Association between per- and polyfluoroalkyl substances exposure and thyroid function biomarkers among females attending a fertility clinic

Abstract

Per- and polyfluoroalkyl substances (PFAS) exposure was associated with changes in thyroid function in pregnant mothers and the general population. Limited such evidence exists in other susceptible populations such as females with fertility problems. This cross-sectional study included 287 females seeking medically assisted reproduction at a fertility clinic in Massachusetts, United States, between 2005–2019. Six long-alkyl chain PFAS, thyroid hormones, and autoimmune antibodies were quantified in baseline serum samples. We used generalized linear models and quantile g-computation to evaluate associations of individual PFAS and their total mixture with thyroid biomarkers. Most females were White individuals (82.7%), had graduate degrees (57.8%), and nearly half had unexplained subfertility (45.9%). Serum concentrations of all examined PFAS and their mixture were significantly associated with 2.6% to 5.6% lower total triiodothyronine (TT3) concentrations. Serum concentrations of perfluorononanoate (PFNA), perfluorodecanoate (PFDA), and perfluoroundecanoate (PFUnDA), and of the total mixture were associated with higher ratios of free thyroxine (FT4) to free triiodothyronine (FT3). No associations were found for PFAS and TSH or autoimmune antibodies. Our findings support the thyroid-disrupting effect of long alkyl-chain PFAS among a vulnerable population of subfertile females.

Graphical Abstract

Introduction

Thyroid hormones play major roles in the metabolism, growth, and development of the human body.¹ Thyroid disease is prevalent in the United States (U.S.), with a 7% prevalence for hypothyroidism,² 1% for hyperthyroidism,³ and up to 10% of people have signs of thyroid autoimmunity.⁴ Females are four times more likely to have thyroid disease than males.^{4, 5} The female reproductive system is regulated by thyroid hormones. Thyroid disease has been linked to infertility, implantation failure, miscarriage, and lower ovarian reserve.⁶ The major cause of hypothyroidism and hyperthyroidism in women of reproductive age is thyroid autoimmunity.⁷ The cause of thyroid autoimmunity is multifactorial and to a large extent unclear. However, environmental factors such as endocrine disrupting chemicals exposure are hypothesized in its cause.^{3, 8}

Per- and polyfluoroalkyl substances (PFAS) are man-made chemicals with high production volume worldwide.⁹ They are widely applied in consumer products including cookware and utensils, cosmetics, food packaging materials, and fabric for their water/oil-resistant properties.⁹ These properties on the other hand make PFAS extremely persistent in the environment; many also persist in the human body (half-lives of long alkyl-chain PFAS ranged over 3 years).⁹ PFAS exposure has been shown to have endocrine disrupting effect and can disrupt female reproductive system including dysregulation of female reproductive hormones, abnormal menstrual cyclicity, and prolonged time to pregnancy.¹⁰

The underlying mechanism of the association between PFAS exposure and female reproductive disease is hypothesized to be partially via disruption on thyroid function.¹⁰ In experimental studies, PFAS can accumulate in the thyroid gland, have cytotoxic and genotoxic thyroidal effects, together with alteration of central and peripheral thyroid homeostasis.^{11, 12} Epidemiological studies have reported that PFAS exposure is associated with changes in thyroid hormone levels in the U.S. general population, though the direction of associations is inconsistent.^{13, 14} There is similar evidence linking prenatal PFAS exposure to changes in thyroid biomarkers in pregnant women and their offspring.^{14, 15}

Women who experience difficulties in getting pregnant may have underlying medical susceptibilities that make them more vulnerable to the effects of endocrine disrupting chemicals,¹⁶ and they are also more likely to have thyroid problems.¹⁷ However, to date, no study evaluated whether PFAS can influence thyroid function among women with subfertility. Additionally, investigating effect heterogeneity by subfertility cause is important to identify vulnerable groups and disentangle the mechanisms of PFAS on thyroid functions in relation to fertility. This study aimed to evaluate if serum PFAS concentrations were associated with thyroid function biomarkers including thyroid hormones and thyroid autoimmune antibodies among females attending a U.S. fertility clinic from 2005 to 2019.

Methods

Population

This study included females from the Environment and Reproductive Health (EARTH) Study, which is a prospective preconception cohort that recruited male and female participants who sought fertility evaluation and medically assisted reproduction at the Massachusetts General Hospital Fertility Center, U.S. between 2005 to 2019. A detailed description of the EARTH study can be found elsewhere.¹⁸ Briefly, females aged 18–45 years were eligible to participate in the study. Participants provided detailed information on demographics, socioeconomics, and reproductive history through questionnaires, underwent anthropometric measurements, and provided a non-fasting blood sample at study baseline.

The current study included females who did not report having diagnosis of hypo- or hyper-active thyroid, not using thyroid interfering medication, and had their baseline blood samples quantified for both PFAS concentrations and thyroid function biomarkers. A participant inclusion flowchart is presented in Figure S1.

Ethics approval of the EARTH study was obtained from the Institutional Review Boards at Harvard T.H. Chan School of Public Health. The Centers for Disease Control and Prevention (CDC) determined the analysis of de-identified specimens at the CDC laboratory not to constitute human subjects research.

Exposure assessment

Non-fasting blood samples were collected at study baseline, centrifuged, and separated into serum aliquots which were stored in polypropylene cryovials at − 80 °C. In 2021, we shipped serum samples (of females who still had sufficient bio-banked serum left) overnight on dry ice to Centers for Disease Control and Prevention (CDC) for quantification of nine PFAS, including perfluorohexane sulfonate (PFHxS), linear perfluorooctane sulfonate (n-PFOS), sum of perfluoromethylheptane sulfonate isomers (sm-PFOS), linear perfluorooctanoate (n-PFOA), sum of branched PFOA isomers (sb-PFOA), perfluorononanoate (PFNA), perfluorodecanoate (PFDA), and perfluoroundecanoate (PFUnDA), using online solid phase extraction–high-performance liquid chromatography–isotope dilution tandem mass spectrometry.¹⁹ The limit of detection (LOD) was 0.1 ng/mL for all PFAS. We calculated serum PFOA and PFOS concentrations as the sum of their respective isomers, i.e., PFOA = n-PFOA + sb-PFOA, PFOS = n-PFOS + sm-PFOS, and used these sums in the statistical analyses. Concentrations below LOD were imputed with the LOD divided by the square root of 2.²⁰

Outcome assessment

Baseline serum samples were shipped on dry ice to the Department of Clinical Chemistry, Máxima Medical Center (Veldhoven, the Netherlands) in 2016, for quantification of thyroid function biomarkers [thyroid-stimulating hormone (TSH), free and total thyroxine (FT4, TT4) and triiodothyronine (FT3, TT3)], and two thyroid antibodies [thyroid peroxidase (TPO) and thyroglobulin (Tg) antibodies (TPOAb and TgAb, respectively)], using electrochemiluminescence immunoassays (Cobas^® e601 platform, Roche Diagnostics).²¹ The between-run coefficients of variation for the assays were 2% to 8% for thyroid function biomarkers and 7% to 12% for autoimmune antibodies.²² Autoimmune thyroid antibody was considered positive if >35 IU/mL for TPOAb and >115 IU/mL for TgAb.

Covariates

Demographic and socioeconomic data were self-reported by questionnaires, and included age, education level (less than graduate degree, graduate degree), race/ethnicity (White, Black, Asian, Others), and smoking status (never smoker, current or former smoker). Trained study staff measured the weight and height of participants at baseline. Body mass index (BMI) was calculated as weight in kilograms (kg) divided by square height in meters (m²). Overweight was defined as BMI ≥ 25 kg/m². Couples’ cause of subfertility was determined by the treating physician and categorized as male factor, female factor, or unexplained, based on definitions by the Society for Assisted Reproductive Technology (ART).²³

Statistical analyses

We conducted descriptive analyses for the study population and the serum PFAS concentrations. Spearman correlation coefficients were calculated among the serum concentrations of the six PFAS evaluated (PFHxS, PFOS, PFOA, PFNA, PFDA, PFUnDA). We compared participants’ serum PFAS concentrations across groups with different causes of subfertility (male factor, female factor, unexplained) by analysis of variance.

Single-chemical analyses

We used multivariable linear regression to examine the percent difference (PD) in continuous thyroid biomarker concentrations per doubling increase in serum PFAS concentrations, adjusted for covariates. In the models, PFAS and thyroid biomarker concentrations were log-2 transformed to improve the interpretation of estimates as PDs per doubling of serum PFAS concentrations. We utilized logistic regressions to examine the Odds Ratio (OR) for thyroid autoimmune antibody positivity (i.e., TPOAb and TgAb) in relation to serum PFAS concentrations. All models were adjusted for a priori identified covariates using a directed acyclic graph (Figure S2), which included age, BMI, race/ethnicity, smoking status, education, and cause of subfertility.

Mixture analyses

We additionally utilized mixture analyses to examine the joint effect of the mixture of the six PFAS evaluated (total PFAS mixture) on each outcome. Specifically, we used quantile g-computation (QGC) to obtain the PD in continuous outcome or OR for binary outcome per quartile increase in the total PFAS mixture concentrations with the lowest quartile as the reference group, adjusting for the afore-mentioned covariates.²⁴ QGC also evaluated the contributions of each PFAS within the mixture to the overall mixture effect in either negative or positive direction.²⁴ We complemented QGC with Bayesian Kernel Machine Regression (BKMR). BKMR is a flexible non-parametric mixture method that allowed examination of non-linear relationships and interactions within the mixture.²⁵ BKMR evaluated 1) the dose-response relationship between individual PFAS and outcomes when fixing other PFAS in the mixture at their median concentrations; 2) changes in outcome per 5^th percentile increase/decrease in the total PFAS mixture concentrations from the median concentrations. BKMR also assessed the importance of individual PFAS on the overall joint effect using Posterior Inclusion Probabilities (PIPs) from pariwise variable selection.²⁵ We fit BKMR with 10000 iterations and determined model convergence by traceplots.

Sensitivity analyses

To explore the effect heterogeneity by cause of subfertility, we stratified the analyses by cause of subfertility (male factor, female factor, unexplained) and conducted a likelihood ratio test for the interaction term of PFAS*cause of subfertility, the p value of which was regarded as the p value for effect heterogeneity. We also examined the effect heterogeneity by obesity. We did not perform stratified analyses for binary outcomes because of the relatively small number of cases in the study population (Table 1, n=30 (10.5%) for positive TPOAb and n=16 (5.6%) for positive TgAb).

Table 1.

Characteristics of the study population from the Environmental Health and Reproductive Health (EARTH) Study (2005–2019).

		N = 287
Age, mean (SD)		34.47 (4.06)
BMI, mean (SD)		24.04 (4.27)
Race/Ethnicity, N (%)
	White persons	235 (82.7%)
	Black persons	9 (3.2%)
	Asian persons	31 (10.9%)
	Persons of other race/ethnicity	9 (3.2%)
Smoking status, N (%)
	Never smoker	210 (73.9%)
	Current or former smoker	74 (26.1%)
Graduate degree, N (%)		144 (57.8%)
Subfertility diagnosis, N (%)
	Male factor	63 (22.3%)
	Female factor	90 (31.8%)
	Unexplained	130 (45.9%)
Positive TPOAb,a N (%)		30 (10.5%)
Positive TgAb,b N (%)		16 (5.6%)
Year of recruitment, N (%)
	2005–2009	12 (4.2%)
	2010–2015	230 (80.7%)
	2016–2019	43 (15.1%)

Abbreviations: BMI, body mass index; TPOAb, peroxidase antibody; TgAb, thyroglobulin antibody.

^a.Positive TPOAb was defined as TPOAb > 35 IU/mL.

^b.Positive TgAb was defined as TgAb > 115 IU/mL.

All analyses were conducted in R (version 4.0.3, R Development Core Team 2020). We used multiple imputation with chained equations (MICE) to account for the missingness in covariates (missingness ranged 0.7%–13%, Figure S1) using the ‘mice’ package (version 3.14.0) in R. We used ‘qgcomp’ (version 2.8.5) and ‘qgcompint’ (version 0.6.6) packages for QGC models and ‘bkmr’ (version 0.2.0) package for BKMR models. Our result interpretation was based on the magnitude of point estimates and consistency across single-chemical analyses and mixture analyses, instead of only focusing on the statistical significance (i.e., 95% CI excluding the null).²⁶

Results

Population

A total of 287 females were included in the current study. The mean (SD) age and BMI of the participants was 34.5 (4.1) years and 24.0 (4.3) kg/m² respectively (Table 1). Most of the participants were White individuals (82.7%), never smokers (73.9%), had a graduate degree (57.8%), and almost half of the participants had unexplained subfertility (45.9%) (Table 1). The positivity rates of TPOAb and TgAb were 10.5% and 5.6% respectively. Our study population presented similar distributions of socio-demographic characteristics compared with the original EARTH cohort except that our study population was mainly recruited during 2010–2015 (Table S1).

PFAS distributions

The examined PFAS were detected in over 94% of the samples excepted for the branched PFOA isomers (1.4%). Distributions of serum concentrations of PFDA and PFUnDA were similar with the same median and interquartile range (median: 0.3 ng/mL, IQR: 0.2 – 0.4) which were close to the LOD (0.1 ng/mL) (Table S2). The Spearman correlation coefficients among the six examined PFAS ranged from 0.24 (PFHxS and PFUnDA) to 0.79 (PFUnDA and PFDA) (Figure S3).

Serum PFAS concentrations were, in general, similar among females with male factor or unexplained cause of subfertility, while those with female factor subfertility had lower PFOA and PFOS concentrations than the other two groups, though differences did not reach statistical significance (Table S3).

Association between PFAS and thyroid biomarkers

Single-chemical results

We found that higher serum concentrations of all examined PFAS were associated with a lower mean TT3 concentration (Table 2). Specifically, per doubling increase in serum PFOA, PFNA, PFDA, and PFUnDA concentrations were associated with 4.2% (95% CI: −7.9%, −0.4%), 3.0% (95% CI: −6.0%, −0.0%), 5.6% (95% CI: −8.8%, −2.2%), and 4.0% (95% CI: −6.5%, −1.5%), decreases in TT3 concentration (Table 2). We also found negative associations for PFOS (PD: −3.0%, 95% CI: −6.2%, 0.4%) and PFHxS (PD: −2.6%, 95% CI: −5.3%, 0.2%) with lower TT3 concentrations, though the confidence intervals crossed the null (Table 2).

Table 2.

Percent changes in thyroid biomarkers per doubling of serum per- and polyfluoroalkyl substances (PFAS) concentrations.

		Percent Difference (95% CI)a
Thyroid Stimulating Hormone (TSH)
	PFOA	−4.02 (−13.10, 5.99)
	PFOS	0.56 (−7.74, 9.59)
	PFHxS	−6.80 (−13.23, 0.09)
	PFNA	2.60 (−5.07, 10.88)
	PFDA	−1.10 (−9.63, 8.23)
	PFUnDA	−0.18 (−6.63, 6.71)
	Mixtureb	−2.38 (−9.29, 5.06)
Free Thyroxine (FT4)
	PFOA	0.36 (−2.04, 2.82)
	PFOS	0.65 (−1.44, 2.77)
	PFHxS	0.42 (−1.33, 2.19)
	PFNA	1.67 (−0.23, 3.60)
	PFDA	1.85 (−0.36, 4.10)
	PFUnDA	1.15 (−0.48, 2.81)
	Mixtureb	0.93 (−0.85, 2.75)
Total Thyroxine (TT4)
	PFOA	−2.78 (−6.16, 0.72)
	PFOS	−0.49 (−3.50, 2.62)
	PFHxS	−1.00 (−3.51, 1.58)
	PFNA	−0.77 (−3.49, 2.02)
	PFDA	−1.58 (−4.70, 1.64)
	PFUnDA	−1.37 (−3.69, 1.01)
	Mixtureb	−1.72 (−4.26, 0.88)
Free Triiodothyronine (FT3)
	PFOA	−0.85 (−2.99, 1.34)
	PFOS	0.06 (−1.81, 1.97)
	PFHxS	−0.73 (−2.29, 0.85)
	PFNA	−0.54 (−2.22, 1.17)
	PFDA	−1.31 (−3.24, 0.66)
	PFUnDA	−1.27 (−2.70, 0.18)
	Mixtureb	−0.83 (−2.40, 0.76)
Total Triiodothyronine (TT3)
	PFOA	−4.22 (−7.90, −0.39)
	PFOS	−2.96 (−6.20, 0.39)
	PFHxS	−2.58 (−5.31, 0.23)
	PFNA	−3.03 (−5.95, −0.01)
	PFDA	−5.56 (−8.82, −2.17)
	PFUnDA	−4.04 (−6.51, −1.50)
	Mixtureb	−4.45 (−7.15, −1.67)
FT4 to FT3 ratio
	PFOA	1.22 (−1.36, 3.87)
	PFOS	0.58 (−1.65, 2.86)
	PFHxS	1.16 (−0.71, 3.06)
	PFNA	2.22 (0.18, 4.29)
	PFDA	3.20 (0.83, 5.62)
	PFUnDA	2.45 (0.71, 4.23)
	Mixtureb	1.78 (−0.13, 3.73)

Abbreviations: PFOA, perfluorooctanoate; PFOS, perfluorooctane sulfonate; PFHxS, perfluorohexane sulfonate; PFNA, perfluorononanoate; PFDA, perfluorodecanoate; PFUnDA, perfluoroundecanoate.

^a.Models were adjusted for age (continuous), BMI (continuous), graduate degree (binary), race (categorical), ever smoking (binary), and subfertility diagnosis (categorical).

^b.Estimates obtained by quantile-based g computation.

Serum concentrations of PFNA (PD: 2.2%, 95% CI: 0.2%, 4.3%), PFDA (PD: 3.2%, 95% CI: 0.8%, 5.6%), and PFUnDA (PD: 2.5%, 95% CI: 0.7%, 4.2%) were associated with higher FT4 to FT3 ratios (Table 2).

Associations of PFOA with TT4 concentration (PD: −2.8%, 95% CI: −6.2%, 0.7%) and of PFHxS with TSH concentration (PD: −6.8%, 95% CI: −13.2%, 0.1%) were in a similar negative direction but did not reach statistical significance (Table 2). No associations were found for serum PFAS concentrations and either TPOAb or TgAb positivity (Table S4).

Mixture results

We identified a per quartile increase in the total PFAS mixture was associated with 4.5% (95% CI: −7.2%, −1.7%) lower TT3 concentration in QGC models (Table 2). Consistently, BKMR showed an imprecise decreasing trend in TT3 concentration across per 5^th percentiles increase in the total PFAS mixture (Figure 1e). The estimates within the three mixture percentile groups (i.e., 25^th to 35^th, 40^th to 60^th, 65^th to 75^th) were similar which was due to the limited variation in PFDA and PFUnDA concentrations within each group (Concentrations of both PFDA and PFUnDA: 25^th, 30^th, 35^th=0.2 ng/mL, 40^th, 45^th, 50^th, 60^th=0.3 ng/mL, 65^th, 70^th, 75^th =0.40 ng/mL). BKMR additionally showed negative dose-response relationships for PFHxS, PFDA, and PFUnDA and TT3 concentrations when holding the rest of PFAS in the mixture at their medians (Figure S4), which were consistent with results from single-chemical regressions, though a positive dose-response relationship was found for PFNA and TT3 in contrast with the linear regression results (Figure S4). PFDA and PFUnDA had high contributions to the joint effect on TT3 levels in both QGC and BKMR models (Tables S4 & S5).

Figure 1. Joint effect of per- and polyfluoroalkyl substances mixture and thyroid function biomarkers, results from Bayesian Kernel Machine Regression (BKMR) models.

Abbreviations: PFAS, per- and polyfluoroalkyl substances; TSH, thyroid stimulating hormone; FT4, free thyroxine; TT4, total thyroxine; FT3, free triiodothyronine; TT3, total triiodothyronine.

Note. The estimates in each plot represent difference in outcome per 5^th percentile increase or decrease in the total PFAS mixture concentration, holding the median of the total mixture concentration as the reference group.

QGC model identified an imprecise positive association between the total PFAS mixture and FT4 to FT3 ratio (PD: 1.8%, 95% CI: −0.1%, 3.7%) (Table 2). Consistently, in BKMR models, comparing to when the total PFAS mixture was at the 40^th to 60^th percentiles, the FT4 to FT3 ratio was significantly lower at the 25^th to 35^th percentiles of the mixture and higher at the 65^th to 75^th percentiles of the mixture (Figure 1f). Consistent with linear regressions, BKMR reported positive dose-response relationships for PFDA and PFUnDA with higher FT4 to FT3 ratios (Figure S4), and PFUnDA contributed the most to the joint effect as consistently shown in both QGC and BKMR models (Tables S5 & S6).

BKMR reported a strong negative dose-response relationship for PFHxS and TSH in the context of mixtures (Figure S4), and no clear joint effect of the total PFAS mixture on TSH levels (Figure 1a), both of which were consistent with linear regressions and QGC. We also found an imprecise decreasing trend of TT4 levels (Figure 1c) and FT3 levels (Figure 1d) and an increasing trend of FT4 levels (Figure 1b) across each 5^th percentile increase in the total PFAS mixture. Similar direction of the joint association was found in QGC but estimates were weak and imprecise (Table 2). No joint effect of the PFAS mixture was seen on autoimmune thyroid antibodies (Table S4). No interaction within the mixture was found for any thyroid biomarker (Figures S5–S10).

Sensitivity analyses

After stratifying the analyses by cause of subfertility, the effect estimates for the negative associations between all examined PFAS (except for PFOA) as well as the PFAS mixture with TT3 concentration were larger among females with male factor subfertility, while these associations were attenuated and imprecise among females with female factor or unexplained subfertility (p values for effect heterogeneity ranged from 0.07 to 0.72). The negative association for PFOA and TT3 concentration was similar among those with male factor or unexplained subfertility but the association was null among females with female factor subfertility (Table S7).

In contrast, the positive associations for PFOA, PFNA, PFDA, PFUnDA, and the total PFAS mixture with FT4 to FT3 ratio were stronger among females with unexplained subfertility than the other two groups (p values for effect heterogeneity ranged from 0.02 to 0.38) (Table S7).

We additionally found negative associations for PFOS with lower TT4 level (p value for effect heterogeneity = 0.05), and for PFOS, PFNA, PFDA, and the total PFAS mixture with lower FT3 levels (p values for effect heterogeneity ranged from 0.03 to 0.32) but generally only among females with male factor subfertility. However, we found positive associations for PFOS, PFNA, PFDA, PFUnDA, and the total PFAS mixture with higher FT4 levels only among females with unexplained subfertility (p values for effect heterogeneity ranged from 0.03 to 0.20) (Table S7). No consistent effect heterogeneity was observed for overweight vs. non-overweight females (Table S8).

Discussion

In this cross-sectional study of females attending an infertility clinic, serum concentrations of the six examined PFAS as well as their mixture were associated with lower TT3 concentrations. Although statistically significant, the magnitude of these associations was relatively small, ranging from 3% to 6%. We also found PFNA, PFDA, PFUnDA, and the total mixture were associated with higher FT4 to FT3 ratios. The findings on TT3 were in general stronger among participants with male factor subfertility while the associations on FT4 to FT3 ratio were stronger among females with unexplained subfertility. No associations were seen for PFAS serum concentrations and autoimmune thyroid antibodies.

Most of the literature on PFAS exposure and thyroid function focused on pregnant women and children; a few examined non-pregnant women in the general population.¹⁴ Our findings are consistent with a 2017 meta-analysis which reported that PFOS was associated with higher FT4 and lower TT3 and TT4 levels, and no association was found for TSH in adults, though the meta-analysis did not stratify results by sex.²⁷ Similarly, a more recent study in the Korean general population found that PFAS exposure was associated with higher FT4 level in both females and males, while no association was seen for TSH.²⁸ Additionally, our results are consistent with findings from a pregnant cohort in Sweden where higher prenatal PFAS exposure was associated with lower TT3 concentration and higher FT4 to FT3 ratio in pregnant women, while no association with TSH or autoimmune antibodies. The serum PFAS concentrations of this pregnant cohort were similar or slightly lower than our study population.²⁹ Of note, the effect estimates from these studies were comparable to our estimates (approx. 3–6% change in thyroid hormone levels). Besides, PFAS exposure is linked to hypothyroidism in women from U.S. and Sweden, supporting our findings on thyroid biomarkers.³⁰ However, one study that utilized data from the National Health and Nutrition Examination Survey (NHANES) 2007–2010, which is a U.S. representative population study, reported associations for PFOA with higher TT3, and for PFHxS with higher TT3 and TT4 among women.³¹ Of note, the PFAS concentrations in females of the 2007–2010 NHANES study were almost twice the PFAS concentrations in our study population that collected serum in 2005–2019, which could be one of the reasons for the heterogeneous findings.

T4 accounts for 90% of the thyroid hormones secreted by thyroid glands and 80% of T3 is converted from T4.¹ The lower TT3 concentrations as well as the higher FT4 to FT3 ratios associated with PFAS concentrations as we observed could suggest that PFAS exposure decreases the deiodination of T4 metabolism. A recent epidemiological study of U.S. adolescents showed higher PFAS mixture was associated with higher TT4 level, which could suggest lower conversion from T4 to T3.³² Experimental studies indeed showed that PFAS exposure influences mRNA expression of deiodinase enzymes Type I and III which are in charge of T4 to T3 conversion.^{12, 33} Additionally, PFAS altered the gene expressions in hypothalamus, pituitary, and thyroid, lending genetic/epigenetic support on its disruption of the hypothalamus-pituitary-thyroid axis.³⁴ Several experimental studies also showed that PFAS can bind with thyroid binding proteins, and disrupt the balance of free and bound T3 and T4.³⁵ Another potential pathway of PFAS on the thyroid hormones is through its damage on liver (which was established in previous studies^{36, 37}) where the deiodination of T4 to T3 takes place.

Additionally, we found effect heterogeneity by cause of subfertility on the PFAS and thyroid biomarker associations. Specifically, PFAS exposure was associated with lower TT3 concentration among those with male factor subfertility, and with higher FT4 concentration and FT4 to FT3 ratio among those with unexplained subfertility, both suggesting reduced conversion of T4 to T3 by deiodinase. There is indeed experimental evidence suggesting that the deiodinase deficiency reduced ovulation and fertilization,³⁸ which could be present among participants with female factor or unexplained subfertility. Females with male factor subfertility generally represent women of reproductive age in the general population. It is not clear why PFAS exposure related to different changes in thyroid biomarkers among females with male factor or unexplained subfertility, which suggests potential different mechanisms among women with and without physiological fertility problems and needs further research. The fact that we did not observe obvious associations among those with female factor subfertility and that this group of participants had lower PFAS concentrations than the other two groups may suggest a potential selection bias, where participants with female factor subfertility might have already been treated for overt thyroid problem before our study enrollment, decreasing our power to detect associations in this group. Nevertheless, more research is needed to understand if thyroid biomarker changes related to PFAS exposure mediate the previously observed associations between PFAS exposure and adverse reproductive outcomes, such as lower birthweight.^{10, 32, 36}

The strengths of this study include being the first study to examine PFAS exposure and thyroid function biomarkers among subfertile females. The assessment of subfertility causes is highly reliable based on SART guidelines. We assessed PFDA and PFUnDA which were not commonly examined in previous studies. Nevertheless, we recommend caution when interpreting the PFDA and PFUnDA results because of their relatively low concentrations and narrow distribution ranges compared to the other PFAS examined. The biggest limitation of this study is its modest sample size which precluded us from conducting stratified analyses for the binary outcomes and may limit our power to detect associations for continuous outcomes. The examined PFAS were all long alkyl-chain PFAS. There is evidence showing that short alkyl-chain PFAS may be more dangerous on thyroid than the long-chain PFAS, guaranteeing further evaluation.¹⁴ Thirdly, as discussed above, conditioning on females who showed up in the fertility clinic may lead to selection bias because those who had experience of prolonged time to pregnancy, which was previously related to PFAS exposure,³⁹ may have already sought medical help for underlying thyroid diseases. However, we would expect the bias to have shifted our observed estimates downwards and the true estimates may be more profound than the ones observed in this study. Besides, both PFAS and thyroid function biomarkers were measured in baseline blood samples, from which we cannot establish the temporality of PFAS exposure and thyroid function changes. However, given the long biological half-lives of the examined PFAS,⁴⁰ the PFAS concentrations measured in our study could represent PFAS exposure years ago, proceeding outcome occurrence. Additionally, iodine is important in thyroid function and there may be effect heterogeneity by iodine levels for the PFAS and thyroid biomarker associations which we were unable to examine due to lack of iodine biomarker data. Lastly, we mostly examined white persons with a high educational level, which limits the generalizability of our findings to subfertile populations with different socioeconomic backgrounds.

Conclusion

In this study of females attending a U.S. fertility clinic, we found that serum concentrations of select PFAS and their mixture were associated with lower TT3 levels and higher FT4 to FT3 ratios. Our findings support the thyroid-disrupting effect of PFAS in the vulnerable population with subfertility. Future study is needed to elucidate the mechanisms of PFAS on thyroid functions, which are critical for human reproduction, especially among populations of different susceptibilities.

Highlights.

• This study was the first to examine PFAS exposure with thyroid biomarkers in subfertile women.

• Serum concentrations of all examined PFAS and mixture were associated with lower TT3.

• PFNA, PFDA, PFUnDA, and total mixture concentrations were associated with higher FT4 to FT3 ratio.