Exposure to polybrominated diphenyl ethers (PBDEs) during childhood and adiposity measures at age 8 years

Abstract

Background:

Animal studies suggest polybrominated diphenyl ethers (PBDEs) may be obesogens. However, epidemiologic studies investigating childhood exposure to PBDEs and adiposity are limited, with several reporting an inverse association.

Objectives:

To investigate associations between repeated childhood PBDE concentrations and adiposity measures at age 8 years.

Methods:

We examined 206 children from the Health Outcomes and Measures of the Environment Study, a birth cohort in Cincinnati, OH (2003-2006). Serum PBDEs were measured at ages 1, 2, 3, 5, and 8 years. We used multiple imputation to estimate missing PBDE concentrations. At 8 years, we measured weight, height, waist circumference, and body fat percentage. We used multiple informant models to estimate age-specific associations between PBDEs and adiposity measures.

Results:

We observed significant inverse associations between BDE-153 with all adiposity measures that became increasingly stronger with later childhood measurements. A 10-fold increase in BDE-153 at ages 1 and 8 years was associated with 2% (95% CI −3.9, −0.1) and 7% (95% CI −9.1, −4.7) lower body fat, respectively. No statistically significant associations were found with BDE-28, −47, −99, or −100. Child sex modified some associations; inverse associations between BDE-153 and body fat were stronger among boys, while positive and null associations were noted among girls.

Conclusions:

Childhood BDE-153 concentrations were inversely associated with adiposity measures and these associations became stronger as BDE-153 measurements were more proximal to adiposity measures. Inverse associations could be attributed to reverse causality arising from greater storage of PBDEs in adipose tissue of children with higher adiposity.

1. Introduction

From 1999 to 2016, the prevalence of overweight and obesity increased among U.S. children, especially among young children (Skinner et al. 2018). Obesity is multifaceted and complex, with contributions from genetic, lifestyle, and environmental factors. The prevalence of obesity, which has reached epidemic proportions cannot be fully explained by the classical risk factors. Endocrine disrupting chemicals have been leading candidates for environmental factors that may play a role as “obesogens” that could potentially interfere with metabolism and adipocyte function to increase the risk of childhood obesity (Nappi et al. 2016).

Polybrominated diphenyl ethers (PBDEs) are flame retardants that were incorporated into household furnishings and electronics from the 1970s to early 2000s. Beginning in 2004, they were phased out of use in the U.S. because of increasing concerns about their persistence and developmental toxicity. PBDEs are identified neurotoxicants and have been reported to adversely affect intellectual abilities and behavior (Herbstman and Mall 2014; Lam et al. 2017; Vuong et al. 2017b). However, their potential role in the risk of obesity remains uncertain. Infancy and childhood exposures are of particular concern given that higher concentrations have been reported among these age groups compared to adults (Schecter et al. 2005; Toms et al. 2008; Toms et al. 2009). Animal studies have reported increased weight gain with higher PBDE exposures occurring during the perinatal via oral gavage or intravenous injections (Bondy et al. 2013; Fernie et al. 2006; Suvorov et al. 2009) and postnatal period via oral dose (Dufault et al. 2005; Gee and Moser 2008). Further, toxicological studies have shown that PBDEs interfere with adipogenic pathways, increase adipocyte differentiation, and alter glucose homeostasis (Auwerx 1999; Bastos Sales et al. 2013; Hoppe and Carey 2007; Kamstra et al. 2014).

Epidemiological studies have primarily reported an inverse relationship between childhood PBDE exposures with adiposity measures in children and adolescents, with some indication of potential sexual dimorphism (Darrow et al. 2017; Erkin-Cakmak et al. 2015; Hoffman et al. 2016; Leijs et al. 2017; Windham et al. 2010). Given the limited number of epidemiological studies that have investigated postnatal PBDEs and the lack of studies that have prospectively investigated PBDEs during childhood and adiposity measures, a majority were cross-sectional studies, we examined whether child serum PBDEs measured repeatedly from 1-8 years were associated with adiposity measures at age 8 years.

2. Methods

2.1. Study participants and design

We employed the Health Outcomes and Measures of the Environment (HOME) Study, an ongoing prospective pregnancy and birth cohort (2003-2006, Cincinnati, Ohio, USA), to test whether postnatal PBDEs are associated with increased childhood adiposity. Details on the HOME Study have been described by Braun et al. (2017). Briefly, we enrolled 468 pregnant women at 16±3 weeks who provided informed consent. The participants were from nine collaborating obstetric practices and were ≥18 years, living in a residential home that was built before 1978, and were not HIV positive or taking medications for seizures, thyroid disorders, or chemotherapy/radiation treatment. Of the enrollees, 66 dropped out prior to delivery, 9 delivered sets of twins, and 3 had stillbirths. From the 390 women who delivered liveborn singletons, 206 (53%) were eligible for the present study based on available PBDE measurements during childhood (at any of the 5 follow-up visits) and anthropometry at 8 years. This study was approved by the institutional review board (IRB) at the Cincinnati Children’s Hospital Medical Center (CCHMC); the Centers for Disease Control and Prevention (CDC) and collaborating institutions relied on CCHMC’s IRB as the IRB of record.

2.2. Assessment of childhood PBDEs

We measured serum PBDE concentrations at the CDC using gas chromatography/isotope dilution high-resolution mass spectrometry from blood samples collected at ages 1, 2, 3, 5, and 8 years (Jones et al. 2012; Sjodin et al. 2004). Details regarding quality assurance and limit of detection (LOD) were previously described (Vuong et al. 2017a). We focused on congeners −28, −47, −99, −100, and −153 (as well as their sum: ∑PBDEs), because detection frequencies at ages 1, 2, 3, 5, and 8 years were 96-100%, 86-100%, 82-100%, 66-100%, and 50-100%, respectively (see Supplemental Table S1). PBDE concentrations <LOD were replaced with LOD/√2 (Hornung and Reed 1990). Total serum lipids are based on triglycerides and total cholesterol that were measured using standard enzymatic methods (Phillips et al. 1989), which were used in the lipid adjustment of PBDEs as ng/g lipid. While 206 children had at least one measurement of PBDEs during childhood, only 86 (42%), 69 (33%), and 69 (33%) children had available PBDE concentrations at ages 1, 2, and 3 years, respectively, due to limited serum availability, compared to 140 (68%) at 5 years and 190 (92%) at 8 years. We used multiple imputation with the Markov Chain Monte Carlo method to estimate missing PBDE concentrations for children who had at least one PBDE measurement during childhood. These procedures have been detailed previously for the 100 generated imputed datasets (Vuong et al. 2017c; Vuong et al. 2017d).

2.3. Child anthropometry

We measured height (to the nearest 0.1 cm), weight (to the nearest 0.01 kg), waist circumference (measured as cm), and body fat percentage in triplicate at age 8 years. We used an Ayrton Stadiometer Model S100 to measure children’s standing height without shoes, and with his/her heels against the wall. We used a ScaleTronix Pediatric Scale Model 4802 to measure weight of children clothed in their undergarments. We measured waist circumference using a plastic measuring tape that was placed around the abdomen at the level of the iliac crest. We measured body fat percentage with bioelectrical impedance analysis using the Tanita children’s body fat monitor (Arlington Heights, IL). We calculated age- and sex-specific weight and BMI z-scores using US references from the National Center for Health Statistics (Kuczmarski et al. 2000).

2.4. Statistical analyses

We used multiple informant models to estimate age-specific βs and 95% confidence intervals (CIs) for associations of repeated log₁₀-transformed PBDE concentrations (BDE−28, −47, −99, −100, −153, and ∑PBDEs) during childhood with adiposity measures at age 8 years for each of the 100 imputed datasets. Multiple informant models, which are non-standard versions of generalized estimating equations, allowed us to estimate associations for each of the repeated PBDE measurements and test whether these associations differ from one another and identify potential windows of susceptibility to PBDEs (Sanchez et al. 2011). We calculated average βs and 95% CIs from the 100 results from the imputed datasets; these are presented as age-specific estimates, because several interaction terms (PBDEs×age) were p<0.10 (Beunckens et al. 2008; Shen and Chen 2013).

We examined non-linear exposure-response using separate generalized additive models for PBDEs at 8 years and adiposity measures using the original, non-imputed data. We explored effect measure modification by sex to investigate whether associations between PBDEs and adiposity measures are influenced by sex. To do this, we included a 3-way interaction term between PBDEs×age×sex (p<0.10) and all possible 2-way interaction terms. All models were adjusted for the following based on a review of the literature (categorized as displayed in Table 1): maternal age, race/ethnicity, household income, maternal smoking status and alcohol consumption, maternal depression (Beck et al. 1996), vitamin intake during pregnancy, marital status, whether the child was breastfed, child age, and sedentary measures, including amount of time spent outside and watching TV. We additionally adjusted for maternal prepregnancy weight (for weight z-score outcome), maternal prepregnancy BMI (for outcomes BMI z-score, waist circumference, and body fat percentage), and child sex (for outcomes waist circumference and body fat percentage) depending on the outcome.

Table 1.

Child serum concentrations of ∑PBDEs (ng/g lipid) and child adiposity at 8 years by maternal and child characteristics, HOME Study^a

	∑PBDEs 2 years		∑PBDEs 8 years	BMI z-score	Weight z-score	Body Fat Percentage		Waist Circumference
	n	GM(GSD)	GM(GSD)	Mean±SD	Mean±SD	n	Mean±SD	n	Mean±SD
Maternal age, yearsb,c
<25	58	169.6(2.0)	39.3(2.2)	0.5±1.1	0.5±1.2	57	22.2±7.9	56	60.9±10.5
25-34	117	129.4(2.5)	49.0(2.1)	0.4±0.9	0.4±1.1	116	20.5±5.6	116	60.8±7.2
≥35	31	68.4(2.3)	34.2(2.3)	0.2±0.9	0.3±1.2	31	20.5±6.8	31	61.1±10.5
Race/ethnicityb,d,e,f
Non-Hispanic White	122	99.0(2.3)	44.0(2.2)	0.2±0.9	0.2±1.0	120	19.4±5.0	120	60.4±7.2
Non-Hispanic Black and Others	84	182.0(2.5)	43.1(2.2)	0.7±1.1	0.7±1.3	84	23.2±7.7	83	61.5±10.6
Maternal Educationb,d,f
High school or less	56	190.1(2.3)	41.9(2.1)	0.7±0.9	0.6±1.1	55	23.2±6.6	55	61.9±9.7
Some college/2 yr degree	58	165.7(2.4)	49.3(2.2)	0.1±1.1	0.2±1.2	58	19.7±7.1	57	58.9±8.6
Bachelor’s	55	83.1(2.3)	40.5(2.3)	0.4±0.9	0.5±1.1	55	21.0±5.6	55	62.2±8.9
Graduate or professional	37	85.0(2.0)	42.7(2.0)	0.3±0.8	0.3±0.9	36	19.6±5.9	36	60.4±6.8
Family Incomeb,f
<$40,000	88	195.8(2.3)	45.8(2.2)	0.6±1.1	0.6±1.2	88	22.4±7.8	87	61.3±10.4
$40,000-$79,999	65	122.0(2.3)	44.3(2.2)	0.3±0.9	0.4±1.1	64	20.0±5.4	64	61.0±8.3
≥$80,000	53	64.8(2.0)	39.4(2.2)	0.2±0.8	0.1±0.9	52	19.8±4.9	52	60.0±5.8
Maternal Depressionb,d,f,g
Minimal/mild	184	120.1(2.4)	44.6(2.1)	0.3±1.0	0.4±1.1	183	20.6±6.3	182	60.5±8.5
Moderate/severe	20	204.1(3.1)	36.0(2.7)	0.8±1.1	0.8±1.2	19	25.1±7.3	19	65.2±10.2
Maternal Alcohol Consumption
Never	114	141.2(2.4)	44.3(2.2)	0.4±1.0	0.3±1.2	114	21.1±6.9	114	60.7±9.5
<1 per month	64	110.9(2.4)	42.2(2.0)	0.3±0.9	0.3±1.0	63	20.1±5.0	63	60.0±6.0
≥1 per month	28	111.9(2.7)	44.2(2.3)	0.6±1.0	0.8±0.9	27	22.3±7.6	26	63.7±10.7
Maternal Smokingb
None	170	117.8(2.5)	44.8(2.2)	0.3±1.0	0.4±1.1	169	20.5±6.2	168	60.4±8.1
ETS	19	180.7(1.8)	48.1(2.2)	0.5±1.2	0.4±1.4	19	23.1±7.9	19	61.4±9.0
Active	17	179.3(2.5)	29.7(2.0)	0.9±0.9	0.8±1.2	16	23.4±7.3	16	64.4±13.5
Maternal Vitamin Useb,f
Daily	158	116.4(2.5)	43.0(2.2)	0.3±0.9	0.3±1.1	157	20.3±5.5	157	60.3±7.3
<Daily	34	146.7(2.3)	48.8(2.2)	0.6±1.1	0.7±1.4	33	23.3±8.7	33	63.7±13.9
Never	14	235.8(2.4)	38.8(2.0)	0.6±1.3	0.7±1.2	14	22.9±9.1	13	60.1±7.9
Prepregnancy BMIb,d,e,f,g
Underweight/Normal	103	106.7(2.5)	42.5(2.2)	0.1±1.0	0.1±1.1	101	19.7±5.6	101	59.3±7.1
Overweight	55	145.9(2.5)	47.3(2.2)	0.3±0.9	0.4±1.0	55	20.0±5.0	54	59.5±5.4
Obese	48	156.7(2.3)	42.0(2.0)	1.0±1.0	1.1±1.1	48	24.7±8.4	48	65.7±12.6
Marital statusb,d,f
Married/living with partner	150	99.1(2.2)	43.2(2.1)	0.3±0.9	0.3±1.1	148	20.2±5.7	147	60.4±8.1
Not married, living alone	56	245.8(2.3)	44.6(2.3)	0.7±1.1	0.7±1.2	56	23.1±7.9	56	62.0±10.3
Breastfedb,f,g
No	41	196.7(2.8)	39.7(2.7)	0.5±1.2	0.6±1.4	41	23.0±8.4	41	63.3±12.3
Yes	164	113.8(2.3)	44.7(2.1)	0.3±0.9	0.4±1.0	162	20.5±5.8	161	60.3±7.5
Child Sexf,g
Male	92	129.6(2.2)	44.0(2.3)	0.3±1.0	0.4±1.1	92	18.6±4.6	92	59.4±7.9
Female	114	124.7(2.7)	43.3(2.1)	0.5±1.0	0.4±1.2	112	22.9±7.1	111	62.1±9.3

Abbreviations: GM, geometric mean; GSD, geometric standard deviation; SD, standard deviation.

^aFrequencies may not add to the total number of participants because of missing values.

p< 0.05 for comparisons of PBDE concentrations or adiposity measures within maternal or child characteristics

^b∑PBDEs at 2 years

^c∑PBDEs at 8 years

^dBMI

^eWeight

^fBody fat percentage

^gWaist circumference (two-sided p values using ANOVA or t-test)

2.5. Sensitivity analyses

We conducted several sensitivity analyses. First, all models were re-examined using the original, non-imputed data. Second, we adjusted for maternal intake of fruits and vegetables during pregnancy as a proxy for children’s dietary intake. Third, we eliminated maternal anthropometric measures from the models. Fourth, we used two additional lipid adjustment methods for PBDEs (as pg/g serum or ng/g lipid) that both had serum lipids as a covariate in the model as recommended by O’Brien et al. (2016). We adjusted for gestational PBDEs since we previously reported inverse associations with adiposity measures (Vuong et al. 2015). Lastly, we excluded children with high body fat percentages ≥30% (n=19) and children with low body fat percentages ≤14% (n=13) to determine whether outliers influenced our results.

3. Results

3.1. Study participants

The geometric mean (GM) of ∑PBDEs in the HOME Study children was highest at age 2 years and lowest at age 8 years, with concentrations of 130.5(2.5) and 43.6(2.2) ng/g lipid, respectively (see Supplemental Table SI). Concentrations of ∑PBDEs at age 2 years were significantly higher in children whose mothers were younger, non-Hispanic Black or other racial and ethnic groups, had lower education, moderately/severely depressed, actively or passively exposed to tobacco smoke, did not take daily vitamin supplementation, obese, and not married or living alone (Table 1). Children who were from households with annual incomes <$40,000 and were not breastfed were also more likely to have higher ∑PBDEs at 2 years. We observed lower BMI z-scores and body fat percentages in children that had mothers who were non-Hispanic white and who more than a high school education. Children’s BMI z-scores and body fat percentages were higher among mothers that were moderately/severely depressed, not married/living alone, and who were obese. Children who had a smaller waist circumferences were more likely to be males and breastfed during infancy; they were also more likely to have mothers with minimal/mild depression who were not obese.

3.2. PBDEs and childhood measures of adiposity

We observed lower weight and BMI z-scores at higher concentrations of BDE-153 at ages 2-8 years, with increasingly larger decrements noted with concentrations more proximal to age 8 years (Figure 1). For example, a 10-fold increase in BDE-153 at age 2 years was associated with 0.4 (95% CI −0.7, −0.1) lower BMI z-scores, while a decrease of 1.0 (95% CI −1.4, −0.7) was noted with concurrent BDE-153 concentrations. Childhood BDE-153 was inversely associated with body fat percentage. Ten-fold concentration increases of BDE-153 at ages 1, 2, 3, 5, and 8 years were associated with lower body fat percentages of 2.0% (95% CI −3.9, −0.1), 2.9% (95% CI −4.9, −0.9), 3.6% (95% CI −5.5, −1.7), 5.6% (95% CI −7.8, −3.4), and 6.9% (95% CI −9.1, −4.7), respectively. We observed similar inverse associations between childhood BDE-153 and waist circumference. In particular, a 10-fold increase in BDE-153 at ages 2, 5, and 8 were associated with a decrease of 4.0 cm (95% CI −6.9, 1.1), 7.3 cm (95% CI −10.5, −4.0), and 9.3 cm (95% CI −12.5, −6.1), respectively. In generalized additive models with only exposure at age 8 years, childhood BDE-153 concentrations was inversely associated with all adiposity measures (Figure 2). In contrast, childhood BDE-28, −47, −99, and −100 were not associated with measures of adiposity at age 8 years.

Figure 1.

Estimated differences and 95% CIs in adiposity measures at age 8 years by 10-fold increases in child serum concentrations of polybrominated diphenyl ethers (ng/g lipid), HOME Study. All models were adjusted for maternal age, race/ethnicity, household income, maternal smoking status, maternal alcohol consumption, maternal depression, maternal vitamin use, marital status, whether the child was breastfed, child age, time spent watching television, time playing video games, and time playing outside. Select models were additionally adjusted for maternal pre-pregnancy weight (weight z-score). BMI (BMI z-score, body fat percentage, waist circumference), and child sex (body fat percentage, waist circumference).

Figure 2.

Concurrent serum concentrations of BDE-153 (ng/g lipid) and adiposity measures at 8 years with generalized additive model curve fitting using original, non-imputed data. Solid lines represent the natural cubic spline of the adjusted associations, and dotted lines represent 95% CIs. Distribution of BDE-153 is illustrated by vertical bars on the log₁₀-transformed x-axis. All models were adjusted for maternal age, race/ethnicity, household income, maternal smoking status, maternal alcohol consumption, maternal depression, maternal vitamin use, marital status, whether the child was breastfed, child age, time spent watching television, time playing video games, and time playing outside. Select models were additionally adjusted for maternal pre-pregnancy weight (weight z-score), BMI (BMI z-score, body fat percentage, waist circumference), and child sex (body fat percentage, waist circumference).

3.3. Child sex differences

Child sex modified associations between childhood BDE-153 concentrations and body fat percentages (Figure 3). A 10-fold increase in BDE-153 at 8 years was associated with a decreased body fat percentage of 4.5% (95% CI −7.6, −1.4) in males, while nonsignificant associations were noted in females (β=0.8%, 95% CI −6.2, 7.8) (p_{BDE-153*sex*8years}=0.023), but effect measure modification was not present at age 5 years. Similar associations were noted between males and females with BDE-153 concentrations at ages 1-3 years. Child sex did not modify associations between childhood PBDEs and z-scores for weight or BMI, or waist circumference (results not shown). However, we observed similar patterns of statistically significant inverse associations between BDE=153 and z-scores for weight and BMI and waist circumference in males, but null associations in females (see Supplemental Table S2).

Figure 3.

Estimated differences and 95% CIs in body fat percentage at age 8 years by 10-fold increases in child serum concentrations of BDE-153 and ∑PBDEs (ng/g lipid) by child sex, HOME Study. All models were adjusted for maternal age, race/ethnicity, household income, maternal smoking status, maternal alcohol consumption, maternal depression, maternal pre-pregnancy BMI, vitamin use, marital status, child sex, whether the child was breastfed, child age, time spent watching television, time playing video games, and time playing outside. Asterisks denote statistically significant interaction terms between PBDEs, sex, and age (p<0.100).

3.4. Sensitivity analyses

The associations between childhood PBDEs and adiposity measures using the original, non-imputed data yielded similar significant inverse associations between BDE-153 and adiposity measures, although several associations between earlier windows of exposure (1-3 years) were only marginally significant (see Supplemental Table S3). Adjusting for lipids as a covariate or lipid-standardizing PBDE concentrations resulted in similar conclusions as when we used the original, non-imputed data (results not shown). In addition, we observed a significant inverse association between BDE-100 at age 3 years and body fat percentage. Additional adjustment for maternal intake of fruits and vegetables during pregnancy and prenatal PBDEs did not materially affect our result, nor did the removal of maternal adiposity measures from the models (results not shown). Lastly, removing children with body fat percentages ≥30% yielded similar conclusions for BDE-153, although statistically significant inverse associations were mainly observed with concentrations at 5 and 8 years (see Supplemental Table S4). Similar observations were noted when we excluded children with body fat percentages ≤14%, with mainly statistically inverse associations observed between BDE-153 at ages 5 and 8 years and z-scores for weight and BMI and BDE-153 at ages 2, 3, 5, and 8 years and body fat percentage and waist circumference (see Supplemental Table S5). Statistically significant effect modification by sex was noted between BDE-153 associations at ages 1, 2, and 5 years and body fat percentage; in particular, a 10-fold increase in concurrent BDE-153 concentration was associated with a decrement of 3.9% in body fat (95% CI −7.4, −0.5) in males, but null associations in females (β=1.0%, 95% CI −6.6, 8.7) (results not shown).

4. Discussion

Findings from the present study do not support the hypothesis that postnatal PBDEs are associated with increased childhood adiposity. We observed a pattern of inverse associations between repeated childhood serum BDE-153 concentrations with weight and BMI z-scores, body fat percentage, and waist circumference, which became stronger with BDE-153 measures more proximal to child adiposity assessment.

Epidemiological studies have generally reported inverse relationships between postnatal PBDEs and measures of adiposity in children. Among girls aged 6-8 years from the Breast Cancer and the Environment Research Centers (BCERC) Study, the means of PBDE congeners −47, −100, −153, and −154 were significantly lower among those with higher BMI (Windham et al. 2010). Darrow et al. reported low BMI z-score was also a significant predictor of higher BDE-153 concentrations among children aged 1-5 years (2017). The Center for the Health Assessment of Mothers and Children of Salinas (CHAMACOS) Study also reported an inverse association between concurrent serum concentrations of BDE-153 and BMI z-score and waist circumference at age 7 years, with a monotonic decrease in both adiposity measures with increasing quartiles of BDE-153 concentrations (Erkin-Cakmak et al. 2015). In contrast, no relationship was noted between concurrent PBDEs and BMI among adolescents from a cohort in the Netherlands (Leijs et al. 2017).

Our study corroborates prior reports of postnatal PBDEs and childhood adiposity, and is congruent with previous findings from the HOME Study examining PBDEs during gestation. Previously, prenatal BDE-153 was reported to be inversely associated with adiposity measures in the HOME Study (Vuong et al. 2016). However, concentrations of prenatal PBDEs from the CHAMACOS Study had null and positive associations with BMI, respectively (Erkin-Cakmak et al. 2015).

The lipophilic nature of PBDEs and their pharmokinetic properties may have led to the observed results through reverse causation. It has been postulated that lipophilic compounds, including PBDEs and organochlorines, would be lower in concentrations for individuals with more adipose tissue, because the total storage compartment of adipose tissue would in effect render the serum concentrations to be “diluted” (Chevrier 2013; Glynn et al. 2003). Previously, individuals who experienced weight loss had significantly higher plasma and tissue concentrations of organochlorines (Chevrier et al. 2000; Imbeault et al. 2001; Kim et al. 2011; Walford et al. 1999).

There is no evidence from this study that postnatal PBDEs, aside from BDE-153, are associated with altered adiposity outcomes in childhood. BDE-153 has been uniquely identified as having an inverse association with adiposity measures in three other U.S. cohorts (Darrow et al. 2017; Erkin-Cakmak et al. 2015; Hoffman et al. 2016) and in NHANES (Lim et al. 2008); no relationship was noted with other congeners. BDE-153 has a higher fat deposition, considerably longer half-life, and is the more difficult to metabolize and excrete than BDE-47, −99, and −100 (Geyer et al. 2004; Staskal et al. 2006). In addition, Malarvannan et al. (2018) reported a significant increase in serum BDE-153 concentrations after weight loss among 94 Belgian obese adolescents even though a non-significant trend was observed with BDE-47 and −100, suggesting possible congener-specific variability in the pharmacokinetic processes controlling metabolism and storage.

If the true association between BDE-153 and adiposity measures is indeed an inverse one, the mechanisms involved in reduction of adipose tissue is as yet unknown. Animal studies on PBDEs’ role as potential obesogens are not conclusive, with some reporting weight gain (Bondy et al. 2013; Dufault et al. 2005; Fernie et al. 2006; Gee and Moser 2008; Suvorov et al. 2009) while others have reported either weight loss or no association (Daubie et al. 2011; Ta et al. 2011; Talsness et al. 2008). Because PBDEs are structurally similar to PCBs, they may also disrupt the GH/IGF-1 axis to delay growth (Ahmed 2013).

We found that sex modified the relationship between BDE-153 and body fat percentage, with significant reductions in body fat percentage in males, but null associations in females. No other study has reported significant sex interactions between postnatal PBDEs and childhood adiposity. However, Hoffman et al. (2016) noted similar inverse associations between PBDEs (except for BDE-153) and weight-for-height z-scores in males and positive associations among females upon stratification, but PBDEs were measured in breastmilk. For gestational PBDEs and childhood adiposity, the CHAMACOS Study found sex modified the relationship between PBDEs and BMI, with positive associations among males and inverse associations in females (Erkin-Cakmak et al. 2015). It is unclear whether the relationship between postnatal PBDEs and childhood adiposity is sexually dimorphic given the limited number of studies to examine this and the disparate findings thus far. It is plausible that PBDEs have sexually dimorphic biological processes since they are agonists of estrogen receptors and antagonists of androgen receptors (Meerts et al. 2001; Stoker et al. 2005). The hypothalamic-pituitary-gonadal (HPG) axis is active during the first months of infancy and reactivates at the onset of puberty in a gender-specific pattern (Chellakooty et al. 2003). Prepubertal children are sensitive to sex steroid actions, because normal endogenous sex steroid levels are low (Aksglaede et al. 2006). Small variations result in shifts in total hormone activity and may subsequently result in phenotypic effects as sex steroids have fat mobilizing properties. Given the sex steroid environment and receptor distribution in adipose tissue are distinct between males and females, it is conceivable that PBDEs’ mechanistic actions regarding adiposity vary by sex (Grun and Blumberg 2007). It is unclear what biological mechanisms are involved in potential sex differences in the reverse causality hypothesis between childhood PBDEs and measures of adiposity.

However, trajectories of fat and lean body mass vary by age and sex, with a divergence in the gain of fat mass between the sexes around age 5 (Weber et al. 2012). Higher gains in fat mass among females could result in lower PBDE concentrations as PBDEs would have more adipose tissue to partition in to. Further, the increase in lean body mass observed in males could also contribute to the higher concentrations of PBDEs. Thus, the findings of effect modification by sex between PBDEs and body fat percentage may be partially explained by the trajectories of fat and lean body mass between sexes rather than sexual dimorphism.

There were several strengths of our study. First, we employed a longitudinal design, which allowed for the use of repeated measures of postnatal PBDEs from infancy to prepubescence. Second, we adjusted for an extensive array of covariates, including sociodemographic factors, child sedentary measures, maternal depression, maternal prepregnancy BMI and weight, whether the child was breastfed, and prenatal PBDEs. Third, we used imputed PBDE concentrations to increase the precision of our estimates and conclusions were similar when we examined the original, non-imputed data. Finally, we modeled PBDEs using different lipid adjustment methods.

This study also had some limitations. First, we lacked data on dietary intake. However, we adjusted for maternal intake of fruits and vegetables during pregnancy as a proxy for dietary intake of children in a sensitivity analysis and found no difference in our results. Second, we did not have information on body fat distribution at earlier ages, which prevented us from analyzing it as both an outcome and an intermediate for body fat percentage at age 8 years. Third, our findings may not be generalizable to the US population. HOME Study children at age 8 years had a prevalence of overweight and obesity of 17% and 2%, respectively, compared to 33% and 19% among children ages 6-8 years in NHANES 2015-2016 (Skinner et al. 2018). Selection bias is also a concern as children who were excluded due to missing information (n=184) were statistically more likely than those included (n=206) to have mothers who were married or living with a partner and were more likely to have mothers with blood lead levels that were higher (0.78 ug/dL versus 0.69 ug/dL) (see Supplemental Table S6). However, they were similar in several maternal and child characteristics, including maternal prepregnancy BMI, age, race/ethnicity, income, education, and child sex.

5. Conclusions

We found significant inverse associations of childhood BDE-153 concentrations with z-scores for weight and BMI, body fat percentage, and waist circumference at ages 8 years. Our findings are consistent with other epidemiological studies identifying inverse associations between childhood BDE-153 and adiposity. Based on PBDEs’ physiochemical properties, reverse causality may have resulted in the observed inverse associations.

Highlights.

• Findings do not support postnatal PBDEs are associated with increased adiposity

• BDE-28, −47, −99, −100, and ∑PBDEs were not associated with adiposity at 8 years

• There was a pattern of inverse associations between BDE-153 and adiposity measures

• Child sex modified BDE-153 associations, with decreases in body fat % in males, but not in females

• Reverse causality may have resulted in inverse associations