Association Between Gestational PFAS Exposure and Children’s Adiposity in a Diverse Population

Abstract

Perfluoroalkyl substances (PFAS) are widely distributed suspected obesogens that cross the placenta. However, few data are available to assess potential fetal effects of PFAS exposure on children’s adiposity in diverse populations. To address the data gap, we estimated associations between gestational PFAS concentrations and childhood adiposity in a diverse mother-child cohort. We considered 6 PFAS in first trimester blood plasma, measured using ultra-high-performance liquid chromatography with tandem mass spectrometry, collected from non-smoking women with low-risk singleton pregnancies (n=803). Body mass index (BMI), waist circumference (WC), fat mass, fat-free mass, and % body fat were ascertained in 4–8 year old children as measures of adiposity. We estimated associations of individual gestational PFAS with children’s adiposity and overweight/obesity, adjusted for confounders. There were more non-Hispanic Black (31.7%) and Hispanic (42.6%) children with overweight/obesity, than non-Hispanic white (18.2%) and Asian/Pacific Islander (16.4%) children (p<0.0001). Perfluorooctane sulfonate (PFOS; 5.3 ng/mL) and perfluorooctanoic acid (2.0 ng/mL) had the highest median concentrations in maternal blood. Among women without obesity (n=667), greater perfluoroundecanoic acid (PFUnDA) was associated with their children having higher WC z-score (β=0.08, 95%CI: 0.01, 0.14; p=0.02), fat mass (β=0.55 kg, 95%CI: 0.21, 0.90; p=0.002), and % body fat (β=0.01%; 95%CI: 0.003, 0.01; p=0.004), although the association of PFUnDA with fat mass attenuated at the highest concentrations. Among women without obesity, the associations of PFAS and their children’s adiposity varied significantly by self-reported race-ethnicity, although the direction of the associations was inconsistent. In contrast, among the children of women with obesity, greater, PFOS, perfluorononanoic acid, and perfluorodecanoic acid concentrations were associated with less adiposity (n=136). Our results suggest that specific PFAS may be developmental obesogens, and that maternal race-ethnicity may be an important modifier of the associations among women without obesity.

Introduction

The growing prevalence of child obesity has raised alarm among clinical and public health professionals (Bell et al., 2011; Ogden et al., 2014). Approximately 18.5% of US children were obese in 2015–2016, with a disproportionate impact among non-Hispanic Black (22.0%) and Hispanic (25.8%) children (Hales et al., 2017). Childhood overweight and obesity are associated with greater long-term risks for developing chronic diseases and experiencing psychosocial distress (Dietz, 1998; Halfon et al., 2013; Lakshman et al., 2012), and increases the risk of adult obesity and its associated health risks (Gordon-Larsen et al., 2010). Thus, identifying modifiable risk factors is key to developing interventions for helping to prevent childhood obesity.

The etiology of the childhood obesity epidemic is complex and multifactorial (Keith et al., 2006; Sahoo et al., 2015; Tavalire et al., 2020), driven by genetic factors (Brandkvist et al., 2019), changes in diet (St-Onge et al., 2003), and diminished physical activity (Hills et al., 2011). However, accumulating evidence suggests that environmental factors are also important (Brandkvist et al., 2019). Specifically, gestational exposure to exogenous agents that impact basal metabolic rate, energy storage, and appetite satiety, so-called “environmental obesogens,” potentially increases the risk for obesity (Heindel and Blumberg, 2019; Nadal et al., 2017; Regnier and Sargis, 2014; Veiga-Lopez et al., 2018). Perfluoroalkyl substances (PFAS), a family of widely distributed and highly persistent industrial chemicals (Andrews and Naidenko, 2020; Hu et al., 2016) that cross (Chen et al., 2017a; Mamsen et al., 2019) and accumulate in the placenta (Bangma et al., 2020), may be obesogenic (Egusquiza and Blumberg, 2020; Rahman et al., 2019). Nearly all women had measurable PFAS levels in samples collected during US biomonitoring studies (CDC, 2019), including during pregnancy (Woodruff et al., 2011).

Despite compelling experimental data of developmental obesogenic PFAS effects (Nadal et al., 2017), the epidemiologic evidence has been inconsistent to date. Epidemiologic studies have reported both greater childhood adiposity (Braun et al., 2016; Chen et al., 2019;Gyllenhammar et al., 2018; Karlsen et al., 2017; Lauritzen et al., 2018; Mora et al., 2017), protective (Hartman et al., 2017), and null associations (Andersen et al., 2013; Manzano-Salgado et al., 2017; Martinsson et al., 2020) associated with gestational PFAS exposure. Still, a recent meta-analysis of 10 prospective epidemiologic studies concluded that gestational perfluorooctanoic acid (PFOA) exposure was associated with greater adiposity among children, and that differences in the magnitude of the associations found in European, North American, and Asian study populations suggested modification by ethnicity (Liu et al., 2018). Previous studies have also suggested modification of PFAS-child adiposity associations by sex (Chen et al., 2019; Maisonet et al., 2012; Mora et al., 2017; Starling et al., 2019) and given the link to child adiposity, maternal parity might also modify the associations (Gaillard et al., 2014).

To date, US studies have been conducted in mostly white populations, potentially limiting the generalizability of the results to diverse communities, where the effects are likely to be most pronounced, given co-exposures (Geronimus, 1996; Hicken et al., 2012; Morello-Frosch and Shenassa, 2006). However, we are aware of no studies to date that have comprehensively assessed associations between gestational PFAS exposure and children’s adiposity in a diverse population. The present study is a prospective investigation of gestational PFAS exposure and adiposity in a diverse cohort of US children designed to address this data gap. We hypothesized a priori, that greater maternal plasma PFAS concentrations would be associated with greater adiposity among offspring at 4–8 years of age, and that the association would vary by maternal obesity and race-ethnicity.

Methods

Study Population:

The current study involves participants originally enrolled in the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) Fetal Growth Studies - Singletons cohort. That study enrolled 2802 women aged 18 to 40 years at 12 US clinical centers (Supplementary Figure S1) from July 2009 to January 2013, between gestational weeks 8–13 (Grewal et al., 2018). Participants included 2334 low-risk pregnant women without pre-pregnancy obesity (BMI 19.0–29.9 kg/m²) and a separate cohort of 468 pregnant women with pre-pregnancy obesity (BMI 30.0–45.0 kg/m²). Women self-identified their race-ethnicity as non-Hispanic Black (NHB), non-Hispanic white (NHW), Hispanic, or Asian/Pacific Islander. Women who smoked cigarettes, used recreational drugs, or regularly consumed alcohol were excluded (Grewal et al., 2018).

The Environmental Influences on Child Health Outcomes (ECHO-FGS) Study, consists of 1116 mother-child pairs recruited from May 2017 through April 2019 who participated at 10 of the original 12 study sites, which included 2373 women from the original cohort through delivery (Supplementary Figure S1). In total, 803 children ages 4 to 8 years completed an in-person ECHO-FGS exam and had data required to contribute to the current analysis (Figure 1). ECHO-FGS study participants included in the current analysis (n=803) were similar to other participants in the original cohort (n=1570) with respect to child sex, birth year, maternal education level. and maternal age (all p>0.15). In contrast, in person study visit participant women, were more likely to identify as NHB or NHW and less likely to identify as Hispanic or Asian/Pacific Islander than other participants from the original cohort (p<0.0001). Written informed consent was obtained from the parent or legal guardian of each enrolled child depending on age and state regulations child assent was obtained. The study was approved by the Medical University of South Carolina Institutional Review Board (IRB) and a Central IRB at Columbia University Medical Center and by all the participating sites.

Figure 1. Flow diagram of ECHO-FGS study participants available for analysis.

Abbreviations: BMI, body mass index; PFAS, perfluoroalkyl substances

Measures of Adiposity:

Anthropometry measurements were collected using a standardized protocol. Child’s height was measured on a flat non-carpeted surface using a wall mounted stadiometer. Children were instructed to remove footwear and to stand erect, looking straight ahead. The headboard was lowered to the top of the head and the height was read (0.01 in.) after inhalation. Children’s weight was measured on a hard flat surface using an electronic scale with shoes removed, and the value read from the scale to the nearest 0.1 lb. Waist circumference (WC) was also measured using a non-stretch measuring tape placed over the skin while the child was standing. The tape was wrapped around the waist, parallel to the floor, until snug at the end of a normal expiration, and the value read to the nearest 0.1 cm. The average of two measures was taken for each metric, with a third measure if the first two differed by a preset amount (i.e., 0.25 in. for height, 1 lb. for weight, and 0.5 cm for WC).

We calculated children’s BMI, a measure of total adiposity, as body mass divided by height squared in kg/m². We operationalized BMI as continuous age and sex-specific percentiles of the 2000 US Centers for Disease Control (CDC) growth reference charts, expressed as z-scores (Kuczmarski et al., 2000). We further categorized children’s BMI as obese (≥95^th %tile), overweight (85^th-<95^th %tiles), and normal/underweight (<85^th %tile) according to CDC definitions (Barlow, 2007). We also operationalized WC, a measure of central adiposity, as continuous percentiles of a nationally representative age and sex-specific sample of US children, expressed as z-scores (Cook et al., 2009).

Whole body bioelectrical impedance analysis (BIA) was completed by n=611 children, using an RJL Systems Quantum series Bioelectric Impedance Analyzer (RJL Systems, Inc. Clinton Township, MI USA) (Talma et al., 2013). Children with implanted electrical devices, unable to remain inactive for several minutes, or who refused to participate did not complete BIA. With shoes and socks removed, children were instructed to lie supine and after calibration, electrodes wired to the instrument were placed on hands and feet. The resistance and reactance values were used to estimate whole body fat mass (kg), fat free (muscle) mass (kg), and % body fat values based on the manufacturer’s equations. Three measures were taken which were then averaged.

Exposure Assessment:

PFAS were measured in maternal plasma collected at 8–13 weeks gestation. PFAS determination was conducted at the Wadsworth Center, New York State Department of Health (Albany, NY USA) for women without obesity (n=667) and using a similar method for women with obesity (n=136) after the laboratory moved to the Department of Pediatrics, New York University School of Medicine (New York, NY USA). Sulfated PFAS, including perfluorohexane sulfonate (PFHxS), perfluorooctane sulfonate (PFOS), perfluorooctane sulfonamide (PFOSA), and perfluorodecane sulfonate (PFDS), and perfluoroalkyl carboxylic acids, including perfluoroheptanoic acid (PFHpA), perfluorooctanoic acid (PFOA), perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA), perfluoroundecanoic acid (PFUnDA), and perfluorododecanoic acid (PFDoDA), were determined using analytical methods that have been previously described in detail (Buck Louis et al., 2018; Honda et al., 2018). Briefly, following solid phase extraction, PFAS were quantified using ultra-high-performance liquid chromatography (UHPLC; Waters Acquity I Class, Milford, MA USA) coupled to an electrospray ionization triple quadrupole tandem mass spectrometer (ESI-MS/MS; AB SCIEX API 5500, Framingham, MA USA) in multiple reaction monitoring mode under negative ionization. The limits of quantification (LOQ), were defined as 10 times the lowest accepted calibration standard on a 14-point matrix-matched calibration curve for each PFAS. This hybrid method has excellent sensitivity, repeatability (inter-day coefficients of variation (CVs) 2.8–6.9%), accuracy (recovery 88.7–117%), precision (intra-day CVs 1.3–5.4%), and range of quantification (0.05–0.09 to 80–200 ng/mL).

The laboratory successfully participates in the CDC and Arctic Monitoring Assessment Program round robin laboratory PFAS proficiency testing programs. Plasma cotinine, a stable nicotine metabolite (Benowitz et al., 2009), was determined using UHPLC with ESI-MS/MS as previously described (Buck Louis et al., 2018). Instrument reported data were used for values less than quantiation limits without imputation to minimize bias in regression models (Cole et al., 2009; Schisterman et al., 2006).

Covariates:

Sociodemographic, reproductive/obstetric, and lifestyle factors from the original study and the ECHO-FGS were selected a priori for adjustment as confounding variables based on literature evidence for causal associations with both gestational PFAS exposure and children’s body composition. We extracted covariate data from interviewer-administer study questionnaires, including maternal age at the time of biospecimen collection (years), self-reported pre-pregnancy BMI (kg/m²) (Catalano et al., 2009), education (dichotomized as completion of secondary/high school or not) as an indicator of socioeconomic status (Galobardes et al., 2006), self-reported race-ethnicity (NHB, NHW, Hispanic, or Asian/Pacific Islander), plasma cotinine concentration as an indicator of environmental tobacco smoke exposure (ng/mL) (Sagiv et al., 2015), and child’s sex (Mamsen et al., 2019) and age at examination (years). We did not adjust for study site given strong correlations with plasma PFAS concentrations. Maternal obesity at study enrollment was defined based on BMI dichotomized as ≥30 kg/m² vs. <30 kg/m².

Statistical Analysis:

We examined and characterized the distributions of all variables. We also assessed these measures stratified by maternal race-ethnicity and maternal obesity. We tested differences in maternal and children’s characteristics and adiposity across maternal race-ethnicity using Student T-tests and x²-tests. We natural log transformed gestational PFAS concentrations to approximate a normal distribution and stabilize variances prior to statistical analysis.

We implemented generalized linear models to estimate associations between individual gestational PFAS concentrations and continuous or categorical measures of child’s adiposity as outcomes. Given differences in the enrollment criteria and PFAS analysis for women without and with obesity, we stratified the multivariate analysis according to maternal obesity. PFAS were rescaled and log transformed to express effect estimates per difference in interquartile range (IQR). For continuous measures, we used multiple linear regression models, including BMI-for-age/sex percentile (z-score), WC-for-age/sex percentile (z-score), fat mass (kg), fat-free mass (kg), and % body fat. Multinomial logistic regression models were used for categorical outcomes, and model coefficients exponentiated to estimate odds ratios (OR) with corresponding 95% confidence intervals (CI). We incorporated and retained quadratic terms with p<0.10 to test for non-linear dose-response associations between individual PFAS and children’s measures of adiposity. We tested for statistical interactions among women without obesity (n=667), using cross-product terms between log-transformed PFAS with maternal race-ethnicity, parity, and child’s sex in additional regression models, retaining non-linear dose-response associations indicated by the main effects models. Models were further stratified by maternal race-ethnicity, parity, and child’s sex to interpret significant interactions. We additionally adjusted the main models for breastfeeding (i.e., breastfed exclusively for ≥6 months) in a sensitivity analysis of n=621 women without obesity and n=129 women with obesity (Kingsley et al., 2018).

We considered p<0.05 and p<0.10 as statistically significant for two-tailed tests of main effects and interactions, respectively. We did not correct for multiple statistical tests, to maximize sensitivity for detecting modest associations (Goldberg and Silbergeld, 2011). We excluded participants with missing covariates. SAS v9.4 (SAS Institute, Inc. Cary, NC USA) and R statistical software (R Foundation for Statistical Computing, Vienna, Austria) were used for the analysis.

Results

Participant Characteristics and Children’s Adiposity:

Table 1 describes the distribution of maternal and children’s characteristics and adiposity measures, overall and by maternal race-ethnicity. Asian women had lower pre-pregnancy BMI than others and fewer were overweight or had obesity. In contrast, most Hispanic women had overweight or obesity, and were more likely to be parous than others. The distribution of maternal race-ethnicity also varied by study site. Child’s sex was similar according to maternal race-ethnicity, although children of NHW women tended to be older on average than their counterparts (7.5 vs. 6.8 years overall), and children of NHB women were least likely to have been breastfed exclusively for ≥6 months (15.4% vs. 22.1% overall). Children of Hispanic women had the highest average BMI (17.6 vs. 16.8 kg/m² overall) and BMI z-score (69.8% vs. 62.4% overall), with the greatest proportion of overweight (19.1% vs. 15.7% overall) and obesity (23.5% vs. 12.8% overall). Similarly, children of Hispanic women had greater WC, % body fat, and fat mass than other children.

Table 1.

Distribution of characteristics and adiposity measures among ECHO-FGS participants, overall and by race-ethnicity

Characteristics	Overall (n=803)	NHW (n=236)	NHB (n=259)	Hispanic (n=204)	Asian/PI (n=104)	P-valuea
Mothers -
Age (years), mean (SD)	28.4 (5.7)	31.1 (4.2)	24.9 (5.5)	28.5 (5.8)	30.9 (4.5)	<0.0001
Pre-pregnancy BMI (kg/m2), mean (SD)	25.6 (5.3)	25.2 (4.9)	26.2 (5.9)	26.6 (5.5)	22.9 (2.8)	<0.0001
Pre-pregnancy BMI, n (%):
Normal weight (18–24.9 kg/m2)b	450 (56.0)	141 (59.8)	134 (51.7)	98 (48.0)	77 (74.0)	<0.0001
Overweight (25–29.9 kg/m2)	217 (27.0)	58 (24.6)	76 (29.3)	57 (27.9)	26 (25.0)
Obese (≥30 kg/m2)	136 (16.9)	37 (15.7)	49 (18.9)	49 (24.0)	1 (1.0)
Education, n (%):
Less than or up to high school	228 (28.4)	10 (4.2)	112 (43.2)	88 (43.1)	18 (17.3)	<0.0001
More than high school	575 (71.6)	226 (95.8)	147 (56.8)	116 (56.9)	86 (82.7)
Plasma cotinine (ng/mL), geometric mean (SD)	0.001 (0.44)	0.0004 (0.20)	0.02 (1.2)	0.0004 (0.19)	0.0001 (0.09)	0.10
Parity, n (%):
Nulliparous	363 (45.2)	103 (43.6)	133 (51.4)	69 (33.8)	58 (55.8)	0.0002
Parous	440 (54.8)	133 (56.4)	126 (48.6)	145 (66.2)	46 (44.2)
Study/enrollment site, n (%):
Christiana Care Health Systems (DE)	239 (29.8)	91 (38.6)	88 (34.0)	43 (21.1)	17 (16.4)	<0.0001
Northwestern University (IL)	105 (13.1)	54 (22.9)	17 (6.6)	18 (8.8)	16 (15.4)
Medical University of South Carolina (SC)	109 (13.6)	65 (27.5)	37 (14.3)	3 (1.5)	4 (3.9)
University of Alabama - Birmingham (AL)	113 (14.1)	2 (0.9)	102 (39.4)	8 (3.9)	1 (1.0)
Columbia University (NY)	96 (12.0)	5 (2.1)	7 (2.7)	76 (37.3)	8 (7.7)
Long Beach Memorial Medical Center (CA)	49 (6.1)	12 (5.1)	6 (2.3)	15 (7.4)	16 (15.4)
New York Hospital - Queens (NY)	36 (4.5)	0	1 (0.4)	2 (1.0)	32 (30.8)
University of California - Irvine/FVH (CA)	21 (2.6)	1 (0.4)	0	13 (6.4)	7 (6.7)
Saint Peter’s University Hospital (NJ)	35 (4.4)	6 (2.5)	1 (0.4)	26 (12.8)	3 (2.9)
Children -
Sex, n (%)
Female	382 (47.6)	136 (44.9)	126 (58.6)	101 (49.5)	49 (47.1)	0.80
Male	421 (52.4)	130 (55.1)	133 (51.4)	103 (50.5)	55 (52.9)
Exclusively breastfed, n (%)c,d	167 (22.1%)	62 (26.3%)	40 (15.4%)	34 (16.7%)	31 (29.8%)	0.004
Age (years), mean (SD)	6.8 (1.0)	7.5 (0.8)	6.3 (0.9)	6.9 (0.9)	6.1 (1.0)	<0.0001
BMI (kg/m2), mean (SD)	16.8 (2.6)	16.5 (1.9)	16.7 (2.7)	17.6 (3.0)	16.0 (2.3)	<0.0001
BMI-for-age/sex (%tile), mean (SD)	62.4 (28.4)	60.3 (24.9)	61.9 (30.5)	69.8 (28.3)	54.0 (27.5)	<0.0001
BMI, n (%):
Underweight (<5th %tile)	25 (3.1)	2 (0.9)	13 (5.0)	7 (3.4)	3 (2.9)	<0.0001
Normal weight (5th–85th %tile)	549 (68.4)	191 (80.9)	164 (63.3)	110 (53.9)	84 (80.7)
Overweight (85th–95th %tile)	126 (15.7)	30 (12.7)	44 (17.0)	39 (19.1)	13 (12.5)
Obese (≥95th %tile)	103 (12.8)	13 (5.5)	38 (14.7)	48 (23.5)	4 (3.9)
WC (cm), mean (SD)	58.4 (7.8)	59.5 (5.9)	56.0 (8.0)	61.8 (8.4)	55.8 (6.6)	<0.0001
WC-for-age/sex (%tile), mean (SD)	56.4 (29.6)	58.4 (24.6)	46.8 (31.7)	68.3 (26.9)	52.2 (26.8)	<0.0001
% Body fat, mean (SD)	17.2 (8.3)	15.5 (5.5)	15.6 (8.4)	20.3 (8.7)	19.9 (10.4)	<0.0001
Fat mass (kg), mean (SD)	4.7 (3.3)	4.3 (2.4)	4.4 (3.6)	6.0 (3.9)	4.3 (2.8)	<0.0001
Fat-free mass (kg), mean (SD)	20.9 (3.9)	22.2 (2.9)	20.7 (3.5)	21.3 (4.4)	17.1 (3.7)	<0.0001

Abbreviations: BMI, body mass index; FVH, Fountain Valley Hospital; NHB, non-Hispanic black; NHW, non-Hispanic white; PI, Pacific Islander; SD, standard deviation; WC, waist circumference

^ap-value for difference between maternal race-ethnicity groups

^bincludes n=6 underweight women (17.8–18.5 kg/m²)

^cExclusive breastfeeding for ≥6 month

^dn=47 missing

Distribution of Maternal Gestational PFAS:

Table 2 shows the distributions of plasma PFAS concentrations, overall and by maternal race-ethnicity. PFOS, a long chain perfluoroalkyl sulfonic acid, had the highest median concentration (5.3 ng/mL) and greatest variability, followed by PFOA (2.0 ng/mL), PFHxS (0.9 ng/mL), PFNA (0.8 ng/mL), PFDA (0.3 ng/mL), and PFUnDA (0.2 ng/mL). There were few PFOSA (0.4%), PFDS (9.3%), PFHpA (45.6%), and PFDoDA (53.1%) values above the LOQ, which we did not consider further. NHWs had higher median PFHxS (1.4 ng/mL), PFOS (7.1 ng/mL), PFOA (2.8 ng/mL), and PFNA (1.0 ng/mL) concentrations than the other racial-ethnic groups. Supplementary Figure S2 shows the distributions of plasma PFAS concentrations among women with and without obesity. There was little consistency in the differences, although the median PFHxS level among women with obesity was nearly two times greater than in women without obesity. Supplementary Table S1 shows weak to strong positive pairwise correlations for the gestational PFAS concentrations, including women with and without obesity. We found a similar pattern of positive pairwise correlations when we limited the analysis to 667 women without obesity (data not shown).

Table 2.

Distribution of maternal gestational plasma PFAS (ng/mL) among ECHO-FGS participants, overall and by race-ethnicity

PFAS	n (%) >LOQ	Mean	SD	Minimum	5th %tile	25th %tile	Median	75th %tile	95th %tile	Maximum
Overall (n=803)
PFHxS	803 (100)	1.4	1.6	0.1	0.3	0.5	0.9	1.5	3.8	19.2
PFOS	803 (100)	6.7	4.7	0.2	2.1	3.7	5.3	8.4	15.3	40.3
PFOA	803 (100)	2.5	2.6	0.3	0.7	1.3	2.0	3.0	6.0	57.4
PFNA	802 (99.9)	1.0	0.8	0.1	0.3	0.6	0.8	1.2	2.2	10.7
PFDA	762 (94.9)	0.4	0.5	−0.2	0.1	0.2	0.3	0.4	1.0	13.0
PFUnDA	721 (89.8)	0.3	0.7	0.0	0.0	0.1	0.2	0.3	1.0	14.2
non-Hispanic Black (n=259)
PFHxS	259 (100)	1.4	1.8	0.1	0.3	0.6	1.0	1.5	3.5	19.2
PFOS	259 (100)	6.9	4.6	1.0	2.0	4.0	5.5	8.4	15.3	31.5
PFOA	259 (100)	1.9	1.2	0.4	0.6	1.1	1.6	2.4	4.3	7.9
PFNA	259 (100)	0.9	0.9	0.2	0.3	0.5	0.7	1.2	2.1	8.0
PFDA	244 (94.2)	0.3	0.4	0.0	0.1	0.1	0.2	0.4	1.1	2.1
PFUnDA	236 (91.1)	0.4	0.7	0.0	0.0	0.1	0.2	0.4	1.2	6.4
non-Hispanic white (n=236)
PFHxS	236 (100)	2.0	1.9	0.2	0.5	0.9	1.4	2.4	5.8	12.6
PFOS	236 (100)	8.5	5.6	1.2	2.8	4.6	7.1	10.6	20.7	40.3
PFOA	236 (100)	3.6	4.1	0.4	1.0	1.9	2.8	4.0	8.1	57.4
PFNA	235 (99.6)	1.1	0.6	0.1	0.4	0.7	1.0	1.4	2.4	3.3
PFDA	225 (95.3)	0.4	0.3	−0.1	0.1	0.2	0.3	0.5	1.2	2.1
PFUnDA	215 (91.1)	0.3	0.2	0.0	0.0	0.1	0.2	0.4	0.7	1.5
Hispanic (n=204)
PFHxS	204 (100)	0.9	0.7	0.1	0.2	0.4	0.7	1.1	2.2	4.2
PFOS	204 (100)	4.4	2.3	0.4	1.7	2.7	3.9	5.2	9.0	13.9
PFOA	204 (100)	2.1	1.3	0.3	0.5	1.3	1.8	2.6	4.3	9.0
PFNA	204 (100)	1.0	1.0	0.1	0.3	0.6	0.7	1.0	2.1	10.7
PFDA	192 (94.1)	0.3	0.9	0.0	0.1	0.2	0.2	0.3	0.6	13.0
PFUnDA	168 (82.4)	0.3	1.0	0.0	0.0	0.1	0.2	0.2	0.5	14.2
Asian/PI (n=104)
PFHxS	104 (100)	0.7	0.5	0.1	0.3	0.4	0.6	0.9	1.7	2.9
PFOS	104 (100)	6.7	4.0	0.2	2.1	3.8	5.7	8.5	13.3	22.3
PFOA	104 (100)	2.4	2.2	0.6	0.8	1.4	1.9	2.8	4.2	18.8
PFNA	104 (100)	1.1	0.8	0.3	0.4	0.7	0.9	1.3	2.6	6.1
PFDA	101 (97.1)	0.4	0.3	−0.2	0.1	0.2	0.3	0.6	1.2	1.7
PFUnDA	102 (98.1)	0.5	0.6	0.0	0.1	0.2	0.3	0.7	1.4	3.8

Abbreviations: LOQ, limit of quantification; NHB, non-Hispanic black; NHW, non-Hispanic white; PFAS, perfluoroalkyl substances; PFDA, perfluorodecanoic acid; PFHxS, perfluorohexane sulfonate; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctane sulfonate; PFUnDA, perfluoroundecanoic acid; PI, Pacific Islander; SD, standard deviation

Associations Between Gestational PFAS and Children’s Adiposity of Women Without Obesity:

Table 3 shows the results of linear regression models with individual gestational plasma PFAS as predictors of children’s continuous adiposity measures, among 661 women without obesity, adjusted for confounding variables. Higher gestational PFUnDA concentration, a long chain perfluoroalkyl carboxylic acid, was associated with greater WC z-score (β=0.08; 95% CI: 0.01, 0.14; p=0.02), fat mass (β=0.55 kg; 95% CI: 0.21, 0.90; p=0.002), and % body fat (β=0.01; 95% CI: 0.003, 0.01; p=0.004). However, the PFUnDA association with fat mass attenuated at the highest concentrations, according to a parabolic dose-response relationship suggested by the statistically significant beta estimate (β=−0.04 kg; 95% CI: -0.09, 0.01 p=0.099) for PFUnDA squared (Supplementary Figure S3). Higher gestational PFOS concentration was also associated with greater % body fat, albeit of “borderline” statistical significance (β=0.01%; 95% CI: −0.001, 0.02; p=0.08). These results were similar in a sensitivity analysis adjusting for exclusive breastfeeding ≥6 months (data not shown).

Table 3.

Adjusted associations between log-transformed maternal gestational plasma PFAS (ng/mL), per interquartile range, and measures of children’s adiposity at 4–8 years of age among ECHO-FGS participants without obesity

PFAS	BMI, Z-score (n=661)			WC, Z-score (n=659)			Fat Mass, kg (n=493)			Fat-free mass, kg (n=493))			% Body fat (n=493)
	β	95% CI Lo	95% CI Hi	β	95%CI Lo	95%CI Hi	β	95% CI Lo	95% CI Hi	β	95% CI Lo	95% CI Hi	β	95% CI Lo	95% CI Hi
PFHxS	−0.06	−0.17	0.05	−0.06	−0.15	0.04	−0.08	−0.42	0.25	−0.04	−0.34	0.26	−0.003	−0.01	0.01
PFOS	−0.06	−0.17	0.05	−0.05	−0.15	0.05	0.19	−0.16	0.55	−0.06	−0.38	0.25	0.01	−0.001	0.02
PFOA	−0.05	−0.16	0.05	0.03	−0.06	0.13	0.14	−0.19	0.47	0.07	−0.22	0.36	0.003	−0.01	0.01
PFNA	−0.02	−0.12	0.07	0.06	−0.02	0.14	0.20	−0.08	0.49	0.16	−0.09	0.41	0.004	−0.004	0.01
PFDA	−0.03	−0.10	0.04	0.01	−0.05	0.08	0.08	−0.15	0.31	0.06	−0.14	0.26	0.002	−0.004	0.01
PFUnDA	0.01	−0.06	0.08	0.08 a	0.01	0.14	0.55 b	0.21	0.90	0.10	−0.09	0.29	0.01 b	0.003	0.01
PFUnDA2	-	-	-	-	-	-	−0.04 c	−0.09	0.01	-	-	-	-	-	-

Abbreviations: BMI, body mass index; PFAS, perfluoroalkyl substances; PFDA, perfluorodecanoic acid; PFHxS, perfluorohexane sulfonate; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctane sulfonate; PFUnDA, perfluoroundecanoic acid; WC, waist circumference.

^ap<0.05

^bp<0.01

^cp<0.10 for Type-3 test of PFAS × PFAS

Note: Mean difference (95% confidence interval) for each interquartile range of natural log transformed plasma PFAS, adjusted for maternal age (years) at the time of specimen collection, pre-pregnancy BMI (kg/m²), education (dichotomous), race (categorical), plasma cotinine (ng/mL), and child’s sex and age at examination. Bold typeface is statistically significant.

Table S2 shows the results of multinomial logistic regression models using individual PFAS as predictors of children’s categorical overweight and obesity. We found mixed associations without statistical significance among women without obesity. These results were similar in a sensitivity analysis also adjusting for exclusive breastfeeding ≥6 months (data not shown).

We tested for statistical interactions and then explored heterogeneity by maternal race-ethnicity using stratified models of 661 women without obesity in Table 4. We detected interactions of race-ethnicity with perfluoroalkyl sulfonic acids, including PFHxS and PFOS, and with perfluoroalkyl carboxylic acids, including PFOA, PFNA, PFDA, and PFUnDA. WC z-score and PFOA (p=0.01) and PFNA (p=0.05). PFNA (p=0.03), PFDA (p=0.05), and curvilinear PFUnDA (p=0.001) interacted with race-ethnicity on fat mass. Race-ethnicity also interacted with PFHxS (p=0.01), PFOS (p=0.01), PFNA (p=0.01), and PFUnDA (p=0.001) on fat-free mass. Finally, race-ethnicity interacted with PFOS (p=0.03), PFNA (p=0.01), PFDA (p=0.001), and PFUnDA (p=0.002) for % body fat. We found mixed associations between gestational PFAS and children’s adiposity measures across maternal raceethnicity strata. Generally, higher PFAS concentrations tended to be associated with greater adiposity among NHB and Asian/Pacific Islanders, although with less adiposity among NHW and Hispanics. However, the directions of the effect estimates were inconsistent.

Table 4.

Adjusted associations between log-transformed maternal gestational plasma PFAS (ng/mL), per interquartile range, and measures of children’s adiposity at 4–8 years of age among ECHO-FGS participants without obesity, by maternal race-ethnicity

	BMI, Z-score			WC, Z-score			Fat Mass, kg			Fat-free mass, kg			% Body fat
PFAS	β	95% CI Lo	95% CI Hi	β	95%CI Lo	95%CI Hi	β	95% CI Lo	95% CI Hi	β	95% CI Lo	95% CI Hi	β	95% CI Lo	95% CI Hi
NHB	(n=208)			(n=207)			(n=141)			(n=141)			(n=141)
PFHxS	−0.15	−0.35	0.06	−0.09	−0.29	0.10	−0.43	−1.11	0.26	−0.28 c	−0.82	0.26	−0.01	−0.03	0.004
PFOS	−0.01	−0.23	0.22	−0.11	−0.32	0.10	0.002	−0.67	0.68	0.06 c	−0.47	0.59	−0.001 c	−0.02	0.02
PFOA	−0.12	−0.37	0.12	0.04 c	−0.19	0.27	−0.07	−0.83	0.69	−0.10	−0.70	0.50	−0.01	−0.03	0.01
PFNA	−0.06	−0.27	0.15	0.11 c	−0.09	0.31	0.51 c	−0.14	1.17	0.51 c	−0.01	1.02	0.0004 c	−0.02	0.02
PFDA	−0.06	−0.23	0.10	−0.07	−0.22	0.08	0.08 c	−0.41	0.58	0.16	−0.23	0.54	−0.001 c	−0.01	0.01
PFUnDA	−0.02	−0.16	0.13	0.07	−0.07	0.20	−0.41 c	−1.24	0.41	0.48 c	0.15	0.81	0.003 c	−0.01	0.01
PFUnDA2	-	-	-	-	-	-	0.24 c	0.04	0.44	-	-	-	-	-	-
NHW	(n=198)			(n=198)			(n=165)			(n=165)			(n=165)
PFHxS	−0.06	−0.18	0.07	−0.02	−0.14	0.09	0.06	−0.30	0.42	0.33 c	−0.09	0.75	−0.004	−0.01	0.01
PFOS	−0.18a	−0.34	−0.03	−0.11	−0.25	0.03	−0.07	−0.54	0.40	0.46 c	−0.09	1.00	−0.01 c	−0.02	0.01
PFOA	−0.12	−0.25	0.02	−0.03 c	−0.15	0.09	−0.15	−0.53	0.23	0.37	−0.08	0.81	−0.01	−0.02	0.002
PFNA	−0.16a	−0.31	−0.002	−0.05 c	−0.19	0.09	−0.28 c	−0.73	0.18	0.47 c	−0.05	0.99	−0.01 c	−0.02	−0.0002
PFDA	−0.11	−0.22	0.002	−0.05	−0.15	0.05	−0.19 c	−0.54	0.15	0.21	−0.19	0.62	−0.01 c	−0.02	0.001
PFUnDA	−0.06	−0.22	0.10	−0.03	−0.17	0.12	−0.50 c	−1.21	0.22	0.34 c	−0.21	0.90	−0.001 c	−0.01	0.01
PFUnDA2	-	-	-	-	-	-	0.36 c	−0.03	0.76	-	-	-	-	-	-
Hispanic	(n=152)			(n=151)			(n=111)			(n=111)			(n=111)
PFHxS	0.01	−0.33	0.36	−0.23	−0.51	0.06	−0.05	−1.18	1.08	−0.98 c	−1.89	−0.06	0.01	−0.02	0.04
PFOS	−0.06	−0.37	0.25	−0.14	−0.41	0.12	−0.10	−1.24	1.04	−0.41 c	−1.35	0.53	0.003 c	−0.03	0.03
PFOA	−0.04	−0.31	0.22	−0.19 c	−0.40	0.03	0.08	−0.81	0.98	−0.14	−0.88	0.61	0.003	−0.02	0.02
PFNA	0.01	−0.18	0.20	−0.05 c	−0.21	0.10	−0.07 c	−0.70	0.56	0.16 c	−0.36	0.68	−0.005 c	−0.02	0.01
PFDA	−0.02	−0.17	0.14	−0.02	−0.15	0.11	−0.12 c	−0.61	0.37	0.11	−0.29	0.52	−0.01 c	−0.02	0.01
PFUnDA	0.02	−0.14	0.19	0.04	−0.10	0.17	0.24 c	−1.05	1.52	0.17 c	−0.25	0.59	−0.003 c	−0.02	0.01
PFUnDA2	-	-	-	-	-	-	−0.03 c	−0.15	0.09	-	-	-	-	-	-
Asian/PI	(n=103)			(n=103)			(n=76)			(n=76)			(n=76)
PFHxS	0.12	−0.30	0.54	0.16	−0.22	0.55	0.10	−1.27	1.46	−0.10 c	−1.47	1.27	0.01	−0.04	0.06
PFOS	0.01	−0.26	0.27	0.01	−0.23	0.25	0.62	−0.17	1.41	−1.05 c	−1.82	−0.28	0.03 c	0.01	0.06
PFOA	0.05	−0.21	0.32	0.16 c	−0.09	0.40	0.50	−0.34	1.33	−0.12	−0.96	0.73	0.02	−0.01	0.05
PFNA	0.01	−0.22	0.23	0.05 c	−0.16	0.25	0.07 c	−0.56	0.70	−0.61 c	−1.23	0.01	0.01 c	−0.01	0.03
PFDA	0.02	−0.17	0.21	0.06	−0.12	0.23	0.23 c	−0.36	0.82	−0.38	−0.96	0.21	0.01 c	−0.01	0.03
PFUnDA	0.01	−0.14	0.16	0.04	−0.10	0.18	1.25 c	0.43	2.07	−0.49 c	−0.90	−0.08	0.01 c	−0.0004	0.03
PFUnDA2	-	-	-	-	-	-	−0.23 c	−0.40	−0.07	-	-	-	-	-	-

Abbreviations: BMI, body mass index; PFAS, perfluoroalkyl substances; PFDA, perfluorodecanoic acid; PFHxS, perfluorohexane sulfonate; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctane sulfonate; PFUnDA, perfluoroundecanoic acid; PI, Pacific Islander; WC, waist circumference

^ap<0.05

^bp<0.01

^cp<0.10 for Type-3 test of PFAS × race-ethnicity

Note: Mean difference (95% confidence interval) for each interquartile range of natural log unit greater plasma PFAS, adjusted for maternal age (years) at the time of specimen collection, pre-pregnancy BMI (kg/m²), education (dichotomous), and child's sex and age at examination. Bold typeface is statistically significant.

We also explored heterogeneity by parity using stratified models and tested for statistical interactions among 661 women without obesity in Table S3. Greater maternal PFDA, a perfluoroalkyl carboxylic acid, was associated with a higher BMI z-score (p=0.06), fat mass (p=0.04), and % body fat (p=0.03) among parous women, but not among nulliparous women for whom the associations were inverted or close to the null. Parous women were less likely to have breastfed exclusively for ≥6 months than nulliparous women (16.7% vs. 24.9%, p<0.0001). There was no significant heterogeneity by child’s sex (Table S4).

Associations between Gestational PFAS and Children’s Adiposity in Women with Obesity:

Table 5 shows the results of linear regression models with individual plasma PFAS as predictors of continuous adiposity measures among the 136 women with obesity, adjusted for confounding variables. Paradoxically, BMI z-score was lower in association with greater maternal urinary PFOS (β=−0.27; 95% CI: −0.53, −0.01; p=0.04) and PFDA (β=−0.23; 95% CI: −0.40, −0.07; p=0.01). Greater PFOS (β=−1.56 kg; 95% CI: −2.92, −0.20; p=0.03), PFNA (β=−1.19 kg; 95% CI: −2.19, −0.18; p=0.02), and PFDA (β=−1.02 kg; 95% CI: −1.84, −0.21; p=0.01) were associated with a smaller fat mass. Lower % body fat was similarly associated greater PFOS (β=−0.03%; 95% CI: −0.06, −0.003; p=0.03), PFNA (β=−0.02%; 95% CI: −0.04, −0.002; p=0.03), and PFDA (β=−0.02%; 95% CI: −0.04, −0.004; p=0.02). The results were similar in a sensitivity analysis adjusting for exclusive breastfeeding ≥6 months (data not shown), although the PFUnDA-fat mass association became statistically significant (β=−0.70; 95% CI: −1.36, −0.03; p=0.04).

Table 5.

Adjusted associations between log-transformed maternal gestational plasma PFAS (ng/mL), per interquartile range, and measures of children’s adiposity at 4–8 years of age among ECHO-FGS participants with obesity

PFAS	BMI, Z-score (n=136)			WC, Z-score (n=136)			Fat Mass, kg (n=95)			Fat-free mass, kg (n=95)			% Body fat (n=95)
	β	95% CI Lo	95% CI Hi	β	95%CI Lo	95%CI Hi	β	95% CI Lo	95% CI Hi	β	95% CI Lo	95% CI Hi	β	95% CI Lo	95% CI Hi
PFHxS	0.01	−0.22	0.24	0.16	−0.09	0.40	0.63	−0.68	1.93	0.66	−0.45	1.77	0.01	−0.02	0.04
PFOS	−0.27 a	−0.53	−0.01	−0.24	−0.52	0.04	−1.56 a	−2.92	−0.20	−0.79	−1.98	0.39	−0.03 a	−0.06	−0.003
PFOA	0.05	−0.24	0.34	0.17	−0.14	0.48	−0.03	−1.53	1.48	0.46	−0.82	1.73	−0.01	−0.04	0.03
PFNA	−0.15	−0.36	0.05	−0.04	−0.26	0.18	−1.19 a	−2.19	−0.18	−0.40	−1.28	0.48	−0.02 a	−0.04	−0.002
PFDA	−0.23 b	−0.40	−0.07	−0.13	−0.30	0.05	−1.02 a	−1.84	−0.21	−0.59	−1.30	0.12	−0.02 a	−0.04	−0.004
PFUnDA	−0.13	−0.27	0.01	−0.02	−0.18	0.13	−0.65	−1.33	0.02	−0.37	−0.95	0.21	−0.01	−0.03	0.002

Abbreviations: BMI, body mass index; PFAS, perfluoroalkyl substances; PFDA, perfluorodecanoic acid; PFHxS, perfluorohexane sulfonate; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctane sulfonate; PFUnDA, perfluoroundecanoic acid; WC, waist circumference

^ap<0.05

^bp<0.01

Note: Mean difference (95% confidence interval) for each interquartile range of natural log transformed plasma PFAS, adjusted for maternal age (years) at the time of specimen collection, pre-pregnancy BMI (kg/m²), education (dichotomous), race (categorical), plasma cotinine (ng/mL), and child’s sex and age at examination. Bold typeface is statistically significant.

Table S5 shows the multinomial logistic regression results with individual PFAS as predictors of overweight and obese children, adjusted for confounders. The ORs for overweight were consistently elevated in association with higher plasma PFAS among women with obesity, although the estimates were imprecise. The results were again similar in a sensitivity analysis adjusting for exclusive breastfeeding ≥6 months (data not shown).

Discussion

Key Findings:

In this diverse mother-child cohort, higher gestational plasma PFUnDA concentration, a perfluoroalkyl carboxylic acid, was associated with greater children’s WC z-score, fat mass, and % body fat among women without obesity. Paradoxically, both perfluoroalkyl sulfonic acids and perfluoroalkyl carboxylic acids were associated with less adiposity among the children of women with obesity. We did not detect associations of childhood adiposity with gestational exposure to PFOA, or other PFAS, although there was a suggestive association of PFOS, a perfluoroalkyl sulfonic acid, with greater % body fat. We found significant differences in children’s measures of adiposity by maternal race-ethnicity. Hispanic women’s children had the greatest adiposity and Asian/Pacific Islander women’s children the least, with the exception of % body fat, where children of Asian/Pacific Islander women were similar to children of Hispanic women. NHW women generally had greater plasma PFAS concentrations than other groups, including PFHxS, PFOS, PFOA, and PFNA.

We also detected interactions between PFAS and children’s adiposity measures by maternal race-ethnicity, including both perfluoroalkyl sulfonic acids and perfluoroalkyl carboxylic acids, especially for fat mass, fat-free mass, and % body fat, suggesting differential susceptibilities, but the effect estimates were inconsistent. In the absence of previous data about the potential influence of race-ethnicity, we did not have a stronger a priori expectation for different PFAS or race-ethnicity groups. These hypothesis-generating results require confirmation. We found stronger associations among the children of parous compared to nulliparous women, but the direction of associations with gestational PFAS exposure were similar among female and male children. Overall, our results suggest that gestational PFUnDA was obesogenic among non-obese mothers, and that race-ethnicity was a modifying factor.

Maternal Gestational PFAS Exposure:

We found higher PFAS concentrations among NHW, than NHB, Hispanic, and Asian women. Other studies have reported higher PFAS concentrations among white and Asian women with pregnancies relative to Black and Hispanic women, including perfluoroalkyl sulfonic acids, PFHxS and PFOS, as well as the perfluoroalkyl carboxylic acids PFOA and PFNA (Kato et al., 2014; Sagiv et al., 2015), although the differences have been inconsistent (Park et al., 2019). Overall, our participants had similar or modestly higher median plasma PFAS than US women 18–40 years of age in 2009–2012 (CDC-NCHS, 2021), including serum PFHxS (0.80 ng/mL), PFOS (5.09 ng/mL), PFOA (1.98 ng/mL), PFNA (0.85 ng/mL), PFDA (0.20 ng/mL), and PFUnDA (0.10 ng/mL). The similar levels between our study population and US women suggest that our sample was representative of background PFAS exposures in the US.

Associations Between Gestational PFAS and Children’s Adiposity:

Previous epidemiologic studies have reported associations of gestational PFAS exposure with greater adiposity among children, which are similar to our results among women without obesity. Early pregnancy maternal PFOA, PFNA, and PFDA were related to greater ponderal index and PFNA with BMI z-score at 18 months of age, in a large Danish study, although not for PFHxS or PFOS, or with WC and % body fat (Jensen et al., 2020). In contrast, a study of mother-offspring pairs from Norway and Sweden found higher BMI z-score and risk for overweight at 7 years of age in association with greater mid-pregnancy PFOS, but not PFOA (Lauritzen et al., 2018). Serum PFHxS, PFOS, and PFOA were associated with greater BMI z-score and risk for overweight among Faroese children at 18 months, and for PFOA at 5 years, without associations for PFNA and PFDA (Karlsen et al., 2017). However, biospecimens were collected 2 weeks postpartum in that study, which may have misclassified some participants. The prior studies discussed above have generally found associations of gestational PFOS, PFOA, and PFNA with children’s adiposity, though exposure levels were mostly higher than ours. We did not detect associations for PFOA and PFNA, although we found positive relationships for gestational PFUnDA with children’s adiposity and suggested a positive association of PFOS with children’s body fat among women without obesity.

Also like our findings, previous epidemiologic studies discussed below have reported null or protective (i.e., inverse) associations between certain gestational PFAS exposure and children’s adiposity. Mid-pregnancy PFHxS, PFOS, PFOA, and PFNA were unrelated to total % body fat measured by dual x-ray remission absorptiometry (DXA) in British girls at 9 years of age, although PFHxS, PFOS, and PFOA were associated with lower trunk fat, BMI, and WC (Hartman et al., 2017). Similarly, no consistent association was reported for early pregnancy PFHxS, PFOS, PFOA, and PFNA with children’s overweight at 4 years of age (i.e., BMI >18 kg/m²) in a large case-control study from Sweden, with protective effects for some quartiles of PFHxS and PFNA (Martinsson et al., 2020). Similar to our results, a large Spanish study also had null associations of first trimester maternal PFHxS, PFOS, PFOA, and PFNA with BMI and WC at 4 years and 7 years of age (Manzano-Salgado et al., 2017). A large Danish investigation reported null associations of early gestational PFOS and PFOA with BMI, WC, and overweight and obesity at 7 years of age, at higher levels of exposure than our own (Andersen et al., 2013). More recently, an exposome-based study in six European countries also found no association of prenatal PFAS exposure with adiposity among 4–6 year old children (Vrijheid et al., 2020).Similar to previous studies we did not identify consistent differences in the associations for perfluoroalkyl sulfonic acids and perfluoroalkyl carboxylic acids.

In contrast to several previous epidemiologic investigations, we found no evidence of sex-specific associations between gestational PFAS and children’s adiposity. A US study found positive associations of mid- to late-gestation maternal PFOA and PFNA with adiposity in male children at 5 months of age, but negative associations of PFHxS and PFOS in females; the serum PFAS exposure levels were approximately half of our plasma levels (Starling et al., 2019). Greater skinfold thicknesses, BMI, and DXA-assessed fat mass were associated with higher first trimester PFHxS, PFOS, PFOA, and PFNA among female, but not male children at approximately 7.7 years, in a large US study, at exposure levels greater than ours (Mora et al., 2017). Second trimester gestational PFOS was also associated with greater weight-for-age among singleton females at 20 postnatal months in a British study, although with null associations for PFHxS and PFOA (Maisonet et al., 2012). In contrast, a Chinese investigation, reported inverse associations of umbilical cord blood PFUnDA, a surrogate for prenatal exposure, and % body fat measured by BIA among female children at 5 years of age, and a positive association of PFNA with % body fat among males (Chen et al., 2019). We found that higher maternal plasma PFUnDA was associated with greater adiposity at 4–8 years, without a sex interaction, and with null associations for PFNA, although PFAS concentrations were higher in the Chinese study.

The timing of prenatal PFAS exposure and the timing of childhood assessment may be critical to the identification of effects on children’s adiposity. Previous studies reported associations between gestational PFAS, primarily PFOS and PFOA, measured in maternal serum, fetal cord blood, and neonatal blood spots with children’s weight gain and growth over time, however these results were also discordant (Braun et al., 2016; Chen et al., 2017b; De Cock et al., 2014; Shoaff et al., 2018; Tanner et al., 2020; Yeung et al., 2019). A future analysis integrating longitudinal adiposity measures among children is necessary for a more definitive interpretation of our results.

Potential Biologic Mechanisms:

PFAS are structural homologs of free fatty acids that bind and activate peroxisome proliferator-activated receptors (PPARs), including the alpha (PPARα) and gamma (PPARγ) isoforms (Buhrke et al., 2015, 2013; Rosen et al., 2017; Vanden Heuvel et al., 2006; Zhang et al., 2014). PPARs are nuclear transcription factors that regulate cell and tissue development, differentiation, and metabolism. PPARs are involved in adipogenesis, adipocyte physiology, and energy regulation (Egusquiza and Blumberg, 2020; Griffin et al., 2020). PPAR binding might alter developmental adipogenesis and adipocyte programming, leading to obesogenic effects. In vitro, PFOS and PFOA promoted adipocyte differentiation from human mesenchymal stem cells (Liu et al., 2019). PFHxS, PFOS, PFOA, and PFNA increased the number of adipocytes and total lipid concentration, and upregulated expression of lipid-related proteins, including PPARα and PPARγ, in a 3T3-L1 pre-adipocyte cell model (Watkins et al., 2015; Yamamoto et al., 2015). Gestational exposure to PFOA led to greater midlife weight gain among female mice treated at the lowest observed effect level, but without an effect among mice treated as adults (Hines et al., 2009). PFOS and PFOA have also been linked to the expression of genes involved in lipid metabolism, including PPARα, in a US study population (Fletcher et al., 2013). However, no experimental studies have reported similar effects for PFUnDA to date. Other mechanisms, such as disrupted sex hormone, leptin, and adiponectin signaling, may also contribute to PFAS’s obesogenic effects, and account in part for the differences in parous and nulliparous mothers and the divergent results in mothers with and without obesity (Heindel and Blumberg, 2019)

Study Strengths and Limitations:

Our study has several strengths, including a prospective design incorporating biomarkers of gestational PFAS exposure, large sample size, a diverse study population, and clinically ascertained outcomes. We selected PFAS for statistical analysis with at least 60% of values above the LOQ, to ensure that most measured values fell within the laboratory’s range of quantitation, a more conservative strategy than using the less rigorous limit of detection (LOD) (Keith et al., 1983). However, our study design also has several limitations. We had small numbers of women with obesity and of clinical events (i.e., child’s obesity and overweight), and limited numbers when stratified by women’s race-ethnicity. Still, ours is the first study to report interaction of gestational PFAS exposure and children’s adiposity by race-ethnicity, which requires further investigation to better characterize these disproportionate effects. We also stratified the analysis according to women without obesity and women with obesity. Our results suggested that the children of women without obesity may be more susceptible to gestational PFAS and that associations may differ in parous and nulliparous women. We measured children’s adiposity using routine anthropometric techniques and standardized protocols. However, while BIA has good reliability its validity is limited in children and adolescents relative to gold-standard DXA (Talma et al., 2013). Furthermore, missing values for BIA, which was poorly tolerated by some younger children, may have introduced a selection bias if associated with both adiposity and gestational PFAS exposure. A future study using greater resolution body composition data, such as skin folds and DXA, and incorporating measures of maternal body composition as predictors of offspring’s adiposity, will be necessary for more definitive results.

Our exposure assessment may have also introduced limitations. We estimated gestational PFAS exposure using an early pregnancy biomarker, which is a potentially more accurate dose estimate at critical developmental windows than previous investigations using late gestation and cord blood. Plasma volume expansion is associated with decreased PFAS concentrations over pregnancy (Kato et al., 2014). During pregnancy, plasma volume expansion is detectable by 6 weeks gestation but remains modest (<10%) throughout the first trimester (Vricella, 2017). As such, we believe that plasma volume expansion was unlikely to meaningfully impact our results, which were similar to the general US population. We estimated PFAS exposure at a single time point, which may have misclassified some participants, with bias towards the null hypothesis. However, most PFAS have long in vivo half-lives, measured in years, so the risk of misclassification was again likely modest (Li et al., 2018). We also adjusted for breastfeeding as a surrogate for postnatal PFAS exposure with similar results (Kingsley et al., 2018). However, we were unable to integrate data on children’s consumption of foods predictive of PFAS exposure (e.g., packaged foods including ice cream and carbonated beverages, fish, meat, and fruit juices), which might bias associations if also associated with gestational PFAS exposure (Jain, 2018; Seshasayee et al., 2021).

While we estimated associations of 6 widely distributed PFAS, we did not measure newer short chain PFAS, such as PFBS, which have also been associated with children’s adiposity (Chen et al., 2019). Furthermore, associations were not estimated for PFOSA, PFDS, PFHpA, and PFDoDA due to few values above the LOQs. The PFAS analysis was completed by a single laboratory, but women without obesity and women with obesity were analyzed at different times, before and after a move, respectively. Though some differences in PFAS may reflect laboratory factors, the laboratory has participated in external quality assurance schemes for more than 10 years and plasma concentrations were not systematically different between women, so the impact was likely modest. Still, the inability to combine women without obesity and with obesity in a single analysis likely limited our statistical power. We conducted multiple statistical tests without correction to maximize sensitivity for detecting modest associations, so some findings may have occurred by chance (Rothman, 2014), requiring confirmation in a future investigation. Finally, we estimated associations for individual PFAS, but we plan to perform a “mixtures-based” approach to simultaneously integrate potential additive, synergistic, and non-linear effects among multiple PFAS in a future analysis (Hamra and Buckley, 2018).

Conclusions:

We found that gestational exposure to PFUnDA in healthy women without obesity was associated with children’s adiposity at levels similar to the general US population. The associations demonstrated between gestational PFAS exposure and children’s adiposity varied by maternal obesity, maternal race-ethnicity, and maternal parity. Our results add to a growing literature implicating select PFAS as developmental obesogens and suggests specific populations at greater risk. However, given the restrictive study inclusion criteria these results should be generalized to other populations with caution. These results require confirmation ideally employing a longitudinal study design with a more rigorous assessment of body composition, enhanced exposure assessment, and a comprehensive statistical analysis to definitively identify and intervene on vulnerable and susceptible population subgroups.

Highlights:

• PFAS were measured in pregnant mothers and adiposity assessed in their children.

• Gestational PFUnDA was associated with children’s adiposity at 6–8 years of age.

• Associations of PFAS and children’s adiposity varied by maternal race-ethnicity.

• Obesogenic effects of gestational PFAS may differ by maternal race-ethnicity.