Exposure profiles of legacy and emerging per- and polyfluoroalkyl substances in follicular fluid and their impacts on assisted reproductive outcomes: A multicenter study in China

Abstract

Per- and polyfluoroalkyl substances (PFAS) are persistent environmental contaminants with potential reproductive toxicity, yet their profiles in follicular fluid and impacts on human fertility remain unclear. Given the importance of diet as an exposure pathway, we aimed to characterize PFAS in follicular fluid, examine associations with reproductive outcomes, and explore dietary contributors across regions in China. In this multicenter study involving 1301 women undergoing assisted reproductive technology in eight administrative divisions, 20 PFAS were quantified in follicular fluid. Reproductive outcomes included live birth, pregnancy loss, and early measures of ovarian hormones, endometrial thickness, and oocyte or embryo development. Associations were estimated using center-specific regression models combined by meta-analysis, with false discovery rate (FDR) correction. Both legacy and emerging PFAS were detected, with perfluorooctane sulfonate (PFOS) and perfluorooctanoic acid contributing 19%–40% and 23%–37% of total concentrations, respectively, and 6:2 chlorinated polyfluorinated ether sulfonate (6:2 Cl-PFESA) 11%–19%. Concentrations were higher in eastern and southern centers. Higher PFOS concentrations were significantly associated with an increased risk of biochemical pregnancy loss after FDR correction (RR per natural-log-unit increase: 2.10; 95% CI: 1.28, 3.45). Several legacy PFAS were linked to lower luteinizing hormone, elevated progesterone, and thinner endometrium on the day of human chorionic gonadotropin. Comparable associations were observed for emerging PFAS. Dietary intake of aquatic food and other animal-derived foods correlated positively with PFAS concentrations. These findings highlight PFAS accumulation in the ovarian microenvironment and potential disruption of early reproductive processes, underscoring the need for exposure mitigation through modifiable dietary pathways.

Graphical abstract

Highlights

• •Both legacy and emerging PFAS were detected in human follicular fluid.
• •Higher PFAS levels were observed in eastern and southern regions of China.
• •PFOS exposure increased biochemical pregnancy loss risk in ART women.
• •PFAS disrupted ovarian hormones and endometrial readiness during stimulation.
• •Aquatic and animal-derived foods were key dietary contributors to follicular fluid PFAS concentrations.

1. Introduction

Per- and polyfluoroalkyl substances (PFAS) have been widely used in industrial applications and consumer products for their water- and oil-repellent properties [1,2]. Legacy compounds, such as perfluorooctane sulfonate (PFOS) and perfluorooctanoic acid (PFOA), are highly persistent, bioaccumulative, and toxic [1,3,4]. Although phased out globally, they remain detectable in biological and environmental matrices [2,5,6]. Meanwhile, so-called emerging alternatives, including the increasingly used 6:2 chlorinated polyfluorinated ether sulfonate (6:2 Cl-PFESA) [7], were originally introduced as substitutes for regulated legacy PFAS but have been in use for over a decade and are now widely detected in environmental and human samples. This evolving exposure profile underscores the need for continued monitoring of both legacy and emerging PFAS [3,8].

Growing evidence suggests that PFAS can act as endocrine-disrupting chemicals (EDCs), interfering with hormonal regulation and adversely affecting female reproductive health [[9], [10], [11]]. Among legacy PFAS, PFOS and PFOA have been the most extensively studied, with epidemiological and experimental evidence linking them to disrupted folliculogenesis, impaired steroidogenesis, menstrual irregularities, and increased risks of subfertility and pregnancy loss, although empirical findings have been inconsistent across populations and exposure levels [[9], [10], [11], [12], [13]]. In contrast, the reproductive toxicity of emerging PFAS in humans remains largely unexplored. Notably, 6:2 Cl-PFESA has demonstrated reproductive and developmental toxicity similar to that of PFOS in both in vivo and in vitro studies [7,14], underscoring the need for further studies in humans.

Previous population-based studies have predominantly assessed PFAS exposure in blood or urine [10]. Follicular fluid offers a more biologically relevant matrix that directly reflects exposure in the ovarian microenvironment [15]. However, data on PFAS profiles in follicular fluid remain scarce and are largely derived from small-scale, single-center studies [[16], [17], [18], [19], [20], [21]]. Assisted reproductive technology (ART) provides a unique opportunity to evaluate a wide range of reproductive endpoints that are otherwise difficult to assess in the general population, including reproductive hormone levels, endometrial thickness, oocyte yield, fertilization rates, embryo quality, and both early and late pregnancy outcomes. Only a few small-scale studies have explored the relationship between PFAS concentrations in follicular fluid and reproductive outcomes of women undergoing ART, yielding inconsistent results [[19], [20], [21], [22], [23], [24]].

In addition to regional environmental contamination, individual lifestyle factors, particularly dietary habits, are recognized as modifiable contributors to PFAS accumulation [10,25,26]. Dietary intake is estimated to account for more than 90% of total PFAS exposure, surpassing inhalation and dermal absorption [27]. However, associations between diet and PFAS concentrations in follicular fluid remain poorly studied [17], limiting our understanding of exposure sources and potential intervention pathways.

We conducted a large multicenter study of women undergoing ART in geographically diverse regions in China [28]. Our objectives were to: (1) characterize the exposure profiles of legacy and emerging PFAS in follicular fluid across geographically diverse regions; (2) examine the associations between PFAS exposure and ART outcomes; and (3) identify dietary factors associated with PFAS concentrations to inform potential strategies aimed at mitigating PFAS-related reproductive risks.

2. Materials and methods

2.1. Study design and study population

We recruited women scheduled to undergo ART between February 2017 and January 2019 from ten medical centers across eight administrative divisions in China, covering eastern, southern, southwestern, central, and northern regions. The ten centers included: the Affiliated Suzhou Hospital of Nanjing Medical University in Suzhou, Jiangsu Province (Jiangsu-Suzhou, East China), Women's Hospital of Nanjing Medical University in Nanjing, Jiangsu Province (Jiangsu-Nanjing, East China), the Sixth Affiliated Hospital of Sun Yat-sen University in Tianhe district, Guangzhou, Guangdong Province (Guangdong-Tianhe, South China), the First Affiliated Hospital of Anhui Medical University in Anhui Province (Anhui, East China), Sun Yat-sen Memorial Hospital of Sun Yat-sen University in Yuexiu district, Guangzhou, Guangdong Province (Guangdong-Yuexiu, South China), Peking University Third Hospital in Beijing municipality (Beijing, North China), Reproductive and Genetic Hospital of CITIC-Xiangya in Hunan Province (Hunan, Central China), West China Second University Hospital of Sichuan University in Sichuan Province (Sichuan, Southwest China), Maternal and Child Health Hospital of Guangxi Zhuang Autonomous Region (Guangxi, South China), and the Third Affiliated Hospital of Zhengzhou University in Henan Province (Henan, Central China).

Upon enrollment, ART participants completed standardized, interviewer-administered baseline questionnaires, and were prospectively followed throughout the whole fertility treatment process, including ovarian response and endometrial thickness monitoring, oocyte retrieval, embryo transfer, serum β-human chorionic gonadotropin (β-hCG) testing, ultrasound confirmation of clinical pregnancy, prenatal visits during early, middle, and late pregnancy, and delivery. We collected participants’ follicular fluid samples collected on the day of oocyte retrieval at every cycle. Clinical information was abstracted from electronic medical records. The study protocol was approved by the Institutional Review Board of Nanjing Medical University ([2014]248), and written informed consent was obtained from all participants.

Women were eligible if they (1) completed at least one ART cycle by December 2019; (2) provided follicular fluid samples during their first oocyte retrieval cycle; (3) were ≤37 years of age, with a spouse aged ≤40 years; (4) had normal karyotypes for both partners; (5) did not use donor oocytes or sperm, and had a spouse without azoospermia; (6) completed baseline questionnaires; and (7) resided within the administrative division where their treatment center was located. Among the 3229 eligible participants, all women from centers with 300 or fewer eligible individuals were included, as most participating centers fell within this range. For centers with more than 300 eligible participants (Jiangsu-Suzhou and Henan), 150 women were randomly selected from each center to maintain balanced weighting across centers and ensure stable center-specific estimates. This sampling strategy yielded a final analytic sample of 1301 women (Fig. S1). To assess potential selection bias, we compared the characteristics between the study population and the remaining eligible women from centers with more than 300 participants, and observed generally similar characteristics between the two groups (Table S1).

2.2. Sample collection and PFAS assessment

Ovarian follicular fluid samples were collected on the day of oocyte retrieval after ovarian stimulation. Pooled follicular fluid from follicles measuring 14–24 mm in diameter was collected under ultrasound guidance and transferred into sterile polypropylene containers. Samples visibly contaminated with blood or flushing media were excluded. The remaining follicular fluid was centrifuged at 4000 rpm for 10 min at 4 °C to remove cellular debris, and the supernatant was aliquoted into sterile polypropylene tubes and stored at −80 °C until analysis.

Twenty target PFAS were quantified in follicular fluid, including 16 legacy PFAS [perfluorobutanoic acid (PFBA), perfluoropentanoic acid (PFPeA), perfluorohexanoic acid (PFHxA), perfluoroheptanoic acid (PFHpA), PFOA, perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA), perfluoroundecanoic acid (PFUnDA), perfluorododecanoic acid (PFDoDA), perfluorotridecanoic acid (PFTriDA), perfluorotetradecanoic acid (PFTeDA), perfluorobutane sulfonate (PFBS), perfluoropentane sulfonate (PFPeS), perfluorohexane sulfonate (PFHxS), perfluoroheptane sulfonate (PFHpS), PFOS] and four emerging PFAS [4:2 Cl-PFESA, 6:2 Cl-PFESA, 8:2 Cl-PFESA, 6:2 polyfluoroalkyl phosphate diester (6:2 diPAP)]. For sample preparation, 200 μL of follicular fluid was transferred into 1.5-mL centrifuge tubes. A mixture of isotopically labeled internal and extraction standards (5 μL, 100 ppb) was added, followed by 1 mL of acetonitrile. The samples were vortexed, sonicated for 15 min, shaken at 900 rpm for 15 min, and then centrifuged at 14,000 rpm for 15 min. The supernatant was transferred to a 15-mL tube. The extraction process was repeated with an additional 1 mL of acetonitrile, and the combined supernatants were evaporated to dryness under a stream of nitrogen. The residue was reconstituted in 200 μL of 50% methanol, shaken at 2000 rpm for 2 min, and centrifuged at 4400 rpm for 15 min. A 100-μL aliquot was transferred to an autosampler vial for instrumental analysis.

Concentrations of 20 PFAS were determined using ultra-performance liquid chromatography (UPLC) coupled with triple-quadrupole mass spectrometry (API 5500, AB SCIEX, Framingham, MA, USA), operated in negative electrospray ionization mode with multiple reaction monitoring. Chromatographic separation was achieved using a Poroshell 120 EC-C18 column (100 mm × 3 mm, 2.7 μm; Agilent, USA), with a binary mobile phase consisting of 5 mM ammonium acetate in water (A) and 5 mM ammonium acetate in methanol (B) at a flow rate of 0.4 mL/min (Table S2).

Rigorous quality control procedures were applied, including the injection of one extraction blank and two quality control (QC) samples at low and high concentrations every 20 samples. No PFAS were detected in the blanks. Intra- and inter-day coefficients of variation for all analytes were ≤15.3%. Matrix-spiked recoveries, absolute recoveries, and matrix effects were evaluated, within the acceptable ranges of 70%–120% for most PFAS except PFTeDA and 6:2 diPAP. Limits of quantification (LOQs) were defined as the concentration corresponding to either a signal-to-noise ratio of 10 in the matrix or the lowest calibrator with accuracy within ±20%, corrected for sample dilution or concentration. LOQs ranged from 0.01 to 0.05 ng/mL (Table S3).

2.3. ART outcome assessment

The primary reproductive outcomes were live birth and pregnancy loss at different stages, including implantation failure, biochemical pregnancy loss, and clinical pregnancy loss. Live birth was defined as the delivery of at least one live-born infant after 28 weeks of gestation. Implantation failure was defined as a negative serum β-hCG result following embryo transfer. Biochemical pregnancy loss was defined as a positive β-hCG result without subsequent ultrasound-confirmed clinical pregnancy. The rate of biochemical pregnancy loss was calculated among women with positive β-hCG results. Clinical pregnancy loss was defined as pregnancy loss after ultrasound-confirmed clinical pregnancy, and the clinical pregnancy loss rate was calculated among women with confirmed clinical pregnancy.

Secondary outcomes included early reproductive outcomes related to ovarian stimulation response and oocyte or early embryo development. Outcomes related to ovarian stimulation response included serum luteinizing hormone (IU/L), estradiol (pg/mL), and progesterone (ng/mL) levels, as well as endometrial thickness (mm) measured on the day of hCG administration in fresh cycles with egg retrieval—a critical time point that reflects ovarian hormonal responses to stimulation and the endometrial readiness for implantation [29,30]. Oocyte or early embryo development outcomes included total oocyte yield (number), oocyte maturation rate (%), fertilization rate (%), good-quality embryo rate (%), and blastocyst formation rate (%). These outcomes were calculated according to standardized definitions: oocyte maturation rate as the number of mature oocytes divided by total oocytes retrieved; fertilization rate as the number of two-pronuclei (2 PN) zygotes divided by mature oocytes; good-quality embryo rate as the number of high-grade embryos divided by cleaved embryos; and blastocyst formation rate as the number of blastocysts formed divided by embryos cultured.

2.4. Dietary factors and other covariates

Prior to ART initiation, standardized interviewer-administered questionnaires were used to collect information on sociodemographic characteristics (e.g., age, education, household income, and spouse's age), dietary intake, lifestyle behaviors (e.g., tobacco use and alcohol intake), anthropometric measures (weight and height), and clinical history (e.g., parity and reproductive disorders). Dietary intake was assessed using a previously validated semi-quantitative food frequency questionnaire (FFQ) that covered 25 dietary groups (Table S4) [31]. Participants reported both the frequency and amount per serving for each food item. The estimated daily intake (g/d) was calculated by multiplying the reported consumption frequency and the amount per serving.

Clinical characteristics related to ART included infertility diagnoses [e.g., polycystic ovary syndrome (PCOS), ovarian dysfunction, endometriosis, and sperm disorder], antral follicle count (AFC), cycle type (fresh vs. frozen), number of embryos transferred, and ovulation induction regimen [gonadotropin-releasing hormone (GnRH) agonist, GnRH antagonist, or other regimens]. Women with multiple infertility diagnoses were included in each relevant subgroup analysis.

2.5. Statistical analyses

Concentrations of individual PFAS below the LOQ were imputed as LOQ/2 [32]. We calculated the summed concentrations of 16 legacy PFAS (ΣLegacy PFAS), four emerging PFAS (ΣEmerging PFAS), and all 20 PFAS (ΣAll PFAS) for subsequent analyses. To characterize the distribution of PFAS concentrations, we reported detection rates (DRs, defined as the proportion of samples with values above the LOQ), geometric means (GMs), and percentile values. Only compounds with an overall DR > 80% were included in further regression analyses. Median concentrations were used to construct PFAS composition profiles stratified by study center. Spearman correlation coefficients were calculated to assess inter-correlations among individual PFAS.

Due to skewed distributions, PFAS concentrations were natural-log (ln)-transformed prior to analyses. Poisson regression models were used to examine associations between ln-transformed PFAS concentrations and pregnancy outcomes, including live birth and pregnancy loss at different stages (implantation failure, biochemical pregnancy loss, and clinical pregnancy loss), in centers with more than five outcome events. Risk ratios (RRs) and 95% confidence intervals (CIs) were reported. Generalized linear models were used to evaluate associations with continuous reproductive outcomes, including serum hormone levels (luteinizing hormone, estradiol, and progesterone) and endometrial thickness on hCG day. Quasi-Poisson regression models were used for count and proportion outcomes, including total oocyte yield, oocyte maturation rate, fertilization rate, good-quality embryo rate, and blastocyst formation rate. Regression coefficients were converted to percent changes using the formula: percentage change = (e^β − 1) × 100% [33]. Center-specific estimates were combined using random-effects meta-analysis implemented via the metafor package in R. To account for multiple testing, false discovery rate (FDR)-adjusted P values were calculated using the Benjamini-Hochberg procedure. Linear trends across tertiles of PFAS concentrations were tested to evaluate dose-response relationships.

Confounders were selected based on prior literature and included age, spouse's age, body mass index (BMI), education, household income, parity, tobacco use, and clinical factors such as PCOS, ovarian dysfunction, endometriosis, sperm disorder, AFC, cycle type, number of embryos transferred, and ovulation induction regimen, where applicable. Due to the small number of tobacco users, this variable was not included in multivariable models. Sensitivity analyses were performed by excluding women of advanced reproductive age (≥35 years). To assess potential effect modification by ovulation induction regimen (GnRH agonist vs. antagonist), we included interaction terms between PFAS exposure and regimen and conducted regimen-stratified analyses using data pooled from the ten study centers with adjustment for study center.

To address potential collinearity arising from co-occurring and highly correlated PFAS exposures, we applied Bayesian kernel machine regression (BKMR) to evaluate the joint effects of PFAS mixtures on reproductive outcomes using pooled data, including PFAS that showed associations at P < 0.10 in single-compound analyses [34]. All PFAS exposure variables were scaled and centered prior to analysis. BKMR models were used to estimate the overall joint effect of the PFAS mixture as well as the relative importance of individual PFAS components, quantified by posterior inclusion probabilities (PIPs). Models were fitted using 20,000 iterations of the Markov chain Monte Carlo (MCMC) sampler.

Finally, Spearman correlation analyses were performed to assess the associations between the daily intake of specific dietary groups and PFAS concentrations across study centers. All statistical analyses were performed using R version 4.3.2 (R Foundation for Statistical Computing), with a two-sided P value < 0.05 considered statistically significant.

3. Results

3.1. Baseline characteristics of study population

The study included 1301 women undergoing ART across ten medical centers in eight administrative divisions of China. The mean age of the women was 30.9 years, and the mean age of their spouses was 32.4 years. Most participants had a normal BMI (68.8%) and at least 12 years of education (71.9%). Approximately 60% reported an annual household income ≥100,000 Chinese yuan (CNY), and 90.8% were nulliparous. Only 1.9% (n = 25) reported ever using tobacco, and 2.7% (n = 35) reported consuming alcohol more than once per month (Table 1). Substantial heterogeneity in demographic and clinical characteristics was observed across centers. Mean age ranged from 29.8 years in the Jiangsu-Nanjing center to 32.7 years in the Beijing center. The proportion of participants with less than 12 years of education varied widely, from 7.9% in the Beijing center to 52.4% in the Guangxi center. Similarly, low-income proportions were highest in Guangxi and Henan centers.

Table 1.

Characteristics of study population across 10 centers [n (%), mean ± SD].

Characteristics	All	Jiangsu- Suzhou	Jiangsu- Nanjing	Guangdong-Tianhe	Anhui	Guangdong-Yuexiu	Beijing	Hunan	Sichuan	Guangxi	Henan
No. of women	1301	150	137	105	73	55	63	216	207	145	150
Baseline characteristics
Age, year	30.9 ± 3.1	30.3 ± 3.2	29.8 ± 3.3	31.6 ± 2.8	30.6 ± 3.3	32.2 ± 3.1	32.7 ± 2.6	30.7 ± 2.9	31.4 ± 2.7	31.1 ± 3.2	30.8 ± 3.0
Spouseʼs age, year	32.4 ± 3.6	31.6 ± 3.4	31.2 ± 3.9	33.1 ± 3.3	31.9 ± 3.6	33.3 ± 3.7	34.4 ± 2.8	32.2 ± 3.5	33.1 ± 3.4	32.7 ± 4.1	31.8 ± 3.3
BMI, kg/m2
<18.5	135 (10.4)	13 (8.7)	15 (10.9)	9 (8.6)	6 (8.2)	7 (12.7)	6 (9.5)	19 (8.8)	29 (14.0)	26 (17.9)	5 (3.3)
18.5–23.9	895 (68.8)	104 (69.3)	87 (63.5)	72 (68.6)	48 (65.8)	41 (74.5)	38 (60.3)	166 (76.9)	152 (73.4)	101 (69.7)	86 (57.3)
24–27.9	212 (16.3)	28 (18.7)	26 (19.0)	16 (15.2)	15 (20.5)	6 (10.9)	13 (20.6)	29 (13.4)	24 (11.6)	16 (11.0)	39 (26.0)
≥28	59 (4.5)	5 (3.3)	9 (6.6)	8 (7.6)	4 (5.5)	1 (1.8)	6 (9.5)	2 (0.9)	2 (1.0)	2 (1.4)	20 (13.3)
Education, year
<12	363 (27.9)	52 (34.7)	30 (21.9)	32 (30.5)	19 (26.0)	13 (23.6)	5 (7.9)	55 (25.5)	32 (15.5)	76 (52.4)	49 (32.7)
≥12	936 (71.9)	98 (65.3)	107 (78.1)	72 (68.6)	54 (74.0)	42 (76.4)	57 (90.5)	161 (74.5)	175 (84.5)	69 (47.6)	101 (67.3)
Household income, CNY
<50,000	141 (10.8)	6 (4.0)	14 (10.2)	5 (4.8)	11 (15.1)	4 (7.3)	1 (1.6)	19 (8.8)	3 (1.4)	48 (33.1)	30 (20.0)
50,000–100,000	377 (29.0)	40 (26.7)	33 (24.1)	32 (30.5)	28 (38.4)	10 (18.2)	8 (12.7)	62 (28.7)	35 (16.9)	65 (44.8)	64 (42.7)
100,000–200,000	449 (34.5)	53 (35.3)	54 (39.4)	32 (30.5)	26 (35.6)	25 (45.5)	15 (23.8)	88 (40.7)	93 (44.9)	24 (16.6)	39 (26.0)
≥200,000	317 (24.4)	51 (34.0)	36 (26.3)	34 (32.4)	8 (11.0)	16 (29.1)	24 (38.1)	47 (21.8)	76 (36.7)	8 (5.5)	17 (11.3)
Parity
Nulliparous	1181 (90.8)	137 (91.3)	126 (92.0)	97 (92.4)	67 (91.8)	50 (90.9)	63 (100.0)	199 (92.1)	199 (96.1)	120 (82.8)	123 (82.0)
Multiparous	120 (9.2)	13 (8.7)	11 (8.0)	8 (7.6)	6 (8.2)	5 (9.1)	0 (0.0)	17 (7.9)	8 (3.9)	25 (17.2)	27 (18.0)
Tobacco use	25 (1.9)	0 (0.0)	6 (4.4)	3 (2.9)	1 (1.4)	1 (1.8)	2 (3.2)	7 (3.2)	2 (1.0)	2 (1.4)	1 (0.7)
Alcohol intake	35 (2.7)	0 (0.0)	4 (2.9)	12 (11.4)	0 (0.0)	1 (1.8)	3 (4.8)	3 (1.4)	0 (0.0)	10 (6.9)	2 (1.3)
ART characteristics
Infertility diagnoses
PCOS	126 (9.7)	1 (0.7)	16 (11.7)	11 (10.5)	8 (11.0)	10 (18.2)	10 (15.9)	16 (7.4)	23 (11.1)	12 (8.3)	19 (12.7)
Ovarian dysfunction	170 (13.1)	14 (9.3)	11 (8.0)	21 (20.0)	17 (23.3)	8 (14.5)	22 (34.9)	16 (7.4)	26 (12.6)	24 (16.6)	11 (7.3)
Endometriosis	163 (12.5)	25 (16.7)	4 (2.9)	18 (17.1)	9 (12.3)	10 (18.2)	5 (7.9)	23 (10.6)	35 (16.9)	22 (15.2)	12 (8.0)
Sperm disorder	575 (44.2)	113 (75.3)	63 (46.0)	54 (51.4)	29 (39.7)	36 (65.5)	31 (49.2)	18 (8.3)	121 (58.5)	56 (38.6)	54 (36.0)
AFC	16.5 ± 9.0	16.5 ± 6.7	15.2 ± 4.9	14.9 ± 8.6	13.3 ± 6.1	17.6 ± 8.7	10.5 ± 4.9	23.5 ± 14.2	14.8 ± 6.5	14.1 ± 6.3	16.9 ± 6.7
Cycle type
Fresh	790 (60.7)	80 (53.3)	5 (3.6)	63 (60.0)	3 (4.1)	42 (76.4)	61 (96.8)	180 (83.3)	145 (70.0)	110 (75.9)	101 (67.3)
Frozen	511 (39.3)	70 (46.7)	132 (96.4)	42 (40.0)	70 (95.9)	13 (23.6)	2 (3.2)	36 (16.7)	62 (30.0)	35 (24.1)	49 (32.7)
No. of embryos transferred
One	455 (35.0)	63 (42.0)	28 (20.4)	34 (32.4)	57 (78.1)	12 (21.8)	3 (4.8)	45 (20.8)	43 (20.8)	109 (75.2)	61 (40.7)
Two or more	846 (65.0)	87 (58.0)	109 (79.6)	71 (67.6)	16 (21.9)	43 (78.2)	60 (95.2)	171 (79.2)	164 (79.2)	36 (24.8)	89 (59.3)
Ovulation induction regimen
GnRH-a	923 (70.9)	91 (60.7)	16 (11.7)	74 (70.5)	59 (80.8)	39 (70.9)	30 (47.6)	191 (88.4)	167 (80.7)	117 (80.7)	139 (92.7)
GnRHant	329 (25.3)	40 (26.7)	116 (84.7)	20 (19.0)	13 (17.8)	15 (27.3)	33 (52.4)	19 (8.8)	40 (19.3)	28 (19.3)	5 (3.3)
Others	49 (3.8)	19 (12.7)	5 (3.6)	11 (10.5)	1 (1.4)	1 (1.8)	0 (0.0)	6 (2.8)	0 (0.0)	0 (0.0)	6 (4.0)

AFC, antral follicle count; BMI, body mass index; GnRH-a, gonadotropin-releasing hormone agonist; GnRHant, gonadotropin-releasing hormone antagonist; PCOS, polycystic ovary syndrome; SD, standard deviation.

The prevalence of infertility diagnoses also differed by center: ovarian dysfunction ranged from 7.3% (Henan) to 34.9% (Beijing), and endometriosis from 2.9% (Jiangsu-Nanjing) to 18.2% (Guangdong-Yuexiu). Regarding ART procedures, 60.7% of participants underwent fresh embryo transfer, ranging from 3.6% in the Jiangsu-Nanjing center to 96.8% in the Beijing center. Single embryo transfer was performed in 35.0% of cycles, most frequently in Anhui (78.1%) and Guangxi (75.2%) centers. Use of a GnRH agonist protocol varied widely, from 11.7% in Jiangsu-Nanjing to 92.7% in Henan.

3.2. Characteristics of PFAS exposure

All 20 target PFAS were detected in follicular fluid samples. Nine legacy PFAS and two emerging PFAS had detection rates >80% and were included in further analyses (Table S5). The detection rates were generally consistent across centers. Among the remaining compounds, PFHxA had a detection rate of 88.7% only in the Jiangsu-Suzhou center, whereas PFHpA was detected in >80% of samples from several centers in the east (Jiangsu-Suzhou, Jiangsu-Nanjing, Anhui), south (Guangdong-Tianhe, Guangdong-Yuexiu), and north area (Beijing) in China, suggesting regional heterogeneity in the occurrence of specific PFAS. Among emerging PFAS, 6:2 diPAP had a detection rate of 94.3% in the Guangdong-Tianhe center, much higher than in other centers (Table S6), indicating potential local sources or usage patterns.

The overall median concentrations were 10.78 ng/mL for ΣAll PFAS, 8.86 ng/mL for ΣLegacy PFAS, and 1.68 ng/mL for ΣEmerging PFAS. PFOS, PFOA, and the emerging alternative 6:2 Cl-PFESA were the most abundant compounds, with median levels of 3.06, 2.84, and 1.58 ng/mL, respectively (Table S5). Marked regional differences were observed: Jiangsu-Suzhou center had the highest median concentration of ΣAll PFAS (42.22 ng/mL), more than twice that of Jiangsu-Nanjing (19.34 ng/mL), which had the second-highest level. Higher levels were also seen in centers from Guangdong and Anhui provinces (Fig. 1A). Similar regional trends were observed for ΣLegacy PFAS (Fig. 1B), ΣEmerging PFAS (Fig. 1C), and most individual compounds (Fig. S2).

Fig. 1.

Concentration distributions and composition profiles of PFAS in follicular fluid samples across 10 centers. (A) Total concentrations of all 20 PFAS combined (ΣAll PFAS). (B) Total concentrations of 16 legacy PFAS combined (ΣLegacy PFAS). (C) Total concentrations of 4 emerging PFAS combined (ΣEmerging PFAS). (D) Composition profiles of PFAS. (A−C) Box plots were colored blue for centers where PFOA predominated, and red for those where PFOS predominated. Concentration distributions of individual PFAS are shown in Fig. S2, and summary statistics are provided in Table S6. Jiangsu-SZ, Jiangsu-Suzhou; Jiangsu-NJ, Jiangsu-Nanjing; Guangdong-TH, Guangdong-Tianhe; Guangdong-YX, Guangdong-Yuexiu; 6:2 diPAP, 6:2 polyfluoroalkyl phosphate diester; Cl-PFESA, chlorinated polyfluorinated ether sulfonate; PFAS, per- and polyfluoroalkyl substances; PFBA, perfluorobutanoic acid; PFBS, perfluorobutane sulfonate; PFDA, perfluorodecanoic acid; PFDoDA, perfluorododecanoic acid; PFHpA, perfluoroheptanoic acid; PFHpS, perfluoroheptane sulfonate; PFHxA, perfluorohexanoic acid; PFHxS, perfluorohexane sulfonate; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctane sulfonate; PFPeA, perfluoropentanoic acid; PFPeS, perfluoropentane sulfonate; PFTeDA, perfluorotetradecanoic acid; PFTriDA, perfluorotridecanoic acid; PFUnDA, perfluoroundecanoic acid.

Composition profiles further demonstrated distinct regional patterns in dominant PFAS species. Legacy PFAS made up 80%–88% of total PFAS across all centers. The two legacy PFAS, PFOS and PFOA, accounted for 19%–40% and 23%–37% of the total PFAS, respectively, and were dominant in most centers, except in Jiangsu-Suzhou, where PFHxS contributed the highest proportion (23%). PFOA predominated in centers from Jiangsu, Beijing, Sichuan, and Henan, while PFOS dominated in the remaining centers, indicating region-specific PFAS signatures. 6:2 Cl-PFESA consistently ranked third, accounting for 11%–19% of the total PFAS across centers (Fig. 1D).

Spearman correlations showed moderate-to-strong intercorrelations (r = 0.47–0.94) among individual PFAS. Notably, several legacy long-chain perfluorocarboxylic acids (PFCAs; e.g., PFNA, PFDA, PFUnDA, PFDoDA) were strongly correlated (r ≥ 0.76), and PFOS was strongly correlated with its alternative, 6:2 Cl-PFESA (r = 0.83; Fig. S3).

3.3. Associations between PFAS exposure and pregnancy outcomes

The overall live birth rate in this study was 50.5%, with specific rates ranging from 41.5% to 66.7% across centers (Table S7). In each center, none of the associations between PFAS and live birth were statistically significant after FDR correction (Table S8). In the meta-analyses across all ten centers, PFAS exposure was likewise not significantly associated with live birth (Fig. 2), and the heterogeneity across centers was minimal (I² = 0−14%). Notably, four legacy PFAS (PFDA, PFDoDA, PFHpS, and PFOS) were positively associated with biochemical pregnancy loss, with the association for PFOS remaining significant after FDR correction (RR per ln-unit increase: 2.10; 95% CI: 1.28, 3.45). Similar associations were also noted for ΣLegacy and ΣAll PFAS, although these did not withstand FDR correction (Fig. 2, Table S9). No PFAS was significantly associated with implantation failure or clinical pregnancy loss (Fig. 2), with low heterogeneity across centers (I² = 0–44%; Tables S10–S11).

Fig. 2.

Associations between ln-transformed PFAS concentrations in follicular fluid samples and pregnancy outcomes among women undergoing ART, estimated from meta-analyses. All analyses were adjusted for age, spouse's age, BMI, education, household income, parity, PCOS, ovarian dysfunction, endometriosis, sperm disorder, cycle type, number of embryos transferred, and ovulation induction regimen. The results with statistical significance (FDR-adjusted P values < 0.05) are indicated with an asterisk (∗) in red. Associations between ln-transformed PFAS concentrations in follicular fluid and pregnancy outcomes among women undergoing ART across centers are reported in Tables S8–S11. ART, assisted reproductive technology; CI, confidence interval; FDR, false discovery rate; ln, natural log; RR, risk ratio; ΣAll PFAS, all 20 PFAS combined; ΣLegacy PFAS, 16 legacy PFAS combined; ΣEmerging PFAS, 4 emerging PFAS combined.

Consistently, when parametrized in tertiles, PFDA, PFOS, and ΣLegacy PFAS demonstrated significant dose–response relationships with the increased risk of biochemical pregnancy loss (P for trend <0.05; Fig. S4). Sensitivity analyses excluding women with advanced reproductive age yielded consistent results for PFHpS (RR: 1.96; 95% CI: 1.05, 3.65) and PFOS (RR: 2.12; 95% CI: 1.10, 4.09), suggesting that the main findings were robust (Table S12). Stratified and interaction analyses showed no evidence of effect modification by ovulation induction regimen (GnRH agonist vs. antagonist) for biochemical pregnancy loss or other pregnancy outcomes (all P for interaction >0.05; Table S13).

We further fitted the BKMR model to evaluate the joint effect of PFAS mixture (PFNA, PFDA, PFUnDA, PFDoDA, PFTriDA, PFHpS, and PFOS) on the risk of biochemical pregnancy loss (Fig. S5). The overall mixture effect suggested a positive but statistically non-significant association with the risk of biochemical pregnancy loss. All individual components of the PFAS mixture demonstrated considerable importance of the association (PIP > 0.50).

3.4. Associations between PFAS exposure and early reproductive outcomes

PFAS exposure was significantly associated with altered ovarian stimulation parameters (Fig. 3A). Two legacy long-chain PFCAs (PFDoDA, PFTriDA) were associated with reduced luteinizing hormone levels on the day of hCG administration, with PFDoDA showing a significant decrease after FDR correction (% Change: −8.28%; 95% CI: −13.27, −3.00). In contrast, higher concentrations of PFTriDA were associated with increased estradiol levels on hCG day (% Change: 9.78%; 95% CI: 0.78, 19.59), while it did not pass FDR correction. Several legacy PFAS and ΣLegacy PFAS were significantly associated with higher progesterone levels (% Change: 5.44% to 8.61%, e.g., PFNA, PFDA, PFUnDA, PFDoDA, PFTriDA, PFOS) and reduced endometrial thickness (% Change: −3.23% to −5.44%, e.g., PFOA, PFHxS, PFHpS) on hCG day after correction. Comparable associations were observed for the emerging alternative 6:2 Cl-PFESA (% Change: 6.29% for progesterone, −2.88% for endometrial thickness) and ΣEmerging PFAS (5.96% and −2.83%, respectively). Heterogeneity for these outcomes was generally low to moderate (I² = 0–63%; Tables S14–S17).

Fig. 3.

Associations between ln-transformed PFAS concentrations in follicular fluid samples and early reproductive outcomes among women undergoing ART, estimated from meta-analyses. (A) Outcomes related to ovarian stimulation response included luteinizing hormone on hCG day, estradiol on hCG day, progesterone on hCG day, and endometrial thickness on hCG day in fresh cycles with egg retrieval. (B) Oocyte or early embryo development outcomes included total oocyte yield, oocyte maturation rate, fertilization rate, good-quality embryo rate, and blastocyst formation rate. Analyses for luteinizing hormone on hCG day, estradiol on hCG day, and progesterone on hCG day were adjusted for age, BMI, education, household income, parity, PCOS, ovarian dysfunction, and ovulation induction regimen. Analyses for endometrial thickness on hCG day were adjusted for age, BMI, education, household income, parity, endometriosis, and ovulation induction regimen. Analyses for total oocyte yield and oocyte maturation rate were adjusted for age, BMI, education, household income, parity, PCOS, ovarian dysfunction, AFC, and ovulation induction regimen. Analyses for fertilization rate, good-quality embryo rate, and blastocyst formation rate were adjusted for age, spouse's age, BMI, education, household income, parity, PCOS, ovarian dysfunction, sperm disorder, AFC, and ovulation induction regimen. The results with statistical significance (FDR-adjusted P values < 0.05) are indicated with an asterisk (∗) in red. Associations between ln-transformed PFAS concentrations and early reproductive outcomes among women undergoing ART across centers are reported in Tables S14–S22. AFC, antral follicle count; hCG, human chorionic gonadotropin.

Meanwhile, meta-analyses showed that several legacy PFAS (PFOA, PFHxS, PFHpS) and ΣLegacy PFAS were significantly associated with a 5.49%−6.67% increase in total oocyte yield after FDR correction (Fig. 3B). Moderate-to-high heterogeneity was observed across centers, with larger effect estimates in Anhui and Guangdong-Yuexiu (Table S18). However, no significant associations were found for oocyte maturation rate, fertilization rate, good-quality embryo rate, and blastocyst formation rate (Fig. 3B), with low-to-moderate heterogeneity (I² = 0–46%; Tables S19–S22).

Dose-response trends across PFAS tertiles were generally consistent with the main findings for early reproductive outcomes (Figs. S6-S7). Sensitivity analyses confirmed the robustness of findings, though several associations lost significance after FDR correction (Table S23). Ovulation induction regimen modified the associations between several PFAS exposures and luteinizing hormone and progesterone levels, as well as endometrial thickness on the hCG day, with larger effect estimates observed among women undergoing GnRH antagonist protocols (Table S24). No effect modification was observed for oocyte or early embryo development outcomes, except for 8:2 Cl-PFESA in relation to oocyte maturation rate (Table S25).

Overall associations between PFAS mixtures and early reproductive outcomes are shown in Fig. S8. Higher PFAS mixture concentrations were associated with reduced endometrial thickness on the hCG day and increased total oocyte yield, with PFOA contributing most strongly to endometrial thickness (PIP = 0.98) and PFHpS to oocyte yield (PIP = 0.61). For luteinizing hormone and progesterone on the hCG day, mixture-response relationships were non-linear, with higher PFAS mixture levels associated with lower luteinizing hormone and higher progesterone at the upper exposure range, although these associations were not strictly monotonic and were accompanied by wide credible intervals, with contributions from multiple PFAS components.

3.5. Associations between dietary factors and PFAS concentrations

The distributions of dietary factors across centers were summarized in Table S26. Generally, dietary factors exhibited weak-to-moderate correlations with PFAS concentrations (r = −0.4 to 0.4; Figs. 4 and S9). In particular, the estimated daily intake of aquatic products (e.g., shrimp and crab, fish, mollusks, and shellfish) showed consistently positive associations with ΣAll PFAS concentrations in most centers (seven to nine centers), with the largest correlation coefficients reaching 0.3–0.4. Significant correlations were observed with shrimp and crab in six centers, fish in six centers, mollusks in two centers, and shellfish in three centers. These associations were largely consistent across both legacy and emerging PFAS. Additionally, the intake of animal organs and blood products was also positively associated with levels of ΣLegacy PFAS, ΣEmerging PFAS, and ΣAll PFAS in eight to nine centers, with significant correlations in one to two centers.

Fig. 4.

Correlations between dietary factors and the concentration of ΣAll PFAS, ΣLegacy PFAS, and ΣEmerging PFAS in follicular fluid samples among women undergoing ART across centers. Different colors represent Spearman correlation coefficients. P values < 0.05 were indicated with an asterisk (∗). Correlations between dietary factors and the concentration of individual PFAS are shown in Fig. S9.

4. Discussion

In this large multicenter study of women undergoing ART across geographically diverse regions in China, both legacy and emerging PFAS were detected in follicular fluid, with PFOS, PFOA, and 6:2 Cl-PFESA being the most abundant compounds. Substantial inter-center variation in PFAS concentration and composition was observed, particularly between industrialized eastern/southern regions and other areas. While no significant associations were identified with live birth, implantation failure, or clinical pregnancy loss, higher PFOS concentrations were significantly associated with an increased risk of biochemical pregnancy loss. In addition, elevated legacy PFAS concentrations were found to be significantly linked to altered ovarian hormonal responses, including reduced luteinizing hormone and increased progesterone, and reduced endometrial thickness. Notably, comparable associations were also observed for emerging PFAS, particularly with progesterone levels and endometrial thickness. Higher legacy PFAS exposure was additionally associated with increased total oocyte yield, but no significant associations were observed with oocyte maturation, fertilization, embryo quality, or blastocyst formation. Furthermore, we identified dietary factors, especially consumption of aquatic and other animal-derived foods, as significant contributors to follicular fluid PFAS concentrations, with consistent associations observed across study centers.

To our knowledge, this is the first multicenter biomonitoring study of PFAS in follicular fluid, a medium that reflects direct ovarian exposure. The follicular fluid concentrations of legacy PFCAs in our study were higher than those reported in previous small-scale studies. In comparison, we observed lower concentrations of both legacy and emerging PFAS than those in another study of 124 Chinese women [19]. The detection of substantial PFAS levels in ovarian compartments across diverse regions underscores widespread exposure in reproductive-age women in China, even after the global phase-out of such compounds. Notably, the overall PFAS composition profile in follicular fluid—characterized by the predominance of PFOS and PFOA—was broadly consistent with patterns reported in serum and plasma in population-based biomonitoring studies [[19], [20], [21]]. However, evidence from prior ART studies suggests that blood-follicular transfer is compound-specific rather than uniform, with some PFAS exhibiting only moderate correlations between serum and follicular fluid [19]. These findings indicate that follicular fluid does not simply mirror systemic exposure, but also reflects selective accumulation within the ovarian microenvironment.

A key strength of this multicenter study is the ability to characterize pronounced regional heterogeneity in both PFAS concentrations and composition profiles. Centers in the Yangtze River Delta (e.g., Jiangsu) and Pearl River Delta (e.g., Guangdong) had higher median PFAS concentrations in follicular fluid, aligning with known hotspots of PFAS production, use, and discharge [3,35,36]. This east-south gradient also mirrors regional exposure patterns previously reported in blood-based biomonitoring studies in China [3]. In addition to industrial sources, higher intake of aquatic products in eastern and southern coastal regions may further contribute to elevated PFAS exposure, as indicated by our dietary analyses, while regional differences in drinking water infrastructure and consumer product use may also play a contributory role [37]. Distinct region-specific PFAS signatures were also observed. PFOA contributed more heavily to total PFAS concentrations in Jiangsu, Beijing, Sichuan, and Henan, consistent with historical patterns of fluoropolymer manufacturing and secondary contamination from consumer products. In contrast, PFOS dominated the PFAS mixture in southern centers such as Guangdong and Guangxi, reflecting persistent environmental contamination in the Pearl River Delta [3,35,38]. While PFOS and PFOA were dominant in most regions, PFHxS contributed the largest proportion in Suzhou, likely due to its historical use as a PFOS replacement in local industries [39]. Notably, 6:2 Cl-PFESA and 8:2 Cl-PFESA—two sulfonated alternatives to PFOS—were detected in over 80% of follicular fluid samples, and 6:2 Cl-PFESA consistently ranked as the third most abundant PFAS across all centers, reflecting industrial shifts following regulatory restrictions on legacy compounds. Given its persistence and bioaccumulation potential [7,36,40], our findings underscore the need to broaden the surveillance scope beyond legacy PFAS.

Our finding that higher follicular fluid legacy PFAS concentrations, particularly PFOS, were associated with an increased risk of biochemical pregnancy loss warrants cautious interpretation, given the biological plausibility and consistency with some previous studies. This observation adds to the growing body of evidence suggesting that PFAS exposure may impair very early pregnancy establishment [9,13,19]. In the C8 Health Project, which investigated a population with exceptionally high environmental exposure to PFOA and PFOS through contaminated drinking water, elevated preconception serum PFOS concentrations were linked to increased odds of miscarriage [13]. Biochemical pregnancy loss is a sensitive marker of implantation failure or early embryonic demise and may be particularly susceptible to endocrine and immune disruptions at the maternal-fetal interface [41,42]. Evidence from smaller ART cohorts has been inconsistent, with studies in China and Sweden reporting no significant associations between PFAS exposure and pregnancy outcomes [19,21]. Such discrepancies likely reflect differences in exposure levels, outcome definitions, and biological matrices. Mixture-based BKMR analyses showed a positive, though non-significant, joint association between combined PFAS exposure and biochemical pregnancy loss, supporting the relevance of cumulative exposure in very early pregnancy. By directly quantifying PFAS in follicular fluid, our study uniquely captured local ovarian exposure, providing more biologically relevant evidence of potential risks.

The underlying biological mechanisms linking PFAS exposure to adverse pregnancy outcomes remain incompletely understood. Experimental studies suggest that PFAS may disrupt folliculogenesis, steroid hormone production, and endometrial receptivity—key processes underlying implantation and early embryonic development [10,11,43,44]. Consistent with these pathways, we observed that higher legacy PFAS exposure was associated with reduced luteinizing hormone levels, elevated estradiol and progesterone levels, and thinner endometrial lining—endocrine changes that may reflect dysregulated granulosa cell function and impaired endometrial preparation. Although modest in magnitude, these changes may still represent biologically meaningful perturbations of early reproductive processes, particularly in the context of the widespread and cumulative nature of PFAS exposure, rather than direct predictors of ART success [29,30].

Notably, these findings contrast with prior endocrine disruption studies reporting predominantly suppressive effects of PFAS on steroidogenesis under physiological conditions [10,11,45]. In the context of controlled ovarian stimulation, however, PFAS-induced dysregulation may manifest differently due to pharmacologically elevated gonadotropin levels and altered follicular dynamics. We further observed effect modification by ovulation induction regimen, with stronger PFAS-associated alterations in luteinizing hormone, progesterone levels, and endometrial thickness among women undergoing GnRH antagonist protocols. Compared with agonist regimens, antagonist protocols are characterized by more abrupt suppression of endogenous gonadotropin secretion and altered pituitary-ovarian feedback, potentially rendering the endocrine milieu more susceptible to environmental perturbations [46]. In this setting, PFAS-related endocrine disruption may be amplified, leading to more pronounced hormonal and endometrial responses. Experimental evidence suggests that PFAS can interfere with early folliculogenesis by impairing communication between oocytes and granulosa cells, inducing oxidative stress, and modulating peroxisome proliferator-activated receptor (PPAR) signaling, thereby disrupting ovarian steroidogenesis and potentially altering gonadotropin receptor expression and downstream luteinizing hormone signaling [10]. Such perturbations could contribute to compensatory changes in progesterone production and a hormonal milieu less favorable for endometrial proliferation and receptivity, consistent with the observed reduction in endometrial thickness [29,30]. Importantly, emerging PFAS demonstrated biological effects comparable to those of legacy compounds, consistent with recent in vivo and in vitro evidence for 6:2 Cl-PFESA-related endocrine and developmental toxicity [7,14]. Supporting the single-compound findings, mixture-based BKMR analyses showed concordant overall patterns for endometrial thickness, suggesting that combined PFAS exposures may contribute to these early endocrine-endometrial perturbations.

In addition to hormonal alterations, we observed that higher legacy PFAS concentrations were associated with increased oocyte yield, while no associations were found for oocyte maturation, fertilization, embryo quality, or blastocyst formation. This pattern suggests that PFAS may promote follicular recruitment or responsiveness without improving oocyte or embryo competence, which is consistent with prior findings showing increased follicle numbers but poorer embryo quality in some studies [21,24]. Given the observational nature of this study, these associations should be interpreted as indicative rather than causal. Mechanistically, PFAS may influence early folliculogenesis and follicle selection by interfering with granulosa cell function, steroidogenic signaling, and metabolic or inflammatory pathways, thereby altering follicular sensitivity to exogenous gonadotropins during ovarian stimulation [10,21]. We observed moderate heterogeneity across centers for total oocyte yield, with larger effect estimates in certain centers, such as Anhui and Guangdong. These centers are located in regions with relatively higher levels of industrialization and PFAS exposure, which may contribute to more pronounced ovarian responses under stimulation. In addition, ovaries with intrinsically altered or dysregulated responsiveness, such as those observed in women with ovarian dysfunction or PCOS [18], may be more susceptible to PFAS-related perturbations, potentially leading to increased follicular recruitment without corresponding improvements in oocyte or embryo competence. Together, these results highlight the complexity of PFAS reproductive effects and underscore the need for future mechanistic and experimental studies to clarify whether and how PFAS exposure influences folliculogenesis, oocyte competence, and ovarian endocrine regulation, and to determine the clinical relevance of these associations for ART success.

Our study also provides novel evidence linking dietary intake to PFAS levels in follicular fluid. Aquatic foods, particularly shrimp, crab, fish, and mollusks, were consistently associated with higher legacy PFAS concentrations across multiple centers, reinforcing their role as major dietary sources of exposure [17,47]. These findings are in line with previous population-based biomonitoring studies showing similar associations between aquatic food consumption and PFAS concentrations in serum and plasma [48,49], reflecting bioaccumulation along aquatic food chains. The intake of animal organs and blood products was also positively associated with legacy PFAS levels in several regions, likely reflecting PFAS bioaccumulation in protein-rich tissues such as liver and blood. Notably, dietary intake patterns associated with legacy PFAS were also linked to emerging PFAS levels, reinforcing the likelihood of co-exposure through shared food chains or packaging sources [25]. Taken together, the concordance of dietary-PFAS associations observed in follicular fluid and blood-based matrices supports the biological plausibility of our findings and suggests that diet-related PFAS exposure in systemic circulation is also reflected in the ovarian microenvironment, highlighting diet as an important and potentially modifiable determinant of PFAS exposure.

Our study has several strengths. We directly quantified a broad panel of legacy and emerging PFAS in follicular fluid using highly sensitive and validated analytical methods. The prospective design, comprehensive collection of lifestyle and clinical data, and systematic evaluation of multiple clinically meaningful assisted reproductive outcomes enhanced the robustness of our findings. The multicenter recruitment across geographically diverse regions further improved external validity and generalizability.

Certain limitations should be acknowledged. First, although we measured a wide range of PFAS, we cannot exclude the potential influence of unmeasured analogues or co-exposures to other environmental chemicals such as phthalates or bisphenols, or the possibility that combined or mixture effects may contribute to the observed reproductive outcomes. Second, residual confounding remains possible despite extensive adjustment for factors including spouse age and sperm disorders, due to other unmeasured or incompletely characterized covariates. Third, PFAS concentrations were measured at a single time point, which may not fully reflect cumulative or long-term exposure and could be influenced by short-term variability in exposure sources or ART-related hormonal environments. Nonetheless, any resulting exposure misclassification is likely to be non-differential with respect to outcome status. Future studies incorporating repeated PFAS measurements across multiple ART cycles and biological matrices are warranted to better characterize intra-individual variability and cumulative exposure. Finally, the small number of pregnancy loss events in some centers may have limited statistical power for center-specific analyses.

5. Conclusions

Both legacy and emerging PFAS were commonly detected in the ovarian microenvironment of women undergoing ART in China, with substantial regional variation in concentration and composition. Higher PFOS concentrations were significantly associated with an increased risk of biochemical pregnancy loss, while several legacy PFAS were linked to altered ovarian hormonal responses, thinner endometrial lining, and increased oocyte yield, but not with oocyte maturation, fertilization, embryo quality, or blastocyst formation. Emerging PFAS also accounted for a meaningful share of total ovarian PFAS exposure levels and showed comparable associations with these early reproductive endpoints, highlighting the need for ongoing monitoring of both legacy and emerging PFAS in relation to reproductive health. Importantly, the consistent associations observed with dietary intake of aquatic products and other animal-derived foods point to potential exposure pathways, which may inform precautionary dietary counseling for women undergoing ART and underscore the importance of environmental and public health policies aimed at reducing PFAS contamination in key food sources. These findings support the need for large-scale, longitudinal studies to confirm causality and inform targeted prevention strategies.

CRediT authorship contribution statement

Hong Lv: Conceptualization, Formal analysis, Funding acquisition, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing. Shujuan Ma: Investigation, Resources, Writing – original draft, Writing – review & editing. Xiaohong Li: Investigation, Resources, Writing – original draft, Writing – review & editing. Qingxia Meng: Investigation, Resources, Writing – original draft, Writing – review & editing. Yichun Guan: Investigation, Resources, Writing – original draft, Writing – review & editing. Zhazheng He: Validation, Writing – original draft, Writing – review & editing. Bo Zhang: Investigation, Resources, Supervision. Xiufeng Ling: Investigation, Resources, Supervision. Xiaoyan Liang: Investigation, Resources, Supervision. Yunxia Cao: Investigation, Resources, Supervision. Chan Tian: Investigation, Resources, Supervision. Qingxue Zhang: Investigation, Resources, Supervision. Yangqian Jiang: Data curation, Methodology, Validation. Lei You: Data curation, Funding acquisition, Methodology, Validation. Yuanyan Dou: Data curation, Methodology, Validation. Jinghan Wang: Data curation, Methodology, Validation. Kang Ke: Data curation, Methodology, Validation. Xin Xu: Data curation, Methodology, Validation. Ganchong Liao: Data curation, Methodology, Validation. Yufan Jin: Data curation, Methodology, Validation. Kun Zhou: Formal analysis, Methodology, Visualization. Xiaoyu Liu: Formal analysis, Methodology, Visualization. Xiumei Han: Formal analysis, Methodology, Visualization. Bo Xu: Formal analysis, Methodology, Visualization. Tao Jiang: Formal analysis, Methodology, Visualization. Jiangbo Du: Formal analysis, Methodology, Visualization. Hongxia Ma: Funding acquisition, Project administration, Supervision, Writing – review & editing. Guangfu Jin: Project administration, Supervision, Writing – review & editing. Yankai Xia: Project administration, Supervision, Writing – review & editing. Jiong Li: Funding acquisition, Project administration, Supervision, Writing – review & editing. Hongbing Shen: Project administration, Supervision, Writing – review & editing. Yuan Lin: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing – review & editing. Jiayin Dai: Conceptualization, Methodology, Resources, Supervision, Writing – review & editing. Zhibin Hu: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing – review & editing.

Declaration of competing interests

The authors have declared no conflicts of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article: