Association of emerging alternatives to per- and polyfluoroalkyl substances with polycystic ovary syndrome among Chinese women: a case-control study.

Table  1 presents the demographic and menstrual characteristics of 94 PCOS patients and 81 controls. The average age and age at menarche of PCOS cases (27.45 years and 12.97 years, respectively) were significantly lower than those of controls (33.89 years and 13.45 years, respectively). The BMI of PCOS cases was significantly higher than that of controls (24.42 kg/m² vs. 22.77 kg/m²). The prevalence of menstrual abnormalities was significantly higher in PCOS cases (86.17%) than in controls (38.27%). Table 1 Characteristics of the study population ( N  = 175). Covariate Cases ( N  = 94) Controls ( N  = 81) N (%) or Mean (SD) N (%) or Mean (SD) Age (years) a 27.45 (6.28) 33.89 (5.48) BMI (kg/m 2 ) a 24.42 (5.35) 22.77 (3.66) Age at menarche (years) a 12.97 (4.92) 13.45 (1.45) Marital status b Unmarried 56 (59.57) 14 (17.29) married 38 (40.43) 64 (79.01) divorced 0 (0.00) 3 (3.70) Monthly personal income(RMB/person)  10,000 b 8 (8.51) 26 (32.10) Educational level Undergraduate degree or higher 9 (9.57) 2 (2.47) Associate degree 13 (13.83) 10 (12.35) High school graduate 18 (19.15) 22 (27.16) Middle school graduate 54 (57.45) 47 (58.02) Smoking 3 (3.19) 0 (0.00) Alcohol consumption 11 (11.70) 7 (8.64) Menstrual abnormality b 81 (86.17) 31 (38.27) Abnormal Menstrual volume 26 (27.66) 24 (29.63) Abbreviations : BMI, body mass index; SD, standard deviation. a. P < 0.05, Mann-Whitney U test. b. P < 0.05, Pearson Chi-square test. Characteristics of the study population ( N  = 175). Abbreviations : BMI, body mass index; SD, standard deviation. a. P < 0.05, Mann-Whitney U test. b. P < 0.05, Pearson Chi-square test. Significant differences in marital status were observed between the two groups across three categories: single, married, and divorced. Additionally, a significant disparity in monthly income was observed between the two groups, particularly among individuals earning less than 3,000 RMB or more than 10,000 RMB. We found that the detection rates of 8 PFAS were below 70%, resulting in their exclusion from statistical analysis 29 . Table 2 presents the concentrations of the 17 PFAS with detection rates exceeding 70%. The 17 PFAS are categorized as follows: (1) 4 emerging PFAS alternatives: FTSA (6/2), F-53B (6/2), F-53B (8/2) and HFPO-DA (GenX); (2) One PFAS precursor: FBSA; (3) 6 long-chain legacy PFAS: PFDA, PFDoA, PFNA, PFOA, PFOS, and PFNS; and (4) 6 short-chain PFAS: PFBS, PFHpA, PFHxS, PFPeA, PFPeS, and PFHpS. PFOA and PFOS were the predominant PFAS in human serum, exhibiting median concentrations of 5.50 ng/mL and 4.85 ng/mL, respectively (Table 2 ). The Mann-Whitney U test revealed significant differences in the concentrations of 14 PFAS between the PCOS cases and the controls, excluding PFDoA, PFHxS, and PFPeA. Notably, the median concentrations of PFNS and F-53B (8/2) in the PCOS cases (0.01 ng/mL and < 0.01 ng/mL, respectively) were lower than those in the controls (0.02 ng/mL and 0.08 ng/mL, respectively). Table 2 Serum concentrations of PFAS ( N  = 175). PFAS DetectionRate Median (IQR) P -Value (ng/mL) (%) Cases ( N  = 94) Controls ( N  = 81) Total ( N  = 175) FTSA (6/2) 98.28 0.06 (0.05,0.09) 0.05 (0.04,0.06) 0.05 (0.04,0.07) < 0.01 F-53B (6/2) 99.43 1.15 (0.66,3.10) 0.82 (0.48,1.57) 1.02 (0.53,2.15) 0.01 FBSA 94.86 0.29 (0.25,0.34) 0.16 (0.11,0.26) 0.25 (0.16,0.31) < 0.01 HFPO-DA (GenX) 96.00 0.63 (0.48,0.85) 0.13 (0.06,0.31) 0.48 (0.14,0.75) < 0.01 PFBS 98.29 0.19 (0.13,0.32) 0.13 (0.08,0.19) 0.16 (0.11,0.25) < 0.01 PFDA 100.00 1.10 (0.63,2.46) 0.64 (0.44,0.96) 0.83 (0.48,1.42) < 0.01 PFDoA 88.00 0.28 (0.15,0.50) 0.26 (< 0.01,1.00) 0.28 (0.09,0.65) 0.54 PFDS 63.43 0.02 (0.01,0.07) < 0.01 (< 0.01,<0.01) 0.01 (< 0.01,0.03) < 0.01 PFHpA 94.86 0.30 (0.19,0.47) 0.20 (0.09,0.28) 0.24 (0.15,0.36) < 0.01 PFHxA 69.14 1.35 (0.01,2.34) 1.73 (0.01,10.35) 1.37 (0.01,3.36) 0.03 PFHxDA 15.43 0.01 (0.01,0.01) 0.01 (0.01,0.01) 0.01 (0.01,0.01) 0.77 PFHxS 100.00 0.53 (0.36,0.88) 0.52 (0.34,0.77) 0.52 (0.35,0.81) 0.32 PFNA 100.00 1.41 (0.72,2.64) 0.58 (0.37,0.80) 0.81 (0.52,1.50) < 0.01 PFOA 99.43 6.05 (4.08,8.95) 4.83 (3.48,6.21) 5.50 (3.73,7.72) < 0.01 PFOdA 18.86 < 0.01 (< 0.01,<0.01) < 0.01 (< 0.01,<0.01) < 0.01 (< 0.01,<0.01) 0.07 PFOS 100.00 5.14 (3.09,11.13) 4.40 (2.59,6.48) 4.85 (2.89,8.21) 0.01 PFPeA 100.00 1.30 (0.69,2.09) 0.99 (0.44,2.40) 1.24 (0.60,2.16) 0.30 PFTeDA 46.86 < 0.01 (< 0.01,0.09) 0.04 (< 0.01,0.08) < 0.01 (< 0.01,0.09) 0.13 PFTrDA 64.57 0.38 (0.25,0.59) < 0.01 (< 0.01,0.03) 0.14 (< 0.01,0.42) < 0.01 PFUdA 66.29 0.17 (< 0.01,0.52) 0.03 (< 0.01,0.06) 0.05 (< 0.01,0.23) < 0.01 PFPeS 82.86 0.08 (0.03,0.20) 0.01 (< 0.01,0.04) 0.04 (0.01,0.12) < 0.01 PFHpS 89.14 0.07 (0.03,0.14) 0.04 (0.01,0.09) 0.05 (0.02,0.13) 0.01 PFNS 76.57 0.01 (< 0.01,0.03) 0.02 (0.01,0.04) 0.01 (< 0.01,0.03) < 0.01 DONA 68.57 0.11 (< 0.01,0.30) 0.03 (0.01,0.06) 0.04 (< 0.01,0.15) < 0.01 F-53B (8/2) 71.43 0.00 (< 0.01,0.02) 0.08 (0.02,0.23) 0.02 (< 0.01,0.09) < 0.01 Abbreviations : DetectionRate: PFAS concentrations ≥ limit of detection (LOD); IQR, interquartile range. LOD (ng/mL): FTSA (6/2) (0.004), F-53B (6/2) (0.001), FBSA (0.001), HFPO-DA (GenX) (0.002), PFBS (0.001), PFDA (0.004), PFDoA (0.003), PFHpA (0.011), PFHxS (0.003), PFNA (0.004), PFOA (0.002), PFOS (0.003), PFPeA (0.017), PFPeS (0.002), PFHpS (0.002), PFNS (0.001), F-53B (8/2) (0.002). Mann-Whitney U test. Serum concentrations of PFAS ( N  = 175). Abbreviations : DetectionRate: PFAS concentrations ≥ limit of detection (LOD); IQR, interquartile range. LOD (ng/mL): FTSA (6/2) (0.004), F-53B (6/2) (0.001), FBSA (0.001), HFPO-DA (GenX) (0.002), PFBS (0.001), PFDA (0.004), PFDoA (0.003), PFHpA (0.011), PFHxS (0.003), PFNA (0.004), PFOA (0.002), PFOS (0.003), PFPeA (0.017), PFPeS (0.002), PFHpS (0.002), PFNS (0.001), F-53B (8/2) (0.002). Mann-Whitney U test. Spearman correlation analysis indicated significant correlations ( p  < 0.05) between the concentrations of 14 PFAS, with coefficients ranging from − 0.33 to 0.85 (Fig.  1 ). Specifically, F-53B (6/2) demonstrated a strong positive correlation with PFOS (coefficient = 0.84), while PFNA exhibited a positive correlation with PFDA (correlation coefficient = 0.85). Notably, PFPeS exhibited a significant negative correlation with PFNS, with a correlation coefficient of -0.33. Fig. 1 Pearson correlation coefficients between ln-transformed serum PFAS concentration. Pearson correlation coefficients between ln-transformed serum PFAS concentration. The association between 14 PFAS and PCOS was assessed using the Firth penalized maximum likelihood estimation method, with adjustments for covariates. The results demonstrated that the concentrations of 12 PFAS were significantly positively correlated with an increased odds of PCOS (Fig.  2 and Table  3 ). Among these, HFPO-DA (GenX) had the highest OR (OR: 9.26; 95% CI: 4.16, 20.59), while PFHpA had the lowest (OR: 1.66; 95% CI: 1.08, 2.55). Additionally, PFNS (OR: 0.55; 95% CI: 0.36, 0.84) and F-53B (8/2) (OR: 0.40; 95% CI: 0.25, 0.63) were significantly negatively correlated with the odds of PCOS. Monte Carlo simulation-based empirical power analysis showed that most PFAS with significant associations, including FTSA (6/2), F-53B (8/2), HFPO-DA (GenX), FBSA, PFDA, PFNA, and PFOA, achieved adequate statistical power (≥ 80%). In contrast, F-53B (6/2), PFOS, PFNS, PFBS, and PFHpA exhibited limited power (< 80%), and their estimates should therefore be interpreted with caution (Table  3 ). Table 3 Associations of ln-transformed serum PFAS concentrations with PCOS in Firth penalized maximum likelihood model a . PAFS Coefficient OR 95% CI b P -Value Empirical power (%) Emerging PFAS alternatives FTSA (6/2) 0.6966 2.01 (1.28, 3.15) < 0.01 84.74 F-53B (6/2) 0.5631 1.76 (1.16, 2.67) 0.01 72.66 F-53B (8/2) -0.9150 0.40 (0.25, 0.63) < 0.01 98.60 HFPO-DA (GenX) 2.2257 9.26 (4.16, 20.59) < 0.01 100.00 Precursor FBSA 1.3422 3.83 (1.79, 8.16) < 0.01 99.84 Long-chain legacy PFAS PFDA 0.7697 2.16 (1.38, 3.37) < 0.01 93.26 PFNA 1.6575 5.25 (2.74, 10.03) < 0.01 100.00 PFOA 0.7314 2.08 (1.32, 3.27) < 0.01 86.58 PFOS 0.5120 1.67 (1.11, 2.52) 0.02 64.84 PFNS -0.6038 0.55 (0.36, 0.84) 0.01 78.04 Short-chain PFAS PFBS 0.6066 1.83 (1.19, 2.83) 0.01 76.56 PFHpA 0.5073 1.66 (1.08, 2.55) 0.02 68.20 PFPeS 1.3208 3.75 (2.17, 6.47) < 0.01 100.00 PFHpS 0.5498 1.73 (1.15, 2.60) 0.01 73.72 Abbreviations : a : Model were adjusted for Age, body mass index (BMI), age at menarche, marital status, monthly personal income, education level, smoking and drinking history, menstrual abnormalities, and menstrual volume irregularities. b : 95% CI: 95% confidence interval. Associations of ln-transformed serum PFAS concentrations with PCOS in Firth penalized maximum likelihood model a . Abbreviations : a : Model were adjusted for Age, body mass index (BMI), age at menarche, marital status, monthly personal income, education level, smoking and drinking history, menstrual abnormalities, and menstrual volume irregularities. b : 95% CI: 95% confidence interval. Fig. 2 Odds ratios of PFAS associated with PCOS. Model were adjusted for Age, body mass index (BMI), age at menarche, marital status, monthly personal income, education level, smoking and drinking history, menstrual abnormalities, and menstrual volume irregularities. Odds ratios of PFAS associated with PCOS. Model were adjusted for Age, body mass index (BMI), age at menarche, marital status, monthly personal income, education level, smoking and drinking history, menstrual abnormalities, and menstrual volume irregularities. Spearman correlation analysis revealed multicollinearity among the 14 PFAS. Therefore, Lasso regression was employed to select PFAS and construct a multifactorial exposure model to examine the association between PFAS and PCOS ( Table S3 ). Without adjusting for covariates, Lasso regression selected 7 PFAS for model construction, 6 of which had positive coefficients: FTSA (6/2), FBSA, HFPO-DA (GenX), PFHpA, PFNA, and PFPeS. Among these, HFPO-DA (GenX) had the highest coefficient (1.7536). Additionally, the coefficient for F-53B (8/2) was negative (-1.4880). After adjusting for all covariates, the results of the Lasso regression variable selection remained consistent, with HFPO-DA (GenX) retaining the largest coefficient (1.8580) and F-53B (8/2) showing a negative coefficient (-1.4787). Table S4 presents the coefficients obtained from ridge regression and elastic net. Both models produced patterns consistent with the Lasso results. HFPO-DA (GenX) remained the PFAS with the largest positive coefficient (ridge: 0.9173; elastic net: 1.7342), whereas F-53B (8/2) consistently exhibited a negative association (ridge: -0.9523; elastic net: -1.5124). FTSA (6/2), FBSA, PFHpA, PFNA, and PFPeS retained positive coefficients. Overall, the ridge and elastic net results closely paralleled the Lasso findings, supporting the robustness of the identified associations between PFAS exposures and PCOS. Based on the results from the Lasso regression model, we selected 7 PFAS to construct a mixed exposure model to assess the relationship between PFAS and PCOS. These selected PFAS included FTSA (6/2), HFPO-DA (GenX), F-53B (8/2), FBSA, PFNA, PFHpA, and PFPeS. The exposure-response plots for each PFAS are shown in Fig.  3 . where serum concentrations of other PFAS were held at their median values, the concentrations of FTSA (6/2), HFPO-DA (GenX), FBSA, PFNA, PFHpA, and PFPeS were positively linearly associated with the odds of PCOS, while F-53B (8/2) exhibited a negative linear association with PCOS odds. These findings were consistent with the results of the multifactorial exposure model. Fig. 3 Dose-response relationship of individual PFAS on PCOS odds.The BKMR Model were adjusted for Age, body mass index (BMI), age at menarche, marital status, monthly personal income, education level, smoking and drinking history, menstrual abnormalities, and menstrual volume irregularities. Dose-response relationship of individual PFAS on PCOS odds.The BKMR Model were adjusted for Age, body mass index (BMI), age at menarche, marital status, monthly personal income, education level, smoking and drinking history, menstrual abnormalities, and menstrual volume irregularities. Figure  4 . A illustrates the combined associations of PFAS with PCOS. We observed a linear increase in PCOS odds with rising levels of the mixture when the concentrations of all PFAS were maintained at their specified percentiles (from 0.25 to 0.75, with an interval of 0.05), compared to their median values. The results presented in Fig.  4 B indicated that FTSA (6/2), HFPO-DA (GenX), FBSA, PFNA, PFHpA, and PFPeS were positively correlated with the odds of PCOS, while F-53B (8/2) was negatively correlated with PCOS odds. Among these, HFPO-DA (GenX) contributed most significantly to the odds of PCOS, as indicated by the largest absolute coefficient value. These results are consistent with multifactorial exposure model. Finally, the bivariate exposure-response functions of the 7 PFAS were estimated ( Fig. S2 ). The results indicated that no significant interactions were observed among the 7 PFAS. Complementing the BKMR analysis, WQS regression was performed to further quantify the relative contributions of individual PFAS to PCOS. Table S5 and Fig. S3 presents the weights assigned to each PFAS within the mixture. FBSA contributed the most to the overall mixture effect (weight = 0.38), followed by PFPeS (0.19), PFHpA (0.13), FTSA (6/2) (0.15), and HFPO-DA (GenX) (0.14). Associations between PFAS index and PCOS are summarized in Table  4 . The overall WQS index was strongly associated with PCOS (OR: 8.27; 95% CI: 4.03–16.94, p  < 0.01). Subgroup index for PFAS alternatives (OR: 5.59), short-chain PFAS (OR: 5.13), and long-chain PFAS (OR: 4.19) were all positively associated with PCOS (all p  < 0.01), indicating that emerging alternatives and short-chain PFAS were the primary drivers of the mixture effect. Together, the WQS findings corroborate the BKMR results, highlighting that PFAS mixtures, particularly emerging alternatives and short-chain PFAS, are positively associated with PCOS odds. Table 4 Associations of PFAS index with PCOS in Firth penalized maximum likelihood model a . PFAS index Coefficient OR 95% CI b P -Value WQS index 2.1122 8.27 (4.03, 16.94) < 0.01 WQS-alternatives index 1.7217 5.59 (2.96, 10.57) < 0.01 WQS-long index 1.4338 4.19 (2.42, 7.26) < 0.01 WQS-short index 1.6344 5.13 (2.67, 9.85) < 0.01 Abbreviations: a : Model were adjusted for Age, body mass index (BMI), age at menarche, marital status, monthly personal income, education level, smoking and drinking history, menstrual abnormalities, and menstrual volume irregularities. b : 95% CI: 95% confidence interval. Associations of PFAS index with PCOS in Firth penalized maximum likelihood model a . Abbreviations: a : Model were adjusted for Age, body mass index (BMI), age at menarche, marital status, monthly personal income, education level, smoking and drinking history, menstrual abnormalities, and menstrual volume irregularities. b : 95% CI: 95% confidence interval. Fig. 4 Joint effect of the PFAS on PCOS using Bayesian kernel machine regression (BKMR) model. Adjusted variables included age, BMI (body mass index), age at menarche, marital status, monthly personal income, educational level, smoking, alcohol consumption, menstrual abnormality, abnormal menstrual volume. (A) Overall effect of the PFAS (estimates and 95% credible interval); This plot showed the estimated change in the odds of PCOS by comparing a particular percentile (ranging from the 25th percentile to 75th percentile) of 7 PAFSs levels with their median value (the 50th percentile). (B) Single PFAS association (estimate and 95% credible intervals). This plot showed the effects of single PFAS by comparing the 75th of the PFAS concentrations with its 25th percentile, when concentrations of all the other PFAS were held at either the 25th (red line), 50th (green line), or 75th percentile (blue line). The BKMR Model were adjusted for Age, body mass index (BMI), age at menarche, marital status, monthly personal income, education level, smoking and drinking history, menstrual abnormalities, and menstrual volume irregularities. Joint effect of the PFAS on PCOS using Bayesian kernel machine regression (BKMR) model. Adjusted variables included age, BMI (body mass index), age at menarche, marital status, monthly personal income, educational level, smoking, alcohol consumption, menstrual abnormality, abnormal menstrual volume. (A) Overall effect of the PFAS (estimates and 95% credible interval); This plot showed the estimated change in the odds of PCOS by comparing a particular percentile (ranging from the 25th percentile to 75th percentile) of 7 PAFSs levels with their median value (the 50th percentile). (B) Single PFAS association (estimate and 95% credible intervals). This plot showed the effects of single PFAS by comparing the 75th of the PFAS concentrations with its 25th percentile, when concentrations of all the other PFAS were held at either the 25th (red line), 50th (green line), or 75th percentile (blue line). The BKMR Model were adjusted for Age, body mass index (BMI), age at menarche, marital status, monthly personal income, education level, smoking and drinking history, menstrual abnormalities, and menstrual volume irregularities.

This case-control study investigated the relationship between PCOS and exposure to PFAS mixture. The study population included women who attended the Obstetrics and Gynecology outpatient clinics at Suzhou Hospital of Nanjing Medical University, Jiangsu province in China between September 2019 and May 2024. Participants were aged 18–40 years and had a menstrual history of at least two years. A total of 94 infertile women diagnosed with PCOS were assigned to the PCOS case group. The diagnosis of PCOS, based on the 2003 revised Rotterdam criteria, requires at least two of the following features: clinical and/or biochemical hyperandrogenism, oligoanovulation, and polycystic ovarian morphology, after excluding other causes of hyperandrogenism 22 , 23 . Only newly diagnosed patients or those who had discontinued hormonal medications for at least three months prior to hormone testing were included. The control group comprised 81 women with regular menstrual cycles, normal gonadal hormone levels, and no clinical or biochemical signs of hyperandrogenism. All controls underwent transvaginal or transabdominal pelvic ultrasound to confirm normal ovarian morphology. Women with a history of hormonal medication use or with conditions affecting endocrine or ovarian function-such as hyperprolactinemia, non-classic congenital adrenal hyperplasia, thyroid dysfunction, androgen-secreting tumors, or Cushing’s syndrome-were excluded. For all participants, serum hormone levels were measured during the early follicular phase (days 2–5 of the menstrual cycle) at the time of recruitment, ensuring comparability between groups. All participants provided written informed consent prior to their inclusion in the study. The research protocol was approved by the Ethics Committees of Suzhou Hospital of Nanjing Medical University (Approval number: KL901295 ), with all procedures adhering to the principles of the Declaration of Helsinki. A trained research assistant assisted participants in completing a standardized questionnaire to collect personal information. The questionnaire gathered information on demographic characteristics (age, residence, marital status, occupation, education level, personal monthly income), lifestyle behaviors (alcohol consumption, smoking history), and menstrual and reproductive history (age at menarche, menstrual cycle and duration, menstrual volume, pregnancy history, and gynecological conditions, etc.). Participants’ medical information was obtained from medical records, including height and weight, as well as other medical examinations (gynecological exams). Previous research and a directed acyclic graph were used to generate a set of minimal sufficient adjustment variables for confounding control ( Fig. S1 ) 5 , 19 , 24 . Age, body mass index (BMI), age at menarche, marital status, monthly personal income, education level, smoking and drinking history, menstrual abnormalities (defined as menstrual cycles shorter than 21 days or longer than 35 days), and menstrual volume irregularities (defined as menstrual flow less than 20 mL or more than 80 mL per cycle) were included as covariates in the statistical analysis. A 5 mL venous serum sample was collected from each participant, centrifuged at 4000 rpm for 10 min, and the serum was separated and stored at -80 °C until analysis. 25 PFAS of serum were quantitatively analyzed, including: 6:2 fluorotelomer sulfonate (FTSA(6/2)), F-53B(6/2), perfluoro-1-butane-sulfonamide (FBSA), HFPO-DA(GenX), perfluorobutane sulfonic acid (PFBS), perfluorodecanoic acid (PFDA), PFDoA, perfluorodecane sulfonic acid (PFDS), perfluoroheptanoic acid (PFHpA), perfluorohexanoic acid (PFHxA), perfluorohexadecanoic acid (PFHxDA), perfluorohexane sulfonate acid (PFHxS), perfluorononanoic acid (PFNA), PFOA, perfluorooctadecanoic acid (PFOdA), PFOS, perfluoropentanoic acid (PFPeA), perfluorotetradecanoic acid (PFTeDA), perfluorotridecanoic acid (PFTrDA), perfluoroundecanoic acid (PFUdA), perfluoropentanesulfonic acid (PFPeS), perfluoroheptanesulfonic acid (PFHpS), perfluorononanesulfonic acid (PFNS), 4,8-dioxa-3 H-perfluorononanoic acid (DONA), and 8: 2 chlorinatedpolyfluorinated ethersulfonate (F-53B(8/2)). 13 mixed isotopic internal standards were used: 13C4-PFBA, 13C5-PFPeA, 13C5-PFHxA, 13C4-PFHpA, 13C8-PFOA, 13C9-PFNA, 13C6-PFDA, 13C7-PFUdA, 13C2-PFDoA, 13C2-PFTeDA, 13C3-PFBS, 13C3-PFHxS, and 13C8-PFOS, from TRC, Canada. Other reagents included methanol (HPLC grade, Fisher Scientific, USA), methyl tert-butyl ether (MTBE), tetrabutylammonium hydrogen sulfate (TBAS, HPLC grade, ROE Scientific, USA), ammonium formate (HPLC grade, ROE Scientific, USA), and sodium bicarbonate (NaHCO3, analytical grade, East China Pharmaceutical Co.). A 200 µL aliquot of uniformly mixed serum was transferred into a 15 mL centrifuge tube. Next, 0.5 ng of isotopically labeled internal standards, 1 mL of 0.5 M TBAS, 2 mL of sodium bicarbonate (NaHCO 3 ) buffer (pH 10), and 4 mL of MTBE were added. The mixture was vortexed for 10 min, ultrasonically extracted at 25 kHz for 15 min, and centrifuged at 12,000 rpm for 10 min. The supernatant was carefully transferred into a new 15 mL centrifuge tube. The residue was re-extracted with an additional 4 mL of MTBE. Both extracts were combined and evaporated to dryness under nitrogen at 40 °C. The dried residue was reconstituted in 200 µL of a 1:1 methanol-water solution. LC-MS/MS analysis was conducted using a UPLC-AB Sciex Triple Quad 5500 MS system, which included a Triple Quad 5500 mass spectrometer and an Exion UPLC system (both AB Sciex, USA) with an electrospray ionization source. Data acquisition and processing were performed using SCIEX Analyst software. Instrument conditions were as follows: an X-Bridget™ BEH C 18 column (100 mm × 2.1 mm, 2.5 μm, Waters, USA) was used, with a column temperature set to 35℃. The injection volume was 10µL. The mobile phase consisted of A (0.2 mM ammonium formate solution) and B (methanol). The gradient elution program was: 0–0.5 min, 20% B; 0.5–7 min, 20–100% B; 7–10 min, 100% B; 10–10.5 min, 20% B; 10.5–15 min, 20% B. Electrospray ionization (ESI) was operated in negative ion mode (ESI-). The ion source interface voltage was set to 4.5 kV. The nebulizer gas was nitrogen at a flow rate of 3 L/min. The drying gas was nitrogen (10 L/min), and the heated gas was air (10 L/min). The collision gas was argon. The desolvation capillary temperature was set to 250 °C, the heating block temperature to 350 °C, and the interface temperature to 250 °C. The multi-reaction monitoring (MRM) parameters are detailed in Table S1 . In accordance with the U.S. Environmental Protection Agency (EPA) recommendations for minimizing background contamination, venous blood samples were collected using pre-cleaned, PFAS-free sampling materials. All procedures strictly followed EPA guidelines for sample integrity and trace-level PFAS determination (EPA Method 537.1 and Method 533). To ensure robust QA and QC for the target analytes, each analytical batch included two ultrapure water extraction blanks and two matrix-spiked recovery samples to monitor potential contamination and analyte loss during the extraction process. All samples, including blanks, spiked recoveries, and field samples, were fortified with 2.5 ng of isotopically labeled internal standards (IS) prior to extraction. For the spiked recovery samples, additional unlabeled calibration standards (CS; 1 ng and 10 ng) were added to evaluate extraction efficiency and method reproducibility (Table S2 ). The extraction, storage, and analytical conditions for blanks and spiked samples were identical to those applied to field samples. For compounds lacking isotopically labeled analogs, a chemically related surrogate IS with the closest retention time was selected (Table S1 ). The method detection limit (MDL) was defined as the mean concentration measured in blank samples plus three times the standard deviation, while the method quantification limit (MQL) was defined as the mean blank concentration plus ten times the standard deviation (Table S2 ). All concentrations exceeding the linear calibration range were reanalyzed after appropriate dilution and subsequently confirmed. PFAS concentrations measured in field blanks were consistently below the limits of detection (LOD). The Mann-Whitney U test was used for continuous variables, and the Pearson chi-square test was applied for categorical variables to compare the baseline characteristics of the study population. The PFAS concentrations in the PCOS case and control groups were summarized using the median and interquartile range (IQR). The LOD for the 25 PFAS are presented in Table S2 . Values below the LOD were substituted with the square root of the LOD divided by 2. The serum PFAS concentrations were transformed using the natural logarithmic (LN) function to reduce the influence of outliers, followed by standardization to ensure data consistency. The Mann-Whitney U test was used to analyze differences in PFAS concentrations between the case and control groups. Collinearity among the PFAS exposure variables was preliminarily assessed using Pearson correlation analysis. Highly correlated variables (correlation coefficient > 0.8) were considered indicative of potential collinearity 25 . Odds ratios (OR) and 95% confidence intervals (CI) for each PFAS were estimated using the Firth penalized maximum likelihood method (Firth PML) to assess the association between individual PFAS concentrations and PCOS with adjustment for all covariates. The Firth PML is an improved technique designed to overcome the limitations of classical maximum likelihood estimation (MLE). It addresses challenges such as unstable or non-existent estimates in statistical models, commonly caused by small sample sizes, overfitting, or complete separation. This method is widely applied in logistic regression. Its core principle introduces a penalization term to correct the bias in maximum likelihood estimation, improving the robustness and reliability of parameter estimates. Because the analysis involved multiple covariates and logistic regression with potentially non-standard covariate distributions, and closed-form power formulae may be unreliable under such complexity, empirical statistical power for each PFAS was estimated using a Monte Carlo simulation-based approach 26 , 27 . For each exposure, 5,000 simulated datasets were generated based on the observed covariate structure and fitted with the same Firth penalized logistic regression model. Empirical power was defined as the proportion of simulations in which the association reached statistical significance at α = 0.05. A power value ≥ 80% was considered indicative of sufficient sample size, whereas lower values suggested limited power and the need for cautious interpretation 28 . In addition, the Lasso regression model was used to construct a multifactorial exposure model for selecting relevant variables and addressing multicollinearity among independent variables. Ridge regression and elastic net were further applied as complementary approaches. Ridge regression shrinks all coefficients toward zero to address multicollinearity, whereas elastic net combines Lasso and ridge penalties (α = 0.5) to balance variable selection and coefficient shrinkage, with the regularization parameter selected by cross-validation. The Bayesian kernel machine regression (BKMR) was utilized to evaluate the combined effects of PFAS mixture on PCOS after adjusting for all covariates. BKMR is a flexible kernel-based approach capable of addressing high correlations and interactions within chemical mixtures, while accommodating nonlinear dose-response relationships. BKMR was applied to evaluate: (1) the dose-response relationship between individual PFAS concentrations and PCOS, with other PFAS in the mixture fixed at their median values; and (2) the joint dose-response relationship between total PFAS mixture concentrations and PCOS, by comparing estimated PCOS values corresponding to a 5% increase or decrease from the median mixture concentration (used as a reference). To complement the BKMR analysis and to quantify the relative contribution of individual PFAS within the mixture, we additionally applied the Weighted Quantile Sum (WQS) regression approach. All PFAS concentrations were standardized using Z-score normalization, and each exposure was categorized into quartiles for WQS index construction. Covariates, including marital status, monthly personal income, education level, smoking and drinking behaviors, menstrual abnormalities, and menstrual volume irregularities, were converted into categorical variables prior to modeling. The WQS model incorporated 1,000 bootstrap iterations to estimate mixture weights under a non-negative constraint, assuming a positive direction of association based on prior evidence. The resulting weights were used to derive the overall PFAS mixture index (WQS index), as well as three subgroup index representing PFAS alternatives (WQS-alternatives; FTSA (6/2), HFPO-DA (GenX), and F-53B (8/2)), long-chain PFAS (WQS-long; PFNA), and short-chain PFAS (WQS-short; PFHpA and PFPeS). Each index was calculated by multiplying the quartile score of each component PFAS by its corresponding weight and summing the weighted values. Associations between each WQS-derived index and PCOS were subsequently evaluated using Firth penalized logistic regression, with adjustment for all covariates. All statistical analyses were performed using SPSS (version 22) and R Studio (version 4.3.3), with P  < 0.05 considered statistically significant.

This case-control study revealed a significant association between PFAS exposure and PCOS in infertile women of reproductive age. FTSA (6/2), HFPO-DA (GenX), FBSA, PFNA, PFHpA, and PFPeS were positively associated with the odds of PCOS, whereas F-53B (8/2) showed a negative association. Specifically, HFPO-DA (GenX) exhibited the highest contribution. To the best of our knowledge, this study is the first to identify associations between PCOS and the emerging PFAS alternatives FTSA (6/2), F-53B (8/2), and its precursor FBSA, among 25 PFAS. Furthermore, it also highlights the positive effect of HFPO-DA (GenX) on PCOS. Our findings offer new insights into the impact of PFAS on PCOS development. Among the 25 PFAS analyzed, PFOA and PFOS had the highest concentrations, with median values of 5.50 ng/mL and 4.85 ng/mL, respectively, and detection rates of 99.43% and 100%. These findings align with previous studies 5 , 19 . The detected concentrations exceeded those reported by Wang et al. (median PFOA and PFOS levels: 4.82 ng/mL and 3.98 ng/mL, respectively, both with a 100% detection rate) but were lower than those in Zhan et al.’s study (median levels: 8.52 ng/mL and 5.07 ng/mL, both with a 100% detection rate). Among the emerging PFAS alternatives, F-53B (6/2) and F-53B (8/2) were detected at 1.02 ng/mL and 0.02 ng/mL, respectively, closely aligning with Zhan et al.’s findings. This consistency indicates the reliability of our analytical method and the stability of the results. However, we observed the median HFPO-DA (GenX) concentration of 0.48 ng/mL, approximately 16 times higher than that reported by Zhan et al. Two factors may explain this discrepancy. First, differences in study populations may be a contributing factor: Zhan et al. sampled populations from Shanghai, Shandong, and Zhejiang, while our study focused on Suzhou, Jiangsu Province. Regional variations in study populations may reduce the impact of extreme exposure values, leading to results that approximate the average across multiple locations. Second, Suzhou is a major industrial city in the Yangtze River Delta, with highly developed sectors such as electronics manufacturing, textile dyeing, and chemical industries 30 . The use of PFAS-containing materials, such as waterproof coatings, stain-resistant treatments, semiconductor components, metal plating, and surfactants, may increase industrial emissions. Consequently, these emissions may elevate environmental PFAS exposure. These findings highlight the necessity of region-specific regulations to mitigate PFAS pollution effectively. As emerging PFAS alternatives, HFPO-DA (GenX) and FTSA (6:2) have garnered significant attention for their toxic effects. Multiple lines of evidence indicates significant positive associations between exposure to FTSA (6:2), HFPO-DA (GenX), and FBSA and the odds of PCOS. Previous studies have demonstrated that HFPO-DA (GenX), a typical alternative for PFOA, possesses toxicity and bioaccumulation properties equal to or even exceeding those of PFOA 31 , 32 . Limited epidemiological evidence has indicated that environmental exposure to F-53B (6/2), HFPO-DA (GenX), PFDoA, and PFOS is associated with an increased odds of PCOS, particularly in obese or overweight women 5 . The positive association between HFPO-DA (GenX) and PCOS is mechanistically plausible. HFPO-DA (GenX) disrupts the expression of key hormones and receptors involved in hypothalamic-pituitary-ovarian (HPO) axis signaling, thereby increasing the risk of PCOS 33 , 34 . Studies on PFAS exposure in human JEG-3 placental trophoblast cells reveal that 24-hour exposure to HFPO-DA (GenX) significantly alters placental gene expression. The dysregulated genes play essential roles in pathways regulating normal placental development and function 35 . Animal studies confirm that gestational exposure to HFPO-DA (GenX) causes intrauterine and postnatal growth retardation in SD rats, with placental inflammation potentially mediated through the Rap1 signaling pathway 36 . As a common alternative to PFOS, epidemiological evidence on the reproductive toxicity of FTSA (6/2) are limited, with most research focusing on its hepatotoxicity and lipid metabolism disruption. Zhang et al. reported that FTSA (6/2) induces oxidative stress and inflammatory responses in zebrafish embryos, potentially causing immunotoxicity via the TLR/NOD-MAPK pathway 37 . A mouse study further highlighted the bioaccumulation potential of FTSA (6/2), along with increases in liver weight, inflammation, and necrosis. Additionally, FTSA (6/2) was found to upregulate PPAR γ and associated proteins 38 . These findings also provide molecular evidence supporting the positive association between FTSA (6/2) and PCOS in our findings. FBSA, another PFOS alternative and PFAS precursor, demonstrates toxicity via mechanisms similar to PFOS 39 , 40 . Currently, epidemiological evidence linking FBSA to PCOS is lacking. Mahoney et al. found that FBSA and PFOS exhibit synergistic toxicity at high concentrations and specific ratios (1:1 or 1:3) 41 . This suggests that the toxicity of FBSA may result from its synergistic interactions with other PFAS. At the molecular level, a developmental toxicity study on zebrafish embryos revealed that FBSA induced abnormal morphology and significantly disrupted normal gene expression 40 . These findings suggest that the reproductive toxicity of FBSA is plausible. F-53B (8/2), as an emerging PFAS alternative, currently lacks evidence linking it to PCOS. Epidemiological research by Zhan et al. identified no significant correlation between F-53B (8/2) and increased PCOS odds 5 . A cohort study on PFAS and adverse pregnancy outcomes suggested that F-53B (6/2) and F-53B (8/2) could act as reproductive toxicants in humans 42 . However, another study reported no association between F-53B (6/2) or F-53B (8/2) and adverse birth outcomes 43 . These conflicting findings may result from variations in PFAS exposure levels, population demographics, and exposure pathways. Our findings indicated that F-53B (8/2) was negatively correlated with the odds of PCOS. In this study, the mean concentration of F-53B (8/2) was higher in the control group (0.08 ng/mL) than in the case group (< 0.01 ng/mL), which may explain its negative association with PCOS. Another possible explanation is that the association may be non-monotonic. At low or moderate levels of exposure, the relationship may be weak or even negative; however, a significant positive association may emerge once the exposure reaches or exceeds a certain threshold 5 . Nonetheless, further research is needed to evaluate the bioaccumulative potential and developmental toxicity of F-53B (8/2). PFNA, a conventional long-chain PFAS, has been widely studied for its toxic effects. Epidemiological studies indicate that PFNA exposure increases the risk of endometriosis in women 44 . Toxicological studies have shown that PFNA is linked to immunotoxicity, hepatotoxicity, developmental toxicity, and reproductive toxicity. Furthermore, in vitro studies have shown that PFNA disrupts oocyte maturation by impairing spindle assembly, mitochondrial function, and inducing oxidative stress, DNA damage, and early apoptosis 45 . Consistent with these findings, our results indicate that PFNA exhibited the strongest and most robust positive association with PCOS among long-chain PFAS, and remained the only long-chain PFAS retained in LASSO regression after accounting for correlations among PFAS. Furthermore, the BKMR analysis excluded certain long-chain legacy PFAS, including PFOA and PFOS. This exclusion does not imply that our findings indicate no association between PFOA, PFOS, and PCOS. On the contrary, the results from the Firth penalized maximum likelihood estimation method demonstrated a significant positive association between PFDA, PFNA, PFOA, PFOS, PFNS and the odds of PCOS in the long-chain legacy PFAS ( Table  3 ). However, due to multicollinearity among independent variables, LASSO regression was employed for variable selection. This analysis suggested that including additional variables improved the model fit. Consequently, the potential impact of long-chain legacy PFAS, such as PFOA and PFOS, on PCOS should not be overlooked. PFHpA and PFPeS are classified as short-chain PFAS, with PFHpA serving as an alternative to long-chain PFAS. An epidemiological study on PFAS and female reproductive hormones identified a significant impact of PFHpA on total testosterone levels in women 46 . Another population study suggested that preconception exposure to 6:2 fluorotelomer phosphate diester (6:2 diPAP) and PFHpA could impair fertility 47 . Although no direct evidence exists on the reproductive toxicity of PFPeS, a zebrafish study indicated that, among sulfonic aliphatic PFAS, chemical potency correlates with carbon chain length but not developmental neurotoxicity. Exposure to PFPeS, PFHxS, PFHpS, or PFOS was associated with shared toxicity phenotypes, including body axis and swim bladder defects and hyperactivity 48 . Additional evidence is needed to establish a definitive association between PFPeS and PCOS. This study has two important implications. First, our findings suggest that exposure to PFAS mixtures is associated with increased odds of PCOS, particularly with emerging PFAS alternatives. These results emphasizes the significant public health risks associated with the potential reproductive toxicity of emerging PFAS alternatives. Given the widespread environmental exposure to these chemicals, especially in regions with intensive industrial activities, their potential impact on female reproductive health warrants urgent attention. At the individual level, adopting safer consumer practices and undergoing regular health monitoring can help reduce exposure to emerging PFAS alternatives, particularly HFPO-DA (GenX), thereby mitigating potential health risks. Furthermore, our findings indicate that the exposure levels and detection rates of emerging PFAS alternatives remain high in Suzhou, Jiangsu Province, China. The specific details have been discussed in Sect. 4.2. These findings suggest substantial regional differences in exposure levels to emerging PFAS alternatives. Therefore, from a public health perspective, region-specific measures should be implemented to effectively regulate emerging PFAS alternatives and reduce local exposure levels. This study has several strengths. First, all PCOS cases were diagnosed by physicians based on clinical criteria, ensuring accurate diagnosis and consistent assessment. Additionally, self-reported menstrual volume was incorporated to address potential confounding related to menstruation. Second, both PCOS cases and healthy controls were nulliparous, eliminating potential confounding from childbirth on serum PFAS concentrations. Third, this study further explored the association between emerging PFAS alternatives and PCOS. Lastly, the LASSO model effectively addressed multicollinearity among PFAS mixtures, and the BKMR model identified key pollutants and predicted the nonlinear association between PFAS mixtures and PCOS odds. There are also several limitations to our study. First, the case-control design limits causal inference, as serum samples were collected after PCOS onset. This raises the possibility of reverse causation between PFAS exposure and PCOS development. Second, limited data availability and small sample size prevented participant matching between the PCOS and control groups, potentially introducing confounding bias. Third, the study population was restricted to women seeking care in obstetrics and gynecology clinics, which may result in selection bias. Last, short-chain PFAS with rapid metabolism and short half-lives may have been underestimated in serum relative to actual exposure levels. Future research should aim to expand sample size and include urinary PFAS measurements for a more accurate evaluation of their reproductive toxicity.

This study demonstrates an association between PFAS mixture exposure and increased risk of PCOS in women of reproductive age. FTSA (6/2), HFPO-DA (GenX), FBSA, PFNA, and PFPeS were positively associated with the odds of PCOS, whereas F-53B (8/2) showed a negative association. Specifically, HFPO-DA (GenX) exhibited the most significant contribution. These findings advance our understanding of the reproductive toxicity of emerging PFAS alternatives and provide critical evidence to guide public health policies on PFAS regulation.

Per- and polyfluoroalkyl substances (PFAS) are synthetic chemicals recognized for their environmental persistence. Due to their water- and oil-repellent properties, PFAS are extensively used in numerous consumer products 1 . PFAS exhibit high stability in both the environment and the human body, primarily due to the robust carbon-fluorine 2 . Specifically, traditional long-chain PFAS are known for their prolonged biological half-lives 1 , 3 . For example, perfluorooctane sulfonic acid (PFOS) and perfluorooctanoic acid (PFOA) are among the most prevalent long-chain PFAS, exhibiting biological half-lives of 5.4 years and 3.8 years, respectively 4 , 5 . Due to increasing public health concerns, many developed countries and regions have restricted or banned long-chain perfluorosulfonates 6 , 7 . This has indirectly led to the shift of long-chain PFAS production to emerging alternatives and short-chain PFAS, such as perfluorobutyric acid (PFBA) and chlorinated chlorinated perfluoroalkyl ether sulfonic (e.g., F-53B), which are short-chain homologs 8 , 9 . Although these alternatives are considered less bioaccumulative, recent studies have raised concerns regarding their potential endocrine-disrupting and reproductive toxicities. Experimental evidence suggests that both legacy and alternative PFAS can interfere with hormone secretion and follicular development, potentially impairing reproductive function 10 – 12 . Despite these findings, epidemiological evidence remains limited and inconsistent, particularly regarding the effects of alternative PFAS 13 , 14 . Polycystic ovarian syndrome (PCOS) is a prevalent endocrine disorder among women of reproductive age and one of the leading causes of infertility. Approximately 80% of women with anovulatory infertility can be attributed to PCOS 15 . The etiology of PCOS is multifactorial, involving genetic and environmental contributors, with endocrine-disrupting chemicals (EDCs) increasingly implicated as potential risk factors 16 . In vitro and in vivo studies indicate that PFAS may alter cytokine expression and disrupt the hypothalamic-pituitary-gonadal axis, suggesting plausible biological pathways linking PFAS exposure to PCOS-related phenotypes 11 , 17 . Currently, epidemiological research on the association between PFAS exposure and PCOS remains limited, and existing findings are inconsistent 18 – 20 . Research specifically focusing on emerging PFAS alternatives and their relationship with PCOS is even scarcer. To date, only one study suggests that 6:2 chlorinated polyfluoroalkyl ether sulfonic acid (F-53B(6/2)), perfluoro-2-propoxypropanoic acid (HFPO-DA (GenX)), PFOS, and perfluorododecanoic acid (PFDoA) are associated with an increased odds of PCOS 5 . However, a major limitation of current research lies in the insufficient quantification of individual PFAS species and the lack of consideration of their mixture effects. As PFAS production increasingly shifts to developing countries, especially China 21 , and the prevalence of polycystic ovary syndrome (PCOS) continues to rise globally, there is an urgent need for comprehensive studies evaluating the effects of emerging PFAS alternatives on female reproductive health. Moreover, PFAS exposure in the general population typically occurs as complex mixtures rather than individual compounds. However, most existing studies have overlooked these co-exposure patterns and their potential combined effects 19 . To address these knowledge gaps, this study aims to investigate the association between exposure to emerging PFAS alternatives and PCOS risk, and also evaluate the combined effects of PFAS mixtures.

Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2