Individual and mixtures of PFAS during pregnancy are associated with maternal cardiometabolic outcomes during pregnancy

Abstract

Introduction

Per- and polyfluoroalkyl substances (PFAS) are endocrine-disrupting chemicals and widespread environmental contaminants. PFAS cross the placental barrier and reach the developing fetus with potential impacts on many organ systems. There are no studies of PFAS in residents of central Arkansas despite reports of environmental contamination in the region. We aimed to quantify PFAS concentrations in repeated serum samples from participants living in central Arkansas and to investigate relationships with maternal cardiometabolic health during pregnancy.

Methods

Participants were enrolled during early pregnancy in a longitudinal study (NCT01131117) from 2010 to 2014. PFAS concentrations were measured in serum from each trimester (first trimester n = 282, second trimester n = 217, and third trimester n = 195). PFAS were compared across pregnancy. Linear and linear-mixed effects models were used to investigate relationships between trimester-specific PFAS levels, as single exposures, and maternal outcomes. Effects of PFAS as an exposure mixture were estimated using quantile g-computation.

Results

Six PFAS were detected in more than 70% of the maternal serum samples: PFOS, PFOA, PFBS, PFHxS, PFNA, and PFHxA. In adjusted linear-mixed models and quantile g-computation models, maternal serum PFAS levels were significantly negatively associated with triglycerides [effect estimates (β)= -16.29; 95% confidence interval (CI)= -24.95, -7.63], total cholesterol (β= -12.77; 95%CI= -19.80, -5.74), low-density lipoproteins (β= -10.83; 95%CI = -16.72, -4.93), high-density lipoproteins (β= -4.10; 95%CI= -6.23, -1.96), and pulse (β= -1.60; 95%CI= -2.85, -0.35). Maternal serum PFAS, as a mixture, was not associated with maternal diastolic blood pressure, but separately, PFAS_sum, PFOS, PFOA and PFNA had significant positive associations.

Conclusion

This study evaluated PFAS exposures during pregnancy in central Arkansas. We show that PFAS exposure during pregnancy influences maternal cardiometabolic outcomes and a case in point that future studies are needed to determine the impact on maternal health and to investigate potential interventions to limit the effects of PFAS exposure during pregnancy.

Introduction

Environmentally ubiquitous per- and polyfluoroalkyl substances (PFAS) [1, 2] are a large group of man-made compounds that are widely used in both industrial and consumer products [3, 4] that may be contributing to the obesity epidemic [5, 6] and poorer cardiometabolic health [7]. Their ubiquitous use and high stability have led to widespread environmental contamination and many sources of human exposure. In a nationally representative survey, the 1999–2000 National Health and Nutrition Examination Survey (NHANES), over 98% of adult Americans had detectable PFAS levels in serum [8]. Geographic location is a major determinant of PFAS exposure [9], and the central Arkansas region has been reported to have groundwater contamination at the Little Rock Air Force Base [10]. To our knowledge, there has been very little environmental monitoring and no published studies of PFAS exposure in people living in the region that are likely to have higher exposure than other US regions.

A life stage that is of particular concern for PFAS exposure is pregnancy. This is partly because PFAS have been shown to cross the placental barrier and reach the developing fetus [11] with the potential to influence many developing fetal organ systems [12, 13]. Maternal PFAS concentrations, which can be modulated by income, race, and parity [14], have been associated with preeclampsia [15, 16], gestational diabetes [17,18,19], altered glucose and insulin regulation [18, 19], and circulating lipids [19, 20], demonstrating a potentially important role for gestational PFAS exposure in maternal health and pregnancy outcomes. Yet, results are still mixed. A systematic review demonstrated that commonly assessed PFAS may be associated with risk for preterm birth, miscarriage, and preeclampsia, although these associations tended to be close to the null, and no associations were observed with gestational diabetes mellitus or pregnancy-induced hypertension via meta-analyses [21]. Subsequent birth cohort studies suggest that increasing concentrations of several PFAS as a mixture, which is a newer approach to estimate overall effect of PFAS exposure, were associated with elevated diastolic blood pressure and increased odds for gestational hypertension [22]. Overall, these findings provide some evidence that gestational PFAS exposures are associated with maternal health or subclinical indicators (lipid and glucose homeostasis) during pregnancy, but additional research in this area is needed. While gestational PFAS levels clearly exhibit some relationships with maternal health, it is important to clarify whether there are trimester-specific effects or whether repeated exposure and outcome assessments yield similar results to those observed in individual trimesters.

Indeed, it is also unknown if there is a critical window of exposure during pregnancy where PFAS has the strongest effect on maternal health. Some studies show that maternal lipid levels have stronger associations with first trimester PFAS concentrations, as opposed to second or third trimester concentrations [23]. However, another study found that PFAS concentrations assessed in the third trimester were associated with higher concentrations of total cholesterol and triglycerides [24]. Discordant results may be explained by differences in lifestyle that can affect circulating levels of PFAS [25], or changes in exposure patterns across pregnancy. Additionally, because PFAS have long half-lives, particularly legacy PFAS like PFOS and PFOA, concentrations across different time points in pregnancy tend to be strongly correlated, but most PFAS levels also appear to decrease in concentration with increasing gestational time [26]. These changes in PFAS across pregnancy are likely due to the deposition of PFAS in the placenta or intrauterine environment [12], as well as hemodynamic changes that occur throughout the gestational period [27].

Taken together, the current literature suggest that specific geographical regions may increase individual exposure, specific windows of exposure during development may heighten effects, and that both mother and fetus are likely affected by prenatal exposure. However, it is unclear what are the effect sizes of these exposures and what developmental windows are critical to focus interventions due to the lack of longitudinal studies with multiple sampling and deep phenotyping. This study will address these challenges quantifying PFAS concentrations in repeated serum samples from the first, second, and third trimesters of pregnancy and test for relationships with maternal cardiometabolic health in a longitudinal cohort birth cohort from central Arkansas.

Methods

Cohort information

Participants from the central Arkansas area were enrolled in a longitudinal study between 2010 and 2014 (www.clinicaltrials.gov, ID # NCT01131117). The cohort consisted of 300 dyads (pregnant women and offspring) who were recruited prior to gestation week 10, second parity, singleton pregnancy, and greater than 21 years of age. Exclusion criteria included pre-existing medical conditions, sexually transmitted infections, medical complications, or medications during pregnancy known to influence fetal growth (including gestational diabetes, thyroid disease, pre-eclampsia, etc.) and conceptions aided with fertility treatment. Pregnancies were uncomplicated with no medical diagnoses or medications known to affect fetal or infant growth. All mothers were non-smokers and reported no alcohol use during pregnancy. Infants included in the analyses were full-term (≥ 37 weeks’ gestation) and had no medical conditions and were not exposed to medications known to influence growth or development of the fetus.

Self-reported variables

Participants self-reported maternal date of birth, maternal education level, household income, delivery method, child date of birth, and child sex. Gestational age at birth was calculated from the estimated due date obtained by early ultrasound. Maternal age was calculated using self-reported maternal date of birth.

Maternal anthropometric and gestational weight gain

Maternal anthropometric measurements were obtained during each trimester’s visit in duplicate using standardized techniques and duplicates were averaged within each trimester to gain a more robust measure for analysis. Maternal weight was measured to the nearest 0.1 kg using a tared standing digital scale (SECA Corp, Ontario, Canada). Maternal height was measured to the nearest 0.1 cm standing against a wall-mounted stadiometer (Perspective Enterprises, Portage, MI). Body mass index at enrollment was calculated (BMI: weight [kg]/height [m]²) for all participants. Gestational weight gain was estimated by subtracting weight at gestation week 36 from enrollment weight and adjusted using the 2009 Institute of Medicine (IOM) (now National Academies of Science, Engineering and Medicine) guidelines [28].

Maternal vitals

Maternal blood pressure and pulse were measured during each trimester’s visits using an electronic blood pressure monitor (Omron Intellisense Blood Pressure Monitor, Omron Healthcare, Inc., Hoffman Estates, IL) in duplicates following standardized techniques, and were averaged within each trimester.

Maternal serum

Maternal blood samples were collected following an overnight fast in each trimester (trimester 1 between < 10 or at 12 weeks, trimester 2 at 24 weeks, trimester 3 between 30 and 36 weeks). For 290 of the 300 participants, at least one blood sample was collected. Serum was isolated and stored at -80 °C until analysis. Total cholesterol, triglycerides, high-density lipoprotein, low-density lipoprotein, and glucose were measured in duplicate from the trimester’s visit sample using an RX Daytona clinical analyzer (Randox Laboratories-US limited; Kearneysville, WV). Insulin was measured using a multispot assay kit (MesoScale Diagnostics, Rockville, MD). Serum creatinine and albumin concentrations were measured using a Roche COBAS Integra 400 Plus (Roche Diagnostics Corp) to estimate glomerular filtration rate (eGFR) using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) calculation [29].

Per- and polyfluoroalkyl substances quantification

Total PFAS were quantified in maternal serum collected at each trimester after acetonitrile and methanol treatment to denature protein using Ultra-High-Performance Liquid Chromatography coupled electrospray ionization tandem mass spectrometry and adaptations of previous published methods [25, 30,31,32,33]. The following PFAS were measured: Perfluorobutane sulfonate (PFBS), Perfluoropentane sulfonic acid (PFPeS), Perfluorohexane sulfonate (PFHxS), perfluoroheptanesulfonic acid (PFHpS), Perfluorooctane sulfonic acid (PFOS), perfluorononanesulfonic acid (PFNS), Perfluorodecanesulfonic acid (PFDS), Perfluoro-n-pentanoic acid (PFPeA), Perfluorohexanoic Acid (PFHxA), Perfluoroheptanoic acid (PFHpA), Perfluorooctanoic acid (PFOA), Perfluorononanoic acid (PFNA), Perfluorodecanoic acid (PFDA), Perfluoroundecanoic acid (PFUnDA), Perfluorododecanoic acid (PFDoA), Perfluorotridecanoic acid (PFTrDA), Perfluorotetradecanoic Acid (PFTeDA), Hexafluoropropylene oxide dimer acid (HFPO-DA), 4,8-dioxa-3 H-Perfluorononanoic Acid (DONA), 11 -Chloroeicosafluoro-3-oxaundecane-1 -sulfonic acid (11CI-PF30UdS), Perfluorooctanesulfonamide (FOSA), N-Methylperfluorooctanesulfonamidoacetic acid (N-MeFOSAA), N-ethyl perfluorooctane sulfonamido acetic acid (N-EtFOSAA), 4:2 Fluorotelomer sulfonic acid (4:2 FTS), 6:2 Fluorotelomer sulfonic acid (6:2 FTS), 8:2 Fluorotelomer sulfonic acid (8:2 FTS), and 9-Chlorohexadecafluoro-3-oxanonane-1 -sulfonic acid (9CI-PF30NS). Materials used as calibrators, controls, and mass-labeled internal standards where available were purchased from Wellington Laboratories (Guelph, Ontario, Canada). The limit of detection (LOD) for all PFAS was 1.0 ng/ml and concentrations below the LOD were replaced with LOD/√2 [34]. Total PFAS exposure (PFAS_sum) was calculated as the sum of the masses of all measured PFAS.

Statistical analyses

Maternal socio-demographic characteristics were summarized using arithmetic mean and standard deviation (SD) for continuous variables, while counts and percentages were used for categorical variables. Trimester-specific maternal outcomes or exposure levels were averaged to estimate levels reflective of the pregnancy-average level. Maternal serum PFAS concentrations were natural logarithmic transformed to account for their skewed distributions; summarized using geometric mean and geometric SD and compared to PFAS levels among NHANES (2011–2012 and 2013–2014) pregnant participants [35, 36]. A Kruskal-Wallis test was used to compare PFAS measurements across the three trimesters followed by Wilcoxon Rank Sum tests to compare concentrations between trimesters. Loess curves were used to observe univariate trends between PFAS burden and maternal hemodynamic measures (albumin and eGFR) in any of the trimesters.

We examined the relation between maternal outcomes (systolic and diastolic blood pressure, pulse, triglycerides, cholesterol, high-density lipoprotein, or low-density lipoprotein) and gestational PFAS levels (as measured in maternal serum) using three modeling approaches. First, we used linear regressions to test for trimester-specific (or pregnancy-average) associations by separately regressing a trimester-specific (or pregnancy-average) maternal outcome on a single, trimester-specific (or pregnancy average) PFAS level. To understand the effects across the gestational period, we regressed each maternal outcome on each PFAS level across all three trimesters using linear mixed-effect models with random intercepts for each participant. Last, we estimated PFAS-mixture effects with quantile g-computation (QGComp), which estimates the joint effects of an exposure mixture while not assuming directionality of effects and yields an effect estimate for the overall mixture [37]. We tested whether a simultaneous 25% increase in all six PFAS levels was associated (1) with each trimester-specific maternal outcome using qgcomp.noboot, and (2) with each maternal outcome across pregnancy using each participant as a clustering variable and 500 bootstrap iterations using qgcomp.boot. For all of the above analyses, we utilized three different adjustment strategies: unadjusted models; adjusted models that included maternal BMI, age, gestational weight gain, income, race, and education as covariates; and then additional adjusted models that included the previous covariates as well as eGFR and albumin. Potential confounders were identified a priori as factors that may influence PFAS exposures and maternal outcomes.

Statistical analysis was performed using R version 4.3.1 or above [38]. Linear mixed-effect modeling was performed using lme4 version 1.1–34 and residual maximum likelihood estimates [39]. Quantile g-computation was conducted using qgcomp version 2.16.1 [40, 41]. Statistical significance was defined based on p < 0.05.

Results

A total of 290 participants provided serum samples during the first, second and/or third trimester of gestation (Table 1). On average, women were 30 years of age, most were White (88%) and attained a college degree (65%). By design, women were of BMI 18.5–35 kg/m², and most women gained weight in excess of the National Academies of Science, Engineering, and Medicine weight gain recommendation (53%) and birthed vaginally (65%). Infants (43% female) were born on average at 39 weeks gestation.

Table 1 Cohort description

Of the 27 PFAS measured in the 805 repeated samples, all were detected in at least one of the samples during pregnancy, but only six PFAS [PFOS (N = 803 above LOD), PFOA (N = 765 above LOD), PFBS (N = 746 above LOD), PFHxS (N = 681 above LOD), PFNA (N = 659 above LOD), and PFHxA (N = 634 above LOD)] were detected in more than 70% of samples across all the sampling time points (Supplemental Table 1) so we focused on these going forward. We observed general trends of decreasing PFAS concentrations with increasing gestational time, and these trends were more apparent for PFAS with higher overall concentrations (Fig. 1A). Mean PFOS and PFHxS levels over pregnancy in our cohort (N = 290) were similar to the corresponding levels among pregnant participants from NHANES (2011–2012 and 2013–2014 cycles; N = 35; Supplemental Fig. 1), but mean PFNA and PFOA levels were significantly higher in NHANES. In total, only 16 samples (15 individual participants) were identified above the “enhanced clinical care” threshold of 20 ng/ml [42] for the sum of PFAS during pregnancy. Thirteen of these fifteen samples with PFAS > 20 ng/ml were from the first trimester assessments. PFAS were weakly and negatively correlated with maternal albumin and eGFR (Supplemental Fig. 2).

Fig. 1

Comparison of natural log-transformed PFAS concentrations (ng/ml) in maternal serum samples at trimester 1 (T1), trimester 2 (T2) and trimester 3 (T3) for PFAS that were detected in at least 70% of samples in our cohort (A, top panel) and maternal outcomes (B, bottom panel)

During pregnancy, we did not observe any noticeable changes in the levels of maternal glucose, HOMA or insulin, high- and low-density lipoproteins, systolic and diastolic blood pressure however, maternal triglycerides [mean (SD) in first trimester = 89.6 mg/dl (34.4); second trimester = 153 mg/dl (52.1); and third trimester = 189 mg/dl (60.4)], total cholesterol [mean (SD) in first trimester = 178 mg/dl (31.1); second trimester = 246 mg/dl (45.7); and third trimester = 253 mg/dl (40.5)], and pulse [mean (SD) in first trimester = 74.2 bpm (8.54); second trimester = 80.3 bpm (8.08); and third trimester = 83.9 bpm (9.04)] increased over time (Fig. 1B).

In unadjusted and adjusted linear mixed-effect models considering each maternal serum PFAS as time-varying (Fig. 2 and Supplemental Table 2), neither the sum nor individual PFAS was significantly associated with maternal systolic blood pressure (SBP) during pregnancy but, PFAS_sum, PFOS, PFOA and PFNA were statistically significantly positively associated with diastolic blood pressure (DBP) during pregnancy. Individually, PFAS_sum, PFOS, PFOA, PFNA, PFHxS, and PFHxA during pregnancy were significantly negatively associated with pulse. PFAS_sum and all well-detected PFAS, other than PFBS, were individually negatively associated with triglycerides, total cholesterol, high-density and low-density lipoprotein during pregnancy. Of all the individual PFAS, the greatest negative effect was related to maternal serum PFOS levels [triglycerides effect estimates (β)= -14.01, 95% confidence interval (CI)= -24.38, -3.64; total cholesterol β= -13.08, 95%CI= -21.56, -4.60; low-density lipoprotein β= -10.88, 95%CI= -17.77, -3.98 and high-density lipoprotein β= -5.88, 95%CI= -8.35, -3.42]. No associations with maternal glucose, insulin, or HOMA-IR were observed. When we considered the mixture effect of all six maternal serum PFAS levels on each maternal outcomes using quantile g-computation, we found that during pregnancy, higher exposure levels were significantly associated with decreased maternal triglyceride (β= -16.29; 95%CI= -24.95, -7.63) the most followed by total cholesterol (β= -12.77; 95%CI= -19.80, -5.74), low-density lipoproteins (β= -10.83; 95%CI = -16.72, -4.93), high-density lipoproteins (β= -4.10; 95%CI= -6.23, -1.96), and pulse (β= -1.60; 95%CI= -2.85, -0.35) (Fig. 2 and Supplemental Table 3).

Fig. 2

Individual and mixture effects of the natural log-transformed gestational PFAS (ng/ml), as measured in maternal serum, on maternal outcomes over time, estimated using linear mixed-effect models or quantile g-computation and adjusting for maternal body mass index, age, gestational weight gain, income, race, and education. ln: natural log transform; CI: confidence interval

To ascertain whether there are specific critical windows of vulnerability, we explored trimester-specific effects of PFAS on each maternal outcome. Across trimesters, we did not observe any robust monotonic effects of maternal serum PFAS levels individually or as a mixture (Fig. 3 and Supplemental Tables 4 and 5).

Fig. 3

Individual and mixture effects of the natural log-transformed gestational PFAS (ng/ml), as measured in maternal serum, on maternal outcomes in each trimester, estimated using linear regression models or quantile g-computation and adjusting for maternal body mass index, age, gestational weight gain, income, race, and education

However, we did note some significant effects: In trimester 1, higher maternal serum PFBS (β= -1.75; 95%CI= -2.97, -0.53) and PFHxS (β= -0.94; 95%CI= -1.84, -0.03) levels were individually related with lower trimester 1 maternal glucose levels whereas, PFHxA was associated with lower maternal triglycerides in trimester 1 (β= -5.66; 95%CI= -10.69, -0.63). Significant positive associations were also in trimester 2. Specifically, higher maternal serum PFOS in trimester 2 was related to increased insulin (β = 52.10; 95%CI = 1.94, 102.26), triglycerides (β = 16.13; 95%CI = 2.50, 29.76), and HOMA for insulin resistance (β = 0.31; 95%CI = 0.04, 0.56) whereas higher maternal serum PFOA had small positive effects on maternal glucose (β = 1.51; 95%CI = 0.27; 2.76) in trimester 2. In trimester 3 specifically, maternal serum PFOS was negatively associated with high-density lipoprotein (β= -5.36; 95%CI = -9.72, -0.99). As a mixture, higher maternal serum PFAS levels also significantly reduced maternal glucose levels in the first trimester (β=-2.31; 95%CI = -3.99, -0.63) and significantly increased maternal pulse in the second trimester (β = 1.98; 95%CI = 0.04, 3.93) indicating two potential specific windows of vulnerability. Partial effects also reflected some of the individual PFAS effects. (Supplemental Fig. 3).

We also explored whether adjusting for albumin levels or eGFR altered the observed associations and found that most results were not changed by the inclusion of either of these variables in any of the models (Supplemental Tables 2–5) suggesting that the relationships between PFAS levels during pregnancy and maternal cardiometabolic outcomes are not driven by changes in these hemodynamic markers.

Discussion

In this longitudinal analysis, PFAS concentrations across each trimester of pregnancy were associated with maternal cardiometabolic outcomes in a cohort located in central Arkansas. We demonstrated that hemodilution and/or renal function did not alter the associations between PFAS and maternal outcomes, similar to findings described in a study of 340 pregnant African American women in Atlanta, Georgia⁴³.

Importantly, to our knowledge, this is the first biomonitoring assessment of PFAS levels in humans in central Arkansas. In our cohort, all 27 measured PFAS were detectable, but only PFOS, PFOA, PFBS, PFHxS, PFNA, and PFHxA were detected in more than 70% of the samples. Given the PFAS exposures are highly geographically driven [9], and there is known environmental contamination at three Arkansas sites [43] (especially with PFHxS), it is important to study populations with unique sources of exposure and vulnerabilities to outcomes and have not been included in prior investigations [9]. According to the Arkansas Maternal and Perinatal Outcomes Quality Review Committee, compared to the United States national average, Arkansas has higher infant mortality, more preterm births, more infants born with low birth weight, and fewer women who received prenatal care during their first trimester [44]. Arkansas also has a higher prevalence of obesity, cardiovascular disease, strokes, and diabetes than the United States as a whole [45]. Our study provides an important contribution about environmental contaminants that have the potential to contribute to these outcomes but have been previously unreported in this population.

To our knowledge, three studies have reported PFAS levels from multiple trimesters, namely, a cohort of 163 African American women for whom serum was collected at two-time points between 2014 and 2020 in Atlanta [46]; a cohort of 128 racially and ethnically diverse women for whom serum was collected at two timepoints in Northern California between 2014 and 2017 [47]; and a cohort of 118 Asian women in Beijing, China, in 2017 for whom serum was collected at each trimester [23]. Generally, our results are similar to what they reported, although their detection of PFHxS (> 98%), PFDA (> 60%) and PFNA (> 97%) was higher [24] than in the Arkansas cohort. In an Atlanta cohort, PFOA tended to decrease over the pregnancy period, while all other PFAS detected increased across pregnancy [46], whereas the Beijing, Northern California and our Arkansas cohorts demonstrated a decrease or no change in PFAS over time in pregnancy. These differences may be attributed to the study design where PFAS were quantified at two or only one time point per participant during pregnancy in the Atlanta study; and given the distinct differences in geographic region and racial and ethnic representation of the cohort, there are likely differences in exposure sources for these two populations [46].

The PFAS concentrations measured in our cohort are in line with prior studies investigating PFAS exposure during pregnancy. In a Norwegian cohort of serum collected between 2003 and 2004 in 891 women during mid-pregnancy²⁰, they detected PFOA, PFNA, PFHxS, PFHpS, PFOS, PFDA as reported in the Arkansas cohort at concentrations that were within the same magnitude, except for PFOS that was lower in Arkansas (range: 2.85–3.59 ng/ml in Arkansas and 5th -95th percentiles: 6.90-24.34 in Norway). They also reported PFUnDA above the limit of quantification in 94% of their samples, whereas PFUnDA was below the limit of quantification in all our samples. Similar findings were found in a Spanish cohort of 1240 women with serum samples collected between 2003 and 2008 during the first trimester of pregnancy with detectable PFOA (2.31 ng/ml), PFOS (5.77 ng/ml), PFHxS (0.55 ng/ml) and PFNA (0.64 ng/ml) [19]. Likewise, serum samples collected during the first trimester in a cohort of 318 women in Denmark collected during 2010–2012 had similar findings to those in Arkansas, except for a higher percentage of women with detectable PFHxS (99%) in the Denmark cohort [18]. Similar results have been reported by other groups [14, 15, 48, 49]. Most cohorts reported similar species detection and magnitude of concentrations as the Arkansas cohort. Differences identified are likely due to the time frame during which the samples were collected and geographical differences. The PFAS detected in our cohort are broadly consistent with the historical use of PFAS in aqueous film firefighting foams, which were originally PFOA- and PFOS-based but then superseded by PFHxS and newer compounds, including fluorotelomers [50].

We found that maternal serum PFAS concentrations during pregnancy were associated with decreased triglyceride, total cholesterol, high- and low-density lipoproteins, which is in contrast with prior studies demonstrating an association between higher serum PFAS concentrations and higher triglycerides, total cholesterol, high-density lipoprotein [51, 52] and low-density lipoprotein [23]. The cohort studied by Cinzori et al. was similar in size and demographic characteristics, however maternal PFAS levels were only assessed during the second trimester [51]. The review by Ho et al. included studies that observed both negative and positive associations between PFAS and maternal lipids and concluded that the relationships were generally positive [52]. The study by Hu et al. was completed in China which highlights the importance of both location and demographic characteristics in investigating relationships between PFAS and health outcomes [23]. We also found that serum PFAS concentrations during pregnancy were associated with higher DBP and lower pulse during pregnancy. These findings align with some prior research that has observed increasing gestational concentrations of PFOA and PFOS are associated with higher DBP [22, 53]; however, one larger study also identified positive associations with SBP [53], which was not observed in our study potentially due to sample size. Another study based in China observed an inverse relationship showing that higher PFAS concentrations during pregnancy were associated with lower DBP and SBP [54]. Study design, time frame of exposure to the PFAS and measures of blood pressure (health care provider clinic or research facility) likely lead to these differences in findings and suggest that more carefully designed longitudinal, prospective cohorts are needed to evaluate the effects of PFAS on maternal cardiometabolic outcomes.

Studies focused on gestational diabetes mellitus (GDM) have demonstrated that in women with high risk for GDM, PFHxS and PFNA concentrations during pregnancy were significantly associated with glucose homeostasis and insulin sensitivity [18, 47, 55, 56]. We observed lower glucose levels associated with an increase in the PFAS mixture and with increasing PFHxS and PFBS, but only in trimester 1, and no significant associations were observed at other time points in our study. While our results indicate a potential critical window of vulnerability, the lack of trends during pregnancy and inconsistent results with the other study could be partly because our study primarily consisted of healthy, non-smoking, second parity women.

Our study has several notable attributes. Serial sampling across trimesters allowed detailed evaluation of changes during pregnancy and critical period of exposure for the mother. It is also strengthened by the rich maternal measured cardiometabolic outcome data and the unique cohort that offers maternal exposures and outcomes longitudinally rather than cross-sectionally. Our study had a moderate sample size compared to other published studies which likely limited our ability to detect small effects. Yet, we used precise measurements of outcomes at a designated research facility which likely decreased variance and measurement error in traditional monitoring that relies on health care records. Another limitation is the timeframe of the study with samples collected from 2010 to 2014, which does not reflect the more recently emergent PFAS. However, given the environmental ubiquity of PFAS contamination and their resistance to degradation, it is important to continue to monitor and study legacy PFAS which will likely continue to expose human populations for years to come. Finally, the mechanisms of toxicity across PFAS may be similar, although future research is needed to explore individual PFAS mechanisms of toxicity.

Conclusion

To our knowledge, this study is the first biomonitoring and evaluation of PFAS exposures in pregnant women living in central Arkansas. Out of 27 PFAS measured, all were detected and only PFOS, PFOA, PFBS, PFHxS, PFNA, and PFHxA were present in more than 70% of the samples. In this cohort, higher maternal serum PFAS levels during pregnancy were significantly associated with decreased maternal triglycerides, total cholesterol, low- and high-density lipoproteins, and pulse during pregnancy but with increased maternal diastolic blood pressure suggesting effects on maternal cardiovascular health. Future studies are needed to confirm the effects of PFAS and other environmental contaminants on maternal health using longitudinal cohort study designs. Further, lifestyle intervention may affect PFAS associations with health outcomes²⁵ and should also be explored.