Per- and polyfluoroalkyl ether acids in well water and blood serum from private well users residing by a fluorochemical facility near Fayetteville, North Carolina

Abstract

Background:

A fluorochemical facility near Fayetteville, North Carolina, emitted per- and polyfluoroalkyl ether acids (PFEAs), a subgroup of per- and polyfluoroalkyl substances (PFAS), to air.

Objective:

Analyze PFAS in private wells near the facility and in blood from well users to assess relationships between PFEA levels in water and serum.

Methods:

In 2019, we recruited private well users into the GenX Exposure Study and collected well water and blood samples. We targeted 26 PFAS (11 PFEAs) in water and 27 PFAS (9 PFEAs) in serum using liquid chromatography-mass spectrometry. We used regression modeling to explore relationships between water and serum PFAS. For the only PFEA detected frequently in water and serum, Nafion byproduct 2, we used generalized estimating equation (GEE) models to assess well water exposure metrics and then adjusted for covariates that may influence Nafion byproduct 2 serum concentrations.

Results:

We enrolled 153 participants ages 6 and older (median = 56 years) using 84 private wells. Most wells (74%) had ≥6 detectable PFEAs; median ∑PFEAs was 842 ng/L (interquartile range = 197–1760 ng/L). Low molecular weight PFEAs (PMPA, HFPO-DA [GenX], PEPA, PFO2HxA) were frequently detected in well water, had the highest median concentrations, but were not detectable in serum. Nafion byproduct 2 was detected in 73% of wells (median = 14 ng/L) and 56% of serum samples (median = 0.2 ng/mL). Cumulative dose (well concentration × duration at address) was positively associated with Nafion byproduct 2 serum levels and explained the most variability (10%). In the adjusted model, cumulative dose was associated with higher Nafion byproduct 2 serum levels while time outside the home was associated with lower levels.

Impact:

PFAS are a large class of synthetic, fluorinated chemicals. Fluorochemical facilities are important sources of environmental PFAS contamination globally. The fluorochemical industry is producing derivatives of perfluoroalkyl acids, including per- and polyfluoroalkyl ether acids (PFEAs). PFEAs have been detected in various environmental samples but information on PFEA-exposed populations is limited. While serum biomonitoring is often used for PFAS exposure assessment, serum biomarkers were not good measures of long-term exposure to low molecular weight PFEAs in a private well community. Environmental measurements and other approaches besides serum monitoring will be needed to better characterize PFEA exposure.

Introduction

Per- and polyfluoroalkyl substances (PFAS) are a class of >12,000 synthetic, organic chemicals that consist of a fully or partly fluorinated carbon chain connected to different functional groups [1]. Because the carbon-fluorine bond is chemically and thermally stable, thousands of PFAS have applications in a wide range of industries, from industrial mining to food production and fire-fighting foams [2]. PFAS production started in the 1940s and early production focused on eight-carbon perfluoroalkyl acids (PFAAs), specifically perfluorooctanesulfonic acid (PFOS) and perfluorooctanoic acid (PFOA). However, beginning in 2001, PFOS, PFOA, and higher molecular weight homologs were phased out of production in the United States and Europe by several major producers due to epidemiologic evidence that exposure is associated with adverse health effects [3,4,5].

The fluorochemical industry is producing shorter chain/lower molecular weight PFAS and PFAS that include chemical modifications such as per- and polyfluoroalkyl ether acids (PFEAs). PFEAs contain one or more ether-oxygen linkages in the carbon chain. Hexafluoropropylene oxide dimer acid (HFPO-DA or “GenX”), a six-carbon monoether, is produced as a replacement for PFOA for use as a polymer processing aid. Importantly, some PFEAs are not produced for commercial purposes but are generated as byproducts of fluoropolymer manufacturing (e.g., Nafion byproducts) [6] and can be released to the environment through air emissions and wastewater discharges [7]. Large fluoropolymer manufacturing facilities in the United States [7,8,9], the Netherlands [10], and China [11], are known sources of PFEA environmental contamination. Additionally, several PFEAs including GenX have been detected in various water matrices such as rainwater [12], groundwater [13], surface water [14,15,16], and municipally treated drinking water [16], indicating that PFEAs are widely distributed across the globe. Emerging PFAS may have similar health endpoints as PFOA and PFOS [17]. Exposure information on PFEA-exposed human populations is limited [11, 18, 19] and, as a result, PFEA water-blood associations are not well described.

To investigate PFEA water-blood associations, it is helpful to have individual-level drinking water measurements and individual-level serum measurements and for there to be a wide range of concentrations in both drinking water and serum. Of previous studies that showed significant associations between PFAS contaminated drinking water levels and PFAS serum levels, few (i.e., [20, 21]) had paired drinking water and serum measurements. In studies of municipal water systems, researchers have tested for significant differences in blood PFAS concentrations across populations served by differently contaminated water districts [22]. Within water districts with known PFAS contamination, researchers have looked at the strength of the relationship between blood PFAS concentration and surrogates of drinking water exposure (e.g., years of residence) [22, 23] and how the strength of the relationship changed with increasing daily water consumption rate [23]. On the other hand, studies of drinking water exposure in private well communities have the potential for probing a wider range of household-level PFAS concentrations than studies in communities receiving municipal water and, thus, a wider range in individual-level serum measurements. For example, investigation of PFOA in a private well community yielded an estimate for a serum:drinking water concentration ratio for PFOA (114, unitless) [20]. Similarly, a private well study may be useful to evaluate water-blood relationships for PFEAs.

In 2017, the North Carolina (NC) Department of Environmental Quality (NC DEQ) began investigating GenX contamination of private wells near the Fayetteville Works fluorochemical manufacturing facility in Fayetteville, NC, and this investigation is ongoing. Fayetteville Works began fluorochemical production in 1980 and, for approximately 40 years, the facility released PFAS, including a suite of PFEAs generated as byproducts of production, to the environment through air emissions [24] and discharge of manufacturing wastewater to the Cape Fear River [6, 15, 16]. As of October 2022, ~10,000 wells near the facility had been sampled. Approximately 3550 wells (36%) did not have detectable PFAS or PFAS ≥ 10 ng/L, ~4640 wells (46%) had any PFAS (except GenX)≥ 70 ng/L, and ~1820 (~18%) had GenX ≥ 10 ng/L [25]. High levels of GenX (up to 4000 ng/L) were found in some private wells. Public awareness about GenX contamination from Fayetteville Works started in June 2017 and is ongoing. In September 2017, NC officials directed Chemours to provide bottled water to area households with GenX in wells above 140 ng/L, based on the company’s test results. In June 2022, the United States Environmental Protection Agency (US EPA) set the final lifetime health advisory level (HAL) for GenX in drinking water at 10 ng/L [26]. GenX was included as part of a hazard index calculation for a suite of PFAS (also PFNA, PFHxS, and PFBS) as part of US EPA’s proposed National Primary Drinking Water Regulation [27]. We established the GenX Exposure Study in 2017 to characterize PFAS exposure in Cape Fear River Basin communities impacted by PFAS contamination of drinking water. In 2019, we recruited individuals who lived near the Fayetteville Works facility and used private wells into the GenX Exposure Study and analyzed for GenX and other PFEAs in paired well water and serum samples. Here, we report on the investigation of the relationships between PFEA levels in well water and serum.

Materials and methods

Study design and population

In February 2019, we recruited private well users residing near the Fayetteville Works fluorochemical manufacturing facility in North Carolina to participate in the GenX Exposure Study. Private well users were selected from a list of ~1000 addresses with wells already tested for GenX by the Chemours Company or NC DEQ [25] by June 2018. The list contained information on homeowner name, well address, and sometimes phone numbers, and well water concentration; no demographic data were included. We split the list into two groups: (1) wells with GenX at or above 140 ng/L, the NC provisional drinking water health goal at the time (n = 238), and (2) those with lower levels of GenX in wells (n = 975). We randomized each list and then called an equal number of homes from the two groups with the goal of having a sample with approximately half of the study participants in each group. We purposely sampled over the range of exposure to allow us to assess variability in this population. We continued recruitment until we reached our target number of volunteers.

Interested individuals were screened over the phone for eligibility and scheduled for a clinic event to be held at a local health department. To be eligible for the study, individuals had to have lived in the Gray’s Creek, NC, area at the address associated with the sampled well since July 2016 and be 6 years of age or older. Up to four individuals per household were allowed to participate. Individuals were excluded if HIV positive, Hepatitis C positive, or pregnant. People enrolled in the study at the clinic event. All participants provided written informed consent; minors provided assent and parents of minors provided permission for their child to participate. Participants also agreed to have well water samples collected by study staff following the clinic visit.

Data collection

Biological sample collection

During a clinic event, we administered a questionnaire, collected biological samples (blood and urine), and measured height and weight; we used a similar protocol in Wilmington, NC, in 2017–2018 [19]. The questionnaire collected information on demographics, drinking water habits, residential history, health history, and potential PFAS exposures. Each participant provided a spot urine sample. A non-fasting blood sample for each participant was collected by a phlebotomist. For participants who were ≥11 years of age, four tubes of blood (two red-top tubes for serum, two EDTA tubes for whole blood or plasma) were collected. For children 6–10 years of age, two red-top tubes for serum were collected. Serum tubes were held at room temperature for 30–45 minutes and then spun at 1300 × g for 10 min in a Sorvall RT 600D centrifuge at room temperature. Serum was aliquoted into transfer tubes. One EDTA tube was processed for plasma; the remainder was saved as whole blood. Urine and blood samples were stored on dry ice and transported to East Carolina University (Greenville, NC) and stored at −80 °C. A 2-mL aliquot of serum was shipped on dry ice to the US EPA in Research Triangle Park, NC, and then stored at −80 °C until PFAS analysis.

Environmental sample collection

Home visits occurred approximately two weeks following the clinic visit. During a home visit, two water samples (well and tap) and one dust sample were collected by trained study staff. After brief flushing, a 135-mL well water sample was collected in a high-density polyethylene bottle from a spigot at the wellhead, if accessible, or another outdoor spigot. Some well samples were collected in duplicate to verify the reproducibility of measurements. Study staff asked participants questions about in-home well water treatment, well age, well type, well screen depth, and frequency of well use and whether there was recent chlorination of the well. The geocoordinates of each well were measured using Trimble^® Geo 5 T handheld units. In each house, study staff collected a 135-mL first-draw tap water sample (i.e., no flushing before sample collection) from the kitchen faucet. All water samples were acidified with nitric acid and stored at room temperature until PFAS analysis. In addition to water samples, study staff collected a dust sample from the kitchen floor using Eureka Mighty-Mite vacuum cleaners fitted on the inlet hose with a DUSTREAM^™ dust collector filter (Indoor Biotechnologies). The dust samples were stored at room temperature at NC State University.

Questionnaire variables

We used questionnaire data for potential covariates in the association between PFEA levels in well water and serum, including the type of water the participant is currently drinking (well water, bottled water, or other), whether a participant eats home grown vegetables (yes or no), the amount of time spent outside the home per week, and whether they ever worked at the Fayetteville Works facility (yes or no). Some of these covariates reflect the potential for other, non-drinking water exposures to PFEAs for individuals living nearby the fluorochemical facility. We created three-level drinking water consumption values (0–3, 4–5, or >5 12-ounce glasses per day) for adults based on participants’ questionnaire responses. The questionnaire asked “How many 12-ounce glasses of water do you drink in a day at home?” and specified that estimates should include juices or other cold or hot drinks that require the addition of water. This question was not asked of children. For children, we used US EPA mean estimates for combined direct and indirect water ingestion for ages 6 - <21 years [28]; consequently, all children were assigned to the first drinking water category [0–3 12-ounce glasses per day].

PFAS Analyses

This manuscript focuses on PFAS results of the well water and blood serum analyses only.

We analyzed for PFAS in water and serum samples using high performance liquid chromatography-tandem mass spectrometry. For compounds, for which mass-labeled internal standards were available, an isotope dilution approach was used to quantify PFAS concentrations. If a mass-labeled analog of an analyte was not available, we used a structurally similar mass-labeled internal standard with a similar retention time for quantitation. We analyzed water and serum samples on different instruments using different analytical methods, detailed below. Overall, we targeted 26 PFAS in water and 27 PFAS in serum (Table S1). At the time of analyses, analytical standards for most PFEAs except GenX were not available from commercial suppliers so we prepared calibration standards from aqueous stock solutions (1000 mg/L) acquired from the Chemours Company. The method reporting limit (MRL) for each PFAS ranged by batch. The MRL was calculated as the lowest analytical standard within 70% accuracy of the target concentration. For PFOS, PFOA, and PFHxS, we monitored and summed the signal for the branched and linear isomers and are reporting here total PFOS, PFOA, and PFHxS.

Water

PFAS measurements in well water samples were performed at NC State University using a high-performance liquid chromatograph (Agilent 1260 series) coupled to a triple quadrupole mass spectrometer (Agilent Ultivo). Each sample was injected twice (800 μL) and analyzed at high and low ion source temperature settings to maximize responses for all targeted PFAS [29]. Isotopically labeled internal standards (Table S1) were purchased from Wellington Laboratories (Guelph, ON, Canada) and diluted in methanol. Internal standards were dosed into each water sample in a liquid chromatography vial (targeted internal standard concentrations 500 ng/L, methanol concentration 10% by volume) and then vortexed. Analytes were chromatographically separated using a Zorbax RR Eclipse Plus C18 column (4.6 × 50 mm, 3.5 μm; Agilent). PFAS were detected using electrospray ionization in negative polarity mode and multiple reaction monitoring. Chromatography conditions and ion source parameters have been described previously [29]. Data acquisition and processing were performed using Agilent MassHunter Quantitative Analysis Version B.09.00. Calibration standards were prepared in deionized water acidified with nitric acid to match the sample matrix. All samples were run in one batch.

Calibration verification standards were prepared at 50 ng/L for GenX, PFBA, PFDA, PFNA, PFOA, and PFPeA with analytical standards obtained from a second source (SynQuest Labs, Alachua, FL, USA); all analytes were within 30% of 50 ng/L. Additionally, measurements of PFHxA, PFHpA, PFOA, and PFNA were within 30% of 10 ng/L and 50 ng/L preparations of National Institute of Standards and Technology (NIST) Standard Reference Material (SRM) 8446; however, PFDA measurements were 3 ng/L and 21 ng/L for 10 ng/L and 50 ng/L preparations of the SRM, respectively. Trip blanks, trip spikes, and matrix spikes were included in the analysis. Trip blanks (n = 13) returned concentrations for most analytes below the MRLs; however, PFBA in three blanks ranged from 7 to 9 ng/L (MRL = 5.0 ng/L), PFO4DA in one blank was 6 ng/L (MRL = 2.5 ng/L), PFDA in one blank was 3 ng/L (MRL = 2.5 ng/L), and PFBS in one blank was 6 ng/L (2.5 ng/L). The accuracy for 10 trip spikes prepared at 200 ng/L was within 30% for most analytes; however, PFDA in one trip spike was 31 ng/L, 8:2 FTS in two trip spikes was 85 and 93 ng/L, and PFOS in two trip spikes was 84 ng/L and 88 ng/L. The accuracy of matrix spikes at 100 ng/L was within 30% for most analytes except, in two matrix spikes, recovery of 8:2 FTS was 145% and 158%. Two of the more problematic analytes, PFDA and 8:2 FTS, were ultimately detected in only one well sample and no samples, respectively. The average percent difference between measurements for 17 well samples collected in duplicate ranged from <1 % to 12% across the analytes.

Serum

Serum samples were analyzed for PFAS at the US EPA in Research Triangle Park, NC, following a previously described method [19]. Briefly, 50 μL of serum was added to 10 μL 0.1 M formic acid containing isotopically labeled internal standards (Table S1). The sample was vortex mixed and 350 μL cold (−20 °C) acetonitrile was added to precipitate proteins. The sample was then vortex mixed again and centrifuged at 10,000 rpm for 5 min. A 100-μL aliquot of the acetonitrile supernatant was placed in a liquid chromatography vial with 300 μL 2.5 mM ammonium acetate buffer. Samples were analyzed using a Thermo Vanquish ultra-performance liquid chromatograph coupled to a Thermo Orbitrap Fusion mass spectrometer. Four batches for serum samples were run and batch-specific MRLs were calculated.

Calibration curves were prepared in charcoal stripped fetal bovine serum. NIST 1957 SRM was analyzed as a quality control measure for PFOS, PFOA, PFHxS, and PFNA [3]. Due to low signal-to-noise ratio for PFMOAA, we could not get reliable measurements for this PFEA in serum. Additionally, an interference prevented reliable measurements for PEPA and PFBA. We ultimately obtained measurements for 24 PFAS in serum (Table S1). A follow up investigation was performed on a subset of samples to analyze for 2,2,3,3-tetrafluoro-3-((1, 1, 1, 2, 3, 3-hexafluoro-3-(1, 2, 2, 2-tetrafluoro-ethoxy)propan-2-yl)oxy)propanoic acid (known as Hydro-EVE; CASN 773804-62-9). All samples that we evaluated for Hydro-EVE were non-detects (NDs).

Statistical methods

Summary statistics

We calculated summary statistics for each PFAS in water and serum using results for one blood sample per participant and one well water sample per well. All water samples were analyzed for PFAS in one analytical batch and, therefore, there was only one MRL per PFAS. To calculate summary statistics for PFAS concentrations in water, values less than the MRL were substituted with the MRL divided by $\sqrt{2}$. For serum samples, we replaced values less than the batch-specific MRL with the overall lowest MRL across batches divided by $\sqrt{2}$ to avoid creating an artificial rank ordering of NDs. PFAS detection frequency in serum was calculated based on the batch-specific MRL. These values were used in subsequent analyses. However, when we summed the mass concentration of detectable PFEAs to determine ∑PFEAs in well water, we added 0 to the total for PFEAs that were below the MRL. We used Spearman correlation coefficients to evaluate correlation among PFAS in serum or water. We limited the correlation analysis to those PFAS detected in at least 20% of serum samples or 20% of well water samples.

Often PFAS are classified as being ultra short-chain, short-chain, or long-chain compounds. Chain length naming is challenging for ether-based compounds because it is unclear whether the oxygen atoms should be counted in the chain length. We therefore used molecular weight (a cutoff of 400 g/mol) to distinguish between high and low molecular weight PFAS in our analysis. This cutoff is based on the molecular weight of PFHxS (400.11 g/mol) as the lightest, long-chain perfluoroalkyl acid (PFAA) based on the Organization for Economic Cooperation and Development (OECD) definition for long-chain PFAAs (perfluoroalkyl carboxylic acids [PFCAs] with seven perfluorinated carbon atoms and perfluoroalkyl sulfonic acids [PFSAs] with six perfluorinated carbon atoms) [30].

Regression modeling

To evaluate the relationship between water PFAS concentrations and blood serum PFAS concentrations, we chose well water PFAS (rather than tap water PFAS) as our water measure of interest. While some participants’ homes had point-of-use (POU) water treatment, most POU systems were installed within a year of blood sample collection. Therefore, well water would be more representative than tap water as the long-term PFAS source. We used this measure in all regression models. To estimate the association of well water to serum PFAS concentration, we built linear regression models for the three PFAS detected in >50% of serum samples and >50% of wells: PFOA, PFOS, and Nafion byproduct 2. We natural log-transformed the serum PFAS concentrations and then presented regression model results for well water concentration and well water concentration adjusted for age.

For Nafion byproduct 2, the only PFAS detected frequently in both well water and serum, we further explored the relationship between well water and serum concentrations using generalized estimating equation (GEE) regression modeling. GEE modeling controls for non-independence of participants who live in the same household. Even though GEE does not require the strict assumption of normality, we continued to use log-transformed serum concentrations of Nafion byproduct 2 in GEE models because a substantial percentage of Nafion byproduct 2 was non-detectable (>40%) and serum concentrations were low (<1 ng/L). Two outliers (defined as having a concentration more than three standard deviations from the mean) were removed from the dataset prior to modeling. Upon visual inspection of model residuals, one of the outliers (well water concentration: 165 ng/L, serum concentration: ND) had created an influential leverage point in the model. The second outlier (well water concentration: 55 ng/L; serum concentration: 21 ng/mL) was a participant who had reported working at the Fayetteville Works site for > 20 years at the time of serum collection and worked with Nafion.

In unadjusted GEE models, we first explored different Nafion byproduct 2 well water exposure metrics: well water concentration alone (a continuous variable), well water concentration × daily intake rate (aka daily dose estimate), and two measures of a cumulative dose, one that included daily intake rate (well water concentration × daily intake rate × years lived at current address) and one that did not (well water concentration × years lived at current address). After considering these water variables, we included the following covariates a priori in a final GEE model: ever worked at Fayetteville Works (yes or no), hours spent outside the home (a categorical variable of 1 [0 hours per week], 2 [1–35 h per week], or 3 [>35 h per week]), and eating home grown vegetables (yes or no). The selection of covariates was guided by previous studies of factors associated with serum PFAS in communities with water contamination. Well water concentration, being male, growing one’s own vegetables, and employment at a fluorochemical facility were associated with elevated serum PFOA in a subset of C8 Health Project participants who drank from private wells [20]. We adjusted for race (white or other races) because race has been reported in some studies as a determinant of serum PFAS [31, 32]. While age is often used as a surrogate for duration of exposure, we used well water concentration × years lived at current address as a measure of long term well water exposure because it was more specific to the exposure period than age. Because age was moderately correlated with duration at address (Fig. S1), we did not include age in the final GEE model. We also did not include variables which may have contributed to Nafion byproduct 2 levels in wells (e.g., prevailing wind direction at the well location relative to the fluorochemical manufacturing facility [33]) because Nafion byproduct 2 concentration in well water was already included in the model. We elected to report results of the full model, even for covariates that were not statistically significantly associated with serum Nafion byproduct 2 concentrations.

Results

Study population

In February 2019, we enrolled 153 individuals who used private wells, including 137 adults and 16 children (Table 1), into The GenX Exposure Study. Participants ranged in age from 9 to 85 years (median = 56 years). The participants came from 85 households with 84 private wells (two households shared one well). Figure S2 shows approximate well locations relative to Fayetteville Works fluorochemical manufacturing facility. Participants resided within 9250 m (approximately six miles) of Fayetteville Works. Our targeted recruitment was relatively effective with 54% of households (n = 46) having wells with ≥140 ng/L GenX based on NC DEQ measurements taken between September 2017 and April 2018. The number of participants using each well ranged from 1 to 5 although most households (84%) had only 1 or 2 participants (Table 1). The median years lived at current address was 15 years (interquartile range [IQR] = 8, 27 years) with most people (83%) having lived at their current address for more than five years.

Table 1.

Demographic characteristics of 153 private well users enrolled in the GenX Exposure Study, February 2019.

Characteristic	n	%	Median	25th percentile	75th percentile
Participant private well water GenXa
GenX <140 ng/L	66	43
GenX ≥140 ng/L	87	57
Household private well water GenXa
GenX <140 ng/L	39	46
GenX ≥140 ng/L	46	54
Age (years) (continuous)			56	42	66
Years at current address (continuous)			15	8	27
Adults/child
Adult (≥18 years)	137	90
Child	16	11
Gender
Female	83	54
Male	70	46
Race/Ethnicity
White, Non-Hispanic	127	83
Other	26	17
Hours spent outside home
0 h per week	62	40
1–34 h per week	15	10
≥35 h per week	76	50
Adults reported having ever worked at Fayetteville Works facility
No	125	82
Yes	12	8
Current drinking water source
Bottled	119	78
Well water	24	16
Other (e.g., municipal)	10	7
Eat home grown vegetables
No	66	43
Yes	87	57
Number of households	85
Wells sampledb	84
Number of participants in household
1	39	46
2	32	38
3	8	9
4–5	6	7

^a.Based on North Carolina Department of Environmental Quality GenX measurements in private wells, September 6, 2017 through April 25, 2018.

^b.Two households shared one well

The lag time between cessation of well water consumption and serum sample collection varied across participants. In February 2019, 119 participants (85%) reported drinking bottled water or another type of water (e.g., municipally treated), with the majority (88%) having started bottled water consumption in 2017 (n = 70, 59%) or 2018 (n = 35, 29%). Only 24 of the 153 participants reported currently drinking well water and the median GenX concentration in their 17 wells was 15 ng/L (range = ND-366 ng/L).

Well water concentrations

Most wells (74%) had ≥6 detectable PFEAs; the median ∑PFEAs across wells was 842 ng/L (IQR = 197–1760 ng/L). The PFAS fingerprint in well water was dominated by low molecular weight (<300 g/mol) PFEAs. Perfluoro-2-methoxypropanoic acid (PMPA) was the most frequently detected PFAS (87%) and had the highest median concentration (342 ng/L, IQR = 143–748 ng/L), followed in detection frequency by GenX (85%, median = 107 ng/L), and perfluoro-2-ethoxypropanoic acid (PEPA) (82%, median = 81 ng/L) (Table 2, Table S2). Higher molecular weight compounds were less frequently detected. Perfluoro(3,5,7-trioxaoctanoic) acid (PFO3OA, 312.044 g/mol) was detected in 63% of wells (median = 8 ng/L, IQR = ND-26 ng/L) and perfluoro(3,5,7,9-butaoxadecanoic) acid (PFO4DA, 378.05 g/mol) in 43% of wells generally at levels at or below 10 ng/L (median = ND, IQR = ND-5 ng/L). Only six of the 84 wells had detectable levels of perfluoro(3,5,7,9,11-pentaoxadodecanoic) acid (PFO5DoA, 444.057) (median = ND, IQR = ND-ND, range of detectable values: 3–13 ng/L; MRL: 2.5 ng/L).

Table 2.

Summary statistics for 30 PFAS in well water and blood serum samples from 84 private wells and 153 private well users in the GenX Exposure Study, February 2019.

	Well water (ng/L)				Serum (ng/mL)
	PFAS	MRL (ng/L)	n > MRL (%)	Median	Max	MRLa (ng/mL)	n > MRL (%)	Median	Max
Perfluoroalkyl ether carboxylic acids (PFECAs)	PFMOAA	10	62 (74)	43	258	NM
	PMPA	5	73 (87)	342	1368	0.2–2	0 (0)	ND	ND
	PFO2HxA	5	65 (77)	107	835	0.1–4	11 (7)	ND	0.2
	PEPA	2.5	69 (82)	81	468	NM
	PFO3OA	2.5	53 (63)	8	144	0.4–0.5	1 (1)	ND	0.6
	HFPO-DA (GenX)	2.5	71 (85)	107	918	0.4–2.5	0 (0)	ND	ND
	PFO4DA	2.5	36 (43)	ND	41	0.4–1	5 (3)	ND	1.8
	Hydro-EVE	2.5	18 (21)	ND	6	1	0 (0)	ND	ND
	PFO5DoA	2.5	6 (7)	ND	13	0.4–1	53 (35)	ND	17.3
Polyfluoroalkyl ether sulfonic acids (PFESAs)	NVHOS	2.5	59 (70)	3	11	0.1–0.2	9 (6)	ND	4.9
	Nafion byproduct 2	2.5	61 (73)	14	165	0.1–0.4	85 (56)	0.14	20.8
Perfluoroalkyl carboxylic acids (PFCAs)	PFBA	5	56 (67)	7	38	NM
	PFPeA	2.5	55 (65)	5	85	0.2–2	4 (3)	ND	0.8
	PFHxA	2.5	37 (44)	ND	60	0.1–0.2	10 (7)	ND	5.1
	PFHpA	2.5	22 (26)	ND	19	0.2	44 (29)	ND	0.8
	PFOA	2.5	31 (37)	ND	55	0.2–1	149 (97)	2.7	17.2
	PFNA	2.5	1 (1)	ND	4	0.2	151 (99)	0.9	4.4
	PFDA	2.5	1 (1)	ND	35	0.1–0.4	140 (92)	0.51	4.0
	PFUnDA	NM				0.1–1	64 (42)	ND	1.3
	PFTrDA	NM				0.2–10	6 (4)	ND	0.6
Perfluoroalkyl sulfonic acids (PFSAs)	PFBS	2.5	19 (23)	ND	12	0.1–1	26 (17)	ND	4.8
	PFPeS	2.5	0 (0)	ND	ND	0.1–0.2	26 (17)	ND	0.4
	PFHxS	2.5	4 (5)	ND	26	0.1–2	153 (100)	3.1	23.5
	PFHpS	2.5	0 (0)	ND	ND	0.1–0.2	145 (95)	0.4	2.4
	PFOS	2.5	20 (24)	ND	114	0.4–1	152 (99)	9.3	72.3
	PFNS	NM				0.1–0.2	0 (0)	ND	ND
	PFDS	NM				0.1–0.2	3 (2)	ND	0.2
Fluorotelomer sulfonic acids (FTS)	4:2 FTS	2.5	0 (0)	ND	ND	0.1–0.2	9 (6)	ND	0.7
	6:2 FTS	5	0 (0)	ND	ND	0.2	41 (27)	ND	11.5
	8:2 FTS	5	0 (0)	ND	ND	0.1–0.2	6 (4)	ND	0.3

PFAS are listed in order of increasing average molecular weight within each chemical class.

NM not measured, ND not detected, MRL method reporting limit.

^aThe range in MRL across four analytical batches is shown.

There were several strong (Spearman’s rho >0.70), positive correlations between PFEAs in well water. Well water levels of PEPA were nearly perfectly positively correlated with GenX (rho =0 .98), PMPA (rho = 0.98), and PFMOAA (rho = 0.96) (Table S7). PEPA concentrations in well water were also very strongly correlated to concentrations of other low molecular weight PFAS (NVHOS, rho = 0.89; PFO3OA, rho = 0.88; and PFPeA, rho = 0.86). PFPeA was also strongly correlated to PFBA (rho = 0.86). Nafion byproduct 2 was strongly correlated to PFO2HxA (rho = 0.84), PFO3OA (rho = 0.82), and PFMOAA (rho = 0.81).

In terms of historically used PFAS, perfluorooctanoic acid (PFOA) and perfluorooctanesulfonic acid (PFOS) were occasionally detected in wells. PFOA was detected in 37% of wells and PFOS was detected in 24%. There was considerable variability across wells, particularly for PFOS (range= ND-114 ng/L,). The variability in PFOS concentrations in wells (coefficient of variation [CV] = 3.01) was more than three times the variability in PMPA concentrations (CV = 0.84), for example. Well water levels of PFOA and PFOS were correlated (rho = 0.70). While PFOS was not strongly correlated to any of the other measured PFAS, PFOA was strongly correlated with PFHpA (rho = 0.84) and PFHxA (rho = 0.75).

Serum concentrations

Table 2 shows detection frequency and median and maximum concentrations for PFAS in participants’ serum (see also Table S3). Six PFAS (PFOS, PFHxS, PFOA, PFNA, PFDA, and PFHpS) were detected in nearly all (>90%) participants’ serum. Of the nine PFEAs analyzed in blood, five were detected: Nafion byproduct 2 (56% of participants), PFO5DoA (35%), PFO2HxA (7%), NVHOS (6%), and PFO4DA (3%). Nafion byproduct 2 was detected in 31% of children (Table S4). Correlations in serum were not as high as in well water; the highest pairwise correlation in serum was between PFOS and PFHpS (rho = 0.88) (Table S8). Most correlations in serum were <0.7 and the PFEAs (specifically Nafion byproduct 2 and PFO5DoA which had high enough detection frequencies to assess correlation) had correlations <0.5.

There was little overlap in the PFAS detected in well water and the PFAS detected in serum; low molecular weight PFAS were frequently detected in well water whereas high molecular weight PFAS were frequently detected in serum (Fig. 1). Two low molecular weight PFEAs (PMPA and GenX), which were detected in >75% of wells, were not detected in any blood samples, including samples from 24 participants who reported continuing to drink well water up until the date of blood collection.

Fig. 1: PFAS detection frequency in well water and blood serum.

PFAS detection frequency in well water samples from 84 wells and serum samples from 153 private well users for 13 PFAS measured in both well water and serum samples and which were detected in either >50% of well water samples or >50% of serum samples. PFAS are in order of increasing molecular weight from left to right; chemical structures for PMPA and PFDA are shown above their respective detection frequencies to illustrate the lowest average molecular weight (230.038 g/mol) and highest (514.086 g/mol) among the PFAS shown.

Modeling results

We initially examined the relationship between well water and serum levels for PFOS, PFOA, and Nafion byproduct 2 individually using linear regression models (unadjusted and adjusted for age) (Table S5). While well water concentration was not associated with PFOS or PFOA serum levels, Nafion byproduct 2 concentration in well water was significantly positively associated with Nafion byproduct 2 concentration in serum in both unadjusted and age-adjusted models (R² for model with both age and well water concentration = 0.16). We further explored the relationship between Nafion byproduct 2 concentrations in serum and water using GEE modeling to control for potential within-household correlation and excluding two outliers (n = 151). In the unadjusted GEE model for serum Nafion byproduct 2, a 10 ng/L increase in well water concentration was associated with an 18% increase in serum concentration.

When we evaluated three different metrics of Nafion byproduct 2 exposure from well water, the metric that explained the greatest proportion of Nafion byproduct 2 serum variability was a cumulative dose metric (well water concentration × years lived at current address) (R² = 0.101, p value = 0.02) (Table 3). The incorporation of daily water intake into an exposure metric did not further explain the proportion of serum PFAS variability. Using well water concentration × years lived at current address as the metric for cumulative Nafion byproduct 2 well water dose, we then expanded our GEE model to include potential covariates that may influence serum levels. Figure 2 illustrates effect estimates for the final GEE model adjusted for various factors (the values are presented in Table S6). The median value for cumulative Nafion byproduct 2 dose was 150 ng-year/L (IQR = 48, 523 ng-year/L). Higher values of the cumulative dose were significantly associated with higher serum Nafion byproduct 2 levels (p value = 0.036). A 100 ng-year/L increase in cumulative well water dose, representing, for example, a 100 ng/L increase in concentration and a one-year duration or a 10 ng/L increase in concentration and a 10-year duration, was associated with a 4% increase in Nafion byproduct 2 serum concentration. In terms of non-drinking water exposures, individuals who reported spending time outside the home on a regular basis had significantly lower Nafion byproduct 2 serum levels compared to those who did not. Factors related to higher Nafion byproduct 2 serum levels were eating home grown vegetables and being employed at the Fayetteville Works site, but these values were not statistically significant. Overall, the adjusted GEE model explained 14% of the Nafion byproduct 2 serum variability.

Table 3.

GEE model results of natural log-transformed serum Nafion byproduct 2 concentrations (ng/mL) for 151 private well users in the GenX Exposure Study, February 2019, with different well water exposure metrics.

Well water exposure metric	β (95% CI)	p value	R2
Nafion byproduct 2 concentration (ng/L) in well water	0.0181 (0.0025, 0.0337)	0.023	0.076
Daily dose
Nafion byproduct 2 well concentration (ng/L) × Daily consumption (categorical; 12-oz glasses/day)	0.0059 (0.0004, 0.0115)	0.037	0.056
Cumulative dose
Nafion byproduct 2 well concentration (ng/L) × Daily consumption (categorical; 12-oz glasses/day) × Duration at address (years)	0.0002 (0.0000, 0.0003)	0.06	0.074
Nafion byproduct 2 well concentration (ng/L) × Duration at address (years)	0.0005 (0.0001, 0.0008)	0.02	0.101

Fig. 2: GEE regression model results for serum Nafion byproduct 2 in private well users.

Plot of exponentiated beta (β) coefficients (black circle) and 95% confidence intervals (whiskers) for adjusted GEE model of natural log-transformed serum Nafion byproduct 2 concentrations (ng/mL) for 151 private well users in the GenX Exposure Study, February 2019. Partial β coefficients of the independent variables may be interpreted as % change in serum concentrations of Nafion byproduct 2 per unit of change in the independent variable, calculated as % change = [1 – exp(B)] x 100. The dotted red line shows an exponentiated β of 1. Cumulative well water dose, which was represented as Well concentration (ng/L) × Years at current address, is shown at a smaller scale in the inset figure.

Discussion

We aimed to investigate water-blood associations for PFEAs in this private well community but were constrained by the lack of overlap in frequently detected PFEAs in both water and serum. The combination of (1) some time lag between when bottled water consumption began and serum sample collection occurred and (2) likely short serum half-lives for the low molecular weight compounds relative to higher molecular weight PFAS, probably contributed to the lower serum detection frequencies in our study. The biological half-lives for most PFEAs are unknown but the human half-life of GenX was estimated to be, on average, 3.4 days [34]. In August 2018, GenX was not detectable in serum of 30 people living nearby the Fayetteville Works facility (whose wells had up to 4000 ng/L GenX) [35]. All participants in that investigation had been using bottled water for 4–14 months prior to blood sample collection. Some lag between cessation of exposure and biological sample collection is not uncommon in studies of environmental contamination where, upon discovery of contamination, exposure stops but it takes time to organize biological sample collection [36, 37].

Environmental exposure assessments often use measurements of blood to characterize chemical exposures [38]. Since 1999–2000, assessment of exposure to PFOS, PFOA, and some other PFAS in the US general population has relied on measuring PFAS serum concentrations in participants in the Centers for Disease Control and Prevention National Health and Nutrition Examination Survey (NHANES) [39, 40]. With growing evidence for the widespread presence of low molecular weight PFAS in drinking water [41,42,43], a reliance on serum biomarkers for exposure characterization provides incomplete information. In the private well community in our study, serum biomarkers could not be used to characterize long-term exposure to low molecular weight PFEAs (e.g., PMPA, GenX, PFO2HxA) despite that these compounds were frequently detected in well water. There is a need for identification of other biological matrices that are effective to characterize emerging PFAS exposures. For example, higher detection frequencies in whole blood have been reported for some short-chain PFAS (e.g., PFHxA) from previous comparisons between whole blood and serum for biomonitoring [44]. On the other hand, serum biomonitoring highlighted participants’ exposure to two higher molecular weight PFEAs, Nafion byproduct 2 and PFO5DoA, which were more frequently detected in serum than in well water, possibly because of their longer serum half-lives and bioaccumulation. Serum measurements in repeat samples from 44 people suggested the half-life of Nafion byproduct 2 and PFO5DoA are ~300 and ~380 days, respectively [45]. The analysis of paired well water samples provided important complementary information in our PFAS exposure assessment. Historical exposure estimation based on our well water measurements may be helpful to support an epidemiologic assessment of PFEA exposure.

The Fayetteville Works facility is the only known source of PFEAs in the Cape Fear River Basin region, NC. The PFEAs detected in wells nearby Fayetteville Works are due to air emissions of PFEAs or their acyl fluoride precursors from the facility and subsequent atmospheric deposition and infiltration downward to the groundwater [7, 29]. Indeed, the high correlations we observed among several PFEAs in well water is consistent with a common source. Although there is a paucity of published data on PFEA levels in air samples from nearby Fayetteville Works, results from one study of air samples collected September 2019-March 2020 suggest the facility emitted several PFAS, including PFEAs, to air [24]. In weekly PM_2.5 filter samples, PFBA, PFHxA, PFHxDA, PFOS, PMPA, NVHOS, PFO5DoA, and Nafion byproduct 1 had maximum concentrations > 1 pg/m³ and contributed the most to the total measured PFAS concentration. Nafion byproduct 2 was detected in almost all PM_2.5 samples collected at nearfield sites southwest and northeast of the facility [24]. Nafion byproduct 2 was also detected in pine needles, which serve as passive air samplers, collected within 2 miles of the facility in 2017–2018 [46].

The predominance of low molecular weight PFEAs (PMPA, PEPA, GenX, PFO2HxA, PFMOAA) in well water samples near Fayetteville Works, which is consistent with other PFAS measurement of groundwater near the facility [29, 47], is likely related to their lower retardation in soils and, therefore, greater mobility to groundwater, compared with higher molecular weight compounds [29]. Due to differences in soil retardation factors across PFAS, the PFAS fingerprint in well water is not necessarily reflective of the PFAS fingerprint in air emissions from the facility. Additionally, the PFAS fingerprint in water from private wells near Fayetteville Works had some distinct features compared with municipal water impacted by wastewater discharges from Fayetteville Works; for example, PMPA and PEPA were also detected in the municipal water derived from the lower Cape Fear River but not as the dominant PFAS species [7].

Some of the PFAS we detected in wells may also be related to historic emissions and discharges from the Fayetteville Works facility. PFOA was produced at Fayetteville Works for several years starting in 2002 [48]. The high PFOA-PFHxA and PFOA-PFHpA correlations we observed in wells are consistent with previous research that reported detection of these compounds in a 2006 sample of the facility’s waste ditch, which discharged to the Cape Fear River [39]. PFHpA was the most abundant among the targeted PFAS, but other PFCAs, including PFHxA, PFOA, and PFNA, were also detected in the sample [49]. In 2012, Strynar et al. (2015) collected water samples from the Cape Fear River upstream and downstream of the facility’s wastewater discharge point and reported substantial increases in PFBA, PFPeA, and PFHpA concentrations in the downstream sample [15].

Because the estimated serum half-life of ~300 days for Nafion byproduct 2, the highest molecular weight PFEA detected frequently in wells, is expected to be longer than that of lower molecular weight PFEAs, we were able to detect Nafion byproduct 2 in >50% of private well users’ serum. While serum Nafion byproduct 2 concentrations in private well users were substantially lower than concentrations we previously reported for downriver municipal water users in 2017–2018 (median=2.7 ng/mL, CV = 0.69) [19], there was more variability in serum levels to explain in the private well community (CV = 2.7). In February 2019, we targeted recruitment of approximately half participants with GenX in wells ≥140 ng/L and half with GenX <140 ng/L. Based on PFAS results for 770 wells tested by NC DEQ in 2017–2018, which is the dataset we had access to during our recruitment, ~20% of wells had GenX >140 ng/L compared with ~50% of our participants’ wells, indicating that our sample over-represented wells with higher GenX. Because Nafion byproduct 2 was highly correlated to GenX in well water (rho = 0.76), our recruitment based on GenX levels in wells also worked to give us a cohort with a broader range of Nafion byproduct 2 exposures. The detection frequency and median concentrations of Nafion byproduct 2 in private wells in our study are consistent with measurements of groundwater and streambeds in the region (detected in 78% of samples at a median concentration of 16.8 ng/L) [29], suggesting that our population is representative of the distribution of Nafion byproduct 2 exposure in the area.

When we explored different well water exposure metrics for Nafion byproduct 2, we found that cumulative dose (well water concentration x years lived at address), an integrated measure of exposure, explained more variability in serum concentrations than well water concentration alone. Incorporating daily intake rate into an exposure metric (either to create daily dose or a cumulative dose that included daily intake rate) did not result in a metric that explained more serum variability. In studies of contaminated public water systems, drinking water daily intake rate has been used as a surrogate for drinking water exposure. In 691 people exposed to PFOA through public drinking water, people who reported drinking >1.5 L water daily had about 2-fold increased PFOA plasma levels compared with those who reported drinking <0.25 L/day [50]. In 1,578 people with a contaminated public drinking water supply in New Hampshire, PFHxS serum levels increased with increasing self-reported water consumption but the same relationship was not found for PFOA or PFOS [51], which may indicate there were other sources of PFOA and PFOS exposure besides drinking water. The calculation of daily dose may have been more sensitive to measurement error than cumulative dose due to the significant influence of daily intake rate. The number of 12-oz glasses of water consumed per day may not have been recalled correctly or the response in February 2019 may not have been representative of daily intake over the full exposure period. We also acknowledge that the cumulative dose metric of well water concentration multiplied by years of residence at the current address may have error. Firstly, it is unclear when private wells near Fayetteville Works became contaminated; Fayetteville Works emissions began as early as 1980, but it would have taken some time for PFAS to arrive at individual wells. We also do not know what the PFAS levels in well water were over the 15 years, on average, that our participants reported living at their address, and whether the well water samples we collected in 2019 were a good representation of past exposures. For 82 of the 84 wells in our study, NC DEQ had collected and analyzed water samples for GenX less than two years before we collected samples (median time difference = 14 months, IQR = 12, 16 months). In 59 of the 82 wells (72%), our GenX result was within 50% difference of the NC DEQ result, but in 13 wells (16%) the GenX concentration decreased >50% and in 10 wells the GenX concentration increased >50% (Table S9).

Our adjusted GEE model explained about 15% of variation in Nafion byproduct 2 serum levels in this population of private well users. The adjusted GEE model explained a similar proportion of the variance in serum to what the linear model with age and Nafion byproduct 2 well water concentration predicted. In a model of PFOA serum levels in 17,516 people who participated in the C8 Health Project and who used contaminated public water, the model explained 68% of variance in PFOA serum concentrations, with residence in the contaminated water district accounting for 39% and cumulative years of residence explaining 1.5% of the variance [22]. PFOA serum concentrations were higher (median serum concentrations of 241 and 69 ng/mL PFOA in the two highest-exposed water districts) and PFOA has a longer half-life (2.3 years [52]) than Nafion byproduct 2, which would make serum PFOA levels more stable estimates of PFOA exposure than serum Nafion byproduct 2 levels would be for Nafion byproduct 2 exposure. We had a much smaller sample size (n = 151) than the PFOA study but a similar size to a previous study of PFOA blood and water levels in a private well community [20]. In that study, each 1 ng/L increase in drinking water PFOA concentration was associated with a 0.14 ng/mL [95% confidence interval (CI) = 0.13–0.15] increase in serum concentrations. Additionally, growing one’s own vegetables, being male, and being employed at a fluorochemical facility were associated with elevated serum PFOA levels. We found that time spent outside the home was significantly associated with lower Nafion byproduct 2 serum levels in our population, suggesting that participants were consuming water elsewhere or there may be other household related exposures (e.g., indoor air, dust, or locally caught fish or wildlife) that are important contributors to Nafion byproduct 2 serum concentration other than well water consumption. Nafion byproduct 2 has been detected in serum from fish caught in the Cape Fear River [53]. However, only four of the 153 participants in our study reported eating fish caught from the Cape Fear River so we did not include river fish consumption as a covariate in our model.

Compared with the general US population, private well users residing near Fayetteville Works had elevated serum levels of some higher molecular weight PFAS (PFOS, PFHxS, PFOA, and PFNA) which were produced in the US historically but reportedly have been phased out of production since the early 2000s. The median serum PFOS concentration among adults in our study was about double the median serum PFOS (4.3 ng/mL) and three-fold higher than the median serum PFHxS (1.1 ng/mL) levels for the US population in 2017–2018. The elevated levels are consistent with results reported for 30 private well owners residing near Fayetteville Works in a previous investigation in 2018 [35]; a few of these individuals participated in our study as well. Given that well water concentration of long-chain PFAAs was not significantly associated with serum concentration in samples collected in February 2019, there may have been (1) exposure to higher concentrations in well water in the past or (2) other sources of exposure in the Fayetteville area besides contaminated well water.

Conclusion

Low molecular weight PFEAs from a fluorochemical manufacturing facility were frequently detected in nearby private wells but infrequently detected in serum samples from local groundwater users. Nafion byproduct 2, a higher molecular weight PFEA, was frequently detected in both well water and serum. Cumulative well water dose was associated with significantly higher Nafion byproduct 2 serum levels and time spent outside the home was associated with lower levels. Serum levels of PFOS, PFOA, PFHxS, and PFNA, which were higher than national estimates, were not associated with well water levels. In this community, serum biomarkers were not good measures of long-term exposure to low molecular weight PFEAs found in drinking water.

Data availability

The datasets generated and analyzed during the current study are not publicly available due to human subjects protections of these data. Data may be made available from the corresponding author on reasonable request with IRB approval.