 6.56)
	BDE-99	3.71 (0.54, 6.89)	3.34 (0.11, 6.56)
	BDE-100	3.44 (0.27, 6.62)	3.19 (0.03, 6.35)
	BDE-153	3.61 (0.44, 6.77)	3.35 (0.36, 6.33)
	Σ5BDE	4.85 (1.44, 8.27)	4.21 (0.83, 7.58) b
Self-Monitor	BDE-28	5.44 (1.4, 9.47)	4.15 (0.29, 8.02) a
	BDE-47	3.98 (0.69, 7.27)	2.34 (−0.98, 5.66)
	BDE-99	4.44 (1.25, 7.62)	3.19 (−0.05, 6.44)
	BDE-100	3.16 (−0.04, 6.36)	1.86 (−1.34, 5.06)
	BDE-153	2.25 (−0.96, 5.46)	1.57 (−1.46, 4.6)
	Σ5BDE	4.37 (0.92, 7.82)	2.82 (−0.61, 6.24)
Shift	BDE-28	4.41 (0.3, 8.53)	3.25 (−0.82, 7.33)
	BDE-47	2.38 (−0.99, 5.74)	1.54 (−1.95, 5.02)
	BDE-99	1.89 (−1.38, 5.15)	0.99 (−2.44, 4.43)
	BDE-100	1.94 (−1.32, 5.2)	1.63 (−1.73, 4.99)
	BDE-153	1.81 (−1.44, 5.07)	1.79 (−1.39, 4.97)
	Σ5BDE	2.77 (−0.76, 6.29)	2.05 (−1.55, 5.66)
Emotional Control	BDE-28	6.22 (2.56, 9.87)	5.57 (1.92, 9.22) a
	BDE-47	3.82 (0.81, 6.82)	2.94 (−0.21, 6.09)
	BDE-99	3.56 (0.64, 6.47)	2.94 (−0.15, 6.04)
	BDE-100	2.54 (−0.39, 5.47)	1.78 (−1.27, 4.83)
	BDE-153	1.17 (−1.77, 4.11)	0.66 (−2.24, 3.55)
	Σ5BDE	3.88 (0.73, 7.03)	3 (−0.26, 6.26)
Working Memory	BDE-28	8.17 (4.17, 12.17)	7.27 (3.3, 11.25) a
	BDE-47	4.13 (0.8, 7.45)	3.55 (0.09, 7.01)
	BDE-99	3.4 (0.16, 6.63)	3.05 (−0.36, 6.46)
	BDE-100	3.41 (0.18, 6.64)	3.09 (−0.25, 6.43)
	BDE-153	1.55 (−1.7, 4.8)	1.27 (−1.91, 4.45)
	Σ5BDE	4.27 (0.78, 7.75)	3.67 (0.09, 7.25) b
Plan/Organize	BDE-28	5.81 (2.19, 9.44)	5 (1.37, 8.63) a
	BDE-47	2.81 (−0.18, 5.8)	2.57 (−0.56, 5.7)
	BDE-99	2.36 (−0.55, 5.27)	2.24 (−0.84, 5.32)
	BDE-100	1.85 (−1.06, 4.76)	1.95 (−1.07, 4.97)
	BDE-153	1.16 (−1.75, 4.07)	1.2 (−1.67, 4.06)
	Σ5BDE	2.91 (−0.23, 6.04)	2.68 (−0.55, 5.92)
Task Completion	BDE-28	7.95 (4.02, 11.88)	6.63 (2.78, 10.47) a
	BDE-47	4.3 (1.04, 7.56)	3.65 (0.33, 6.98)
	BDE-99	3.31 (0.13, 6.48)	2.79 (−0.5, 6.07)
	BDE-100	3.42 (0.25, 6.59)	3.35 (0.14, 6.56)
	BDE-153	1.68 (−1.5, 4.87)	1.63 (−1.43, 4.7)
	Σ5BDE	4.38 (0.96, 7.8)	3.8 (0.36, 7.24) b
Behavior Regulation Index	BDE-28	6.63 (2.61, 10.64)	5.36 (1.56, 9.15) a
	BDE-47	4.35 (1.06, 7.64)	3.14 (−0.12, 6.41)
	BDE-99	4.33 (1.14, 7.52)	3.53 (0.33, 6.74)
	BDE-100	3.59 (0.39, 6.79)	2.84 (−0.31, 5.99)
	BDE-153	3.3 (0.1, 6.5)	2.83 (−0.15, 5.81)
	Σ5BDE	5.05 (1.61, 8.48)	3.93 (0.57, 7.3) b
Emotion Regulation Index	BDE-28	5.75 (1.85, 9.65)	4.76 (0.88, 8.64) a
	BDE-47	3.31 (0.11, 6.52)	2.39 (−0.95, 5.73)
	BDE-99	2.92 (−0.19, 6.03)	2.1 (−1.18, 5.38)
	BDE-100	2.4 (−0.71, 5.52)	1.86 (−1.37, 5.08)
	BDE-153	1.65 (−1.47, 4.76)	1.38 (−1.67, 4.43)
	Σ5BDE	3.57 (0.22, 6.93)	2.72 (−0.73, 6.17)
Cognitive Regulation Index	BDE-28	7.76 (4.09, 11.43)	6.69 (3.08, 10.31) a
	BDE-47	3.94 (0.88, 6.99)	3.45 (0.3, 6.59)
	BDE-99	3.18 (0.2, 6.16)	2.86 (−0.24, 5.96)
	BDE-100	3 (0.03, 5.98)	2.92 (−0.11, 5.96)
	BDE-153	1.51 (−1.47, 4.5)	1.42 (−1.47, 4.31)
	Σ5BDE	4.04 (0.83, 7.25)	3.57 (0.32, 6.82) b
Global Executive Composite	BDE-28	7.47 (3.64, 11.3)	6.31 (2.59, 10.03) a
	BDE-47	4.14 (0.97, 7.31)	3.32 (0.1, 6.54)
	BDE-99	3.62 (0.54, 6.71)	3.04 (−0.13, 6.22)
	BDE-100	3.21 (0.12, 6.29)	2.83 (−0.28, 5.94)
	BDE-153	2.14 (−0.95, 5.24)	1.9 (−1.06, 4.85)
	Σ5BDE	4.47 (1.15, 7.79)	3.7 (0.37, 7.03) b

Adjusted for maternal age, maternal education, maternal BDI-II score, PRQ relational frustration score, and maternal serum cotinine.

^aFDR adjusted p-value for adjusted association <0.05;

^bFDR adjusted p-value for adjusted association <0.1

In covariate-adjusted analyses (Table 3, Figure 1), gestational PBDE concentration associations with adolescent-reported T-scores were slightly attenuated from the bivariate models, yet many remained statistically significant. Each tenfold increase in gestational Σ₅BDE concentration was associated with higher adolescent reported BRIEF-2 T-scores for global executive composite [β=3.70 (95%CI: 0.37, 7.03)], behavior regulation index [β=3.93 (95%CI: 0.57, 7.3)], cognitive regulation index [β=3.57 (95%CI: 0.32, 6.82)], inhibit [β=4.21 (95%CI: 0.83, 7.58)], task completion [β=3.80 (95%CI: 0.36, 7.24)], and working memory [β=3.67 (95%CI: 0.09, 7.25)]. Gestational BDE-28 concentrations were also associated with the above scales, as well as emotion regulation index, emotional control, self-monitor, and plan/organize. Several significant associations between gestational BDE-47 and other composites, indices and scales (global executive composite, cognitive regulation index, task completion, inhibit and working memory) were also observed. Of the 25 significant adjusted associations, 10 associations with BDE-28 had FDR adjusted p-values<0.05; an additional 6 associations with Σ₅BDE had FDR adjusted p-value <0.1.

Figure 1.

Adjusted difference in Behavior Rating Inventory of Executive Functioning (BRIEF-2) T-scores per each tenfold increase of gestational serum PBDE concentrations

All models adjusted for maternal age, maternal education, maternal Beck Depression Inventory II (BDI-II) score, PRQ relational frustration score, and maternal serum cotinine.

In covariate-adjusted analyses (Table 4, Figure 1), gestational PBDE concentration associations with caregiver-reported T-scores were significant only for one index and two subscales. Each tenfold increase in gestational BDE concentration was associated with higher behavior regulation index [BDE-99: β=2.89 (95%CI: 0.29, 5.49), BDE-153: β=2.54 (95%CI: 0.14, 4.94) and Σ₅BDE: β=2.78 (95%CI: 0.05, 5.52)), task-monitor (BDE-28: β=3.71 (95%CI: 0.07, 7.36), and self-monitor (Σ₅BDE: β=2.82 (95%CI: 0.01, 5.63)] subscales. The FDR adjusted p-values of these associations were all >0.1.

Table 4.

Differences in caregiver-rated BRIEF-2 T-scores at age 12 years per each tenfold increase of gestational serum PBDE concentrations

		Unadjusted	Adjusted*
BRIEF Scales and Indices	Congeners	Estimate (95%CI)	Estimate (95%CI)
Inhibit	BDE-28	4.3 (0.06, 8.54)	2.85 (−0.47, 6.17)
	BDE-47	3.78 (0.34, 7.23)	2.25 (−0.58, 5.08)
	BDE-99	3.45 (0.1, 6.79)	2.75 (−0.03, 5.53)
	BDE-100	2.85 (−0.5, 6.19)	1.79 (−0.94, 4.52)
	BDE-153	3.18 (−0.15, 6.51)	2.41 (−0.16, 4.98)
	Σ5BDE	4.11 (0.51, 7.72)	2.62 (−0.3, 5.55)
Self-Monitor	BDE-28	4.4 (0.45, 8.34)	2.97 (−0.22, 6.15)
	BDE-47	3.63 (0.42, 6.84)	2.5 (−0.22, 5.22)
	BDE-99	3.19 (0.07, 6.31)	2.84 (0.17, 5.51)
	BDE-100	3.11 (0, 6.22)	2.51 (−0.11, 5.12)
	BDE-153	2.82 (−0.29, 5.93)	2.29 (−0.18, 4.76)
	Σ5BDE	3.91 (0.55, 7.27)	2.82 (0.01, 5.63)
Shift	BDE-28	2.67 (−1.43, 6.78)	0.84 (−2.64, 4.32)
	BDE-47	2.87 (−0.46, 6.2)	0.74 (−2.22, 3.71)
	BDE-99	3.52 (0.3, 6.74)	2.07 (−0.84, 4.99)
	BDE-100	2.57 (−0.66, 5.79)	1 (−1.85, 3.86)
	BDE-153	0.95 (−2.29, 4.18)	0.11 (−2.59, 2.81)
	Σ5BDE	2.86 (−0.64, 6.35)	0.79 (−2.28, 3.85)
Emotional Control	BDE-28	3.52 (−0.68, 7.71)	2.4 (−0.9, 5.71)
	BDE-47	2.37 (−1.05, 5.79)	1.09 (−1.74, 3.92)
	BDE-99	2.25 (−1.07, 5.57)	1.81 (−0.96, 4.59)
	BDE-100	2.28 (−1.03, 5.59)	1.49 (−1.22, 4.21)
	BDE-153	2.26 (−1.04, 5.56)	1.56 (−1.01, 4.12)
	Σ5BDE	2.73 (−0.85, 6.31)	1.5 (−1.42, 4.41)
Initiate	BDE-28	3.57 (−0.35, 7.49)	1.68 (−1.53, 4.89)
	BDE-47	2.55 (−0.64, 5.74)	0.77 (−1.98, 3.51)
	BDE-99	1.98 (−1.13, 5.08)	0.8 (−1.9, 3.5)
	BDE-100	1.65 (−1.45, 4.75)	0.47 (−2.17, 3.11)
	BDE-153	0.43 (−2.67, 3.53)	−0.19 (−2.69, 2.3)
	Σ5BDE	2.23 (−1.12, 5.59)	0.51 (−2.32, 3.35)
Working Memory	BDE-28	4.55 (−0.04, 9.14)	2.09 (−1.79, 5.97)
	BDE-47	4.49 (0.77, 8.21)	1.87 (−1.44, 5.18)
	BDE-99	3.69 (0.06, 7.31)	1.7 (−1.56, 4.95)
	BDE-100	3.2 (−0.42, 6.82)	1.26 (−1.92, 4.44)
	BDE-153	1.08 (−2.55, 4.71)	0.08 (−2.94, 3.09)
	Σ5BDE	3.97 (0.06, 7.88)	1.42 (−2, 4.84)
Plan/Organize	BDE-28	4.34 (0.48, 8.19)	2.57 (−0.62, 5.76)
	BDE-47	3.2 (0.05, 6.34)	1.54 (−1.19, 4.27)
	BDE-99	2.49 (−0.57, 5.54)	1.47 (−1.22, 4.15)
	BDE-100	2.36 (−0.69, 5.41)	1.26 (−1.37, 3.89)
	BDE-153	1.36 (−1.69, 4.41)	0.73 (−1.76, 3.21)
	Σ5BDE	3.15 (−0.15, 6.44)	1.56 (−1.26, 4.38)
Task-Monitor	BDE-28	4.87 (0.79, 8.95)	3.71 (0.07, 7.36)
	BDE-47	3 (−0.33, 6.34)	2.32 (−0.8, 5.45)
	BDE-99	1.43 (−1.82, 4.68)	1.25 (−1.83, 4.34)
	BDE-100	2.01 (−1.23, 5.25)	1.62 (−1.39, 4.63)
	BDE-153	0.56 (−2.68, 3.8)	0.19 (−2.66, 3.05)
	Σ5BDE	2.54 (−0.96, 6.04)	1.79 (−1.44, 5.02)
Organization of Materials	BDE-28	2.98 (−1.05, 7.01)	1.58 (−1.86, 5.02)
	BDE-47	2.73 (−0.54, 6.01)	1.22 (−1.72, 4.15)
	BDE-99	1.93 (−1.26, 5.11)	0.99 (−1.9, 3.88)
	BDE-100	1.96 (−1.21, 5.14)	0.83 (−1.99, 3.65)
	BDE-153	2.27 (−0.89, 5.44)	1.61 (−1.05, 4.27)
	Σ5BDE	2.91 (−0.52, 6.34)	1.45 (−1.58, 4.48)
Behavior Regulation Index	BDE-28	4.48 (0.35, 8.6)	2.94 (−0.17, 6.04)
	BDE-47	3.86 (0.5, 7.21)	2.37 (−0.28, 5.02)
	BDE-99	3.51 (0.25, 6.76)	2.89 (0.29, 5.49)
	BDE-100	3.06 (−0.2, 6.31)	2.1 (−0.45, 4.65)
	BDE-153	3.26 (0.02, 6.51)	2.54 (0.14, 4.94)
	Σ5BDE	4.22 (0.71, 7.73)	2.78 (0.05, 5.52)
Emotion Regulation Index	BDE-28	3.22 (−0.96, 7.4)	1.71 (−1.51, 4.94)
	BDE-47	2.75 (−0.65, 6.15)	0.9 (−1.86, 3.66)
	BDE-99	3.05 (−0.24, 6.34)	2.06 (−0.64, 4.76)
	BDE-100	2.66 (−0.63, 5.95)	1.38 (−1.27, 4.02)
	BDE-153	1.93 (−1.36, 5.22)	1.06 (−1.45, 3.57)
	Σ5BDE	2.99 (−0.56, 6.55)	1.21 (−1.64, 4.06)
Cognitive Regulation Index	BDE-28	4.52 (0.4, 8.64)	2.56 (−0.76, 5.87)
	BDE-47	3.61 (0.25, 6.96)	1.7 (−1.14, 4.53)
	BDE-99	2.63 (−0.64, 5.9)	1.38 (−1.41, 4.17)
	BDE-100	2.59 (−0.66, 5.85)	1.25 (−1.47, 3.98)
	BDE-153	1.36 (−1.9, 4.63)	0.61 (−1.98, 3.19)
	Σ5BDE	3.38 (−0.14, 6.9)	1.52 (−1.41, 4.45)
Global Executive Composite	BDE-28	4.52 (0.46, 8.59)	2.63 (−0.35, 5.61)
	BDE-47	3.66 (0.35, 6.97)	1.78 (−0.78, 4.33)
	BDE-99	3.05 (−0.17, 6.26)	1.97 (−0.54, 4.47)
	BDE-100	2.86 (−0.35, 6.07)	1.58 (−0.88, 4.03)
	BDE-153	2.05 (−1.17, 5.26)	1.24 (−1.08, 3.57)
	Σ5BDE	3.67 (0.21, 7.14)	1.84 (−0.79, 4.48)

Adjusted for maternal age, maternal education, maternal BDI-II score, PRQ relational frustration score, and maternal serum cotinine.

The regression estimates did not change notably with further adjustment of gestational urinary DAP in the models (Supplemental Figure 3). In the two sensitivity analyses 1) excluding participants with extremely high values in PBDE concentrations, and 2) without imputed values for missing PBDE concentrations, regression estimates also remained similar to the primary analysis described above (Supplemental Figures 4 and 5).

Sex-stratified models demonstrated consistent evidence of effect measure modification by adolescent biological sex, especially for caregiver-reported T-scores (Figures 2 and 3, Supplemental Tables 3 and 4). We found significant associations between gestational PBDE concentrations with caregiver-reported BRIEF-2 T-scores across most subscales among males but not females, and most of which were not observed in models analyzing the whole sample. In addition, the estimates of associations for males were consistently and markedly greater in magnitude. For example, each tenfold increase of gestational Σ₅BDE was associated with an increase of 7.38 (95%CI: 3.17, 11.59) points in global executive composite among males, compared to −0.37 (95%CI: −3.60, 2.86) among females. Similar effect modification by adolescent sex also occurred with adolescent-reported BRIEF-2 T-scores, but to a lesser extent than those caregiver-reported.

Figure 2.

Sex-specific adjusted difference in adolescent-reported BRIEF-2 T-scores per each tenfold increase of gestational serum PBDE concentrations

All models adjusted for maternal age, maternal education, maternal BDI-II score, Behavioral Assessment System for Children, Third Edition, Parenting Relationship Questionnaire (PRQ) relational frustration score, and maternal serum cotinine, *EMM p-value < 0.1

Figure 3.

Sex-specific adjusted difference in caregiver-reported BRIEF-2 T-scores per each tenfold increase of gestational serum PBDE concentrations

All models adjusted for maternal age, maternal education, maternal BDI-II score, PRQ relational frustration score, and maternal serum cotinine, *EMM p-value < 0.1

In a secondary analysis examining dichotomous outcomes, we found a consistent pattern of associations linking higher gestational PBDE concentrations with a greater risk for an elevated score in the behavior regulation index for both adolescent-reported BRIEF-2 (BDE-28, -47, -99, and Σ₅BDE) and caregiver-reported BRIEF-2 (BDE-28, -47, -99, and Σ₅BDE). For example, each tenfold increase in gestational BDE-28 was associated with adjusted risk ratios of 2.5 (95%CI: 1.31, 4.75) for the adolescent-reported behavior regulation index and 2.25 (95%CI: 1.07, 4.76) for caregiver-reported behavior regulation index (Supplemental Tables 5 and 6).

We did not observe consistent evidence of effect measure modification by breastfeeding duration (Supplemental Tables 7 and 8). For adolescent-reported BRIEF-2, BDE-28 and BDE-99 were more strongly associated with self-monitor among participants with shorter breastfeeding duration (<16 weeks). However, for caregiver-reported BRIEF-2, associations were stronger and positive rather than negative among participants with longer breastfeeding duration (>=16 weeks), especially in initiate, organization of materials, cognitive regulation index and global executive composite, which were consistently observed with BDE-28, -47, -99, -100, and Σ₅BDE. We also noted an exception with BDE-153, significant positive associations (inhibit and behavior regulation index) were found among those with shorter breastfeeding duration.

Discussion

This study found that gestational PBDE serum concentrations were associated with adolescent self- and caregiver- reported BRIEF-2 T-scores. Notably, adolescents self-reported elevated scores more often than their caregivers. This was expected as adolescent self-reporting of executive function difficulties at higher rates than caregivers or teachers has been previously described (Sutin et al., 2022). We also reported a similar pattern in our prior analysis of gestational PBDE serum concentrations in which adolescents self-reported more internalizing behaviors than were reported by their caregivers (Cecil et al., 2024). For covariate-adjusted analyses, each tenfold increase in gestational Σ₅BDE concentration was associated with an approximate four-point increase in adolescent self-reported BRIEF-2 T-scores for several key scales and indices. The corresponding analyses of caregiver-reported BRIEF-2 T-scores found associations with an approximate three-point increase for select PBDE congeners with the behavior regulation index.

We also noted effect measure modification for males from both adolescent self- and caregiver- reported analyses. For covariate-adjusted analyses, each tenfold increase in gestational PBDE concentration was associated with increases in caregiver-reported BRIEF-2 T-scores for almost all scales for males, with estimates ranging from 3.14 to 8.68. Key composites, such as the global executive composite and cognitive regulation index, along with initiate and plan/organize scales showed associations with all congeners. This finding is consistent with our prior work at the age 12 visit in which we reported effect measure modification upon examining sex-stratified regression models adjusted for covariates using caregiver-reported behaviors on the BASC-3 (Cecil et al., 2024). Here, we similarly reported stronger associations with gestational PBDE concentrations and lower scores on executive functioning among males (BDE-28, -47, -99, -100, Σ₅BDE) (Cecil et al., 2024).

The caregiver findings from the reports by Sagiv et al. (Sagiv et al., 2015) in CHAMACOS and Vuong et al. (Vuong et al., 2018; Vuong et al., 2016) also in the HOME Study are more consistent with our adolescent-reported findings than with our caregiver-reported findings. Prenatal PBDE exposure has been associated with poorer attention at 5–7 years (Eskenazi et al., 2013; Roze et al., 2009) and poorer attention and working memory at 9 and 12 years (Sagiv et al., 2015). Vuong previously employed a multiple informant model to evaluate childhood PBDE concentrations during several exposure assessment windows (1, 2, 3, 5, and 8 years) for participants of the HOME Study (Vuong et al., 2018). Significant adverse associations were reported between a 10-fold increase in concentrations of BDE-28 and BDE-153 at age 8 years with simultaneously measured behavioral regulation, poorer emotional and impulse control using the BRIEF-2. Child sex significantly modified the associations, with significantly poorer executive function among males with higher concurrent BDE-153, and null associations in females. (Vuong et al., 2018) These results contrast the report from CHAMACOS with effects observed only in females. (Sagiv et al., 2015)

Wallenborn et al. (Wallenborn et al., 2024) observed that higher gestational PBDE concentrations were associated with poorer function on the global executive composite at ages 9 and 12 among children in the CHAMACOS cohort who were complementary breastfed for a shorter duration (less than 7 months) compared to those with a longer breastfeeding duration (greater than seven months). In CHAMACOS, nearly all mothers (~96%) initiated breastfeeding. In contrast, we observed stronger and positive associations for caregiver-report with key BRIEF-2 indices and scales with no consistent evidence of effect modification by breastfeeding duration. In the HOME Study, 82% of children were ever breastfed with a median duration of 16 weeks (3.7 months), and only 40% breastfed longer than 7 months. Additional studies would be useful to clarify the potential role of breastfeeding duration in relationship to PBDE exposures and executive function outcomes in offspring.

Our study has several strengths to mention. The foremost strength is the prospective pregnancy birth cohort design with longitudinal follow-up as it allowed us to examine gestational serum PBDE concentrations and subsequent early adolescent executive function. This also provided detailed information on potential covariates to adjust for in our analyses as potential confounding factors. We collected BRIEF-2 ratings from both adolescents and their caregivers, which were not commonly completed in other studies. Self-reporting in adolescence is often missing in the environmental epidemiology literature. Historically, there was a bias as most studies exclusively employed caregiver ratings of child and adolescent behavior. We also examined BRIEF-2 T-scores as dichotomous outcomes in secondary analyses. These analyses provided moderately correlated and consistent findings for higher BRIEF-2 scores reported by adolescents in association with gestational PBDE concentrations. The data collection for this analysis was completed prior to the COVID-19 pandemic.

The study sample reflects a typically developing cohort with rural, suburban and urban residents from a metropolitan community within the Midwest region of the United States. With recruitment beginning in 2003, the gestational PBDE serum concentrations are consistent with values reported from national survey data for the same time period (Woodruff et al., 2011). We observed a moderate-to-high degree of correlation among the individual congener concentrations. This may arise from similar exposure pathways, including hand-to-mouth contact with consumer products and inhalation/ingestion of dust in indoor microenvironments (Petreas et al., 2011). The high correlation among individual congener concentrations may also indicate that we captured the overall PBDE exposure, especially with our summary variable strongly correlated with individual PBDE concentrations. Similar toxicological profiles might be expected with high PBDE congener concentration correlations to produce consistent findings for executive function outcomes. However, the design of the HOME Study required mothers to live in housing constructed prior to 1978, so the findings from these analyses may be less generalizable to other populations. Another limitation is the modest sample size. In addition, we have only one assessment timepoint in early adolescence, and this is a period during which executive function develops most rapidly (Tervo-Clemmens et al., 2023), so follow-up studies will be needed to fully characterize the impact of gestational PBDE exposure at later developmental stages. Caregivers confer genetic risks for executive function along with parenting practices that may contribute to problem behaviors in children (Walkup et al., 2022). However, our adjusted models include parental depressive symptoms and the BASC-3 Parenting Relationship Questionnaire to help account for caregiver symptoms and parenting approaches. We recognized the large number of associations examined in our analysis could lead to increased possibility of chance findings. Therefore, we focused on findings with consistent association patterns when reporting our results and computed the FDR-adjusted p-values to facilitate result interpretation. Finally, newer statistical approaches for estimating the effects of exposure mixtures are available; however, we desired to maintain a consistent approach with prior HOME Study publications to enable direct comparisons and aid interpretation (Braun et al., 2017b; Cecil et al., 2024; Chen et al., 2014; Hartley et al., 2022; Zhang et al., 2017). Future studies employing newer statistical mixture analysis methods (quantile g-computation, Bayesian kernel machine regression (BKMR)) are in progress. However, because of the high correlation of the PBDE compounds and no other chemical classes in the analysis, the mixture statistical methods may not be the best way to analyze the data.

Conclusions

This study provides evidence that gestational PBDE serum concentrations may influence offspring executive function during adolescence as identified from both adolescent- and caregiver-reporting. Effects for males were more pronounced. Individual and public health efforts may need to consider removing products in the home that were produced in the period when PBDEs were commonly applied as flame retardants in the effort to reduce exposures that can cause neurodevelopmental harm and explore strategies to mitigate their impact and optimize executive function development throughout childhood and adolescence.

Highlights.

Gestational PBDEs were associated with poorer executive function in adolescence.

Adolescents’ self-reports on the BRIEF identified more associations than caregivers’.

Higher gestational PBDEs were associated with a greater risk for elevated scores.

Stronger associations were observed for males, especially for caregiver-rated scales.