Plasma Concentrations of Per- and Polyfluoroalkyl Substances and Body Composition From Mid-Childhood to Early Adolescence

Abstract

Context

Per- and polyfluoroalkyl substances (PFAS) may alter body composition by lowering anabolic hormones and increasing inflammation, but data are limited, particularly in adolescence when body composition is rapidly changing.

Objective

To evaluate associations of PFAS plasma concentrations in childhood with change in body composition through early adolescence.

Methods

A total of 537 children in the Boston-area Project Viva cohort participated in this study. We used multivariable linear regression and Bayesian kernel machine regression (BKMR) to examine associations of plasma concentrations of 6 PFAS, quantified by mass spectrometry, in mid-childhood (mean age, 7.9 years; 2007-2010) with change in body composition measured by dual-energy x-ray absorptiometry from mid-childhood to early adolescence (mean age, 13.1 years).

Results

In single-PFAS linear regression models, children with higher concentrations of perfluorooctanoate (PFOA), perfluorooctane sulfonate (PFOS), perfluorodecanoate (PFDA), and perfluorohexane sulfonate (PFHxS) had less accrual of lean mass (eg, −0.33 [95% CI: −0.52, −0.13] kg/m² per doubling of PFOA). Children with higher PFOS and PFHxS had less accrual of total and truncal fat mass (eg, −0.32 [95% CI: −0.54, −0.11] kg/m² total fat mass per doubling of PFOS), particularly subcutaneous fat mass (eg, −17.26 [95% CI −32.25, −2.27] g/m² per doubling of PFOS). Children with higher PFDA and perfluorononanoate (PFNA) had greater accrual of visceral fat mass (eg, 0.44 [95% CI: 0.13, 0.75] g/m² per doubling of PFDA). Results from BKMR mixture models were consistent with linear regression analyses.

Conclusion

Early life exposure to some but not all PFAS may be associated with adverse changes in body composition.

Adolescence is a key time for changes in body composition. Levels of fat and lean mass in adolescence track into adulthood (1, 2), and higher fat mass and lower lean mass have both been associated with increased risk for adverse cardiometabolic health among adolescents (3-5) and adults (6). It is thus critical to identify modifiable factors that may influence changes in body composition across adolescence.

Growing evidence suggests that chemicals in the environment may contribute to changes in body composition (7). For example, per- and polyfluoroalkyl substances (PFAS) are synthetic chemicals used to make many products, such as carpets, clothing, and cookware, stain-repellant and oil-resistant (8). PFAS persist in the environment and are detectable in over 98% of US children (8-10), although PFAS exposures may be modifiable by individual behaviors or through following regulatory actions (11). Studies in both human and mouse models suggest that PFAS may increase fat mass and decrease lean mass by lowering levels of anabolic hormones, such as testosterone and insulin-like growth factor-1 (12), and by increasing inflammatory cytokines (13).

Epidemiologic studies of PFAS and fat mass in adults and children have been mixed, showing positive (14), inverse (15-17), and null associations (18), as recently reviewed (19). However, these studies have been limited by cross-sectional design and by evaluating anthropometric measures such as body mass index (BMI) rather than gold-standard dual-energy x-ray absorptiometry (DXA) measures of fat and lean mass. Furthermore, to our knowledge, no studies have considered associations between PFAS and lean mass, despite biologic plausibility and potential long-term implications for cardiometabolic health (6, 20, 21). Also, prior studies have focused on associations of individual PFAS with body composition without considering PFAS as a chemical mixture, and PFAS are often detected together in the environment and the body.

In the present study, we used longitudinal data to evaluate the associations of plasma concentrations of individual PFAS and the PFAS mixture in mid-childhood with changes in BMI Z-score and DXA measures of fat and lean mass from mid-childhood to early adolescence.

Methods

Study Population and Design

We recruited pregnant women to Project Viva, a longitudinal cohort study of prenatal exposures and child health, during their first prenatal visit at Atrius Harvard Vanguard Medical Associates, a multispecialty group practice in eastern Massachusetts, between 1999 and 2002 (22). Of 2128 live singleton births, 901 children had a follow-up research visit in both mid-childhood (age range, 6-10 years; mean, 7.9 years), and early adolescence (age range, 11-16 years; mean, 13.2 years); 537 had PFAS plasma concentrations and at least one measure of body composition and were included in the present analysis. Among participants who attended the mid-childhood and early adolescent visits, those who were included in this analysis were less likely to be White and female compared with participants excluded from this analysis (Supplemental Table 1 (23)).

The institutional review boards of participating institutions approved Project Viva’s research protocol. The involvement of the Centers for Disease Control and Prevention (CDC) laboratory did not constitute engagement in human subjects research. All mothers provided written informed consent for their child’s participation, and children assented at the early adolescent visit.

PFAS Plasma Concentrations

Staff at the Division of Laboratory Sciences at the CDC quantified 8 PFAS in plasma collected in mid-childhood using online solid-phase extraction with isotope dilution high-performance liquid chromatography–mass spectrometry, as previously described (24). The limit of detection (LOD) was 0.1 ng/mL for all PFAS; we replaced values below the LOD with the $\mathrm{L}\mathrm{O}\mathrm{D}/\sqrt{2}$. We summed the concentrations of the isomers of perfluorooctanoate (PFOA, ie, n-PFOA and the sum of perfluoromethylheptanoates and perfluorodimethylhexanoates [Sb-PFOA]) to obtain total PFOA. We summed the concentrations of the isomers of perfluorooctane sulfonate (PFOS, ie, n-PFOS, sum of perfluoromethylheptane sulfonates [Sm-PFOS], and sum of perfluorodimethylhexane sulfonates [Sm2PFOS]) to obtain total PFOS. We decided a priori to consider PFAS with detectable concentrations in >60% of samples for the current analysis, as described in prior literature (10). The PFAS that met this threshold were n-PFOA, n-PFOS, Sm-PFOS, total PFOA, total PFOS, perfluorodecanoate (PFDA), perfluorohexane sulfonate (PFHxS), N-methyl-perfluorooctane sulfonamido acetate (MeFOSAA), and perfluorononanoate (PFNA).

Measures of Body Composition

Trained research assistants measured child weight (TBF-300 A; Tanita, Arlington Heights, IL) and height (calibrated stadiometer; Shorr Productions, Olney, MD) using standardized protocols. We calculated age- and sex-specific BMI Z-scores using the CDC 2000 growth reference (25).

Research assistants performed whole-body DXA scans (Discovery A model; Hologic Inc., Marlborough, MA) to obtain measures of lean (fat-free) mass as well as total, truncal, subcutaneous, and visceral fat mass. We used the same DXA machine on all participants and calibrated it daily with a standard synthetic phantom to check for machine drift. A trained research assistant checked all scans for positioning, movement, and artifact and defined body regions for analysis. We analyzed data with pediatric software (Hologic Apex 4.0). Intra-rater reliability on a subset of measurements was high (r = 0.99) (26). We calculated height-standardized indices of total fat mass, truncal fat mass, and lean mass as kg/(height[m])² and of visceral and subcutaneous fat mass as g/height[m]². Hologic software estimates visceral fat mass by subtracting subcutaneous fat mass from truncal fat mass at the level of the fourth to fifth lumbar vertebrae, with the lower boundary placed at 1 cm above the iliac crest (3).

We subtracted BMI Z-score and DXA measures in mid-childhood from those in early adolescence to obtain changes in body composition from mid-childhood to early adolescence.

Covariates

At enrollment, mothers reported their race/ethnicity, education, prepregnancy height and weight, smoking habits, annual household income, and the child’s father’s height and weight. We calculated total gestational weight gain by subtracting self-reported prepregnancy weight from the last clinical weight recorded prior to delivery. While self-reported weight may be underestimated, a prior validation study of 170 Project Viva participants with measured prepregnancy weight showed that ranking of weight was preserved, and weight reporting did not differ by BMI or race/ethnicity (27). At 6 and 12 months postpartum, mothers reported breastfeeding duration. In early childhood (mean [SD] 3.2 [0.30] years of age), mothers reported their child’s diet (via a food frequency questionnaire (28)) and race/ethnicity. We substituted maternal race/ethnicity for the 10% of children in the cohort who were missing data on race/ethnicity. At the mid-childhood visit, mothers reported child physical activity, screen time, time spent outdoors, sleep duration, and diet (via a PrimeScreen (29)).

Statistical Analyses

We ran separate unadjusted and covariate-adjusted linear regression models to examine associations between each PFAS and change in each body composition outcome. We evaluated linearity of exposure-outcome associations via spline models, and we log₂-transformed PFAS plasma concentrations to meet the linearity assumptions of linear regression. Outcomes were normally distributed. We additionally checked assumptions of linearity, homoscedasticity, and normality by examining the distribution of the residuals from the final, covariate-adjusted linear regression models against fitted values and normal q-q plots.

In our adjusted linear regression models, we included the following covariates based on prior literature showing an association with PFAS exposure (9) or body composition (30) (Supplemental Fig. 1 (23)): maternal age at study enrollment (continuous), gestational weight gain (continuous), and education (with or without college degree); and child age (continuous), race/ethnicity (White, Black, Asian, Hispanic, or other), sex (male or female), and calendar year of PFAS measurement (2007–2010). We examined but did not include the following covariates that did not materially change the effect estimate: maternal smoking during pregnancy, prepregnancy BMI, paternal BMI, annual household income, and child’s breastfeeding duration, physical activity, screen time, time spent outdoors, and sleep duration. In addition, we found no evidence of confounding by diet as assessed by a food frequency questionnaire in early childhood (ie, adherence to the Youth Healthy Eating Index (31) or a dietary pattern of frequently packaged foods and fish that explained the most variation in PFAS concentrations in this cohort (32)) or by an abbreviated PrimeScreen in mid-childhood (ie, intake of dairy, meat, fish, fast food, or soda). We did not control for measures of body composition in mid-childhood to avoid the overadjustment bias that can occur in cases when the exposure may affect the baseline measure of growth (33). PFAS, which although measured in mid-childhood, have a long (3- to 8-year) half-life and thus reflect exposures over several years prior to “baseline.” Ninety-eight percent of participants had information on all covariates, and we performed complete case analyses.

We used Bayesian kernel machine regression (BKMR) to examine associations of exposure to the PFAS mixture with each body composition outcome, accounting for correlations between PFAS and nonlinearity of exposure-outcome relationships. BKMR permitted us to (1) visualize individual PFAS-body composition associations, holding all other PFAS at the median; (2) estimate an overall mixture effect; (3) identify compounds primarily responsible for the overall effect by comparing posterior inclusion probabilities (PIPs); and (4) detect interactions between mixture components (34-36). To meet assumptions of BKMR, we log₂-transformed, scaled, and centered PFAS plasma concentrations and adjusted all models for the covariates included in the linear regression models above. Due to sparse sample sizes for some race/ethnicity groups, we dichotomized race/ethnicity to White and non-White.

We performed several sensitivity analyses of the single-PFAS linear regression models. We adjusted for maternal prenatal PFAS plasma concentrations (measured at a median 10 weeks of gestation), examined associations of PFAS isomers detectable in >60% of our sample (n-PFOA, n-PFOS, and Sm-PFOS) with body composition, and examined associations of PFAS with subcutaneous and visceral fat, the 2 components of truncal fat mass. Finally, we assessed effect modification by child sex via an interaction term and stratification.

For all analyses, we used R (version 3.6.2; R Development Core Team) (36, 37).

Results

Population Characteristics

Forty-seven percent of children were female, 60% were White, and 67% had college-educated mothers. Table 1 shows mean (SD) of each measure of body composition in mid-childhood and mean (SD) change in each measure of body composition from mid-childhood to early adolescence.

Table 1.

Participant characteristics overall and by quartile of perfluorooctanoate (PFOA) plasma concentration measured in mid-childhood

	Quartile of PFOA
	Overall	Q1 a	Q2 a	Q3 a	Q4 a
	n = 537	n = 140	n = 139	n = 121	n = 137
	Mean (SD) or n (%)
Maternal characteristics
Mother’s age at enrollment (years)	32.2 (5.3)	31.0 (6.1)	32.3 (5.5)	32.4 (5.1)	33.0 (4.4)
College graduate	360 (67%)	66 (47%)	94 (68%)	88 (73%)	112 (82%)
Pregnancy weight gain (kg)	15.6 (5.3)	14.9 (5.8)	15.5 (5.1)	16.1 (5.3)	15.7 (5.0)
Child characteristics
Age at mid-childhood visit (years)	7.9 (0.8)	8.2 (1.0)	7.8 (0.7)	7.9 (0.7)	7.6 (0.5)
Age at early teen visit (years)	13.1 (0.8)	13.1 (0.8)	13.1 (0.8)	13.1 (0.9)	13.0 (0.7)
Female	251 (47%)	66 (47%)	66 (47%)	54 (45%)	65 (47%)
Race/ethnicity
Black	108 (20%)	60 (43%)	25 (18%)	14 (12%)	9 (6.6%)
White	323 (60%)	43 (31%)	82 (59%)	86 (71%)	112 (82%)
Asian	12 (2.2%)	5 (3.6%)	2 (1.4%)	2 (1.7%)	3 (2.2%)
Hispanic	30 (5.6%)	13 (9.4%)	10 (7.2%)	5 (4.1%)	2 (1.5%)
Other	63 (12%)	18 (13%)	20 (14%)	14 (12%)	11 (8.0%)
Year of blood draw
2007	63 (12%)	7 (5.0%)	18 (13%)	14 (12%)	24 (18%)
2008	189 (35%)	23 (16%)	42 (30%)	46 (38%)	78 (57%)
2009	176 (33%)	53 (38%)	55 (40%)	40 (33%)	28 (20%)
2010	109 (20%)	57 (41%)	24 (17%)	21 (17%)	7 (5.1%)
Body composition in mid-childhood
BMI Z-score	0.43 (1.08)	0.77 (1.04)	0.50 (1.09)	0.25 (1.10)	0.18 (0.99)
Total fat mass index b	6.3 (3.1)	7.4 (3.8)	6.3 (3.2)	5.8 (2.5)	5.8 (2.3)
Truncal fat mass index b	2.4 (1.4)	2.9 (1.8)	2.4 (1.5)	2.1 (1.2)	2.2 (1.1)
Lean mass index b	14.9 (2.1)	15.8 (2.1)	15.1 (2.2)	14.6 (2.1)	14.3 (1.6)
Change in body composition from mid-childhood to early adolescence
Change in BMI Z-score	−0.02 (0.60)	0.01 (0.60)	−0.03 (0.6)	0.02 (0.61)	−0.08 (0.64)
Change in total fat mass index b	1.8 (1.8)	2.2 (1.9)	1.8 (1.8)	1.8 (1.7)	1.7 (1.7)
Change in truncal fat mass index b	0.9 (0.9)	1.0 (0.9)	0.9 (0.9)	0.9 (0.8)	0.8 (0.8)
Change in lean mass index b	1.9 (1.3)	2.1 (1.4)	2.0 (1.3)	1.7 (1.3)	1.6 (1.1)

Abbreviations: BMI, body mass index; PFOA, perfluorooctanoate.

^a Range of PFOA plasma concentration by quartile (ng/mL): Q1, [<LOD-3.10]; Q2, [3.10–4.50]; Q3 [4.50–6.10]; Q4, [6.10–13.80]. LOD was 0.1 ng/mL.

^b Measured by dual-energy x-ray absorptiometry in kg/m²

We detected PFOA, PFOS, PFHxS, and PFNA in >99% of children. PFDA was detectable in 88% and MeFOSAA in 65% of samples. PFOS and PFOA had the highest median [interquartile range] plasma concentrations (6.4 [5.9] ng/mL and 4.5 [3.0] ng/mL, respectively), and concentrations of all PFAS were similar to those measured in a representative sample of US children in the National Health and Nutrition Examination Survey (NHANES) during the same time period (9). In our sample, PFAS plasma concentrations were in general moderately correlated (Spearman’s rank correlation coefficients [r_s]: 0.12-0.77), with PFOS and PFOA most strongly correlated (r_s = 0.77, Supplemental Table 2 (23)). Children with PFOA plasma concentrations in the highest quartile were more likely to be younger at the mid-childhood research visit, White, or have a mother who was a college graduate. (Table 1).

Single-PFAS Linear Regression Models

The strength of the associations between plasma concentrations of each PFAS and change in each measure of body composition from mid-childhood to early adolescence were similar for unadjusted (data not shown) and covariate-adjusted linear regression models (Fig. 1 and Supplemental Table 3 (23)). In the covariate-adjusted models, children with higher plasma concentrations of PFOS and PFHxS had less accrual of total fat mass, truncal fat mass, and BMI Z-score, although confidence intervals for BMI Z-score crossed the null. For example, for each doubling of PFOS, change in total fat mass index from mid-childhood to early adolescence was −0.32 (95% CI: −0.54, −0.11) kg/m², change in truncal fat mass index was −0.14 (95% CI: −0.24, −0.03) kg/m², and change in BMI Z-score was −0.06 (95% CI: −0.12, 0.00) kg/m². PFDA and PFNA were associated with greater accrual of total and truncal fat mass, although confidence intervals crossed the null (eg, for each doubling of PFDA, change in total fat mass index was 0.09 [95% CI −0.10, 0.28] kg/m²). Children with higher plasma concentrations of PFOA, PFOS, PFDA, and PFHxS had less accrual of lean mass (eg, for each doubling of PFOA, change in lean mass index was −0.33 [95% CI: −0.52, −0.13] kg/m² from mid-childhood to early adolescence). Plasma concentrations of other PFAS were not associated with change in BMI Z-score, fat mass, or lean mass. (Fig. 1 and Supplemental Table 3 (23)).

Figure 1.

Adjusted^a associations of per- and polyfluoroalkyl substance (PFAS) plasma concentrations measured in mid-childhood with change in A, BMI Z-score; and height-adjusted indices of B, total fat mass; C, truncal fat mass; and D, lean mass from mid-childhood to early adolescence. Abbreviations: BMI, body mass index; CI, confidence interval; MeFOSAA, N-methyl-perfluorooctane sulfonamido acetate; PFDA, perfluorodecanoate; PFHxS, perfluorohexane sulfonate; PFNA, perfluorononanoate; PFOA, perfluorooctanoate; PFOS, perfluorooctane sulfonate. NOTE: All effect estimates and 95% CIs are listed in Supplemental Table 2 (23). ^a Adjusted for maternal characteristics (age at enrollment, education, gestational weight gain) and child characteristics (age at mid-childhood visit, race/ethnicity, sex, and calendar year of blood draw). ^b n = 526. ^c n = 417, measured by dual-energy x-ray absorptiometry in kg/m².

PFAS Mixture Analyses

The directionalities of single-PFAS outcome associations in mixture models were generally consistent with the directionalities we observed in single-PFAS linear regression models. For example, in mixture models accounting for all PFAS, children with greater plasma concentrations of PFOA, PFOS, and PFHxS had less accrual of total and truncal fat mass, whereas children with greater plasma concentrations of PFDA and PFNA had greater accrual of total and truncal fat mass. Children with greater plasma concentrations of PFOA, PFOS, PFHxS, and PFDA had less accrual of lean mass (Fig. 2). We also found that greater exposure to the PFAS mixture was associated with less accrual of lean mass (Fig. 3D). For example, a child at the 75^th percentile of the PFAS mixture (vs the 50^th percentile) had 0.15 kg/m² (95% credible interval: −0.28, −0.01) less accrual of lean mass index from mid-childhood to early adolescence. The effect of the PFAS mixture on lean mass was largely driven by PFOA (PIP 0.77, with PIPs of all other PFAS < 0.25, data not shown). Exposure to the PFAS mixture was not associated with other measures of body composition (Fig. 3A-3C).We found no evidence of interactions between any pairs of PFAS in any of our mixture analyses (data not shown).

Figure 2.

Longitudinal change (95% credible intervals) in A, BMI Z-score; and height-adjusted indices of B, total fat mass; C, truncal fat mass; and D, lean mass per doubling of single-PFAS exposure, with all other PFAS held at median. Abbreviations: MeFOSAA, N-methyl-perfluorooctane sulfonamido acetate; PFDA, perfluorodecanoate; PFHxS, perfluorohexane sulfonate; PFNA, perfluorononanoate PFOA, perfluorooctanoate; PFOS, perfluorooctane sulfonate. NOTE: PFAS concentrations were log₂-transformed and standardized. Longitudinal change in total fat mass index, truncal fat mass index, and lean mass index were measured in kg/m² from mid-childhood to early adolescence.

Figure 3.

Overall mixture effect (and 95% credible intervals) on longitudinal change in A, BMI Z-score; B, total fat mass index; C, truncal fat mass index; and D, lean mass index from mid-childhood to early adolescence, when all exposures are set to a given percentile relative to median. Abbreviations: BMI, body mass index; PFAS, per- and polyfluoroalkyl substances. NOTE: Longitudinal change in total fat mass index, truncal fat mass index, and lean mass index were measured in kg/m² from mid-childhood to early adolescence. Models were adjusted for maternal characteristics (age at enrollment, education, gestational weight gain) and child characteristics (age at mid-childhood visit, race/ethnicity, sex, and calendar year of blood draw).

Sensitivity Analyses

Maternal prenatal PFAS plasma concentrations were weakly correlated with PFAS plasma concentrations measured in mid-childhood (r_s range, 0.04-0.39), and when we additionally adjusted our final multivariable models for maternal PFAS, we found that the associations between greater PFAS in childhood and less accrual of fat and lean mass were slightly stronger (Supplemental Table 3 (23)).

PFAS isomers had similar magnitude and directionality of associations with body composition as their parent compounds. For example, change in lean mass index from mid-childhood to early adolescence was −0.19 (95% CI: −0.34, −0.05) kg/m² for each doubling of n-PFOS and −0.22 (95% CI: −0.37, −0.07) kg/m² for each doubling of Sm-PFOS (Supplemental Table 4 (23)), as compared with −0.21 (95% CI: −0.36, −0.06) kg/m² for each doubling of total PFOS.

Children with higher plasma concentrations of PFOS and PFHxS had less accrual of subcutaneous fat mass (eg, −17.26 [95% CI −32.25, −2.27] g/m² per doubling of PFOS and −12.07 [95% CI −21.70, −2.43] g/m² per doubling of PFHxS). Children with higher plasma concentrations of PFDA and PFNA had greater accrual of visceral fat mass from mid-childhood to early adolescence (eg, 4.44 [95% CI 1.34, 7.55] g/m² per doubling of PFDA and 3.89 [95% CI: 0.70, 7.07] g/m² per doubling of PFNA). Other PFAS were not associated with accrual of subcutaneous or visceral fat mass (Supplemental Table 5 (23)).

We found no effect modification by child sex except in the associations of PFDA with accrual of total (P = 0.008) and truncal (P = 0.027) fat mass. Boys with higher plasma concentrations of PFDA had greater fat accrual (eg, for total fat mass, 0.34 [95% CI: 0.07, 0.61] kg/m² for each doubling of PFDA), whereas there were no associations between PFDA plasma concentrations and fat mass in girls. Although interaction terms were not significant, in sex-stratified analyses, other PFAS-body composition associations were generally stronger in girls (Supplemental Table 3 (23)).

Discussion

In our longitudinal analysis of a large Boston-area cohort, children with higher plasma concentrations of several PFAS, particularly PFOA, had less accrual of lean mass from mid-childhood to early adolescence. Children with higher plasma concentrations of PFDA and PFNA had greater accrual of total and truncal fat mass, driven by changes in metabolically active visceral fat mass. Children with higher plasma concentrations of PFOS and PFHxS had less accrual of total and truncal fat mass, driven by changes in subcutaneous fat mass, which is less metabolically active.

Our study is the first, to our knowledge, to examine associations between PFAS exposure and lean mass. The association of greater PFAS plasma concentrations and less accrual of lean mass is biologically plausible, via PFAS-induced lowering of the anabolic hormones testosterone and insulin-like growth factor-1 (12). This finding is clinically important for both adolescent and adult health. Lower lean mass has been associated with greater risk for metabolic syndrome in adolescents (5). Additionally, lean mass tracks from adolescence to adulthood (2), at which time lower lean mass has been associated with lower bone mineral density (38) and increased risk for metabolic syndrome (39).

Our results related to PFAS and adiposity are more nuanced. We found some PFAS to be associated with greater adiposity and others to be associated with less adiposity, such that the mixture of PFAS did not have an overall impact on adiposity. Our finding that children with higher PFOS and PFHxS had less accrual of total and truncal fat mass aligns with other pediatric studies that have observed higher PFAS plasma concentrations to be associated with lower adiposity (15, 16, 40). In an analysis using data from the 1999–2000 and 2003–2004 NHANES cycles, in which participants had PFAS concentrations higher than those in Project Viva, adolescents with higher serum PFOS, PFOA, and PFHxS concentrations had lower waist circumference (15). This is also in line with some rodent models that showed PFAS administration to result in decreased adipose tissue (41), perhaps via PPARα activation and subsequent interference with fatty acid metabolism (42).

However, our data also suggest that children with higher plasma concentrations of PFDA and PFNA may have greater accrual of fat mass. While confidence intervals of the associations of these PFAS with total and truncal fat mass included the null, the magnitude and directionality was suggestive of an association, and higher plasma concentrations of PFDA were significantly associated with total and truncal fat mass in boys. Adults with higher concentrations of PFNA, PFOA, and PFOS have been shown to have greater adiposity (16, 43, 44). Only one pediatric analysis observed an association between higher plasma concentrations of PFAS and adiposity; in this study, PFAS concentrations were measured in 1997 and were higher than in Project Viva (14). However, most prior studies of PFAS and adiposity in children only considered PFOA and PFOS (14, 18, 40), and PFAS binding may differ by functional group and carbon chain length (45-47). Like PFOA, PFDA and PFNA are both longer-chain perfluorinated carboxylic acids, whereas PFOS and PFHxS are perfluorinated sulfonic acids; additional research is needed to determine if effects on fat mass may be related to the impact of PFAS functional group and chain length on PPAR or other receptors.

The present study further extends the prior literature because, to our knowledge, we are the first to suggest that PFAS that lower adiposity may preferentially affect the subcutaneous fat depot, whereas PFAS that lead to greater adiposity may preferentially affect the visceral fat depot. Higher visceral fat mass has been associated with greater metabolic risk and adipocyte dysfunction (3). Thus, although select PFAS appear to have different effects on fat mass (and lean mass), the net clinical impact may be adverse long-term cardiometabolic health outcomes.

In this study, we present the direct effect of PFAS exposure on body composition, and future studies that follow children through later adolescence may investigate the extent to which puberty (48) or other factors may mediate this relationship. Generalizability is a limitation of our cohort because participants in Project Viva are predominately White and college-educated, and all had health care at enrollment. Because predictors of PFAS and PFAS exposure levels may vary by race/ethnicity or socioeconomic status, future studies would benefit from examining more diverse populations. Generalizability is also limited in that PFAS concentrations measured in 2007–2010 in Project Viva participants represent exposures in the United States during those and earlier years (9, 49), and future studies with more recent PFAS measurements will be able to evaluate more recent exposure patterns. In addition, PFDA plasma concentrations had relatively low variability in Project Viva participants. Few studies have investigated PFDA, and additional research will improve our understanding of the impact of PFDA on health outcomes. Finally, as in all studies of PFAS, we are unable to determine an exact vulnerable window of exposure because the PFAS studied have a 3- to 8-year half-life (50), so PFAS measured in mid-childhood reflect exposures throughout early life. Strengths of our study include a large cohort with availability of multiple confounding variables, an environmental mixtures approach, gold-standard DXA measures of body composition, and longitudinal body composition data across adolescence when levels of fat and lean mass are increasing and track into adulthood.

Conclusion

In a large prospective US cohort, children with higher plasma concentrations of PFAS had less accrual of lean mass from mid-childhood to early adolescence. While children with higher plasma concentrations of some PFAS (ie, PFOS and PFHxS) had less accrual of fat mass, particularly subcutaneous fat mass, children with higher concentrations of other PFAS (ie, PFDA and PFNA) had greater accrual of visceral fat mass. Thus, early life exposure to some but not all PFAS may be associated with adverse changes in body composition.

Glossary

Abbreviations

BKMR

Bayesian kernel machine regression

BMI

body mass index

DXA

dual-energy x-ray absorptiometry

LOD

limit of detection

MeFOSAA

N-methyl-perfluorooctane sulfonamido acetate

PFAS

per/polyfluoroalkyl substances

PFDA

perfluorodecanoate

PFHxS

perfluorohexane sulfonate

PFNA

perfluorononanoate

PFOA

perfluorooctanoate

PFOS

perfluorooctane sulfonate

PIP

posterior inclusion probability

Sm-PFOS

sum of perfluoromethylheptane sulfonates

Additional Information

Disclosures: The authors declare they have no actual or potential competing financial interests.

Data Availability

Restrictions apply to the availability of some or all data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.