Prenatal exposure to per- and polyfluoroalkyl substances (PFAS) and their influence on inflammatory biomarkers in pregnancy: Findings from the LIFECODES cohort

Abstract

Background: Per- and polyfluoroalkyl substances (PFAS) are fluorinated chemicals linked to adverse pregnancy and birth outcomes. However, the underlying mechanisms, specifically their effects on maternal inflammatory processes, are not well characterized.

Objective: We examined associations between prenatal PFAS exposure and repeated measures of inflammatory biomarkers, including C-reactive protein (CRP) and four cytokines [Interleukin-10 (IL-10), IL-1β, IL-6, and tumor necrosis factor-α (TNF-α)].

Methods: We analyzed data from 469 pregnant women in a nested case-control study of preterm birth at Brigham and Women's Hospital in Boston, Massachusetts (2006-2008). We measured nine PFAS in early pregnancy plasma samples (median gestation: 10 weeks), with inflammatory biomarkers measured at median gestations of 10, 18, 26, and 35 weeks. We used linear mixed models for repeated measures and multivariable regression for visit-specific analysis to examine associations between each PFAS and inflammation biomarker, adjusting for maternal demographics, pre-pregnancy BMI, and parity. We examined the effects of PFAS mixture using sum of all PFAS (∑PFAS) and quantile-based g-computation approaches.

Results: We observed consistent inverse associations between most PFAS and cytokines, specifically IL-10, IL-6, and TNF-α, in both single pollutant and mixture analyses. For example, an interquartile range increase in perfluorooctanesulfonic acid was associated with -10.87 (95% CI: -19.75, -0.99), -13.91 (95% CI: -24.11, -2.34), and -8.63 (95% CI: -14.51, -2.35) percent change in IL-10, IL-6, and TNF-α levels, respectively. Fetal sex, maternal race, and visit-specific analyses showed associations between most PFAS and cytokines were generally stronger in mid-pregnancy and among women who delivered males or identified as African American.

Conclusions: The observed suppression of both regulatory (IL-10) and pro-inflammatory (TNF-α) cytokines suggests that PFAS may alter maternal inflammatory processes or immune functions during pregnancy. Further research is needed to understand the effects of both legacy and newer PFAS on inflammatory pathways and their broader clinical implications.

Keywords: C-reactive protein; Cytokines; Inflammation; LIFECODES; PFAS; Pregnancy.

Fig. 1.

Associations between prenatal PFAS exposure and inflammatory biomarkers in LIFECODES. Analysis was based on the complete case data (N = 1485 observations from 448 participants). Linear mixed-effects models with random intercept were used to account for the correlation between repeated measures outcomes. Both PFAS and inflammatory biomarker concentrations were log-transformed prior to modeling. All models were adjusted for maternal age, prepregnancy BMI, race, education level, insurance status, and parity. Effect estimates were obtained as % change in biomarker concentrations per IQR increase in PFAS concentrations. All models were weighted using inverse probability weights (1.1 for preterm cases and 2.95 for controls) to account for the unequal selection of preterm cases in the case-control study. The vertical lines represent the null effect. * represents associations with 95% CIs that do not include null. Note: CRP, C-reactive protein; IL, interleukin; TNF, tumor necrosis factor; IQR, interquartile range.

Fig. 2.

Associations between prenatal PFAS exposure and inflammatory biomarkers in LIFECODES, stratified by fetal sex. Analysis was based on the complete case data (N = 829 observations from 245 participants for males and 656 observations from 203 participants for females). Linear mixed-effects models with random intercept were used to account for the correlation between repeated measures outcomes. Both PFAS and inflammatory biomarker concentrations were log-transformed prior to modeling. All models were adjusted for maternal age, prepregnancy BMI, race, education level, insurance status, and parity. Effect estimates were obtained as % change in biomarker concentrations per IQR increase in PFAS concentrations. All models were weighted using inverse probability weights (1.1 for preterm cases and 2.95 for controls) to account for the unequal selection of preterm cases in the case-control study. The vertical lines represent the null effect. * represents associations with 95% CIs that do not include null. Note: CRP, C-reactive protein; IL, interleukin; TNF, tumor necrosis factor; IQR, interquartile range.

Fig. 3.

Cross sectional associations between prenatal PFAS exposure and visit-specific inflammatory biomarker concentrations among all participants in LIFECODES. Analysis was based on the complete case data (N = 390, 382, 360, and 353 participants for visits 1–4, respectively). Multivariable linear regression models were used to assess relationships between early pregnancy PFAS levels and visit-specific plasma biomarker concentrations. All models were adjusted for maternal age, race, prepregnancy BMI, education level, insurance status, and parity. Effect estimates were obtained as % change in biomarker concentrations per IQR increase in PFAS concentrations. All models were weighted using inverse probability weights (1.1 for preterm cases and 2.95 for controls) to account for the unequal selection of preterm cases in the case-control study. The vertical lines represent the null effect. * represents an association with 95% CIs that does not include null. Note: CRP, C-reactive protein; IL, interleukin; TNF, tumor necrosis factor; IQR, interquartile range.

Fig. 4.

Associations between PFAS mixture and inflammatory biomarkers in LIFECODES using the summed PFAS approach. Visit 1–4 indicates the repeated measures analysis of biomarkers, while the individual visits indicate the visit-specific analysis. Summed PFAS was obtained by adding concentrations of seven PFAS with a detection rate >70% (PFOA, PFOS, PFNA, PFHxS, PFDA, PFUA, and MPAH). Linear mixed-effects models with random intercept were used for repeated measures analysis, while multivariable linear regression models were used for visit-specific analysis. All models were adjusted for maternal age, race, prepregnancy BMI, education level, insurance status, and parity. Effect estimates were obtained as % change in biomarker concentrations per IQR increase in sum of all PFAS concentrations. All models were weighted using inverse probability weights (1.1 for preterm cases and 2.95 for controls) to account for the unequal selection of preterm cases in the case-control study. The vertical lines represent the null effect. * represents associations with 95% CIs that do not include null. Note: CRP, C-reactive protein; IL, interleukin; TNF, tumor necrosis factor; IQR, interquartile range.

Fig. 5.

Associations between PFAS mixture and inflammatory biomarkers in LIFECODES using quantile-based g-computation (qgcomp) approach. Visit 1–4 indicates the repeated measures analysis of biomarkers, while the individual visits indicate the visit-specific analysis. For Visit 1–4, the bootstrapped implementation of qgcomp (2500 Monte Carlo simulations and 2500 bootstrapped samples) with linearity assumption was used to obtain robust standard errors. For visit-specific analysis, the non-bootstrapped implementation of qgcomp with linearity assumption was implemented. All models were adjusted for maternal age, race, prepregnancy BMI, education level, insurance status, and parity. Effect estimates were obtained as % change in biomarker concentrations per quartile increase in all PFAS levels. All models were weighted using inverse probability weights (1.1 for preterm cases and 2.95 for controls) to account for the unequal selection of preterm cases in the case-control study. The vertical lines represent the null effect. * represents associations with 95% CIs that do not include null. Note: CRP, C-reactive protein; IL, interleukin; TNF, tumor necrosis factor.