Prediction of endometrial tissue PFAS concentrations using serum levels, demographic, and reproductive factors: Findings from the Investigating Mixtures of Pollutants and Endometriosis in Tissue (IMPLANT) study

Abstract

Background:

Per- and polyfluoroalkyl substances (PFAS) accumulate in reproductive tissues, yet most studies rely on serum measurements. It remains unclear whether serum concentrations adequately reflect levels in target tissues. We evaluated whether serum PFAS predicts endometrial tissue concentrations and examined associations with endometriosis.

Methods:

The Investigating Mixtures of Pollutants and Endometriosis in Tissue study included 433 reproductive-aged women undergoing laparoscopy or laparotomy in Salt Lake City, Utah, and San Francisco, California. Nine PFAS—perfluorooctane sulfonic acid (PFOS), perfluorohexane sulfonic acid (PFHxS), perfluorooctanoic acid (PFOA), perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA), perfluoroundecanoic acid (PFUnDA), perfluorododecanoic acid (PFDoDA), perfluoroheptanoic acid (PFHpA), and perfluorooctane sulfonamide (PFOSA)—were quantified in matched serum (ng/mL) and endometrial tissue (ng/g) using high-performance liquid chromatography with tandem mass spectrometry. Values below the limit of detection were multiply imputed. Linear regression estimated log-transformed tissue concentrations from serum PFAS and covariates. Logistic regression estimated odds ratios (ORs) for endometriosis. Composite variables were created by summing log-transformed concentrations of short-chain PFAS (<8 carbons: PFHpA, PFHxS, PFOA, and PFOS) and long-chain PFAS (≥8 carbons: PFNA, PFDA, PFUnDA, PFDoDA, and PFOSA).

Results:

Serum PFAS was moderately associated with tissue levels, with the strongest performance for PFOS (R² = 0.46) and PFHxS (R² = 0.37). Model fit was low for longer-chain compounds (e.g., PFUnDA R² = 0.09, PFDoDA R² = 0.07). Serum PFOA (OR = 3.38, 95% CI = 1.08, 10.59) and PFNA (OR = 3.12, 95% CI = 1.15, 8.45) were significantly associated with endometriosis, whereas corresponding tissue-based estimates were attenuated. The serum long-chain composite was associated with increased odds of endometriosis (OR = 2.74, 95% CI = 1.15, 6.59), while the short-chain composite showed no association.

Conclusions:

Serum PFAS minimally reflect endometrial tissue burden, with variability by chain length. While serum remains practical for large studies, tissue-based measures may provide more accurate estimates of exposure in tissue. These findings inform PFAS–endometriosis research and may extend to other gynecologic conditions.

What this study adds

This study is the first to directly compare serum and endometrial tissue polyfluoroalkyl substance (PFAS) concentrations in relation to endometriosis. Using prediction models and logistic regression in a surgically confirmed population, we show that serum PFAS levels moderately predict tissue concentrations, with variability by compound. Odds ratios for endometriosis differed by PFAS type and biospecimen, with some tissue-based estimates notably larger in magnitude. These findings demonstrate the potential of tissue-specific measurement and have implications for future etiologic research on endocrine-disrupting chemicals and endometriosis.

Introduction

Per- and polyfluoroalkyl substances (PFAS) are a large class of synthetic chemicals used in industrial and consumer products.^1,2 Their strong carbon-fluorine bonds make them highly resistant to degradation, resulting in significant environmental persistence.^3,4 PFAS are widespread in the environment and frequently detected in air, water, and soil. Human exposure primarily occurs through contaminated drinking water, food consumption, and occupational exposure.⁵ In the United States, PFAS were detected in approximately 98% of adults between 2003 and 2014.⁶ Although PFAS exposure is widespread in the United States, the levels of exposure are not uniform across individuals. Factors such as income, race and ethnicity, marital status, and age play a significant role in determining PFAS burden among women in the United States.^7,8 These differences may reflect variations in environmental exposures, occupational settings, use of consumer products,^9,10 and disparities in contaminated drinking water sources.¹¹

PFAS are typically measured in serum, which is easy to collect and reflects recent exposures. However, research suggests that PFAS also accumulates in various tissues, where concentrations may reflect longer-term exposures, depending on the PFAS half-life, protein-binding properties, and tissue characteristics.^12,13 Despite the widespread usage of serum as the proxy for exposure, it is unclear how well serum concentrations reflect PFAS in tissues that are relevant to specific health outcomes. This uncertainty raises concern that studies relying on serum measurements may introduce measurement error compared with levels at the relevant target tissue, potentially underestimating or mischaracterizing associations between PFAS and tissue-specific endometriosis.

Beyond their persistence, PFAS have been increasingly recognized as endocrine disruptors with potential implications for reproductive health. Although mechanisms are not fully understood, PFAS may disrupt estrogen signaling and contribute to metabolic and inflammatory pathways that affect reproductive outcomes.^14–16 Studies suggest that there is an association between PFAS and decreased fertility, increased endometriosis, altered pubertal timing, as well as an increase in hormonally sensitive cancers like breast and endometrial cancer.^17–20

Given its hormonal sensitivity and inflammatory nature, endometriosis is a particularly compelling condition to examine the potential effects of PFAS. Endometriosis is a chronic, estrogen-dependent inflammatory condition affecting ~10% of the female population at reproductive age.^21,22 It is characterized by the growth of endometrial tissue outside of the uterus. The current gold standard for diagnosis is laparoscopy, which is invasive, costly, and often delayed. The invasiveness also complicates the measurement of environmental exposures within the target tissue. If serum or another more accessible biospecimen could reliably reflect PFAS concentrations in endometrial tissue, it would improve the feasibility of studying these exposures.

Despite growing epidemiologic literature relating PFAS to endometriosis, several key gaps remain. Most prior studies have relied exclusively on serum PFAS concentrations, which are valuable for characterizing systemic burden and may capture exposure relevant to endocrine and inflammatory pathways implicated in endometriosis.^23–25 However, PFAS experience specific pharmacokinetics, protein binding, and tissue partitioning, raising concern that serum-based measures may misclassify exposure at the endometrial tissue level.^13,26 The extent to which serum PFAS concentrations serve as appropriate proxies for endometrial tissue exposure, and how any discordance may influence the associations with endometriosis, has not been well characterized.

To explore this, we developed models to estimate PFAS concentrations in endometrial tissue based on serum levels and individual demographic and reproductive characteristics, and compare associations with endometriosis between PFAS measured in serum and tissue. This approach aims to explore whether serum can serve as a reasonable proxy for tissue PFAS in endometriosis, where direct endometrial sampling is not feasible.

Methods

Study design and participants

The Investigating Mixtures of Pollutants and Endometriosis in Tissue (IMPLANT) study was designed to examine the relationship between serum and endometrial tissue PFAS concentrations in women undergoing laparoscopy or laparotomy for any indication. IMPLANT is nested within the operative cohort of the Endometriosis Natural History, Diagnosis, and Outcomes (ENDO) study, which included 433 women recruited from clinical centers in Salt Lake City, Utah, and San Francisco, California, between 2007 and 2009.

Eligible participants were aged 18–44 years, currently menstruating, and had no prior surgically confirmed diagnosis of endometriosis. Exclusion criteria included current pregnancy, breastfeeding within the past 6 months, postmenopausal status, recent use of injectable or systemic hormonal treatments (past 2 years), a history of cancer (except nonmelanoma skin cancer), and inability to provide informed consent.

Covariate collection

Trained interviewers collected demographic, reproductive, and lifestyle data using structured questionnaires. Serum samples were collected at a preoperative visit (approximately 2 months before surgery). Eutopic endometrial tissue was collected intraoperatively using standardized protocols. When ectopic endometrial-like tissue was identified, lesions were excised and biopsied; participants with ectopic tissue also provided eutopic samples. PFAS concentrations were quantified only in eutopic endometrial tissue; ectopic endometriotic lesions were not analyzed for PFAS in the present study.

Endometriosis status was determined at the time of surgery (laparoscopy or laparotomy) and has been described previously.²² Disease severity and subtype were classified by surgeons using the American Society for Reproductive staging system, with participants categorized as having superficial endometriosis, deep infiltrating endometriosis, or ovarian endometrioma. Individuals could have more than one of these subtypes.

Polyfluoroalkyl substances measurement

Serum and endometrial tissue concentrations of nine PFAS compounds: perfluorooctane sulfonic acid (PFOS), perfluoroheptanoic acid (PFHpA), perfluorooctanoic acid (PFOA), perfluorononanoic acid (PFNA), perfluorooctane sulfonamide (PFOSA), perfluorodecanoic acid (PFDA), perfluoroundecanoic acid (PFUnDA), perfluorododecanoic acid (PFDoDA), and perfluorohexane sulfonic acid (PFHxS), were quantified using high-performance liquid chromatography with tandem mass spectrometry (HPLC-MS/MS). Sample collection, handling, storage, preparation, analytical procedures, limits of detection, and quality control have been described in detail previously for the IMPLANT study.²⁷

Briefly, serum samples were collected at baseline, and eutopic endometrial tissue samples were collected at surgery from all participants, regardless of endometriosis status. PFAS were quantified using isotopic dilution with internal standards, with calibration curves demonstrating high linearity and quality control procedures confirming analytical accuracy and precision.

PFAS concentrations below the laboratory-specific limit of detection (LOD) were not reported as numeric values for tissue samples and were provided only as “<LOD” flags. These tissue values were recoded as missing and imputed using multiple imputation by chained equations. In contrast, for serum, concentrations below LOD were machine-reported as low numerical values and were retained as they were reported. Regression analyses were conducted within each imputed dataset, with estimates pooled across imputations using Rubin’s rules.

We summarized PFAS distributions using medians and interquartile ranges (IQRs) by endometriosis status. Wilcoxon rank-sum tests were used to compare distributions, and Tukey’s rule (1.5 × IQR beyond Q1 or Q3) was used to identify outliers. To evaluate concordance between matched serum and endometrial tissue concentrations, we generated scatterplots of log-transformed serum versus tissue PFAS concentrations overall and stratified by endometriosis status and calculated Spearman rank correlations for each PFAS. As a sensitivity analysis, correlations were recalculated after excluding PFAS-specific outliers defined using Tukey’s rule.

Prediction models for tissues polyfluoroalkyl substances using serum concentrations

We evaluated whether serum PFAS concentrations were associated with corresponding tissue levels using linear regression models. For each PFAS, we fit models within each imputed dataset with log-transformed serum concentrations as the primary predictor and natural log-transformed tissue concentrations as the outcome to account for right-skewness. Covariates included age, BMI (continuous), race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, Asian American and Pacific Islander, American Native and Alaska Native, and all other races), parity (nulliparous, primiparous, multiparous, and grand multiparous), menstrual phase (menstrual, follicular, ovulation, and luteal), household income (low [≤$49,999], middle [$50,000–$99,999], high [≥$100,000]), and current smoking status (yes/no). Menstrual phase classification was derived from the reported date of last menstrual period (LMP) and usual cycle length. Ovulation was approximated as occurring 14 days before the anticipated next menses, reflecting the relatively stable luteal phase duration of about 14 days.²⁸ Phases were defined as menstrual (bleeding days), follicular (LMP through the day before ovulation), ovulation (the day of ovulation), and luteal (ovulation through the day before the next menses). We acknowledge that menstrual phase classification based on LMP and cycle length is an approximation, as the fertile window is highly variable across women, even among those with regular cycles.²⁹ Race/ethnicity was included as a proxy for structural and social determinants of both exposure and diagnosis.

Stepwise model selection using Akaike Information Criterion (AIC) was applied, with forced retention of the serum PFAS predictor to maintain alignment with the study aim. Estimates were pooled across imputed datasets using Rubin’s rules. Model performance was assessed using root mean square error (RMSE) and R².

Because serum concentrations are reported in ng/mL and tissue concentrations in ng/g, we conducted sensitivity analyses to assess whether unit differences affected model performance. Specifically, we repeated models using log-transformed, z-score concentrations to place serum and tissue on comparable scales, and we converted tissue concentrations to ng/mL, assuming a soft tissue density of 1.05 g/mL.³⁰ We also repeated these conversions using alternative density values of 1.02–1.08 g/mL to reflect the plausible range reported in prior literature.^30,31

We created composite variables for short-chain PFAS (PFHpA) and long-chain PFAS (PFOA, PFNA, PFDA, PFUnDA, PFDoDA, PFHxS, PFOS, and PFOSA) by summing log-transformed concentrations in serum and tissue. Short-chain PFAS, defined as PFCAs with ≤7 carbons and PFSAs with ≤5 carbons, are less persistent yet remain prevalent in indoor and outdoor environments.³² Long-chain PFAS, defined as PFCAs with ≥8 carbons and PFSAs with ≥6 carbons, are more environmentally stable and bioaccumulative due to their longer hydrophobic chains.³³ Composite models were analyzed using the same procedures as individual PFAS.

This analysis uses matched serum and tissue concentrations to assess the degree to which serum levels reflect target tissue burden. Our framework builds on earlier ENDO Study work, which focused on serum PFAS and endometriosis risk.²³ Unlike Marroquin et al²⁷, whose primary analyses used tissue PFAS concentrations to evaluate associations with endometriosis_,²⁸ our main aim is to assess how well serum PFAS concentrations represent corresponding tissue levels.

While we also present logistic regression models of PFAS and endometriosis status, these serve as secondary, contextual analyses to compare tissue- versus serum-based associations, rather than as the central focus of our study.

To contextualize the prediction results, we examined associations between individual and composite PFAS concentrations (serum and tissue) and endometriosis status using logistic regression. Composite PFAS variables were constructed for short-chain and long-chain subclasses, as described above. Separate models were run for each composite and for each individual PFAS. Covariates were selected a priori based on prior literature and directed acyclic graphs and were retained in final models if they improved model fit based on AIC. Odds ratios (ORs) and 95% confidence intervals (CIs) were pooled across imputations and presented using forest plots.

This study was approved by the George Mason University Institutional Review Board (IRB #1325797-4). All participants provided written informed consent before enrollment.

Results

Demographic characteristics were generally similar by endometriosis status. The exception was parity, which differed significantly, with nulliparity more common among women with endometriosis (53.9%) than among those without (32.9%). Most participants identified as non-Hispanic White (76.0%), and smoking prevalence was low (14.6%). Half of the participants had household incomes between $40,000 and $99,999 (Table 1).

Table 1.

Sociodemographic participant characteristics in the operative cohort of (N = 433), the ENDO study: Endometriosis: Natural History, Diagnosis, and Outcomes

	Operative cohort (n = 433)	Endometriosis n = 181 (4.8%)	No endometriosis n = 252 (58.2%)
Characteristic	x̄ ± sd	x̄ ± sd	x̄ ± sd
Age at visit (years)	32 ± 7	32 ± 7	34 ± 7
BMI (kg/m2)	27.8 ± 7.8	26.3 ± 8.2	29.0 ± 7.0
Race/ethnicity	n (%)	n (%)	n (%)
Non-Hispanic white	329 (76.0)	136 (75.1)	193 (76.6)
Hispanic	57 (13.2)	23 (12.7)	34 (13.5)
Non-Hispanic black	6 (1.4)	1 (0.6)	5 (2.0)
AAPI/ANAIa	24 (5.6)	13 (7.2)	11 (4.4)
Multi-racial	8 (1.9)	6 (3.3)	2 (0.8)
All other race groups	9 (2.1)	0 (1.1)	7 (2.8)
Smoker
Yes	63 (14.6)	20 (11.1)	43 (17.1)
No	368 (85.4)	161 (89.0)	207 (82.1)
Missing	0	0	2
Household income
Low: <$10k to $39,999	122 (29.0)	40 (22.10)	82 (32.5)
Middle: $40k to $99,999	204 (48.5)	91 (50.28)	113 (44.8)
High: $100k	95 (22.6)	44 (24.31)	51 (20.2)
Missing	0	6	6
Parityb
Nulliparity	179 (41.3)	96 (53.04)	83 (32.9)
Primiparity	40 (9.24)	18 (9.94)	22 (8.7)
Multiparity	173 (40.0)	55 (30.39)	118 (46.8)
Grand multiparity	41 (9.5)	12 (6.63)	29 (11.5)
Missing	0	0	0
Menstrual phase
Follicular	71 (23.8)	36 (19.89)	35 (13.9)
Luteal	126 (42.1)	46 (25.41)	80 (31.8)
Menstrual	97 (32.4)	44 (24.31)	53 (21.0)
Ovulation	5 (1.7)	1 (0.55)	4 (1.6)
Missing	0	54	80

^aAsian American and Pacific Islander (AAPI)/American Native and Alaska Native (ANAI).

^bParity definitions: Nulliparity: no previous live births, Primiparity: one previous live birth, Multiparity: two to four previous live births, Grand multiparity: five or more previous live births.

Sd indicates standard deviation.

In serum, women with endometriosis tended to have slightly higher median PFAS concentrations than those without (Table 2). For example, PFOS (7.43 vs. 7.00 ng/mL), PFOA (2.90 vs. 2.37 ng/mL), and PFNA (0.70 vs. 0.60 ng/mL) were modestly elevated among cases. In contrast, PFHxS levels were nearly identical between groups. Other analytes, including PFHpA, PFDA, PFUnDA, PFDoDA, and PFOSA, followed similar patterns with small differences by endometriosis status. Detection frequencies varied by compound, with PFOS, PFOA, and PFHxS consistently detected and long-chain PFAS (PFDoDA and PFHpA) more often below the LOD.

Table 2.

Serum Per-and polyfluoroalkyl substances (PFAS) concentrations in the operative cohort (N = 433), the ENDO study: Endometriosis: Natural History, Diagnosis, and Outcomes

	Operative cohort (n = 433)	Endometriosis (n = 181, 41.80%)	No endometriosis (n = 252, 58.20%)
PFAS metabolites in serum (ng/mL)	Median (IQR)	Median (IQR)	Median (IQR)	% Below LOD
PFOS	7.11 (5.74)	7.43 (6.01)	7.00 (5.27)	0.23%
PFhxS	0.44 (0.42)	0.45 (0.41)	0.43 (0.44)	75.10%
PFOA	2.58 (1.84)	2.90 (2.10)	2.37 (1.73)	0.69%
PFNA	0.65 (0.49)	0.70 (0.47)	0.60 (0.48)	6.00%
PFDA	0.19 (0.15)	0.03 (0.02)	0.18 (0.12)	19.86%
PFUnDA	0.06 (0.12)	0.06 (0.13)	0.06 (0.11)	58.66%
PFHpA	0.05 (0.0)	0.06 (0.11)	0.05 (0.10)	72.52%
PFDoDA	0.03 (0.03)	0.03 (0.02)	0.02 (0.03)	84.76%
PFOSA	0.02 (0.02)	0.03 (0.02)	0.02 (0.02)	1.62%

When directly examining matched serum and tissue concentrations, tissue levels were generally higher across most PFAS (Tables 2 and 3). This pattern suggests potential bioaccumulation within endometrial tissue, particularly for long-chain compounds such as PFHxS and PFUnDA. While tissue PFOS concentrations (median 6.59 ng/g) were similar in magnitude to serum PFOS (7.11 ng/mL), serum concentrations tended to underestimate corresponding tissue values, especially at the higher end of the distribution. These differences emphasize the importance of considering tissue-specific burden, rather than assuming serum levels are sufficient proxies for internal dose.

Table 3.

Tissue per-and polyfluoroalkyl substances concentrations in the operative cohort (N = 433), the ENDO study: Endometriosis: Natural History, Diagnosis, and Outcomes

	Operative cohort (n = 433)	Endometriosis (n = 181, 41.80%)	No endometriosis (n = 252, 58.20%)
PFAS metabolites in tissue (ng/g)	Median (IQR)	Median (IQR)	Median (IQR)
PFOS	6.59 (6.44)	6.72 (7.04)	6.54 (6.27)
PFHxS	0.65 (0.75)	0.61 (0.79)	0.68 (0.73)
PFOA	1.94 (1.72)	1.95 (1.94)	1.93 (1.53)
PFNA	0.58 (0.52)	0.58 (0.54)	0.58 (0.50)
PFDA	0.17 (0.12)	0.17 (0.12)	0.17 (0.11)
PFUnDA	0.16 (0.13)	0.16 (0.13)	0.15 (0.13)
PFHpA	0.16 (0.21)	0.17 (0.18)	0.16 (0.20)
PFDoDA	0.21 (0.49)	0.19 (0.27)	0.22 (0.70)
PFOSA	0.16 (0.18)	015 (0.20)	0.16 (0.17)

Visual inspection of Figures 1 and 2 did not suggest that associations were driven by a small number of influential observations. In sensitivity analyses excluding PFAS-specific outliers, Spearman correlations were materially unchanged, indicating that observed serum-tissue relationships were not driven by extreme values (data not shown).

Figure 1.

Full Cohort Scatterplots of matched serum versus tissue PFAS with linear fits and R².

Figure 2.

Scatterplots of matched serum versus tissue PFAS concentrations comparing women with and without endometriosis, with linear fits.

Prediction models demonstrated that serum PFAS concentrations were only modestly associated with endometrial tissue concentrations (Table 4). The strongest models were for PFOS (R² = 0.46, RMSE = 0.24; predictor: serum PFOS) and PFHxS (R² = 0.37, RMSE = 0.31; predictors: serum PFHxS and BMI). PFOA (R² = 0.32, RMSE = 0.28; serum PFOA) and PFNA (R² = 0.24, RMSE = 0.28; serum PFNA with age) followed. For other analytes, additional covariates such as smoking, BMI, race/ethnicity, parity, and menstrual phase were retained (e.g., PFDA with serum PFUnDA, race/ethnicity, age, and smoking; PFUnDA with serum PFUnDA, smoking, BMI, and menstrual phase; PFHpA with serum PFHpA, race/ethnicity, smoking, and parity; PFOSA with serum PFOSA and smoking). Nonetheless, model fit was weak for PFHpA (R² = 0.09), PFDA (R² = 0.18), PFUnDA (R² = 0.09), PFOSA (R² = 0.03), and PFDoDA (R² = 0.07), indicating that even with covariate adjustment, serum concentrations were insufficient proxies for tissue levels of these compounds. Covariates were selected via stepwise AIC, with the serum PFAS predictor forced into all models.

Table 4.

Summary of predictive models for endometrial tissue per-and polyfluoroalkyl substances concentrations

Chemical	RMSE	R2	Predictors
PFOS	0.24	0.46	Serum_PFOS
PFHxS	0.31	0.37	Serum_PFHxS, BMI
PFOA	0.28	0.32	Serum_PFOA
PFNA	0.28	0.24	Serum_PFNA, age
PFDA	0.26	0.18	Serum_PFUnDA, race/ethnicity, age, smoke
PFUnDA	0.26	0.09	Serum_PFUnDA, smoking, BMI, menstrual phase
PFHpA	0.37	0.09	Serum pFHpA, race/ethnicity, smoking, parity
PFDoDA	0.50	0.07	Serum_PFDoDA
PFOSA	0.35	0.03	Serum_PFOSA, smoking

For each PFAS, the model with the highest R² and lowest RMSE is presented. Predictors were selected using stepwise AIC-based selection from serum PFAS concentrations and covariates including age, BMI, smoking, parity, menstrual phase, race/ethnicity, and income.

Finally, we compared serum- and tissue-based logistic regression models to assess associations with endometriosis (Table 5 and Figure 3). Serum PFOA (OR = 3.38, 95% CI = 1.08, 10.59) and PFNA (OR = 3.12, 95% CI = 1.15, 8.45) were significantly associated with higher odds of endometriosis. Corresponding tissue-based ORs were attenuated and not statistically significant (PFOA: OR = 1.00, 95% CI = 0.47, 2.16; PFNA: OR = 0.88, 95% CI = 0.37, 2.05). PFOS and PFHxS, which had the strongest serum-tissue model fit, were not significantly associated with endometriosis in either matrix. ORs for PFOS were 1.09 (95% CI = 0.42, 2.86) in serum and 0.63 (95% CI = 0.27, 1.22) in tissue. PFHxS estimates were 0.90 (95% CI = 0.39, 2.05) in serum and 0.59 (95% CI = 0.28, 1.22) in tissue.

Table 5.

Comparison of adjusted odds ratios for endometriosis diagnosis by log-transformed PFAS concentrations in serum and endometrial tissue

Chemical	Adjusted ORa	95% CI	Tissue	Serum
PFOS	0.63	0.27, 1.22	1.24	0.61, 1.86
PFHxS	0.59	0.28, 1.2	0.80	0.41, 1.55
PFOA	1.00	0.47, 2.16	3.38	1.08, 10.59
PFNA	0.88	0.37, 2.05	3.12	1.15, 8.45
PFDA	1.54	0.30, 7.89	1.80	0.72, 4.43
PFUnDA	2.10	0.37, 12.01	1.22	0.63, 2.35
PFHpA	2.25	0.88, 5.72	1.25	0.67, 2.35
PFDoDA	1.05	0.17, 6.55	0.96	0.33, 2.79
PFOSA	1.10	0.19, 6.16	1.16	0.44, 3.07
Short-chainb	2.25	0.88, 5.72	1.25	0.67, 2.35
Long-chainc	1.06	0.72, 1.57	2.74	1.15, 6.59

^aAdjusted for age at visit, BMI, race/ethnicity, parity, menstrual phase, smoke, household income.

^bShort-chain PFAS: PFHpA (perfluoroheptanoic acid).

^cLong-chain PFAS: PFOA (perfluorooctanoic acid), PFNA (perfluorononanoic acid), PFDA (perfluorodecanoic acid), PFUnDA (perfluoroundecanoic acid), PFDoDA (perfluorododecanoic acid), PFHxS (perfluorohexane sulfonic acid), PFOS (perfluorooctane sulfonic acid), PFOSA (perfluorooctane sulfonamide).

Figure 3.

Adjusted odds ratios for endometriosis by PFAS compound and biospecimen type.

For PFAS with weaker serum-tissue associations, such as PFHpA, PFDoDA, and PFOSA, estimates were imprecise and not statistically significant across both biospecimens (Table 5). Serum and tissue ORs for PFHpA were 1.25 (95% CI = 0.67, 2.35) and 2.25 (95% CI = 0.88, 5.72), respectively. PFDoDA had ORs of 0.96 (95% CI = 0.33, 2.79) in serum and 1.05 (95% CI = 0.17, 6.55) in tissue. PFOSA showed ORs of 2.49 (95% CI = 0.71, 8.81) in serum and 1.10 (95% CI = 0.19, 6.16) in tissue. PFDA and PFUnDA tissue-based ORs were 1.54 (95% CI = 0.30, 7.89) and 2.10 (95% CI = 0.37, 12.01), respectively, though confidence intervals were imprecise. Results are consistent with prior ENDO Study findings that identified significant associations between serum PFOA and serum PFNA and endometriosis risk.²²

Composite variables of serum and tissue concentrations were created to evaluate associations with endometriosis by PFAS chain length (Table 5). The short-chain composite, which included only PFHpA, was not significantly associated with endometriosis in serum (OR = 1.25, 95% CI = 0.67, 2.35) or tissue (OR = 2.25, 95% CI = 0.88, 5.72). In contrast, the long-chain composite, which included PFOA, PFNA, PFDA, PFUnDA, PFDoDA, PFHxS, PFOS, and PFOSA, was significantly associated with increased odds of endometriosis in serum (OR = 2.74, 95% CI = 1.15, 6.59), whereas the long-chain tissue composite showed no significant association (OR = 1.06, 95% CI = 0.72, 1.57).

Discussion

In this cohort of reproductive-aged women, serum PFAS concentrations were only modestly correlated with endometrial tissue levels, particularly for PFOS and PFOA. These correlations reflect unadjusted serum-tissue relationships and complement multivariable models, which evaluated the explanatory ability of serum PFAS for tissue levels after covariate adjustment. This distinction is important, as serum and tissue reflect different biological compartments and patterns of accumulation rather than one consistently being higher or lower than the other. Although serum remains a practical biospecimen for most epidemiologic studies, our findings suggest that it may not always serve as a reliable proxy for target tissue exposure.

We observed that endometrial tissue PFAS concentrations were often higher than serum concentrations, although the degree of difference varied by compound (Tables 2 and 3). For example, PFHxS (long-chain) and PFUnDA (long-chain) showed noticeably elevated tissue burdens relative to serum, whereas PFOS (long-chain) had comparable levels across matrices. These findings suggest that while tissue accumulation is evident for certain PFAS, the pattern is not determined by chain length alone. Differences in medians and distribution indicate that serum may underestimate tissue burden, particularly at higher exposure levels, emphasizing the need to evaluate both biospecimens when assessing exposure relevant to endometriosis.

PFAS are retained in blood due to their binding affinity to human serum albumin, a transporter protein found in blood.³⁴ Concentrations in blood rely on passive transport, or a single-compartment pharmacokinetic-pharmacodynamic model. Human cadaver studies show that different PFAS accumulate in kidney, liver, lung, and brain tissue at varying concentrations.¹³ At the cellular level, PFAS can diffuse into the phospholipid bilayer,³⁵ may interact with transport processes or accumulate within intracellular compartments, which could contribute to tissue-specific effects.³⁶ Passage into uterine tissue involves transport proteins, and permeability likely varies by PFAS chain length. For example, organic anion transporters are a means of facilitated transport³⁷ through which PFAS can pass,³⁸ and organic anion transporters are found in the endometrium.³⁹ Fatty acid binding protein (FABP) is similarly found in endometrium⁴⁰ and PFAS are ligands of liver fatty acid binding proteins.^41,42 PFOS can increase cell membrane fluidity, which may lead to certain organs becoming more permeable.⁴³ However, PFAS were below the LOD in mouse uterine tissue studies, although other organs contained detectable levels.⁴⁴ Tissue levels may be influenced by factors relevant to endometriosis, such as chronic inflammation and tissue remodeling, which could plausibly affect the local retention of PFAS compounds.³⁶ In addition, PFAS are endocrine disrupting chemicals that may influence hormonal and inflammatory pathways that are central to the pathogenesis of endometriosis. These mechanisms provide biologically plausible explanations for the differences in PFAS concentrations between serum and tissue.

This pattern is evident in previous evidence that PFAS accumulation differs according to chemical structure, functional group, and protein-binding, all of which influence absorption, distribution, and retention in tissues.^12,34 For instance, Ilieva et al²⁶ reported tissue-specific variability in PFAS concentrations, and in vitro work in human hepatocytes has demonstrated that structural differences affect cellular activity, reinforcing the role of PFAS-specific properties in tissue.⁴⁵ Together, these results highlight that serum and tissue reflect distinct aspects of internal dose, and tissue-specific concentrations may provide additional insights into mechanisms of PFAS toxicity.

Prediction models for PFOS and PFOA showed better performance compared with other PFAS. This may be partly due to their higher detection frequencies, broader value ranges, and stronger linear correlations between serum and tissue concentrations.²⁴ In contrast, PFHpA and PFDoDA had poorer performance, which may reflect greater analytic challenges, lower detection frequencies, or true biological variability in tissue retention.

We also observed distinct patterns in the associations of PFAS with endometriosis by chain length. Long-chain PFAS, which in our analysis included PFOA, PFNA, PFDA, PFUnDA, PFDoDA, PFHxS, PFOS, and PFOSA, were more strongly associated with endometriosis than short-chain PFAS (PFHpA). Long-chain PFAS tend to have longer biological half-lives, stronger protein-binding affinity, and greater potential for bioaccumulation in tissues such as the endometrium.³³ These characteristics may also enhance their endocrine-disrupting properties. Our findings are consistent with experimental studies showing increased tissue retention and persistence of long-chain PFAS relative to short-chain analogs.^12,24

Although the primary goal of this analysis was to evaluate concordance between serum and endometrial tissue PFAS concentrations, we also examined associations with endometriosis. In general, ORs estimated from logistic regression models using tissue concentrations followed similar directional trends as those based on serum, though the magnitudes varied. For example, tissue PFUnDA and PFOS were associated with over two-fold increased odds of endometriosis, although CIs were imprecise. Because we used multiple imputation for values below the LOD, our models retained the full sample size and minimized bias. While not all associations reached statistical significance, the consistency in direction underscores the importance of further investigation. Replicating this study in larger and more diverse populations would provide more conclusive evidence for the strong associations we observed in IMPLANT and help confirm the generalizability of our findings.

Scatterplots illustrated moderate linear relationships between serum and tissue PFOS and PFOA concentrations, the compounds with the highest detection frequencies. For some participants, serum concentrations appeared lower than corresponding tissue values, particularly for PFOS, PFOA, and PFNA. This suggests that serum may not always capture the full extent of tissue burden, underscoring the importance of examining both matrices.

This study has several strengths. First, it included both serum and endometrial tissue samples from the same participants, providing the opportunity to examine relationships between PFAS concentrations across biospecimens and their relevance to endometriosis. Second, we applied multiple linear and logistic regression models, adjusting for key demographic and reproductive variables, and used multiple imputation to address values below the LOD.⁴⁶ This approach follows best practices outlined in Harel et al⁴⁷, who caution that complete-case analysis can bias estimates and reduce efficiency, while multiple imputation improves statistical power, enhances precision, and accounts for uncertainty in the final estimates through Rubin’s rules.⁴⁸ Serum was collected approximately 2 months before tissue, but given the long biological half-lives of PFAS, concentrations would be expected to remain stable over this period. PFAS concentrations were quantified using highly sensitive LC-MS/MS, allowing for detection of low-level exposures and reducing the likelihood of exposure misclassification. Lastly, the cohort’s composition (women undergoing laparoscopic surgery for a variety of clinical indications) provided a surgically confirmed population with well-characterized endometriosis, improving diagnostic accuracy and increasing the relevance of our findings.

Several limitations should also be considered. Although we used multiple imputation for values below the LOD, measurement error remains possible, particularly for PFAS with very low detection frequency. Despite the use of imputation, limited variability in some PFAS may have affected model performance and interpretation. The cross-sectional nature of our data means we cannot assess temporality or account for critical windows of exposure that may influence the development of endometriosis. Although we adjusted for several confounders, residual confounding is possible, particularly for variables like diet, consumer product use, or occupational exposures, which were not measured. In addition, the IMPLANT study is nested within a surgically evaluated cohort, and women without endometriosis underwent laparoscopy or laparotomy for other gynecologic indications. As such, the comparison group does not represent a disease-free general population, which may limit generalizability beyond women undergoing surgical evaluation. If PFAS exposures are also associated with other gynecologic conditions present in the comparison group, this could reduce exposure contrast and attenuate observed associations. Variability in analytic performance across compounds may limit generalizability, particularly for PFAS with lower detection rates. Finally, because model performance was evaluated using R² without cross-validation, our estimated R² are overly optimistic. Yet our results demonstrated low predictive ability of serum for tissue PFAS, and cross-validation would not be expected to alter this conclusion.

Our results support the value of tissue-based measurements in environmental epidemiology. While serum remains a widely accessible and valuable biospecimen, it may not always reflect the full extent of PFAS accumulation in endometrial tissue. Direct comparisons between serum and tissue concentrations, as presented here, help our understanding of exposure misclassification and help identify scenarios where tissue measurement is essential. Future work should continue to explore these relationships in larger populations, with careful consideration of chain length, to better characterize how PFAS may influence endometriosis.

Conflicts of interest statement

The authors declare that they have no conflicts of interest with regard to the content of this report.

ACKNOWLEDGMENTS

We gratefully acknowledge the participants and study staff for their commitment to the study.