The study used the data from a large birth cohort study conducted at the Women and Children’s Healthcare Hospital of Linyi, China. The inclusion and exclusion criteria were detailed in our previous study (22). A total of 420 puerperae were enrolled from January 2019 to December 2021. Among them, 265 puerperae with cesarean delivery and 35 puerperae with thyroid diseases were excluded. A puerpera without complete data on PFASs was also excluded. Thus, 119 puerperae who delivered vaginally at 36–42 weeks of gestation were included in this study. The Ethics Committee of our hospital approved this study (approval number: KYL-YXLL-2021015). Informed consent was obtained from all participants.

The clinical characteristics of puerperae, including age, height, weight, body mass index (BMI), education level, times of deliveries, and gestational weeks, were collected. The birth height and birth weight of neonates were also collected.

After vaginal delivery, the umbilical cord blood samples were collected by a trained nurse using disposable syringes under sterile conditions. Then, the samples were centrifuged. The plasma was separated and aliquoted into polypropylene EP tubes using a disposable sterile Pasteur pipette. The umbilical cord plasma was stored at −80°C until chemical analysis.

A total of 29 PFASs were quantified in the umbilical cord plasma using a Waters ACQUITY ultra-performance liquid chromatography (UPLC) system (Waters Corporation, Milford, MA) coupled with a Waters Quattro Premier XE triple quadrupole mass spectrometer: 6:2 Cl-PFESA, perfluorobutanoic acid (PFBA), PFHpS, linear perfluorohexane sulfonic acid (L-PFHxS), PFNA, PFOA, linear PFOS (L-PFOS), sum of 3,4,5 monohydroperfluorooctane sulfonates (Σ3,4,5 m-PFOS), PFUnDA, PFDA, 6-monohydroperfluorooctane sulfonate (6 m-PFOS), perfluorohexanoic acid (PFHxA), PFDoA, perfluorotetradecanoic acid (PFTeDA), 1-monohydroperfluorooctane sulfonate (1 m-PFOS), PFHpA, perfluorobutane sulfonate (PFBS), hexafluoropropylene oxide dimer acid (HFPO-DA), perfluoropentanoic acid (PFPeA), 4:2 fluorotelomer sulfonic acid (4:2 FTS), perfluorooctane sulfonamide iodide (FOSA-I), sodium dodecafluoro-3 h-4,8-dioxanonanoate (NaDONA), perfluoropentane sulfonic acid (PFPeS), sum of dihydroperfluorooctane sulfonates (Σm2-PFOS), 8:2 fluorotelomer sulfonic acid (8:2 FTS), N-methyl perfluorooctane sulfonamidoacetic acid (N-MeFOSAA), 6:2 fluorotelomer sulfonic acid (6,2 FTS), perfluorononane sulfonic acid (PFNS), and N-ethyl perfluorooctane sulfonamidoacetic acid (N-EtFOSAA).

The Waters ACQUITYTM BEH C18 column (50 mm × 2.1 mm, 1.7 μm, Waters, United States) was used for quantification under the following conditions: column temperature was 40°C, flow rate was 0.4 mL/min, injection volume was 10 μL, and injection mode was partial quantification loop mode. The mobile phase A was 2 mmol/L of ammonium acetate aqueous solution, while the mobile phase B was methanol. The gradient elution was divided into the following stages: 0.0–0.5 min, with 40% of phase B; 0.5–12 min, phase B increased from 40 to 90%; 12–14 min, phase B decreased from 90 to 40%; and 14–15.5 min, held at 40% of phase B. Mass spectrometry was performed in the electrospray ionization (ESI) negative ionization mode with a capillary voltage of 3.0 kV, a source temperature of 120°C, a desolvation temperature of 400°C, a desolvation gas flow rate of 800 L/h, a cone-hole gas flow of 150 L/h, and a collision gas flow rate of 0.15 mL/min. Multiple response monitoring (MRM) mode was used for quantitative analysis.

For each batch of 20 samples, two blanks (Milli-Q pure water) were added to monitor the background during the pretreatment process. Before analyzing the samples, methanol was injected to monitor the instrument background. The detection limit of this method was 0.1–5.0 pg./mL, the limit of quantification was 0.2–10.0 pg./mL, and the recovery rate ranged from 70 to 130%.

A total of five thyroid hormones were quantified in the umbilical cord plasma using a Roche Analytics E170 modular analyzer (Roche Diagnostics, Mannheim, Germany): total T3 (TT3), FT3, total T4 (TT4), FT4, and TSH. The intra-assay coefficients of variation were less than 10%, while the inter-assay coefficients of variation were all below 15%.

SPSS version 29.0 (IBM, United States) and R version 4.3.3 were used for data analysis. The clinical characteristics of subjects were described as mean ± standard deviation or numbers with percentage. The Spearman test was used for correlation analyses among PFASs and to determine the correlation between PFASs and thyroid hormones. Due to the high collinearity between PFASs, dimension reduction was completed for the 29 PFASs by elastic network regression. A 20-fold cross-validation method was used, and the loss was calculated using mean-squared error. As a result of dimension reduction, the 29 PFASs were reduced to 9. Furthermore, to explore the independent association between PFASs and thyroid hormones, linear regression models were constructed. The results of dimension reduction and significant results of the Spearman test were included in the model, where the enter method was used. A p-value of <0.050 was indicated as statistical significance.

Initially, 420 puerperae were included, and 300 of them were excluded because they had a cesarean delivery (n = 265) and had thyroid diseases (n = 35). A total of 119 puerperae were enrolled, and their umbilical cord plasma samples were collected. Twenty-nine PFASs and five thyroid hormones were quantified. One patient was excluded from the analysis due to incomplete PFAS data. Subsequently, the Spearman test was applied to analyze the correlation between PFASs and thyroid hormones, and elastic network regression was used to deal with multicollinearity among PFASs through dimension reduction. A linear regression model was constructed to explore the independent association between these PFASs and thyroid hormones (Supplementary Figure 1).

The mean age of puerperae was 28.6 ± 4.6 years. A total of 106 (89.1%) puerperae were primiparity, and the other 13 (10.9%) puerperae were multiparity. The gestational period was 39.3 ± 1.1 weeks. The birth height of neonates was 49.9 ± 0.3 cm, and the birth weight was 3429.1 ± 376.2 g. The detailed clinical information is given in Table 1.

Five thyroid hormones (TT3, FT3, TT4, FT4, and TSH) were quantified in the umbilical cord plasma. The median [quartile 1 (Q1)–quartile 3 (Q3)] levels of TT3, FT3, TT4, FT4, and TSH in the umbilical cord blood were 0.8 (0.7–0.8) nmol/L, 2.1 (2.0–2.3) pmol/L, 98.7 (88.3–111.8) nmol/L, 15.0 (13.8–16.2) pmol/L, and 7.3 (5.7–11.7) μIU/mL, respectively (Table 2).

A total of 29 PFASs were quantified in the umbilical cord plasma. Nineteen PFASs had a detection rate of >50%: 6:2 Cl-PFESA, PFBA, PFHpS, L-PFHxS, PFNA, PFOA, L-PFOS, Σ3,4,5 m-PFOS, PFUnDA, PFDA, 6 m-PFOS, PFHxA, PFDoA, PFTeDA, 1 m-PFOS, PFHpA, PFBS, HFPO-DA, and PFPeA. Among them, PFOA, 6:2 Cl-PFESA, and L-PFOS possessed the highest levels in the umbilical cord blood, with median (Q1–Q3) levels of 3.23 (2.32–4.32), 1.35 (0.84–2.01), and 0.94 (0.63–1.41) ng/mL, respectively (Table 3).

To explore the correlation between PFASs and thyroid hormones, the Spearman test was used. It was found that Σ3,4,5 m-PFOS was negatively correlated with TT3 (r = −0.184, p = 0.045). Furthermore, PFUnDA (r = −0.201, p = 0.029), 6 m-PFOS (r = −0.221, p = 0.016), and 4:2 FTS (r = −0.192, p = 0.036) were negatively correlated with TSH. PFDoA was positively correlated with FT3 (r = 0.209, p = 0.022) but negatively correlated with TSH (r = −0.201, p = 0.028). These findings suggested a complex interplay between specific PFASs and thyroid hormone regulation in the context of the umbilical cord blood. No correlations were found between other PFASs and thyroid hormones in the umbilical cord blood (all p > 0.050) (Table 4).

To explore the relationships among PFASs in the umbilical cord blood, we generated a correlation heatmap that revealed intercorrelations among these compounds (Figure 1). These intercorrelations might introduce multicollinearity, which can complicate the interpretation of statistical models. To address this issue, we used elastic network regression, a technique that combines the strengths of LASSO and ridge regression, to perform dimension reduction on 29 PFASs. Subsequently, nine PFASs were screened out: 6:2 Cl-PFESA, PFBA, L-PFHxS, PFNA, PFOA, 6 m-PFOS, PFHxA, PFTeDA, and N-MeFOSAA (Supplementary Figure 2).

To further elucidate the association between PFASs and thyroid hormones in the umbilical cord blood, we conducted an analysis using linear regression models. Given that the results from elastic network regression did not include all five PFASs that were correlated with thyroid hormones in the Spearman analysis, and to maximize the inclusion of PFASs, we decided to incorporate the results obtained from both the Spearman analysis and the elastic network regression into the construction of linear regression models.

A total of 13 PFASs were enrolled in linear regression analysis: 6:2 Cl-PFESA, PFBA, L-PFHxS, PFNA, PFOA, Σ3,4,5 m-PFOS, PFUnDA, 6 m-PFOS, PFHxA, PFDoA, PFTeDA, 4:2 FTS, and N-MeFOSAA. L-PFHxS (β = 0.557, p = 0.038) and PFNA (β = 0.613, p = 0.045) were independently and positively associated with FT3. However, PFOA was independently and negatively associated with FT3 (β = −0.040, p = 0.002). Σ3,4,5 m-PFOS was independently and negatively associated with TT3 (β = −0.349, p = 0.007). PFDoA was independently and positively associated with TT3 (β = 2.107, p = 0.027) and FT3 (β = 5.254, p = 0.008). These findings indicated that there was a strong interaction between these five PFASs and thyroid hormones in the umbilical cord blood. No associations were found between other PFASs and thyroid hormones in the umbilical cord blood (all p > 0.050) (Table 5).