Effects of per- and polyfluoroalkyl substances on maternal thyroid function: a stratified analysis by maternal iodine status and thyroid autoantibody status

Abstract

Background

Per- and polyfluoroalkyl substances (PFASs) are persistent environmental pollutants with bioaccumulative properties. Emerging evidence indicates that PFAS disrupts thyroid hormone homeostasis, potentially leading to thyroid dysfunction and impairing physiological functions. However, the associations between PFAS exposure, iodine status, and thyroid function during pregnancy remain insufficiently understood.

Methods

A total of 911 pregnant women from a birth cohort in Zhejiang Province were included. Serum concentrations of 13 PFASs and thyroid function parameters were measured during the first antenatal visit. Linear regression, quantile regression, and restricted cubic spline models were used to explore associations between individual PFAS and thyroid hormones. Quantile g-computation was conducted to evaluate the overall effects of PFAS mixtures, while Bayesian kernel machine regression was applied to capture potential nonlinear and interactive effects. Stratified analyses were further conducted to examine potential effect modification by iodine status and thyroid autoantibody status.

Results

PFOS, PFDA, PFNA, PFUnDA, and PFHpS were negatively associated with FT₃ and positively associated with the FT₄/FT₃ ratio (all P < 0.05). The novel PFMOAA showed positive associations with TSH (β = 0.162; 95% CI: 0.069, 0.255) and the TSH/FT₄ ratio (β = 0.006; 95% CI: 0.0002, 0.011). When all 13 PFASs were modeled as a mixture, higher PFAS concentrations were associated with lower FT₃ levels. Similar inverse associations with FT₃ were observed for the mixture of PFASs selected by Least absolute shrinkage and selection operator. Meanwhile, the mixture composed of PFASs identified as significant in single-pollutant models was additionally associated with higher TSH and an increased FT₄/FT₃ ratio. Iodine status and thyroid autoantibody status appeared to modify these associations.

Conclusions

Prenatal exposure to both legacy PFASs (PFOS, PFDA, PFNA, PFDoDA, PFUnDA, and PFHpS) and novel PFAS (6:2 Cl-PFESA) was associated with lower FT₃ levels, while PFMOAA was positively associated with TSH levels. The effects of PFAS mixtures on thyroid function varied according to their specific composition. Thyroid autoantibody status and iodine status may act as effect modifiers of the associations between PFAS exposure and thyroid hormone levels.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12884-026-09599-9.

Introduction

Per- or polyfluoroalkyl substances (PFASs) are synthetic compounds characterized by multiple carbon-fluorine (C-F) covalent bonds, primarily defined by the presence of fully fluorinated methyl (− CF₃) or methylene (− CF₂−) groups, though there are some exceptions [1–3]. These compounds are highly stable due to the strong C-F bonds and exhibit both hydrophobic and oleophobic properties, resulting in widespread use across various commercial applications since the 1950s [4–6]. Consequently, PFASs are now almost ubiquitous in the environment, and living organisms can be exposed to them through multiple routes over the course of their lifetime [7–10]. This global contamination exhibits distinct regional disparities in rapidly industrializing nations [11, 12]. Research on PFAS pollution in China shows that surface water pollution is most severe in East China, followed by Northeast and North China. Sediment pollution is most serious in the east and north, followed by the northeast [13, 14]. Beyond their environmental persistence, accumulating evidence suggests that PFAS can interfere with biological systems [15–18], particularly by disrupting metabolic and endocrine functions [19–21], thereby raising growing public health concerns.

Thyroid-stimulating hormone (TSH) stimulates the thyroid gland to secrete triiodothyronine (T₃) and thyroxine (T₄) [22], which regulate metabolic processes across multiple tissues and physiological systems [23–25]. Thyroglobulin (Tg) serves as the precursor of thyroid hormones (THs) [26], and its serum concentration reflects the synthetic capacity of the thyroid gland. Furthermore, the biologically active free forms —free triiodothyronine (FT₃) and free thyroxine (FT₄) —are the core indicators for clinical assessment of thyroid function [27]. Epidemiological studies indicate a U-shaped relationship between THs levels and adverse pregnancy outcomes. Compared with euthyroid pregnant women, those with hyperthyroidism or hypothyroidism face significantly higher risks of pre-eclampsia and gestational hypertension [28].

The endocrine-disrupting potential of PFAS, especially regarding thyroid function, has attracted increasing attention. A birth cohort study from Ezhou, China showed that multiple PFASs were associated with thyroid hormone levels in pregnant women. Moreover, prenatal exposure to certain PFAS is negatively associated with birth length and head circumference, and FT₃ and FT₄ might partially mediate this association [29]. Toxicological studies have confirmed that PFAS can interfere with thyroid function in animals through the following mechanisms: (1) directly disrupting the physiological function of the thyroid gland; (2) enhancing the metabolic clearance of THs or tissue utilization; (3) reducing circulating TH levels by competitively binding with transthyretin [30–32]. However, information regarding the molecular interactions and toxicological mechanisms of emerging PFAS alternatives remains limited. Furthermore, the potential for synergistic or antagonistic effects within complex PFAS mixtures comprising both legacy and emerging compounds remains understudied. Therefore, further epidemiological studies are required to systematically evaluate the associations between legacy and emerging PFAS exposure and thyroid function during pregnancy.

Iodine is an essential nutrient with a recommended daily intake of 250 µg during pregnancy and is critical for THs synthesis [33, 34]. Studies have shown that fluoride exposure impairs iodine absorption and transport by suppressing the expression of the sodium/iodide symporter (NIS) gene or inhibiting the activity of sodium-potassium adenosine triphosphatase [35]. This suggests that iodine could act as an effect modifier or mediator in the association between exposure to PFAS and thyroid function. However, the role of iodine in the relationship between PFAS exposure and thyroid function remains underexplored. This gap is most evident in studies of pregnant women who are especially vulnerable to both iodine deficiency and thyroid dysfunction [36, 37]. Identifying whether iodine status exacerbates or mitigates the thyrotoxic effects of PFAS is therefore essential for developing targeted public health interventions for pregnant populations.

In this study, we analyzed data from a cross-sectional investigation of 911 pregnant women in China to assess the associations between exposure to 13 PFASs—including both legacy compounds and emerging alternatives—and thyroid hormone levels in pregnant women. Our findings provide novel evidence on the effects of both individual and mixture exposures, while accounting for iodine status and thyroid autoimmunity as potential effect modifiers, thereby offering new insights into the effects of PFAS on maternal thyroid function during pregnancy.

Methods

Study population

This study included pregnant women who underwent routine prenatal examinations at four maternal and child health hospitals in Dongyang, Pujiang, Jiaojiang, and Linhai in Zhejiang Province, China, between April 2022 and December 2024. The inclusion criteria were as follows: (1) maternal age between 18 and 45 years; (2) the first antenatal visit was enrolled before 12 gestational weeks; (3) no self-reported family history of thyroid disease; (4) resided in the local community for at least one year and planning to reside in the next 3 years. The exclusion criteria were: (1) diagnosis of chronic diseases such as heart disease, hypertension, cancer and diabetes; (2) receiving iodine-containing contrast media over the past year. After excluding participants with missing questionnaire data (n = 18) or thyroid hormone measurements (n = 7), a total of 911 eligible pregnant women were included in the final analysis.

Each participant completed a questionnaire that collected information on demographic characteristics (age, ethnicity, education, and occupation), lifestyle factors (passive smoking and alcohol consumption). Height and weight were measured by trained staff at the time of enrollment, and body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters (kg/m²). Fasting blood samples were obtained by venipuncture, and plasma was separated by centrifugation and stored at − 80 °C until analysis. The study protocol was approved by the Ethics Committee of the Zhejiang Provincial Center for Disease Control and Prevention (No. 2020-047-01).

PFAS measurement

A total of 34 PFASs were measured in maternal serum, of which only 13 compounds with detection rates ≥ 40% of the limit of detection (LOD) were included in the statistical analyses. These comprised nine legacy PFAS: perfluorooctanoic acid (PFOA), perfluorooctanesulfonic acid (PFOS), perfluorodecanoic acid (PFDA), perfluorononanoic acid (PFNA), perfluorododecanoic acid (PFDoDA), perfluoroheptanoic acid (PFHpA), perfluoroundecanoic acid (PFUnDA), perfluoroheptanesulfonic acid (PFHpS), perfluorohexanesulfonic acid (PFHxS); and four emerging alternatives: perfluoro-2-methoxyacetic acid (PFMOAA), hexafluoropropylene oxide trimer acid (HFPO-TrA (C9)), 6:2 chlorinated polyfluoroalkyl ether sulfonic acid (6:2 Cl-PFESA), 8:2 chlorinated polyfluoroalkyl ether sulfonic acid (8:2 Cl-PFESA). The detailed detection rates are presented in Table S1.

The concentrations of PFASs were quantitatively measured using AB SCIEX ultrahigh-performance liquid chromatography coupled with triple-quadrupole mass spectrometry (QTRAP 6500 plus; SCIEX, MA, USA) operated in electrospray ionization multiple reaction monitoring mode [38, 39]. The methodologies used to quantify PFAS have been described in previous studies [40]. The limits of detection (LODs, ng/mL) in serum were: 0.01 for PFMOAA, 6:2 Cl-PFESA, and 8:2 Cl-PFESA; 0.02 for PFOA, PFOS, PFDA, PFNA, PFDoDA, PFUnDA, PFHpA and HFPO-TrA (C9); 0.04 for PFHxS; and 0.05 for PFHpS [41]. Concentrations below the LOD were substituted with LOD/√2.

Thyroid hormone measurements

Serum levels of FT₄, FT₃, TSH, Tg, thyroglobulin antibody (TgAb), and thyroid peroxidase antibody (TPOAb) were determined using electrochemiluminescence immunoassay (Cobas^® e411; Roche, Mannheim, Germany). TgAb and TPOAb positivity were defined as TgAb > 115 IU/mL and TPOAb > 34 IU/mL, respectively [42]. In addition, FT₄/FT₃ and FT₄/TSH ratios were calculated as markers of peripheral thyroid hormone metabolism.

Iodine concentration measurements in serum

Serum iodine concentration (SIC) was determined by inductively coupled plasma mass spectrometry (ICP-MS; Agilent 7900, Agilent Technologies, California, USA). Detailed information on the assay methodology has been described elsewhere [43]. Briefly, 100 µL of serum was diluted with 1.9 mL of a solution containing 0.1% Triton X-100 and 1% nitric acid. Quality control samples, including blank, duplicate, and standard reference material (Seronorm Trace Elements Serum L-2, Norway), were included in every batch of 20 samples. Recovery of iodine in standard reference material ranged from 90% to 110%.

Statistical analyses

Descriptive statistics were used to summarize the characteristics of participants, PFAS concentrations, and TH levels. Continuous variables were presented as means ± standard deviations (SD). Concentrations of PFAS, Tg, TSH, FT₄, and FT₃ were natural log-transformed to improve normality and model fit. Spearman rank correlation analysis was used to evaluate the correlation between PFASs.

To evaluate the associations between individual PFASs and thyroid function measures (TSH, FT₄, FT₃, Tg, FT₄/FT₃, TSH/FT₄), multivariable linear regression models were first applied. PFAS concentrations were categorized by quartiles, with the lowest quartile serving as the reference group. For 8:2 Cl-PFESA and HFPO-TrA (C9) with a high proportion of values below LOD (Table S1), exposure levels were classified into three categories: < LOD, LOD to the median, and median to the maximum value. Restricted cubic spline (RCS) models with knots at the 25th, 50th, and 75th percentiles were used to characterize potential nonlinear dose–response relationships. To estimate the joint effects on thyroid hormone levels, quantile g-computation (qgcomp) was further applied under three model specifications: (1) an overall model including all 13 PFASs; (2) a model including only compounds that were significantly associated with outcomes (P < 0.05) in single-pollutant analyses; and (3) a model including compounds selected through LASSO. Finally, Bayesian Kernel Machine Regression (BKMR) was conducted to capture potential nonlinear and interactive effects among PFASs. All models were adjusted for potential confounders, including maternal age, BMI, education, occupation, passive smoking, alcohol consumption and serum iodine.

Two sensitivity analyses were conducted. First, to evaluate potential effect modification by iodine status, SIC was dichotomized at the median of 82.79 µg/L, and stratified regression analyses were performed within each group. Second, to assess whether maternal thyroid autoimmunity influenced the associations, participants were stratified according to their thyroid autoantibody status. Autoantibody positivity was defined as TPOAb > 34 IU/mL and/or TgAb > 115 IU/mL. Regression models were then fitted separately for the positive and negative subgroups.

All statistical analyses were performed using R (version 4.3.1), and two-sided p-values < 0.05 were considered statistically significant.

Results

Population characteristics

Table 1 summarizes the characteristics of the 911 participants. The cohort had a mean age of 28.9 ± 4.4 years and a mean gestational age of 9.2 ± 1.58 weeks. Based on BMI classification, 24.0% of the participants were overweight or obese. The majority were of Han ethnicity (98.0%) and had attained a college degree or higher (62.0%). Regarding occupation, 55.0% were engaged in non-manual labor, and regarding lifestyle factors, 22.0% were exposed to passive smoking while 5.0% reported alcohol consumption. The mean SIC was 84.99 ± 28.06 µg/L, and elevated thyroid antibodies were observed in 8.5% (TPOAb) and 7.5% (TgAb), respectively.

Table 1.

Demographic characteristics of study participants (N = 911)

Characteristic	mean ± SD or n (%)
Maternal age, years	28.92 ± 4.42
Gestational age, weeks	9.20 ± 1.58
SIC, µg/L	84.99 ± 28.06
Maternal BMI, kg/m2
< 18.5	109 (12%)
18.5–23.9	581 (64%)
≥ 24.0	221 (24%)
Ethnicity
Han	895 (98%)
Minority	16 (2%)
Education
Junior high school and lower	132 (14.5%)
Senior high or vocational school	214 (23.5%)
College and higher	565 (62.0%)
Occupation
Non-manual labor	503 (55%)
Manual labor	232 (26%)
Others	176 (19%)
Passive smoking
No	713 (78%)
Yes	198 (22%)
Alcohol consumption
No	868 (95%)
Yes	43 (5%)
TgAb, IU/mL
≤ 115	843 (92.5%)
> 115	68 (7.5%)
TPOAb, IU/mL
≤ 34	834 (91.5%)
> 34	77 (8.5%)

Abbreviations: BMI body mass index, SD Standard deviation, SIC Serum iodine, TgAb thyroglobulin antibody, TPOAb thyroid peroxidase antibody

Table S2 shows the concentrations of PFASs and thyroid function measures in maternal serum. Among the PFASs, PFOS exhibited the highest mean concentration (8.33 ng/mL), followed by PFOA (7.63 ng/mL), 6:2 Cl-PFESA (4.42 ng/mL), PFNA (1.54 ng/mL), PFDA (1.33 ng/mL), and PFUnDA (1.17 ng/mL). In contrast, the remaining PFASs (such as PFHxS, PFMOAA, 8:2 Cl-PFESA, PFDoDA, PFHpS, HFPO-TrA (C9), and PFHpA) had the lowest concentrations, with mean values below 1 ng/mL. As shown in Fig. S1, high positive correlations were observed among the long-chain PFCAs, including PFNA, PFDA, PFUnDA, and PFDoDA, with Spearman’s correlation coefficients (r_s) ranging from 0.82 (between PFDoDA and PFNA) to 0.95 (between PFUnDA and PFDA). Conversely, HFPO-TrA (C9), PFMOAA, and PFHpA showed low correlations with the other 12 PFASs, with r_s ranging from 0.10 to 0.43.

Relationships between individual PFAS exposure and THs

FT₃ was negatively associated with the exposure to PFOS, PFDA, PFDoDA, PFUnDA, PFHpS, and 6:2 Cl-PFESA when evaluated as log-transformed continuous variables or quartiles; however, in the quartile models, these inverse associations were observed in the highest exposure group (Tables S3 and S4). In Table S3, FT₄ was positively associated with PFHxS (β = 0.009; 95% CI: 0.002, 0.016) and negatively associated with 8:2 Cl-PFESA (β = -0.016; 95% CI: -0.027, -0.004). PFDA, PFHpS, PFHxS, PFNA, PFOS, and PFUnDA (all P < 0.05) were positively associated with the FT₄/FT₃ ratio. Additionally, PFMOAA was consistently positively associated with TSH and the TSH/FT₄ ratio (all P < 0.05).

The associations between individual PFAS concentrations and thyroid hormone levels were examined using RCS regression (Figs. S2–S7). The RCS analyses identified significant nonlinear relationships for PFMOAA, including its associations with TSH (P for nonlinearity = 0.005) and the FT₄/FT₃ ratio (P for nonlinearity = 0.001). In addition, U-shaped relationships were observed between FT₃ and PFMOAA (P for nonlinearity = 0.002), as well as between FT₃ and 8:2 Cl-PFESA (P for nonlinearity = 0.018) (Fig. 1).

Fig. 1.

Significant associations between individual PFASs and thyroid indicators identified by restricted cubic spline (RCS) models. Note: P_overall means global association, P_nonlinearity means nonlinear trend test. Covariates adjusted for included age, BMI, ethnicity, education, occupation, passive smoking, alcohol consumption, and serum iodine. Both exposure and outcome measures were log-transformed

Joint effects of PFAS mixtures

The combined associations of PFAS mixtures with thyroid hormone levels were derived through qgcomp assessment, and the resulting estimates are provided in Table 2. The mixtures including either all 13 target PFASs or those selected through LASSO showed a significant inverse association only with FT₃. In these models, the individual PFAS contributing most strongly to the mixture effect differed between mixture specifications (Figs. S8 and S9). When restricting the mixture to PFASs that were individually associated with thyroid hormones in single-pollutant analyses, the mixture was additionally associated with higher TSH and a higher FT₄/FT₃ ratio, alongside lower FT₃. In this model, PFMOAA contributed most to the increase in TSH, PFHxS to the elevation of the FT₄/FT₃ ratio, and PFDoDA to the suppression of FT₃ (Fig. S10).

Table 2.

Generalized componentwise estimation (gcomp) estimates of PFAS mixture effects on thyroid function measures

	Overall Effect β (95% CI)	Selected by Univariate Significance β (95% CI)	Selected by LASSO β (95% CI )
TSH	0.071(-0.012,0.153)	0.036(0.008,0.064)*	0.070(-0.001,0.141)
FT3	-0.011(-0.022,-0.000) *	-0.009(-0.015,-0.004)***	-0.010(-0.019,-0.002)*
FT4	-0.010(-0.020,0.001)	-0.006(-0.014,0.002)	-0.010(-0.020,0.0003)
Tg	-0.033(-0.117,0.050)	N/A	N/A
TSH/FT4	0.002(-0.003,0.007)	0.002(-0.0001,0.003)	0.002(-0.001,0.006)
FT4/FT3	0.002(-0.031,0.035)	0.037(0.020,0.054)***	-0.005(-0.036,0.027)

β estimates derived from generalized componentwise estimation (gcomp). PFAS included in each model were selected based on univariate significance (p < 0.05) and via LASSO regularization

Covariates adjusted for included age, BMI, ethnicity, education, occupation, passive smoking, alcohol consumption, and serum iodine

Both exposure and outcome measures were log-transformed

Bold values indicate statistically significant associations (*: p < 0.05, **: p < 0.01, ***: p < 0.001); N/A denotes that no PFAS variables were retained for Tg in the univariate or LASSO models owing to non-significant associations

As shown in Fig. 2, the overall associations between the PFAS mixture and thyroid hormone levels across quantiles were estimated using BKMR. FT₃ was observed to decrease significantly with increasing levels of the PFAS mixture (Figs. 2, S11 and S12). The FT₄/FT₃ ratio, in contrast, was observed to increase with rising levels of the PFAS mixture (Figs. 2, S11, and S12).

Fig. 2.

Overall effects of the mixture of thirteen PFASs on thyroid function measures in BKMR. Note: Effect estimates across quantiles of the PFAS mixture (x-axis) for the specified outcome (y-axis), with 95% confidence intervals. Models adjusted for serum iodine. Abbreviations: TSH, thyroid-stimulating hormone; FT3, free triiodothyronine; FT4, free thyroxine; Tg, thyroglobulin. Both exposure and outcome measures were log-transformed

Stratified analyses

To evaluate whether maternal iodine modified the associations between PFAS exposure and thyroid hormones, interaction terms between PFAS and SIC were incorporated into the multivariable models (Table S6). Significant negative interaction effects between SIC and 6:2 Cl-PFESA were observed for FT₃ (β = -0.028; 95% CI: -0.053, -0.004) and FT₄ (β = -0.024; 95% CI: -0.048, -0.000). Significant negative interactions between SIC and HFPO-TrA (C9) for FT₃ (β = -0.036; 95% CI: -0.072, -0.001) and between SIC and PFHpA for the FT₄/FT₃ ratio (β = -0.115; 95% CI: -0.214, -0.016) were observed. Stratified analyses by median SIC revealed that, among women with higher SIC (above the median), HFPO-TrA (C9) was inversely associated with FT₃ (β = -0.034; 95% CI: -0.058, -0.009), while 6:2 Cl-PFESA was inversely associated with FT₃ (β = -0.038; 95% CI: -0.056, -0.020) and FT₄ (β = -0.030; 95% CI: -0.048, -0.012). In addition, PFHpA was positively associated with the FT₄/FT₃ ratio among women with lower SIC (below the median).

Table S7 presents the interaction analyses between PFAS exposure and thyroid autoantibody status. Significant negative interactions between thyroid autoantibody status and PFOS, PFDA, PFNA, PFDoDA, PFUnDA, 6:2 Cl-PFESA, and 8:2 Cl-PFESA were observed for TSH, whereas significant positive interactions were observed for FT₃ (all P < 0.05). In addition, significant positive interactions between thyroid autoantibody status and PFNA, PFDA, PFOS, PFDoDA, PFUnDA, HFPO-TrA (C9), and 6:2 Cl-PFESA were observed for the FT₄ (all P < 0.05). Stratified analyses revealed that among antibody-positive women, PFOS, PFDA, PFNA, PFDoDA, PFUnDA, HFPO-TrA (C9), 6:2 Cl-PFESA, and 8:2 Cl-PFESA were inversely associated with FT₃ (all P < 0.05). Additionally, PFOS, PFDA, PFNA, PFDoDA, PFUnDA, 6:2 Cl-PFESA, and 8:2 Cl-PFESA exhibited positive associations with TSH (all P < 0.05). Among antibody-negative women, PFNA was positively associated with FT₄ (β = 0.022; 95% CI: 0.004, 0.041), whereas PFHpA was negatively associated with FT₃ (β = -0.022; 95% CI: -0.040, -0.003).

Discussion

This study explored the associations between PFAS exposure and thyroid function among pregnant women. Higher PFAS concentrations were associated with lower FT₃ levels. Nonlinear associations were observed between PFMOAA and TSH, FT₃, and the FT₄/FT₃ ratio, as well as between 6:2 Cl-PFESA and FT₃. The effects of PFAS mixtures on thyroid function may vary according to their compositional profiles. Additionally, iodine nutritional status and thyroid autoantibody status may act as effect modifiers in these associations.

In our study, higher PFAS concentrations were associated with lower FT₃ levels. Consistent with our findings, a study among pregnant women in northern Norway reported that participants in the highest quartile of PFUnDA exposure had FT₃ concentrations 3% lower than those in the lowest quartile [44]. Similarly, a birth cohort study conducted in Ezhou, China, found that most PFASs were negatively associated with FT₃ levels during pregnancy [29]. In contrast, some studies have reported positive associations of PFOA, PFNA, and PFHxS with FT₃ [45, 46]. These inconsistencies may be related to regional differences in PFAS concentrations. Compared with European and U.S. cohorts [44, 47–50], prenatal PFOS concentrations were lower in our study population. Within China, PFAS concentrations in our study population were lower than those reported in Shanghai (PFOA: 19.41 ng/mL; PFOS: 10.77 ng/mL; PFDA: 2.15 ng/mL) [51] and some regions of Zhejiang Province (PFOA: 14.8 ng/mL; PFHxS: 0.6 ng/mL; PFHpA: 0.10 ng/mL) [52], but higher than those reported in Beijing (PFOA: 1.64 ng/mL; PFOS: 4.41 ng/mL; PFDA: 0.37 ng/mL; PFNA: 0.46 ng/mL) [53]. In addition, prenatal 6:2 Cl-PFESA concentration in our study was lower than that reported in Wuxi (6:2 Cl-PFESA: 6.54 ng/mL), where significant negative associations with FT₄ and TSH were observed, whereas no significant associations were detected in our study [31]. Generally, reductions in circulating FT₃ or FT₄ concentrations stimulate compensatory regulation of the hypothalamic–pituitary–thyroid axis, leading to increased synthesis and secretion of TSH by the pituitary gland to maintain thyroid hormone homeostasis. In the present study, PFMOAA was positively associated with TSH levels. This finding differs from a previous Beijing study of 157 pregnant women, which reported inverse correlations between maternal TSH levels and several PFAS compounds (r = -0.261 to -0.170, all P < 0.05) [53]. Such discrepancies may be due to a smaller sample size compared to ours (157 vs. 911), as well as variations in covariate adjustment and other demographic characteristics.

A non-monotonic dose–response relationship is regarded as a hallmark feature of endocrine-disrupting chemicals [54, 55]. In the present study, nonlinear associations were observed between PFMOAA and TSH, FT₃, and the FT₄/FT₃ ratio, as well as between 6:2 Cl-PFESA and FT₃, suggesting potential non-monotonic effects of specific PFAS compounds on thyroid hormone homeostasis. Similar nonlinear relationships between PFAS exposure and thyroid-related biomarkers have also been reported previously. For example, analysis from the Project Viva cohort identified nonlinear relationships between PFHxS exposure and maternal TSH levels [50]. Furthermore, a study based on the SELMA birth cohort in Sweden found an inverted U-shaped association between maternal PFOS levels and total T₄ concentrations during early pregnancy [56]. Collectively, these findings highlight the potential complexity of PFAS disruption of the endocrine system and underscore the need for further mechanistic investigations.

Iodine is an essential determinant of thyroid hormone synthesis and is critical for T₄ and T₃ production [57, 58]. Previous studies have suggested a U-shaped association between iodine intake and thyroid disorders, indicating that both deficiency and excess may lead to thyroid dysfunction [59]. Pregnant women are particularly vulnerable to iodine deficiency due to increased renal iodine excretion and fetal developmental demands [60, 61]. Serum iodine serves as an effective biomarker for assessing iodine nutrition, exhibiting significantly lower inter-individual variability compared to urinary iodine [62]. Given that the effects of PFAS on the thyroid may be influenced by iodine levels, serum iodine status was incorporated into our models as an effect modifier. Among lower SIC (below the median) group, the positive associations of PFHpA with the FT₄/FT₃ ratio were observed. Previous in vitro studies have suggested that PFAS may inhibit NIS-mediated iodide uptake, a critical step in thyroid hormone biosynthesis [63]. The thyroid-disrupting effects of endocrine-disrupting chemicals may be exacerbated under iodine-deficient conditions [64]. Therefore, inadequate iodine availability may amplify PFAS-induced disruption of iodide transport and further impair thyroid hormone synthesis, potentially increasing susceptibility to thyroid dysfunction during pregnancy [65]. Interestingly, among participants with higher SIC (above the median), HFPO-TrA was inversely associated with FT₃, while 6:2 Cl-PFESA showed an inverse association with FT₃ and FT₄. These findings suggest that adequate iodine status may not attenuate PFAS-related effects. Collectively, our findings demonstrate that iodine status modifies the association between PFAS exposure and thyroid hormone homeostasis during pregnancy.

Thyroid peroxidase is a key enzyme involved in thyroid hormone biosynthesis and can be inhibited by TPOAb, resulting in reduced hormone synthesis [66]. TgAb is another type of autoantibody targeting Tg. Clinically, both TPOAb and TgAb serve as biomarkers for identifying individuals at high risk of hypothyroidism. The presence of either or both antibodies is indicative of autoimmune thyroid disease [67]. Among pregnant women with high TPOAb (≥ 9 IU/mL), concentrations of PFNA, PFOA, and PFOS were positively associated with TSH levels; meanwhile, inverse associations were observed between FT₄ levels and four PFASs: PFHxS, PFNA, PFOA, and PFOS [68]. Based on these associations, Webster et al. proposed the “multiple-hit hypothesis”, indicating TPOAb-positive individuals may be more vulnerable to PFAS-induced changes in THs during pregnancy [69]. To explore this hypothesis, we stratified participants based on their thyroid autoantibody status into antibody-positive and antibody-negative groups. Stratified analyses revealed that within the antibody-positive group, PFOS, PFDA, PFNA, PFDoDA, PFUnDA, HFPO-TrA (C9), 6:2 Cl-PFESA, and 8:2 Cl-PFESA were inversely associated with FT₃; meanwhile, PFOS, PFDA, PFNA, PFDoDA, PFUnDA, 6:2 Cl-PFESA, and 8:2 Cl-PFESA exhibited positive associations with TSH. These findings provide support for the multiple-hit hypothesis. Conversely, a positive association of PFNA with FT₄ and a negative association of PFHpA with FT₃ were observed in the antibody-negative group. While the underlying mechanism for this positive association remains unclear and warrants further investigation, these highlight the critical importance of considering maternal autoantibody status.

PFMOAA is the shortest known chain perfluoroalkyl ether carboxylic acid (PFECA) and is a by-product of fluorochemical production [70]. During oral exposure studies in pregnant and lactating Sprague-Dawley rats, PFMOAA exhibited toxicological effects similar to PFAS carboxylates such as PFOA and HFPO-DA, including increased liver weights and reduced thyroid hormone levels (maternal and pup) [71]. In our study, linear regression revealed a significant positive association between PFMOAA and TSH levels, whereas RCS analysis indicated a nonlinear exposure–response relationship. A plausible mechanism involves the binding affinity of short-chain PFAS to transthyretin (TTR). Short-chain PFAS exhibit stronger binding affinity to TTR than long-chain PFAS [72]. Given TTR’s critical role in transporting THs in the bloodstream [73], PFAS-induced disruption of TTR-hormone binding may lower hormone levels [30, 74], thereby triggering a compensatory increase in TSH secretion by the pituitary gland [75]. However, these mechanistic insights are based on limited evidence, and further animal studies are necessary to elucidate the precise mechanisms underlying the PFMOAA-TSH association.

Analysis of PFAS mixtures revealed that their effects on thyroid function depend on compositional profiles, study design and population characteristics. In our study, mixtures including all 13 PFASs, as well as those derived from LASSO selection, were inversely associated with FT₃. When the mixture was restricted to PFASs identified as significant in single-pollutant models, positive associations with TSH and the FT₄/FT₃ ratio were additionally observed. Zhou et al. reported that combined PFAS exposure was inversely associated with TSH levels, which differed from the associations observed in our study [31]. Furthermore, the Health Outcomes and Measures of the Environment (HOME) Study (Cincinnati, 2003–2006) reported no significant associations between PFAS mixtures and maternal serum thyroid hormone levels [49]. These discrepancies may reflect the complexity of multi-chemical exposures, including differences in PFAS composition, exposure levels, study populations, and analytical approaches across studies [76]. Collectively, these findings highlight the importance of considering both chemical interactions and biological mechanisms when assessing PFAS mixture effects during pregnancy.

Nevertheless, this study has several limitations. First, due to its cross-sectional design, it can only reveal associations between PFAS exposure and thyroid function measures, without establishing causality. Second, certain potential confounders—such as parity and genetic polymorphisms—were not fully accounted for. Third, our study lacks in vitro mechanistic evidence, and the biological pathways underlying the observed associations remain unclear. Previous studies have suggested that PFAS may disrupt thyroid hormone homeostasis through several mechanisms, including impaired iodine uptake via inhibition of NIS [77], interference with thyroid hormone synthesis, disruption of thyroid hormone signaling pathways [78], and changes in peripheral thyroid hormone metabolism via deiodinase enzymes [30, 79]. However, the relative importance of these mechanisms, particularly for emerging PFAS compounds, remains unclear. Finally, further development of toxicity prediction models based on exposure levels is warranted to better evaluate their potential impacts on maternal health.

Conclusion

Prenatal exposure to both legacy PFASs (PFOS, PFDA, PFNA, PFDoDA, PFUnDA, and PFHpS) and novel PFAS (6:2 Cl-PFESA) was associated with lower FT₃ levels, while PFMOAA was positively associated with TSH levels. Nonlinear associations between PFMOAA and 6:2 Cl-PFESA and FT₃ levels were observed. Thyroid autoantibody status and iodine status may modify the associations between PFAS exposure and thyroid hormone levels. Collectively, these findings suggest that PFAS exposure may disrupt peripheral thyroid hormone metabolism during pregnancy, highlighting the need for further mechanistic and longitudinal studies.

Authors’ contributions

W.W., S.Y., and Z.M. conceived and designed the study. S.G. and F.G. performed software-based data processing and analysis. Z.C. and G.M. conducted the formal statistical analyses. X.W., X.L. and Z.M. provided key resources. F.G. and X.C. curated and managed the data. W.W. and S.Y. drafted the manuscript, and S.G., W.Y., R.Q. and Z.M. critically revised it. W.W. and S.Y. prepared the figures and tables. Z.M. supervised the study, managed the project, and secured funding.

Funding

This research was funded by Noncommunicable Chronic Diseases-National Science and Technology Major Project (2024ZD0531600); the Zhejiang Provincial Natural Science Foundation of China (No. LTGY23H24001); the Zhejiang Provincial Project for Medical Research and Health of China (No. 2025KY051); the Zhejiang Provincial Project for Disease Control and Prevention (No. 2025JK176); the Zhejiang Provincial Natural Science Foundation of China (No. LMS25H260006) and the National Natural Science Foundation of China (No. 32400980).

Data availability

Due to policy, the data are not publicly accessible. The data presented in this study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Prior to the initiation of the study, ethical review and approval were obtained from the Ethics Committee of the Zhejiang Provincial Center for Disease Control and Prevention (No. 2020-047-01). All participants provided written informed consent prior to study participation, in line with the approved ethics protocol. The study adhered to the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.