The association between PFAS exposure, menstrual cycle parameters, and reproductive hormones in adolescent girls

Abstract

Context

Perfluoroalkyl and polyfluoroalkyl substances (PFAS), sometimes referred to as “forever chemicals”, are widespread. Certain PFAS exposures have been associated with reproductive abnormalities in women, but limited data exist in adolescents.

Objective

To investigate the relationship between PFAS, menstrual cycle length/variability, and reproductive hormones in adolescents.

Methods

Thirty-eight girls completed menstrual diaries and contributed daily urine samples to measure creatinine (Cr)-corrected luteinizing hormone (LH), estrone-3-glucuronide (E1G), and pregnanediol-3-glucuronide (PdG). Twenty-four PFAS were measured (n = 88 serum samples, average 2.32 samples/participant over the course of 1.20 ± 1.34 years) using mass spectrometry. Linear mixed-effects models were used to assess associations between PFAS and total cycle, follicular phase, and luteal phase lengths and hormone levels. Models were performed unadjusted and adjusted for gynecologic age, body fat percent, and race/ethnicity.

Results

Participants were 12.8 ± 1.0 years old (mean ± SD) with a gynecologic age of 0.3 ± 0.2 years. Most were non-Hispanic White and of normal weight. Four PFAS were consistently detected: perfluorobutanesulfonic acid (PFBS), perfluorohexanesulfonic acid, perfluorooctanoic acid, and perfluorooctanesulfonic acid. There were no associations between PFAS and cycle length or variability. PFBS was positively associated with peak E1G levels (β = .15, 95% CI [0.00, 0.29], P = .049), but no species was associated with peak LH or PdG levels.

Conclusion

PFAS were widely detected in healthy girls. PFBS was positively associated with E1G levels, suggesting potential ovarian effects. Analyses are ongoing to understand PFAS exposure sources and to determine if ongoing exposure may impact reproductive health.

Clinical trial information

Clinicaltrials.gov registration number: NCT02583646.

Per- and polyfluoroalkyl substances (PFAS), also referred to as “forever chemicals”, are highly stable synthetic organic compounds that are ubiquitous in the environment. They are found in drinking water, household and personal care products, air emissions, food, and food packaging (1). A growing body of literature has linked certain PFAS exposures to a variety of adverse health effects, including decreased infant and fetal growth, cancer, obesity, and liver, kidney, and thyroid disease (1-3). The mechanistic basis for PFAS toxicity is complex but appears to involve disruptions in cell signaling (eg, calcium, Wnt, mammalian target of rapamycin (mTOR), and nuclear factor-KB (NF-kB)), mitochondrial function, changes in lipid and amino acid metabolism, and binding to nuclear receptors (eg, peroxisome proliferator-activated receptor α, estrogen receptor, and farnesoid X receptor) (4).

Certain types of PFAS exposure have also been linked to a number of reproductive health issues (5). For example, in adult women, certain PFAS have been associated with irregular menstrual cycles (6) and lower levels of estradiol and progesterone (7, 8). Some PFAS have also been linked to infertility (9), lactation insufficiency (10), polycystic ovarian syndrome (11), and earlier age at menopause (12). In girls, certain PFAS have been associated with later ages at pubarche (13, 14), thelarche (13), and menarche (13-15), changes in breast density (16), and lower estradiol levels (at age 12 years) (13). Importantly, adolescence has been deemed a critical developmental window of susceptibility to reproductive toxicants (17). However, there have been very few studies to investigate the relationship between PFAS levels, menstrual cycle patterns, and reproductive hormone levels in healthy girls during the first few years after menarche.

Given this knowledge gap and that PFAS are known to be widespread in the surface water, groundwater, air, and sediment in central and eastern North Carolina (18), we were particularly interested in investigating the influence of PFAS exposure on menstrual cycle patterns and reproductive hormones in adolescent girls enrolled in the Girl's First Period Study (FIRST). FIRST is a longitudinal study of healthy, early post-menarchal girls residing in central North Carolina. It aims to understand the development of the female reproductive axis from the time of menarche to the establishment of mature ovulatory cycles in young adulthood as well as genetic and environmental contributors (19). Thus, in this substudy, we investigated the association between PFAS levels and menstrual cycle length/variability and reproductive hormones in adolescent girls. We also measured the same PFAS in a cohort of women with regular menstrual cycles from the same region to examine age effects.

Methods

FIRST (NCT02583646) was approved by the NIH Institutional Review Board. Informed assent and consent were obtained from each pediatric participant and her guardian, respectively, and consent was obtained from each adult participant.

Adolescent girls

Participants

Participants were 10.0-14.5 years old and no more than 6 months post-menarche at enrollment. Inclusion criteria included having a healthy body weight (defined as having a body weight >85th% of expected body weight and a body mass index <99th percentile) and normal levels of thyroid hormone, prolactin, and testosterone, as previously described (19, 20). Potential participants who were pregnant, taking or planning to take medication that affects reproductive hormones, and those with a chronic medical condition, a first-degree relative with a pubertal disorder, an excessive exercise routine (defined as running over 20 miles per week or its equivalent), or anemia were excluded (19, 20).

Participants were recruited starting in 2019. Recruitment was accomplished via advertisements in local newspapers and family-focused magazines, postcards via PostcardMania (a direct mail marketing company), hospital and community bulletin boards, radio, the internet, and by partnering with local schools.

Research study visits and at-home urine collection

Clinical visit procedures included a medical history and physical exam, BOD POD® exam (COSMED USA 50.26L, Chicago, IL) to determine body composition, blood and urine sampling for screening laboratory tests, and a transabdominal pelvic ultrasound. Every 4-6 months, participants underwent study visits approximately every 7-10 days during the course of 1 menstrual cycle for blood sampling and pelvic ultrasounds. At home, participants collected daily first-morning urine samples using filter paper strips to measure creatinine (Cr)-corrected luteinizing hormone (LH), estrone-3-glucuronide (E1G), and pregnanediol-3-glucuronide (PdG) and completed an online menstrual diary for the duration of the study (19, 20). A protocol schema is provided in Figure S1 (21).

Adult women

Participants were healthy, reproductive-aged women (28.50 ± 5.10 years old) with a history of regular menstrual cycles. Potential participants who were pregnant, taking or planning to take medication that affects reproductive or metabolic hormones, and those with a chronic medical condition were excluded. Adult participants provided daily first-morning urine samples using filter paper strips to measure reproductive hormones and completed an online menstrual diary for 2 menstrual cycles. These adult participants were also recruited under the FIRST protocol.

Laboratory assays

Reproductive hormones

Urinary Cr-corrected LH, E1G, and PdG were measured via enzyme immunoassay in dried urine strips (ZRT Laboratory, Beaverton, Oregon), as previously described (22). The LH enzyme-linked immunosorbent assays (ELISA) [Cal Biotech; El Cajon, CA, USA; Catalog # LH231F, RRID not available] is a direct assay, while E1G and PdG ELISAs [Arbor Assays; Ann Arbor, MI, USA; RRID: AB_3665778 and RRID: AB_3665779, respectively] are competitive assays. Coefficients of variation and limits of quantification (LOQs) are reported in Table S1 (21).

PFAS measurements

Twenty-four PFAS were measured in serum samples using ultra-high performance liquid chromatography (UHPLC) at the United States Environmental Protection Agency (EPA) (Table 1). Mass labeled internal standards (IS) and native compounds for calibration curves (mass-labeled solution/mixture MPFAC24ES and native solution/mixture PFAC-24PAR) were purchased from Wellington Laboratories (Guelph, Ontario, Canada). A 10-point calibration curve ranging from 0.1 to 75 ng/mL was prepared using the native mix, and quality controls (QCs) were spiked at 2 ng/mL. Samples were extracted alongside blanks, QC samples, and the calibration curve using the following method: 50 μL of sample was added to a 2 mL polypropylene tube with 100 μL of 0.1 M formic acid combined with IS. After a 3-second vortex, samples were allowed to sit at room temperature for 1.5 hours. Five hundred microliters of −20 °C acetonitrile (ACN) was then added to the tube and vortexed for 3 seconds, and samples were centrifuged at 10 000g for 5 minutes at room temperature. The eluent was dried and reconstituted in 100 μL of methanol and vortexed for 30 seconds.

Table 1.

List of 24 targeted PFAS analytes for mass spectrometry analysis

Acronym	DTXSID	Preferred name	CASRN	QC Status	LOQ (ng/mL)	Detection in adolescent samples n = 88	Detection in adolescent participants n = 38	Detection in adults n = 11
PFBA	DTXSID4059916	Perfluorobutanoic acid	375-22-4	Low	0.25	0 (0%)	0 (0%)	0 (0%)
PFPeA	DTXSID6062599	Perfluoropentanoic acid	2706-90-3	Low	0.25	0 (0%)	0 (0%)	0 (0%)
PFHxA	DTXSID3031862	Perfluorohexanoic acid	307-24-4	Low	1	0 (0%)	0 (0%)	0 (0%)
PFHpA	DTXSID1037303	Perfluoroheptanoic acid	375-85-9	Pass	0.25	0 (0%)	0 (0%)	0 (0%)
PFOA	DTXSID8031865	Perfluorooctanoic acid	335-67-1	Pass	0.25	82 (93.2%)	35 (92.1%)	9 (81.8%)
PFNA	DTXSID8031863	Perfluorononanoic acid	375-95-1	High	0.25	0 (0%)	0 (0%)	1 (9.1)
PFDA	DTXSID3031860	Perfluorodecanoic acid	335-76-2	Pass	1	0 (0%)	0 (0%)	0 (0%)
PFUdA	DTXSID8047553	Perfluoroundecanoic acid	2058-94-8	High	0.25	0 (0%)	0 (0%)	0 (0%)
PFDoA	DTXSID8031861	Perfluorododecanoic acid	307-55-1	High	0.25	0 (0%)	0 (0%)	0 (0%)
PFTeDA	DTXSID3059921	Perfluorotetradecanoic acid	376-06-7	Pass	0.25	0 (0%)	0 (0%)	0 (0%)
PFTrDA	DTXSID90868151	Perfluorotridecanoic acid	72629-94-8	Low	0.25	0 (0%)	0 (0%)	0 (0%)
FOSA	DTXSID3038939	Perfluorooctanesulfonamide	754-91-6	Low	2	0 (0%)	0 (0%)	0 (0%)
NMeFOSAA	DTXSID10624392	2-(N-Methylperfluorooctanesulfonamido)acetic acid	2355-31-9	High	0.25	7 (8.0%)	3 (7.9%)	0 (0%)
NEtFOSAA	DTXSID5062760	2-(N-Ethylperfluorooctanesulfonamido)acetic acid	2991-50-6	High	0.1	1 (1.1%)	1 (2.6%)	0 (0%)
PFBS	DTXSID5030030	Perfluorobutanesulfonic acid	375-73-5	Pass	0.25	71 (80.7%)	32 (84.2%)	10 (90.9%)
PFPeS	DTXSID8062600	Perfluoropentanesulfonic acid	2706-91-4	Pass	1	0 (0%)	0 (0%)	0 (0%)
PFHxS	DTXSID7040150	Perfluorohexanesulfonic acid	355-46-4	Pass	0.25	88 (100%)	38 (100%)	11 (100%)
PFHpS	DTXSID8059920	Perfluoroheptanesulfonic acid	375-92-8	Pass	1	0 (0%)	0 (0%)	0 (0%)
PFOS	DTXSID3031864	Perfluorooctanesulfonic acid	1763-23-1	Pass	0.25	87 (98.9%)	37 (97.4%)	10 (90.9%)
PFNS	DTXSID60873010	Perfluorononanesulfonic acid	68259-12-1	Pass	0.25	0 (0%)	0 (0%)	0 (0%)
PFDS	DTXSID3040148	Perfluorodecanesulfonic acid	126105-34-8	Pass	0.25	0 (0%)	0 (0%)	0 (0%)
4:2 FTS	DTXSID30891564	4:2 Fluorotelomer sulfonic acid	757124-72-4	Pass	0.1	0 (0%)	0 (0%)	0 (0%)
6:2 FTS	DTXSID6067331	6:2 Fluorotelomer sulfonic acid	27619-97-2	High	1	0 (0%)	0 (0%)	0 (0%)
8:2 FTS	DTXSID00192353	8:2 Fluorotelomer sulfonic acid	39108-34-4	High	0.1	0 (0%)	0 (0%)	0 (0%)

A DTXSID (DSSTox Substance Identifier) is a unique identifier assigned by the U.S. Environmental Protection Agency (EPA) in its CompTox Chemistry Dashboard (https://comptox.epa.gov/dashboard/). A CASRN (Chemical Abstracts Registry Number) is another unique identifier assigned by the Chemical Abstracts Service (CAS).

Abbreviations: LOQ, limit of quantification; QC, quality control.

Samples were then transferred to polypropylene autosampler vials, and 5 μL was injected onto a Thermo Vanquish UHPLC system equipped with a C18 column (Restek, Raptor C18 column 2.7 μM, 100×3.0 mm) coupled to a TSQ Quantis triple quadruple mass spectrometer. Mobile phase A and B were 95:5 water:acetonitrile with 2 mM ammonium acetate and 95:5 acetonitrile: water with 2 mM ammonium acetate, respectively. The liquid chromatographic gradient can be found in Table S2 (21), and the monitored transitions with matched IS are reported in Table S3 (21). LOQs were calculated for each PFAS as the lowest calibration point to achieve a <30% relative standard deviation that was also 3 standard deviations above method blanks (Table 1).

Statistical analyses

For adolescent participants, univariate linear regression models were estimated to evaluate crude associations between demographic and clinical factors, including age, gynecologic age at baseline, race (Black or African American, Asian, Multiracial, or White [reference group]), ethnicity (Hispanic, non-Hispanic [reference group]), BMI Z-scores, and body fat percentage (determined using BOD POD). Gynecologic age at baseline was included because it strongly affects the outcomes of interest (23).

The correlation between PFAS species was evaluated among adolescent participants' first available samples. PFAS levels in adult and adolescent participants were also compared with those of 225 adult women (aged 19-34 years) and 65 adolescent girls (aged 10-15 years), respectively, from the National Health and Nutrition Examination Survey (NHANES), cycle 2017-2018 (24). NHANES data were estimated using weighted geometric means with 95% CIs.

To evaluate whether PFAS levels changed over the sample collection period among FIRST participants, linear mixed effects models were estimated using the calendar time (years) since the first sample collection as the exposure and the PFAS level as the outcome. These models included a random intercept for each participant and were adjusted for gynecologic age at baseline, body fat percentage (determined using BOD POD), race (White vs non-White), ethnicity (non-Hispanic vs Hispanic), and whether both parents had a college degree. Sensitivity analyses using participant-specific time since enrollment rather than calendar time since initiation of the cohort sample collection were also conducted.

Linear mixed effects models were used to estimate associations between each PFAS and 1) total cycle, follicular phase, and luteal phase lengths and the number of cycle bleed days; and 2) peak LH, E1G, and PdG. A cycle was defined as ovulatory if the PdG value exceeded 2500 ng/mg Cr in the last 17 days of the cycle, as previously described (20). To calculate follicular and luteal phase lengths, a multistep algorithm was used, informed by literature in adult women and adolescents (25-27), to first determine the day of ovulation (Fig. S2) (21). In brief, the LH surge was identified as the first day that the LH value exceeded both 40 ng/mg Cr and the mean + 3 SDs of the values from the previous 5 days and did not occur within 5 days of another LH surge within that cycle. An E1G rise was identified when 2 consecutive days had E1G values > mean + 3 SDs of the previous 5 days, within the window of the identified LH surge + 4 days. The day of ovulation was identified as the day with the highest LH value within this LH surge/E1G rise window. The follicular phase included the day of ovulation. Models for each of these outcomes utilized longitudinally updating PFAS values over time, included a random intercept for each participant, and were adjusted for the number of cycle bleed days (except when this was the outcome variable) in addition to the covariates in the time models. Analyses were also restricted to cycles that started after the first sample was collected, were completed during the observation period, and that had <5 consecutive missing days and <50% overall missingness for urinary hormone levels. Hormone values were natural log-transformed before analysis to improve normality.

Linear regression models were estimated to evaluate the association between each PFAS and cycle length variability (defined as the participant-specific cycle length SD) using the first observed PFAS measurement to ensure that the exposure preceded all cycles contributing to the outcome measure. These models were adjusted for the same set of covariates as the mixed effects models.

For all regression models (linear regression and linear mixed effects), PFAS values below the LOQ were imputed with the LOQ divided by the square root of 2, and all values were natural log-transformed to improve normality.

Results

The 38 adolescent participants were 12.8 ± 1.04 years old (mean ± SD) at enrollment with a gynecologic age of 0.29 ± 0.24 years. Most participants were White (68.4%), non-Hispanic (89.5%), and of normal weight (BMI percentile 5th to ≤85th, 72.1%). The majority (71.1%) of participants had parents who both attended college, and 81.6% had parents who were both employed (Table 2).

Table 2.

Characteristics of adolescent participants in FIRST

	(N = 38)
Age at baseline (years)
Mean (SD)	12.8 (1.04)
Median [Min, Max]	12.8 [9.75, 14.7]
Time since first period, baseline (years)
Mean (SD)	0.286 (0.244)
Median [Min, Max]	0.208 [0.0274, 1.19]
Race
White	26 (68.4%)
Asian	4 (10.5%)
Black or African American	6 (15.8%)
Multiracial	2 (5.3%)
Ethnicity
Not Latino or Hispanic	34 (89.5%)
Latino or Hispanic	4 (10.5%)
BMI Z-score
Mean (SD)	0.593 (0.853)
Median [Min, Max]	0.635 [−1.40, 2.07]
Percent body fat
Mean (SD)	22.2 (8.17)
Median [Min, Max]	20.1 [6.20, 41.7]
Parents both employed
Yes	31 (81.6%)
No	7 (18.4%)
Parents both college-educated
Yes	27 (71.1%)
No	11 (28.9%)
Length of follow-up (years)
Mean (SD)	2.56 (1.29)
Median [Min, Max]	2.68 [0.159, 4.91]
Cycles under follow-up
Mean (SD)	26.0 (15.1)
Median [Min, Max]	26.5 [3.00, 59.0]

Twenty-four PFAS were measured in 88 serum samples (average 2.32 samples/participant over the course of 1.20 ± 1.34 years; Table 1). Four PFAS were detected in 81% to 100% of samples: perfluorobutanesulfonic acid (PFBS), perfluorohexane sulfonate (PFHxS), perfluorooctanoic acid (PFOA), and perfluorooctane sulfonate (PFOS). Our analyses therefore focused on these 4 species (Fig. 1, Fig. S3) (21). Among adolescent participants, PFOA was positively correlated with PFOS and PFHxS but not with PFBS (Fig. 2). In univariate analyses, there were significant racial differences: Asian participants had higher PFOS (β = 1.26, 95% CI [0.61, 1.91], P = .0004) and Black participants had lower PFOA (β = −.61, 95% CI [−1.13, −0.08], P = .02), lower PFHxS (β = −.32, 95% CI [−0.58, −0.06], P = .02), and higher PFBS (β = .39, 95% CI [0.03, 0.75], P = .04) than their White counterparts. There were no ethnicity-based differences, but this analysis was limited by the small number of Hispanic girls (n = 4). PFAS levels were not associated with chronological age or gynecological age at baseline [P > .05 for all PFAS]. Both BMI z-score and total body fat percent were borderline inversely associated with PFOS (BMI z: β = −.26, 95% CI [−0.53, 0.01], P = .06; body fat percent: β = −.03, 95% CI [−0.05, 0.002], P = .07).

Figure 1.

Distribution of 4 PFAS in (A) adolescent (in first sample) and (B) adult subjects shown via boxplots. Boxes represent the interquartile range (IQR), with the line indicating the median. Whiskers extend to 1.5 × IQR width from the box; data outside this range are outliers.

Figure 2.

Pearson correlation coefficients for each pair of 4 PFAS in FIRST adolescent subjects' first available samples.

To understand the regional impact of PFAS exposure as well as potential age effects, the PFAS in adolescents were also compared with 11 regularly cycling adult women from central North Carolina (63.6% White, 72.7% non-Latina, 27.3% normal weight, 10 nulliparous, Table 3) and with that of adolescent girls (aged 10-15 years) and adult women (aged 19-34 years) from across the US who participated in NHANES (cycle 2017-2018) (24). Adolescent participants from the FIRST study tended to have higher PFHxS levels than the NHANES cohort, whereas PFOA and PFOS levels were lower than in the NHANES cohort (Fig. 3). PFHxS levels were also higher in the North Carolina adult participants than in the NHANES adult cohort, whereas PFOA and PFOS levels tended to be lower than in NHANES women (Fig. 3). These analyses were unchanged after restricting both adult cohorts to nulliparous women. PFOA, PFOS, and PFBS (PFBS median [IQR] for adolescents: 0.37 [0.1]; adults 0.36 [0.08]) levels were comparable in North Carolina adolescents and adults, whereas PFHxS levels were higher in adults.

Table 3.

Characteristics of 11 adult women with regular menstrual cycles in FIRST

	(N = 11)
Age (yrs)
Mean (SD)	28.5 (5.1)
Race
White	7 (63.6%)
Asian	1 (9.1%)
Black or African American	3 (27.3%)
Multiracial	0 (0.0%)
Ethnicity
Not Latino or Hispanic	8 (72.7%)
Latino or Hispanic	3 (27.3%)
BMI (kg/m2)
Mean (SD)	24.3 (3.3)
Percent body fat
Mean (SD)	31.9 (6.7)
Age first period
Mean (SD)	12.5 (1.5)
Education level
Some college or less	2 (18.2%)
College degree or more	9 (81.8%)
Employment
Unemployed	3 (27.3%)
Employed	8 (72.7%)

Figure 3.

Comparison of PFAS in adolescents and adults from the current study and that of NHANES. Squares are geometric means and lines indicate 95% CI. Note that PFBS was not measured in NHANES.

Samples included in these analyses were collected between December 2019 and April 2024. Figure S3 (21) displays the levels of 4 PFAS over time in our study population. Over the sample collection period, levels of PFOA (β = −.15, 95% CI [−0.21, −0.08], P < .001) and PFBS (β = −.13, 95% CI [−0.24, −0.01], P = .03) decreased but no change was observed for PFOS (β = .00, 95% CI [−0.08, 0.08], P = .99) or PFHxS (β = .03, 95% CI [−0.05, 0.11], P = .50). Models evaluating time since enrollment identified a comparable pattern, with a decrease in PFOA and PFBS but not PFOS or PFHxS over time.

During the observation period corresponding to PFAS measurements, adolescent participants reported an average of 26.0 (SD 15.1) menstrual cycles. Average cycle length was 33.30 ± 11.23 days, and the average number of bleed days was 5.00 ± 1.71. In the 52.7% of cycles that were ovulatory, the average follicular phase length was 22.2 ± 8.5 days, and the average luteal phase length was 11.3 ± 3.7 days. In anovulatory cycles, peak LH, E1G, and PdG were 62.63 mIU/mg Cr (95% CI: 53.06, 73.93), 60.88 ng/mg Cr (95% CI: 53.16, 69.71), and 1205.61 ng/mg Cr (95% CI: 1072.61, 1355.66), respectively, and in ovulatory cycles, peak LH, E1G, and PdG were 81.06 mIU/mg Cr (95% CI: 68.49, 95.93), 70.70 ng/mg Cr (95% CI: 61.66, 81.06), and 4087.35 ng/mg Cr (95% CI: 3627.55, 4605.44), respectively. No associations were found between PFAS exposure and menstrual cycle phase length (total cycle, follicular phase, and luteal phase), cycle variability, or average number of bleed days (Fig. 4). In the peak hormone models, PFBS was positively associated with peak E1G (β = .15, 95% CI [0.00, 0.29], P = .049) (Fig. 5) but no other significant associations were observed.

Figure 4.

PFAS levels were not associated with menstrual cycle total length or follicular and luteal phase length, cycle length variability, or number of bleed days in adolescent subjects. Squares indicate the value of the beta coefficient and lines indicate 95% CI. All models were adjusted for gynecologic age at baseline, body fat percentage (determined using BOD POD), race (White vs non-White), ethnicity (non-Hispanic vs Hispanic), whether both parents had a college degree, and number of cycle-specific bleed days.

Figure 5.

PFAS levels in relationship to urinary reproductive hormones in adolescent subjects. Squares indicate the value of the beta coefficient and lines indicate 95% CI. PFBS was positively associated with peak E1G (P = .049). All models were adjusted for gynecologic age at baseline, body fat percentage (determined using BOD POD), race (White vs non-White), ethnicity (non-Hispanic vs Hispanic), whether both parents had a college degree, and number of cycle-specific bleed days.

Discussion

The elimination half-lives of certain PFAS are estimated to range from months to years, indicating that some PFAS can accumulate in the body over time with consistent exposure. This accumulation, coupled with existing studies suggesting adverse health effects from exposure to certain PFAS, raise potential public health concerns, particularly those related to the reproductive system (1). Data from NHANES (2013-2014), a large population-based epidemiological study in the US, suggests nearly universal exposure to 4 long-chain PFAS (PFOS, PFOA, PFNA [perfluorononanoic acid], and PFHxS) among the US population (3), and several hot spots of contamination have been identified, including in North Carolina (28). In the current studies, we observed that while PFOA, PFOS, and PFHxS levels in adolescent girls in central North Carolina were comparable to levels reported nationally, PFBS levels were higher than expected (as discussed below). PFBS levels were also positively associated with E1G levels, but none of the PFAS measured were associated with menstrual cycle length or regularity during the early post-menarchal years.

Our study identified a positive association between PFBS levels and urinary estrogens in healthy post-menarchal girls. To our knowledge, a relationship between PFBS and reproductive hormones has not been previously reported among girls or women; with rare exception (29, 30), PFBS was not part of the PFAS panel (13-15, 31, 32) or levels were below detection in most subjects (33-35). A negative relationship between 1 or more other PFAS and serum estradiol has been observed in adult women, including regularly cycling, nulliparous women (7), peri- and menopausal women (36), and women with premature ovarian insufficiency (30, 37), but not in women attending an in vitro fertilization clinic (29, 34). Two recent studies based on data from NHANES (33, 38) and a third study in a Cincinnati-based cohort (13) also reported negative associations between other PFAS and serum estradiol levels in pubertal girls. However, studies in pre-pubertal/early pubertal American girls (14, 39) and in adolescents or young adults in Taiwan (32, 40), Denmark (15), and Norway (41) did not observe an association with estradiol.

The current study did not identify an association between PFAS exposure and LH levels. While the literature on PFAS and LH is somewhat limited, the majority of previous studies have also not identified an association with the exception of 1) 2 studies in adult men residing in either Durham, North Carolina (42) or in the Copenhagen area of Denmark (43) (both with positive associations between PFOA and LH), 2) 1 study in male adolescents living near a chemical factory in Belgium which reported an inverse correlation between LH and PFHxS and PFOA (44), and 3) a study of mother-infant pairs in Shanghai, China which reported an inverse correlation between maternal PFBS and perfluoroheptanoic acid (PFHpA) and LH levels in umbilical cord blood (45). A study of 12-year-old girls (n = 109, 44% post-menarchal) in the Cincinnati, Ohio, region, for example, found no association between prenatal PFOA, PFOS, PFNA, or PFHxS levels and serum LH at age 12 years (13). However, statistical models did not adjust for the relative pubertal delay observed in girls with greater exposure, and serum sampling was not performed on a standardized menstrual cycle day. Another recent study in nearly 500 Norwegian post-menarchal girls reported no association between serum PFHxS, PFOA, PFOS and LH, however, more than one-third of participants were using hormonal contraceptives (41). Additional studies in Taiwanese teenage students (32), 20-year old Danish women (15), Chinese women with or without premature ovarian insufficiency (30), and infants studied during the mini-puberty (46) reported no associations between LH and certain PFAS levels (PFBS, PFHxS, PFOA, PFOS, PFNA, perfluorodecanoic acid [PFDA], and perfluoroundecanoic acid [PFUA]). Of interest, a recent mouse model of adolescent PFAS exposure showed that a PFAS mixture led to higher levels of serum LH, increased protein abundance of ovarian cytochrome P450 17A1 (CYP17A1, a key enzyme involved in androgen biosynthesis in theca cells), and disrupted estrous cyclicity (47). As noted by the authors, this study raises the question of whether prolonged PFAS exposure during adolescence could lead to future hyperandrogenism-related disorders (eg, polycystic ovarian syndrome). While our participants had normal testosterone levels at baseline (an inclusion criterion), with ongoing monitoring of this longitudinal cohort, we will be able to determine the effect of PFAS exposure on reproductive hormones and function at age 18 yrs.

We did not observe an association between the PFAS measured in this study and phase-specific nor overall cycle length. To our knowledge, there have been no previous studies to examine this association in early post-menarchal adolescent girls or in healthy, regularly cycling women. Two retrospective studies in pregnancy cohorts produced conflicting results, with 1 study in Norway showing no association between pre-pregnancy menstrual cycle length and serum PFAS (7 different species measured) (48) and the other study of women in Greenland, Poland, and Ukraine suggesting that higher PFOA levels were linked to longer menstrual cycles (49). Studies in women trying to conceive have also produced conflicting results (6, 50). We did not observe an association between PFAS levels and menstrual cycle irregularity in adolescent girls. An association between certain PFAS levels, including PFOA, PFNA, PFHxS, PFAS, and the sum of all perfluorodimethylhexane sulfonates, and cycle irregularity has been reported in some (6, 51, 52) but not all (48) studies in adult women. Of note, while a recent study of Italian high school girls (53) reported that girls with greater PFOA exposure were more likely to experience irregular cycles, they did not account for differences in gynecological age, a major limitation.

While some toxicological studies in animal models have shown that PFAS mixtures or PFOS alone disrupts estrous cyclicity (47, 54, 55), it remains unclear whether this is due to PFAS action at the level of the brain, ovary, or both. In vivo studies in adolescent and adult mice using a PFAS mixture that included PFOA, PFOS, PFBS, and undecafluoro-2-methyl-3-oxahexanoic acid (47) as well as in vitro studies of PFOS or PFOA in porcine theca and granulosa cells (56) have demonstrated that PFAS alters ovarian steroidogenesis. Chronic, low-dose PFOS exposure in adult female mice led to reduced levels of all reproductive hormones (estradiol, progesterone, LH, FSH, and gonadotropin-releasing hormone [GnRH]), the absence of an LH surge, and fewer kisspeptin neurons in the anteroventral periventricular (AVPV) nucleus. However, this phenotype could be rescued by supplementation with estradiol, suggesting that the neuroendocrine abnormalities were secondary to impaired follicular estradiol biosynthesis (55). Nevertheless, the ability of PFAS, particularly PFOS and PFOA, to cross the blood brain barrier in humans (57) and for PFOS to preferentially accumulate in the hypothalamus in rodents (54) raises the possibility that PFAS may directly affect the neuroendocrine components of the reproductive axis in addition to the ovary.

PFBS is a short-chain PFAS that was developed and used in consumer products beginning in the early 2000's in an effort to replace PFAS with longer half-lives (eg, PFOA, PFOS) and greater bioaccumulation potential (58). PFBS exposure primarily occurs through intake of food and water containing PFBS or PFBS precursors, however, PFBS and its precursors are also present in consumer products (such as perfumes, cosmetics, and detergents) (58). In the current studies, PFBS concentrations were higher than expected based on recent population-based studies. For example, in a 2015 study of >600 adult blood donors from 5 regional centers in the United States (including 1 in Charlotte, NC), less than 10% of serum samples had quantifiable levels of PFBS (59). This trend may suggest that PFBS use in consumer goods has increased over the past years. Also notable are the relatively high levels of PFHxS that were identified among the North Carolina adult women compared to the adolescent participants as well as the NHANES adult and pediatric cohorts. In the USA, the main sources of exposure to PFHxS are contaminated drinking water (particularly near military bases, airports, or firefighting activities where aqueous film-forming foams [AFFF] are used), certain foods (seafood), household dust (because of historical use in stain-resistant carpets and upholstery), and occupational settings (such as construction, firefighting, and manufacturing plants) (60). We do not have the necessary data to identify the exact source(s) of exposure in our adult cohort.

The main strengths of the current studies are the prospective collection of menstrual cycle dates and sampling of multiple PFAS, including PFBS, and reproductive hormones at several time points. We did not have information on the amount of menstrual blood loss and did not collect data on potential sources of PFAS exposure (eg, dietary intake), which may have provided additional context for the observed associations. Lastly, given the relatively small sample size and predominance of White participants of higher socioeconomic standing in a limited geographic area, our findings may not be generalizable to all other populations.

In summary, these studies demonstrate PFAS, and notably PFBS, exposure among healthy post-menarchal girls in central North Carolina. Using models that appropriately controlled for gynecological age, body fat, and race/ethnicity, we found no association between the measured PFAS levels and menstrual cycle characteristics. However, we did observe a positive association between PFBS and peak urinary estrogen metabolites, which, with ongoing exposure, could lead to reproductive health issues such as menorrhagia (via increased stimulation of the endometrial lining), growth of fibroids, or increased breast density and a higher risk of breast cancer. This cohort will be followed through late adolescence to monitor reproductive health and further investigate sources of exposure and potential ways to mitigate exposure during this vulnerable stage of development.

Funding

This research was supported in part by the Intramural Research Program of the National Institutes of Health (NIH) (Z01-ES103315 and ZIC ES103363 to NDS, ZIAES103344 to KKF, and NIEHS P30 ES025128 to SEF). NDS is also supported by the NIH as a Lasker Clinical Research Scholar (1SI2ES025429-01). The contributions of the NIH author(s) were made as part of their official duties as NIH federal employees, are in compliance with agency policy requirements, and are considered Works of the United States Government. However, the findings and conclusions presented in this paper are those of the author(s) and do not necessarily reflect the views of the NIH, the U.S. Environmental Protection Agency, or the U.S. Department of Health and Human Services. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.

Disclosures

None.

Data availability

Datasets generated and/or analyzed during the current study are either included in the published article or are available in SciHub (Doi: 10.23719/d-9gjd) and can also be made available by the corresponding author on reasonable request.