Exposure to Per- and Polyfluoroalkyl Substances and Adiposity at Age 12 Years: Evaluating Periods of Susceptibility

Abstract

Per- and polyfluoroalkyl substances (PFAS) exposure may increase adiposity and obesity risk in children. However, no studies have extended these findings into adolescence or identified periods of heightened susceptibility. We estimated associations of repeated pre- and postnatal serum PFAS concentrations with adolescent adiposity and risk of overweight/obesity. We studied 212 mother–offspring pairs from the HOME Study. We quantified serum concentrations of four PFAS in mothers at ~16 week gestation and their children at birth and ages 3, 8, and 12 years. We assessed adiposity at 12 years using anthropometry and dual-energy X-ray absorptiometry. Using multiple informant models, we estimated covariate-adjusted associations of an interquartile range (IQR) increase in log₂-transformed PFAS for each time period with adiposity measures and tested differences in these associations. Serum perfluorooctanoate (PFOA) and perfluorohexane sulfonate (PFHxS) concentrations during pregnancy were associated with modest increases in central adiposity and risk of overweight/obesity, but there was no consistent pattern for postnatal concentrations. We observed nonlinear associations between PFOA in pregnancy and some measures of adiposity. Overall, we observed a pattern of modest positive associations of gestational PFOA and PFHxS concentrations with central adiposity and the risk of obesity in adolescents, while no pattern was observed for postnatal PFAS concentrations.

Graphical Abstract

1. INTRODUCTION

Per- and polyfluoroalkyl substances (PFAS) are a diverse group of fluorinated chemicals, which have been widely used in industry and consumer products since the 1950s because of their high thermal and chemical stability, hydrophobicity, and lipophobicity.¹ PFAS have been used in some food packaging materials, nonstick cookware, fire-fighting foams, waxes, furniture, and pesticides.^1,2 The production of perfluorooctane sulfonate (PFOS) and perfluorooctanoate (PFOA) was discontinued or reduced in the United States due to a set of voluntary phase-outs beginning in the 2000s.^1,3 However, PFOA, PFOS, perfluorononanoate (PFNA), and perfluorohexane sulfonate (PFHxS) are still detectable in the serum of nearly all American children and adults.^4,5 The major routes of human PFAS exposure include consumption of contaminated food and drinking water, as well as inhalation of indoor air and consumption/inhalation of house dust.^2,5 At least 19 million Americans are exposed to PFAS via drinking water,⁶ highlighting the need to improve our understanding of the impact of PFAS on human health, especially in susceptible populations such as children and pregnant women.

Prenatal exposure to PFAS has been associated with preterm birth,⁷ decreased birth weight,⁸ and altered child growth^9,10 in humans. However, epidemiological studies report mixed findings about the association between early life PFAS exposure and adiposity in children or adults.^11,12 Prenatal exposure to PFAS has been positively associated with adiposity in early to mid-childhood^13–17 and adulthood,¹⁸ while other studies found inverse¹⁹ or null associations.^20,21 Two prospective studies reported that postnatal serum PFAS concentrations were positively associated with adiposity in preschoolers,²² adolescents, and young adults;²³ however, one study reported null associations.²⁴ Cross-sectional studies have reported inconsistent findings.^25–29

To the best of our knowledge, no previous studies have examined the associations between PFAS exposure and adolescent adiposity or identified potential periods of heightened susceptibility during gestation or childhood. Moreover, the majority of published studies used indirect measurements of adiposity, such as weight, body mass index (BMI), and waist circumference, with few studies using direct measures (e.g., dual-energy X-ray absorptiometry (DXA)).^17,20,25,27 Therefore, we estimated the associations of five repeated pre- and postnatal serum PFAS concentrations collected during gestation and the first 12 years of life with detailed measures of adolescent adiposity to evaluate periods of heightened susceptibility to the potential obesogenic effects of PFAS exposure.

2. MATERIALS AND METHODS

2.1. Study Participants.

We used data from the Health Outcomes and Measures of the Environment (HOME) Study, an ongoing pregnancy and birth cohort that recruited pregnant women from nine prenatal clinics affiliated with three hospitals in the Cincinnati, Ohio metropolitan area between March 2003 and January 2006. Pregnant women in the HOME Study had median PFOA concentrations 2 times higher than pregnant women in the general population in the United States. This is possibly because their drinking water was contaminated with PFOA released by DuPont Washington Works plant in Parkersburg, West Virginia. Details of eligibility criteria and recruitment methods have been published previously.³⁰ A total of 468 pregnant women in their first or second trimester enrolled. We conducted longitudinal follow-up among offspring at age 4 weeks and 1, 2, 3, 4, 5, 8, and 12 years.

Among 423 participants eligible for follow-up at age ~12 years (range: 11–14 years), we excluded children who were twins (n = 22) and had congenital anomalies (n = 5). Additional details regarding participant characteristics, follow-up, and study measures at the 12 year visit are described elsewhere.³¹ Among the remaining 396 children, we only included participants who had at least one serum PFAS measure prenatally or in childhood at age 3, 8, or 12 years (n = 389). Of these, 212 children who had at least one measurement of anthropometry or dual-energy X-ray absorptiometry (DXA) and complete information on covariates were included in the final analytic sample.

The research protocols were approved by the Institutional Review Boards (IRBs) at the Cincinnati Children’s Hospital Medical Center (CCHMC) and the participating delivery hospitals prior to enrollment. Brown University and the Centers for Disease Control and Prevention (CDC) IRBs deferred to the IRB from CCHMC. Written informed consent was obtained from mothers at each study visit, and written informed assent was obtained from children at age 12 years.

2.2. PFAS Exposure Assessment.

To assess prenatal PFAS exposure, we quantified serum concentrations of PFOA, PFOS, PFNA, and PFHxS in mothers at ~16 or 26 week gestation or within 48 h of delivery. Given that blood volume may change during pregnancy, we measured PFAS concentrations in the 16 week serum if available (88%). For mothers who did not have 16 week PFAS concentrations due to insufficient volume, we used the 26 week (8%) or delivery sample (4%). For postnatal PFAS exposure assessment, we quantified serum concentrations of these same four PFAS in children at birth (venous cord blood) and ages 3, 8, and 12 years. All serum PFAS concentrations were measured using online solid-phase extraction coupled to high-performance liquid chromatography-isotope dilution tandem mass spectrometry at CDC.³² The limit of detection (LOD) was 0.1 ng/ mL for PFOA, PFNA, and PFHxS and 0.2 ng/mL for PFOS. PFAS were detectable in approximately 98% of serum samples, and we replaced concentrations below LOD with LOD/√2. We summed the concentrations of both linear and branched isomers to obtain the total concentrations of PFOS and PFOA, respectively. Additional information on the quality control tests have been described elsewhere.³³

2.3. Anthropometry and DXA.

When adolescents were on average 12.1 years old, trained staff measured weight, height, and waist and hip circumferences in triplicate following standardized protocols to the nearest 0.01 kg, 0.1 cm, and 0.1 cm, respectively.³¹ Children wore hospital scrubs and had their shoes and head covering off during anthropometric assessments. We calculated body mass index (BMI) in kg/m² from weight and height and obtained age- and sex-specific BMI z-scores based on the CDC growth curve.³⁴ Childhood overweight was defined as BMI z-scores >1, and obesity as BMI z-score >2.

At the same visit, trained technicians measured children’s total and regional fat mass by a whole-body DXA scan with a Hologic Horizon densitometer (model A; Hologic Inc., Marlborough, MA). Percentage regional fat measures were calculated for the trunk, android, and gynoid subregions as the regional fat mass divided by the whole-body fat mass × 100. The whole-body DXA scan also yielded a measure of visceral fat area (cm²), which is highly correlated with visceral fat area estimated by computed tomography.³⁵ A spine phantom was scanned daily to ensure optimal machine performance and calibration as recommended by the manufacturer. Age- and sex-specific fat mass index (fat mass/height², kg/m²) z-scores were calculated based on the reference values generated using the National Health and Nutrition Examination Survey (NHANES) 1999–2004.³⁶ We applied thresholds of >1 and >2 standard deviation (SD) to define overweight and obesity, respectively, using fat mass index z-scores.

2.4. Covariates.

We selected a set of potential confounders that may be associated with both PFAS concentrations and adolescent adiposity using directed acyclic graphs (DAGs) and based on prior knowledge^37,38 (Figures S1 and S2). Trained research staff collected sociodemographic covariates using standardized computer-assisted interviews, including maternal age, race, marital status, and education in the second or third trimester. We collected data on prepregnancy weight and height and parity from medical records. We used standardized interviewer-administered surveys to assess breastfeeding duration.³⁹ We measured serum concentrations of cotinine, a biomarker of tobacco smoke exposure, using high-performance liquid chromatography-tandem mass spectrometry.⁴⁰ Puberty was evaluated using the Tanner stage based on pubic hair development in both sexes at the 12 year study visit using a self-report instrument. For pubic hair growth, stage 1 indicates prepubertal showing no sexual hair at all. Stage 2 indicates the onset of puberty showing sparse growth of long and downy hair with slight pigmentation. At stages 3 and 4, the pubic hair becomes darker and coarser and reaches adult levels at stage 5.⁴¹ In the sensitivity analyses, we also considered adjusting for several additional variables, including birth weight and pregnancy-induced hypertensive disorders, which were assessed using medical records. We also evaluated adolescent physical activity using a validated self-report questionnaire.⁴² To assess adolescent diet quality, we calculated 2015 Health Eating Index scores using data from three 24 h dietary recalls.⁴³

2.5. Statistical Analyses.

We described the univariate characteristics of each PFAS, adiposity measures, and covariates (mean ± standard deviations [SD] or proportions [%]). We also calculated the mean ± SD of total body fat percentage according to the categories of each covariate.

We log₂-transformed PFAS concentrations prior to analysis to reduce the potential influence of outliers. We calculated Spearman correlation coefficients between log₂-transformed PFAS at each time point. In addition, we examined the potential nonlinear association between PFAS concentrations and measures of adiposity at age 12 years using covariate-adjusted natural cubic splines and testing the statistical significance of the spline terms.

We used multiple informant models to estimate covariate-adjusted differences in each adiposity measure with an interquartile range (IQR) increase in log₂-transformed concentrations of the four PFAS (βs and 95% confidence intervals [CIs]) at the five different time points. Moreover, we used these models to statistically test the difference in the PFAS-adiposity association at each time point. Multiple informant models jointly estimate a set of separate multiple regressions, using generalized estimating equations for the parameter estimation, and provide a formal test on the difference in the strength of the associations between PFAS concentrations and adiposity measures across five time points.⁴⁴ Finally, since some previous studies reported sex-specific associations between PFAS and adiposity,^17,18,26,28 we examined the potential modifying effects by child sex using a three-way interaction term of child sex, serum PFAS concentration, and timing of exposure assessment. All models were adjusted for maternal age, race, marital status, education, parity, serum cotinine concentrations, prepregnancy BMI, breastfeeding duration (postnatal PFAS concentrations only), and child age, sex, and pubic hair stage.

We performed all these analyses using SAS (version 9.4, SAS Institute Inc., Cary, NC) and R version 3.2.3 (R Core Team, Vienna, Austria).

2.6. Secondary and Sensitivity Analyses.

We estimated the risk of adolescent overweight or obesity in relation to PFAS concentrations using modified Poisson regression with robust standard errors. Because hypertension is associated with an increased risk of impaired kidney function, which confounds the association between prenatal PFAS and fetal growth,^45,46 we adjusted for the presence of pregnancy-induced hypertensive disorders. In addition, because prior studies reported associations of low birth weight with pediatric obesity,⁴⁷ we repeated our analyses by adjusting for sex-specific birth weight for gestational age z-scores.⁴⁸ Finally, we also adjusted for adolescent physical activity levels and diet quality using a summary physical activity and Health Eating Index score, respectively.

3. RESULTS

The mean age of the 212 children included in this analysis was 12.4 years (SD: 0.7), 55% were girls, and 90% were stage 2+ for pubic hair development (Table 1). Mothers in our study sample were predominately non-Hispanic white (59%), married (63%), and had at least one previous pregnancy (58%). The percentage of mothers with a high school degree or less was about 23%. On average, mothers were age 29.4 (SD: 6.0) years at delivery. Adolescents whose mothers were obese before pregnancy had a higher body fat percentage at age 12. Body fat percentages did not differ substantially according to categories of other covariates. Thirty-four percent of these children were overweight or obese (Table 2). Generally, girls had higher adiposity than boys at age 12 years.

Table 1.

Adolescent Body Fat Percent at 12 Years of Age According to Covariates: The HOME Study (Total N = 212)

characteristics	N (%)	body fat percent (%)a mean (SD)
maternal characteristics
race/ethnicity
white, non-Hispanic	125 (59.0)	32.3 (6.3)
black, non-Hispanic	74 (34.9)	31.7 (7.0)
all others	13 (6.1)	31.5 (7.2)
education
high school or less	49 (23.1)	32.4 (7.6)
tech school/some	60 (28.3)	31.5 (6.3)
college
bachelor’s or more	103 (48.6)	32.1 (6.3)
maternal age (years)
18–25	52 (24.5)	31.1 (7.1)
>25–35	128 (60.4)	32.2 (5.9)
>35	32 (15.1)	32.9 (8.1)
marital status
Yes	134 (63.2)	32.1 (6.4)
No	78 (36.8)	31.9 (7.0)
parity
0	90 (42.4)	32.1 (6.4)
1	69 (32.6)	31.4 (6.3)
2+	53 (25.0)	32.6 (7.3)
serum cotinine (ng/mL)
<0.015	40 (18.9)	32.5 (6.4)
0.015-<3	152 (71.7)	32.0 (6.7)
≥3	20 (9.4)	31.0 (6.0)
prepregnancy BMI (kg/m2)
underweight/normal (<24.9)	113 (53.3)	31.0 (6.0)
overweight (25.0–29.9)	55 (25.9)	32.5 (6.6)
obese (≥30)	44 (20.8)	34.0 (7.6)
child characteristics
sex
females	116 (54.7)	34.0 (6.1)
males	96 (45.3)	29.6 (6.3)
breastfeeding duration (months)
<6	119 (56.1)	31.8 (7.0)
≥6	93 (43.9)	32.2 (6.0)
pubic hair stage
1	22 (10.4)	31.3 (6.8)
2	55 (25.9)	32.3 (6.4)
3	65 (30.7)	32.2 (6.5)
4	43 (20.3)	32.0 (6.8)
5	27 (12.7)	31.4 (7.0)

^aBody fat percent = total body fat mass (g) × 100/total body mass (g).

Table 2.

Distribution of Adolescent Measures of Overall and Regional Adiposity at Age 12 Years: The HOME Study

variable	total N	overall mean (SD) or N [%]	N	boys mean (SD) or N [%]	N	girls mean (SD) or N [%]
overall adiposity
normal/underweight		141 [66.5]		71 [74.0]		70 [60.3]
overweight (>1 SD)		61 [28.8]		23 [24.9]		38 [32.8]
obese (>2 SD)		10 [4.7]		2 [2.1]		8 [6.9]
BMI z-score	212	0.3 (1.1)	96	0.1 (1.2)	116	0.5 (1.1)
fat mass index z-score	206	0.2 (0.8)	93	0.2 (0.7)	113	0.2 (0.9)
body fat percent (%)	206	32.0 (6.6)	93	29.6 (6.3)	113	34.0 (6.1)
central and gynoid adiposity
waist circumference (cm)	212	72.2 (12.3)	96	69.5 (10.4)	116	74.4 (13.4)
hip circumference (cm)	212	84.5 (11.0)	96	81.3 (9.7)	116	87.0 (11.4)
waist-to-hip ratio	212	0.9 (0.06)	96	0.9 (0.06)	116	0.9 (0.06)
visceral fat area (cm2)	206	44.6 (20.2)	93	50.5 (13.9)	113	39.8 (23.2)
trunk fat percent (%)a	206	36.3 (4.0)	93	35.3 (3.5)	113	37.2 (4.2)
android fat percent (%)a	206	5.2 (1.0)	93	5.1 (0.9)	113	5.4 (1.1)
gynoid fat percent (%)a	206	17.0 (1.6)	93	16.4 (1.4)	113	17.4 (1.6)
android/gynoid fat percent ratio	206	0.3 (0.07)	93	0.3 (0.06)	113	0.3 (0.08)

^aPercentage of regional fat mass to total body fat mass. Fat mass index = total body fat mass (kg)/(height × height (m²)).

Median serum concentrations of PFOA, PFOS, PFNA, and PFHxS in pregnancy were 5.3, 13.3, 0.9, and 1.3 ng/mL, respectively (Figure 1 and Table S1). Median maternal serum PFAS concentrations during pregnancy were higher than median concentrations in infants at delivery. When comparing PFAS concentrations across childhood, 3 year PFAS concentrations were the highest. PFAS concentrations did not differ by child sex (Table S1). Serum PFAS concentrations were weakly to moderately correlated with each other in pregnancy (ρ = 0.33–0.60), cord blood (ρ = 0.22–0.59), at age 3 (ρ = 0.26–0.72), at age 8 (ρ = 0.19–0.61), and at age 12 (ρ = 0.39–0.68) (Figure S3A). Overall, for each PFAS, prenatal concentrations were weakly correlated with postnatal concentrations (Figure S3B).

Figure 1.

Violin plots of serum log₂-transformed PFAS concentrations (ng/mL) from pregnancy to age 12 years: the HOME Study. Preg: pregnancy; PFOS: perfluorooctane sulfonate; PFOA, perfluorooctanoate; PFNA: perfluorononanoate; and PFHxS: perfluorohexane sulfonate.

Figure 2 shows the associations of prenatal and postnatal PFAS concentrations with central adiposity measures assessed via anthropometry (Figure 2A) and DXA (Figure 2B), as well as overall adiposity (Figure 2C) at age 12 years. After adjusting for covariates, serum PFOA and PFHxS concentrations during pregnancy were consistently associated with higher levels of central adiposity across all measures but lower levels of gynoid fat percent (Figure 2A,B and Table S2). For instance, each interquartile range increase in log₂-transformed prenatal PFOA concentration was associated with higher waist-to-hip ratio (β = 0.02, 95%CI: 0.00, 0.03), waist circumference (β = 2.0 cm, 95%CI: −0.8, 4.8), visceral fat area (β = 1.4 cm², 95%CI: −2.9, 5.7), trunk fat percent (0.4%, 95%CI: −0.5, 1.4), and android fat percent (β = 0.2%, 95%CI: −0.1, 0.4) but lower gynoid fat percent (β = −0.2%, 95%CI: −0.6, 0.3). There was no pattern of associations between other PFAS during pregnancy and measures of adiposity in these adolescents (Figure 2 and Table S2). Moreover, childhood PFAS concentrations were generally not associated with measures of overall or regional adiposity. Our results suggest that the associations with adiposity potentially differ by exposure periods, although the interaction terms did not reach statistical significance (PFAS × visit interaction p-values ≥0.2) (Table S2). In addition, we observed a pattern of positive associations of serum PFHxS concentrations with all measures of overall adiposity (Figure 2C and Table S2).

Figure 2.

Adjusted associations (β [95% CI]) of 1-interquartile range (IQR) increase in log₂-transformed PFAS concentrations with measures of overall and regional adiposity at 12 years: the HOME Study. Preg: pregnancy; PFOS: perfluorooctane sulfonate; PFOA, perfluorooctanoate; PFNA: perfluorononanoate; PFHxS: perfluorohexane sulfonate; percentage of regional fat mass to total body fat mass. Adjusted for a visit, the interaction between visit and PFAS, maternal age, race, marital status and education, parity, maternal smoking, prepregnancy BMI, and child age, sex, and puberty. Y-axis shows β and 95% confidence interval; X-axis shows the different visits across time points.

We observed evidence suggesting potential nonlinear associations between prenatal PFOA and some measures of adiposity, with nonlinearity p-values ranging from 0.03 to 0.05 for BMI z-score, fat mass index z-score, waist circumference, hip circumference, and visceral fat area (Figure 3). Adjusted mean adiposity measures were positively associated with prenatal PFOA up to the median PFOA concentration and then slightly declined or plateaued at higher concentrations. Figure S4 shows the adjusted natural cubic splines of the associations of all PFAS concentrations during pregnancy with overall (Figure S4A) and central adiposity using anthropometry (Figure S4B) and DXA (Figure S4C,D) at age 12 years. We found that the associations between other PFAS (i.e., PFOS, PFNA, and PFHxS) and adiposity appeared to be linear (nonlinearity p-values >0.05) (Figure S4).

Figure 3.

Adjusted natural cubic splines of the associations of log₂-transformed perfluorooctanoate (PFOA) concentrations during pregnancy with adiposity measures at age 12 years in adolescents from the HOME Study. Adjusted for maternal age, race, marital status and education, parity, maternal smoking, breastfeeding duration (postnatal concentrations only), prepregnancy BMI, and child age, sex, and puberty.

When stratifying by child sex, we found that PFOA and PFHxS concentrations during pregnancy were consistently associated with higher overall and central adiposity across all measures in girls, while no consistent associations were found in boys (Table S3 and Figure S5). However, the three-way interaction terms between PFAS, the timing of the PFAS exposure assessment, and child sex were not significant (p-interaction >0.2) (Table S3).

In secondary analyses, we observed positive associations between prenatal exposure to all four PFAS during pregnancy and the risk of being overweight or obese at age 12; these associations were modestly stronger when using the fat mass index z-score to define overweight or obesity (Table 3). For instance, each IQR increase in prenatal PFAS concentrations was associated with 70% increased risk of childhood overweight or obesity (PFOA: β = 1.7, 95%CI: 0.9, 3.1; PFHxS: β = 1.7, 95%CI: 1.1, 2.7).

Table 3.

Adjusted Relative Risk of Overweight/Obesity with 1-Interquartile Range (IQR) Increase in Log₂-Transformed PFAS Concentrations during Pregnancy^a

	overweight/obesity defined by BMI-z		overweight/obesity defined by FMI-z
PFAS	crude RR (95%CI)	adjusted RR (95%CI)	crude RR (95%CI)	adjusted RR (95%CI)
PFOA	1.30 (0.90, 1.87)	1.37 (0.92, 2.03)	1.53 (0.89, 2.63)	1.66 (0.90, 3.07)
PFOS	0.94 (0.64, 1.37)	1.11 (0.76, 1.63)	1.26 (0.69, 2.32)	1.51 (0.81, 2.85)
PFNA	1.01 (0.66, 1.54)	1.14 (0.74, 1.75)	1.17 (0.68, 2.01)	1.19 (0.67, 2.13)
PFHxS	1.07 (0.82, 1.41)	1.23 (0.93, 1.63)	1.40 (0.94, 2.08)	1.71 (1.08, 2.73)

^aPFOS: perfluorooctane sulfonate; PFOA, perfluorooctanoate; PFNA: perfluorononanoate; PFHxS: perfluorohexane sulfonate. Fat mass index = total body fat mass (kg)/(height × height (m²)). Adjusted for maternal age, race, marital status and education, parity, maternal smoking, prepregnancy BMI, and child age, sex, and puberty.

In sensitivity analyses, we adjusted for physical activity level scores, healthy eating index scores, birth weight for gestational age z-scores, and pregnancy-induced hypertensive disorders. However, these adjustments did not meaningfully change the associations of prenatal PFAS concentrations with adolescent adiposity measures, and in some cases strengthened the associations (Table S4). When mutually adjusting for all prenatal PFAS concentrations in the same model, the associations with PFOA and PFHxS were strengthened, but the association with PFOS became negative (Table S5).

4. DISCUSSION

In these mother–adolescent pairs from the HOME Study, we observed a pattern of modest positive associations of prenatal serum concentrations of PFOA and PFHxS during pregnancy with greater central body fat, as well as the risk of being overweight or obese at age 12. However, there was no clear pattern for postnatal PFAS concentrations. We observed evidence suggesting that prenatal PFOA was nonlinearly associated with some adiposity measures and more strongly associated with greater adiposity in girls than in boys. These results were robust to additional adjustment for physical activity, diet, birth weight, and pregnancy-induced hypertensive disorders.

Prenatal concentrations of PFOS, PFNA, and PFHxS in the HOME Study were similar to those of pregnant women in National Health and Nutrition Examination Survey (NHANES) 2003–2008,⁴⁹ whereas the median PFOA concentrations (5.3 vs 2.2 ng/mL) during pregnancy were higher. We speculate that elevated serum PFOA concentrations in HOME Study women could be caused by exposure to drinking water contaminated by PFOA emitted from the Parkersburg DuPont Washington Works plant.¹⁵

We found that increasing prenatal PFOA and PFHxS concentrations were associated with greater central adiposity in adolescence although the 95% CIs of our estimates included the null value. Our results are consistent with some prior studies in children or young adults,^13–18 although the majority of these studies used indirect measures of adiposity and did not assess body fat distribution. Specifically, three prospective cohort studies reported that prenatal PFAS concentrations were positively associated with BMI among 200 children at age 3–5 years from Sweden,¹³ body fat percent and waist circumference in 204 HOME Study children at age 8 years,¹⁵ and waist circumference and skinfold thickness in 870 children at age 6–10 years from the Greater Boston area.¹⁷ In addition, maternal concentrations of PFAS during pregnancy were associated with increased risk of being overweight/obese in 412 Norwegian and Swedish children at age 5 years,¹⁴ 1022 Ukraine children at age 5–9 years,¹⁶ and 665 young adults at age 20 years from Denmark.¹⁸ In contrast, a Danish birth cohort reported inverse associations between maternal PFAS during pregnancy and BMI at age 7 years (n = 519).¹⁹ Two prospective studies reported no clear associations.^20,21

Inconsistencies across prior studies may relate to the differences in the study designs, concentrations of PFAS, or the timing and method of adiposity measurements. For instance, Andersen et al.¹⁹ used self- or parent-reported weight and height, which could contribute to the misclassification of BMI and subsequently attenuate associations with PFAS.^50,51 Moreover, prenatal PFAS concentrations were lower in Manzano-Salgado’s study²¹ than those reported in our study (e.g., median PFOA: 2.3 vs 5.3 ng/mL; PFHxS: 0.6 vs 1.3 mL). Hartman et al. included a subset of girls who were originally selected for a study on menarche and completed DXA assessments; therefore, their study may be subject to selection bias²⁰.

We found no clear pattern in the associations of postnatal PFAS concentrations at age 3, 8, or 12 years with any measure of body fat distribution in children at age 12 years. Similar to our results, a prospective study of 8764 U.S. adults from a community with drinking water contaminated with PFOA found that early life serum PFOA concentrations (average concentrations at ages 1–3 years) were not associated with the risk of being overweight or obese.²⁴ A cross-sectional study of 499 Danish children aged 8–10 years found that plasma concentrations of PFOA and PFOS were not associated with BMI, waist circumference, or skinfold thickness.²⁹ Some prospective or cross-sectional studies report mixed associations between postnatal PFAS concentrations and childhood adiposity measures.^{22,23,25–28} Inconsistent findings among these studies could be attributed to the use of different study populations and design, inaccurate measurement of adiposity, or varying concentration ranges and timing of PFAS exposure assessment.

In this study, prenatal PFAS concentrations were overall weakly correlated with postnatal PFAS concentrations in our study. Thus, correlations among PFAS may not explain the lack of statistically significant results that identify the specific periods of heightened susceptibility. In addition, we found that some of the point estimates for postnatal PFAS concentration are negative, while their prenatal counterparts are positive. Therefore, even modest correlations are not solely responsible for our lack of statistically significant discernible periods of heightened susceptibility.

Our findings suggest that the developing fetus may be more susceptible to the influence of PFAS exposure on body composition, as compared with postnatal exposures. This is biologically plausible because PFAS can readily pass through the placenta and enter fetal circulation.⁵² In addition, late pregnancy may be a more sensitive period for exposures to perturb body fat development, as the fetus undergoes rapid fetal growth and fat deposition during the third trimester.⁵³ Finally, given that fetal exposure to certain obesogens may alter the epigenome of stem cells, resulting in increased production of adipocytes later in life,^14,54 it is possible that the development of obesity may be programmed as early as during gestation.

While the biological mechanisms underlying the PFAS-adiposity associations are not well established, a few potential pathways have been proposed to explain these associations. PFAS can activate peroxisome proliferator-activated receptors α and γ (PPARα/γ), which can increase lipid metabolism and adipocyte differentiation to stimulate body fat storage.^12,55 PFAS have been found to disrupt thyroid function and possibly cause hypothyroidism (i.e., decreasing total or free thyroxine [T4] and increasing thyroid stimulating hormone [TSH]), which can increase body fat.^56–58 As endocrine-disrupting chemicals, PFAS may directly interact with estrogen receptors or decrease circulating estrogen concentrations,^59,60 and women with low levels of estrogen may have greater visceral fat.⁶¹ Thus, the sex-specific associations of prenatal PFOA with childhood adiposity measures may be explained by the ability for PFAS to interact with gonadal hormone pathways related to adipose tissue development.

We found some evidence suggesting that prenatal PFOA and PFHxS concentrations were more strongly associated with adolescent adiposity than PFOS and PFNA The differential effects of PFAS could be due to the differences in their biological activity. For instance, it is possible that the activation of PPARα may be higher for PFOA than PFOS,⁶² and some studies argue that, unlike PFOA, the toxicity of PFOS may not depend on the activation of PPARα.⁶³ Moreover, the placental transfer and postnatal offspring/maternal blood ratio was higher for PFOA than other PFAS, which could increase the burden of PFOA exposure in early life.⁶⁴

There are some limitations of our study. First, we had a relatively modest sample size, especially for identifying distinct periods of heightened susceptibility and sex-specific associations. Given the movement away from null hypothesis testing and the fact that few of the observed statistically significant findings would not hold after adjusting for multiple comparisons, we emphasized the pattern of consistent associations across each measure of adiposity to interpret our results without relying on the statistical significance of these associations. In addition, we treated our sex-specific analyses as more exploratory in our secondary analyses given the lack of power to detect significant three-way interaction. Longitudinal studies with large samples are needed to improve our understanding of the sex-specific effects on adiposity. Second, it is possible that our results were limited by the potential residual confounding effects. For instance, we did not measure circulating albumin concentrations or glomerular filtration rate during pregnancy. However, prior studies found that adjusting for markers of pregnancy hemodynamics (i.e., glomerular filtration rate and plasma albumin) did not meaningfully alter effect estimates, suggesting that these markers may not confound the associations between PFAS and adiposity.^7,17,65 A large number of studies have indicated that parental BMI may be a strong predictor of offspring adiposity highlighting the genetic component of obesity. However, we did not collect data on paternal body weight, and it seems unlikely that paternal weight could confound the associations between prenatal PFAS during pregnancy and offspring adiposity. Additionally, we did not adjust for maternal gestational weight gain (GWG) in this study because it may be a causal intermediate of the associations between prenatal PFAS and adiposity in children.¹⁷ Furthermore, it has been shown that the impact on offspring adiposity of GWG in addition to the impact of prepregnancy BMI is small.⁶⁶

Despite these limitations, our study has several strengths. First, we assessed body fat composition and distribution using DXA, which provides detailed and accurate measurements of adiposity and reduces potential misclassification of adiposity. Second, we were able to account for several important confounders including prepregnancy BMI, maternal smoking, breastfeeding duration, and adolescent physical activity and diet. Third, we examined the associations of adolescent adiposity with five repeated measures of serum concentrations of PFAS that covered the periods of pregnancy, early and middle childhood, and adolescence. In addition, the use of multiple informant models enabled us to identify the potential periods of heightened susceptibility and reduce the number of fitted regression models.

In conclusion, we found that gestational, but not postnatal, serum concentrations of PFOA and PFHxS were consistently associated with subtle increases in measures of central body fat among HOME Study adolescents. Future studies could confirm these findings in other populations and identify potential biological pathways underlying these associations.