Characterization of Per- and Polyfluoroalkyl Substance (PFAS) concentrations in a community-based sample of infants from Samoa

Abstract

Per- and polyfluoroalkyl substances (PFAS) are persistent contaminants with documented harmful health effects. Despite increasing research, little attention has been given to studying PFAS contamination in low- and middle-income countries, including Samoa. Using data and biosamples collected through the Foafoaga o le Ola (“Beginning of Life”) Study, which recruited a sample of mothers and infants from Samoa, we conducted an exploratory study to describe concentrations of 40 PFAS analytes in infant cord blood collected at birth (n=66) and infant dried blood spots (DBS) collected at 4 months post-birth (n=50). Of the 40 PFAS analytes tested, 19 were detected in cord blood, with 10 detected in >50% of samples (PFBA, PFPeA, PFOA, PFNA, PFDA, PFUnA, PFTrDA, PFHxS, PFOS, and 9Cl-PF3ONS); and 12 analytes were detected in DBS, with 3 detected in >50% of samples (PFBA, PFHxS, and PFOS). PFAS concentrations were generally lower than those reported in existing literature, with the exception of PFHxS, which was detected at higher concentrations. In cord blood, we noted suggestive (p<0.05) or significant (p<0.006) associations between higher PFHxS and male sex; higher PFPeA and residence in Northwest ‘Upolu (NWU) compared to the Apia Urban Area (AUA); lower PFUnA and 9Cl-PF3ONS and greater socioeconomic resources; lower PFOA and higher parity; higher PFDA and higher maternal age; and lower PFUnA, PFTrDA, and 9Cl-PF3ONS and higher maternal BMI. In DBS, we found suggestive (p<0.05) or significant (p<0.025) associations between lower PFBA and residence in NWU versus AUA; lower PFBA and PFHxS and higher maternal age; and higher PFBA and higher maternal BMI. Finally, we observed associations between nutrition source at 4 months and DBS PFBA and PFHxS, with formula- or mixed-fed infants having higher concentrations compared to exclusively breastfed infants. This study represents the first characterization of PFAS contamination in Samoa. Additional work in larger samples is needed to identify potentially modifiable determinants of PFAS concentrations, information that is critical for informing environmental and health policy measures.

Graphical Abstract

1. INTRODUCTION

Per- and polyfluoroalkyl substances (PFAS) are a group of man-made chemicals that emerged in the 1940s. Lauded for their unique properties of resistance to oil, grease, water, and heat, PFAS became ideal candidates for a wide range of applications, such as stain- or water-resistant fabrics/carpeting, fast food packaging, non-stick cookware, and fire-fighting foams.^1,2 While the chemical structures of PFAS are responsible for these desirable properties, they also enable many PFAS to mimic fatty acids.^3,4 As a result, PFAS are bioavailable and many bioaccumulate and interfere with normal physiological processes.⁵

In fact, alarming evidence has linked PFAS exposure to a myriad of negative health effects across the lifespan, including increased cholesterol levels, liver dysfunction, lower birth weight, elevated blood pressure, endocrine disruption, immune hazards, and more.^6,7 The pervasiveness of PFAS contamination and its correlated health impacts is not confined to populated/urban areas but extends to remote locations as well. For example, in the Faroe Islands, PFAS have been detected in human blood and have been associated adverse health effects, particularly immune dysfunction.⁸ This underscores the far-reaching consequences of PFAS contamination, even in geographically isolated regions, emphasizing the urgent need for global PFAS mitigation strategies.

Further, despite a growing body of PFAS research, instances of ‘regrettable substitution’—replacing banned PFAS chemicals with structurally similar, yet unregulated alternatives—pose significant challenges.⁹ This practice complicates monitoring and management efforts, adding layers of complexity to predicting the health consequences of PFAS mixtures, especially given the vast array of PFAS chemicals (thousands in total).¹⁰ While these challenges are pronounced in high-income countries with established regulatory bodies, their impact is magnified on a global scale, reaching even regions with limited resources such as low- and middle-income countries.

Unfortunately, research efforts on PFAS in low- and middle-income countries, including Samoa, have been limited^11–14 despite rapid modernization and increasing prevalence of chronic disease in these regions.^15,16 Understanding if the mixtures of PFAS exposures in Samoa are different than mixtures in other populations with longer histories of development is an essential first step in exploring potential health implications in this setting. In turn, this information is also potentially valuable as a starting point for informing public health policies and interventions. On a broader scale, the exploration of PFAS contamination in Samoa is critical to addressing the global challenge of environmental justice. This is important because communities with limited environmental resources and regulatory frameworks sometimes face disproportionately higher relative risks when it comes to PFAS contamination.

Therefore, the purpose of this exploratory study was to begin to address this critical knowledge gap by characterizing PFAS concentrations in cord blood, collected at birth, and dried blood spots (DBS), collected at 4 months post-birth, in infants from Samoa, and characterize concentrations across a variety of infant, maternal, and household level factors. The focus on infants is particularly significant, as early-life exposure to PFAS may have far-reaching consequences on health and development into adulthood.¹⁷ In light of the potential risks associated with PFAS exposure, understanding the levels of these substances around the world is of paramount importance to support continued research to improve health equity and to inform policy measures to safeguard public health globally.

2. MATERIALS AND METHODS

2.1. Study overview, design, and setting

Data were derived from a longitudinal observational study of mother-infant dyads from Samoa. Participants were recruited through the 2017-2019 Foafoaga o le Ola (“Beginning of Life”) parent study, which focused on identifying factors in pregnancy and the early postpartum period that contributed to non-communicable diseases in the Samoan setting.¹⁸ Dyads (n=161) were enrolled before birth at the antenatal care clinic of the Tupua Tamasese Meaole (TTM) Hospital in Apia, Samoa. At only about 2,800 km², Samoa is home to approximately 222,000 people.¹⁹ Samoa is geographically isolated and made up of two islands including ‘Upolu, where the majority of the population resides, and Savaii (Figure 1). ‘Upolu contains a single urban center, Apia, in which approximately 35,000 residents live, while the rest of the population lives in semi-rural or rural villages. The participants in this study resided in different geographic regions across the island of ‘Upolu. The majority of participants resided in the Apia Urban Area (AUA, urban) and Northwest ‘Upolu (NWU, periurban), while only a handful of participants resided in the Rest of ‘Upolu (ROU, rural) region.

Figure 1.

Map depicting census regions and airports of Samoa.

Created using QGIS. Data sources: Global Map of Samoa © ISCGM / The Ministry of Natural Resources and Environment, Samoa Ministry of Natural Resources (https://github.com/globalmaps/gmws10, polbnda layer) and NextGIS (Airports, Settlements, and Waterways layers). The circle on the map indicates the approximate area in which research participants were resident.

Parent study inclusion criteria for infants were as follows: their mothers were over 18 years old, 35-41 weeks gestation, had uncomplicated singleton pregnancies, planned to give birth at TTM Hospital, and lived within 30-minutes of Apia for easy follow-up. Written informed consent was obtained from all mothers, and the study was approved by Institutional Review Boards at Yale University, the University of Pittsburgh, and the Health Research Committee of Samoa. As part of the Foafoaga o le Ola study, trained bilingual Samoan research assistants assessed the mother-infant dyads at 1-week, 2-months, and 4-months post-birth. Data on behavior, environment, and physical measurements were collected, and biospecimens were collected for later analysis. Additional inclusion criteria for this ancillary study included the availability of stored biospecimens from which to quantify PFAS concentrations, focusing on infant cord blood collected at birth and infant DBS collected at 4 months post-birth.

2.2. Biospecimen collection

At the pre-birth recruitment visit, the medical charts of women who agreed to provide a cord blood sample were flagged in the medical record. Infant cord blood samples were obtained by trained clinical staff immediately post-birth, with approximately 6 mL collected into a syringe, and 3 mL transferred to a serum separator tube. Standard processing procedures were followed, and serum was stored in cryovials at −80°C. Samples were shipped on dry ice to Yale University for storage, and then to the University of Pittsburgh for PFAS measurement.

At 4 months post-birth, infant capillary blood was collected on a Whatman 903 Protein Saver Card by trained research staff using a heel stick procedure. After cleaning the skin, the blood collection card was placed in contact with the heel to allow the blood to saturate the designated circles. The blood spots were air dried and refrigerated at 4°C, and then shipped at room temperature to Yale University for storage, and then to the University of Pittsburgh for PFAS data collection.

2.3. PFAS measurement

We analyzed concentrations (ng/mL) of 40 PFAS analytes (Table S1) in all available cord blood samples (n=66). Quantification was performed based on established methods with some modifications.²⁰ Specifically, 100 μL of serum was transferred into a 2-mL polypropylene tube and then spiked with 24 isotopic PFAS (Wellington Laboratories; the list of compounds and the amount spiked can be found in Table S2) as surrogate standards. Next, 250 μL of 0.1 M formic acid and 750 μL of methanol were added to each tube, followed by vortexing for 10 s and then sonication for 10 min. The supernatant was transferred into a new tube and 750 μL of methanol was added into the sample tube for a second round of extraction. The combined supernatants were concentrated to 100 μL, transferred into LC vials, spiked with injection internal standards of 7 isotopic PFAS (Wellington Laboratories; Table S2), and stored at 4°C for instrumental analysis. PFAS quantification was performed based on methods described in US EPA Draft 2 Method 1633²¹ using a Thermo Scientific TSQ Quantum triple quadrupole mass spectrometer (LC-MS/MS).

Method validation was performed prior to analysis of participant samples by spiking native PFAS (ranging from 0.2 to 2 ng depending on the analyte, Table S1) and isotopic PFAS (Table S2) as surrogates into deionized water; recoveries ranged from 62.8% to 111%. The limits of detections (LOD) for PFAS were determined as the lowest concentrations of spiked matrix that met a signal-to-noise ratio of 3:1 on the instrument and ranged from 0.1 to 2.5 ng/mL (Table S1). Samples were run in batches of approximately 20 samples. Each batch included two laboratory blanks (deionized water for cord blood samples and blank cards for DBS samples) and 1 sample for duplicate. The blanks and duplicates were prepared in the same way as the samples. None of the PFAS analytes targeted in this study were detected in the blank samples. The relative standard deviation of duplicate samples ranged from 3.53-22.6%. The isotope dilution method was used for quantifying PFAS, producing recovery-corrected results. The recoveries of surrogate PFAS ranged from 57.8% to 123%.

In an exploratory arm of the study, 50 purposefully selected DBS cards were chosen based on visual inspection to ensure adequate blood content for analysis. Specifically, cards were selected if at least two full punches (full 6 mm circles) could be obtained to provide sufficient sample for the detection of PFAS concentrations. These uniform punches were taken using a 6 mm QIAGEN UniCore Punch Kit (Fisher Scientific) from the center part of the blood circles. The hole punch instrument was cleaned between punches to prevent contamination across samples. Based on availability, two to five holes were punched out from a single sample. The sample preparation and instrumental analysis of PFAS were identical to those used for the cord blood samples, except that we were not able to include duplicate samples for DBS data due to limited blood volume. Blood volume was estimated based on the assumption that a single 6 mm punch contained 13.2 microliters of blood.²² Analyte-specific concentrations were calculated by dividing the detected mass by 0.0132 mL multiplied by the number of punches.

2.4. Phenotype data

To explore PFAS concentrations across samples, we used phenotype data collected through the Foafoaga o le Ola study.¹⁸ Before birth, a comprehensive set of maternal and household characteristics was collected through medical record review and questionnaires. These characteristics included age, years of education, relationship status, census region of residence, socioeconomic resources, and parity. Social data included the material lifestyle score, a measure of socioeconomic resources commonly used in Samoa. The score is calculated as the number of 15 possible household assets owned by participants (house, fridge, freezer, stereo, portable speaker, television, VCR/DVD, couch, carpet/rugs, washing machine, landline telephone, mobile phone, computer/laptop, electricity, and motor vehicle).^23,24 Infant date of birth and sex were parent-reported at a 1-week assessment. At each visit, maternal height/weight were also collected to calculate body mass index (BMI). Breastfeeding status was reported by mothers at each follow-up assessment, distinguishing between exclusively breastfed and formula/mixed fed infants (combined because of the high rate of exclusive breastfeeding and small sample size). To further describe and characterize the sample, infant birth weight and length were obtained from hospital medical records, and follow-up weight and length were measured using standardized instruments—an infant scale and length board.¹⁸

Of note, dyads were enrolled pre-birth, which resulted in some parents consenting for infant cord blood collection but being lost to follow-up immediately post-birth. Consequently, phenotype data, including infant sex, were missing for some infants. However, to comprehensively characterize PFAS concentrations, data were still collected from these samples.

2.5. Statistical analyses

All statistical and graphical descriptions were conducted using R version 4.2.1.²⁵ Standard descriptive statistics were computed based on each variable’s level of measurement to provide a comprehensive overview of the data. The 25^th, 50^th (median), 75^th, and 95^th percentiles were reported across PFAS data. Sample data distributions and patterns were examined both graphically and statistically. Rain cloud or sina with violin plots, created using the ggplot2 R package²⁶, were used to describe the sample data. Across the majority of analytes, LODs resulted in censored data or “non-detects” below these limits. To visualize data distributions more clearly, values below the LOD were retained on the plots, graphically depicted as LOD/2. Plots were created for all analytes with observations (i.e., >LOD values) for more than a single sample.

We then examined formal differences in cord blood and/or DBS PFAS concentrations by infant sex, census region of residence, family socioeconomic resources, infant nutrition source, and maternal parity, age, and BMI. In analyzing the censored data, we avoided introducing bias through arbitrary estimates (i.e., replacing non-detect values with the LOD or LOD/2) and instead employed more robust approaches that accounted for the underlying distribution of censored data as recommended elsewhere.²⁷ Specifically, we used regression using maximum likelihood estimation implemented via the cencorreg function from the NADA2 R package.²⁸ Given the data skewness, PFAS data were log transformed. We reported $\widehat{\beta }$, regression estimates for natural log transformed PFAS concentrations, and $\widehat{\beta }$ 95% confidence intervals. To aid in interpretation, we also reported $exp(\widehat{\beta })$, the exponentiated regression estimate, and percent change, calculated as $100\ast (exp(\widehat{\beta })-1)$. Except for the sex-focused regression, all models controlled for sex. In addition, DBS analyses controlled for infant age in days. Because maximum likelihood estimation depends on detections to calibrate the model, and because we were working with small sample sizes, these analyses focused on PFAS analytes that were detected in >50% of samples.

Finally, to explore the relationship between PFAS concentrations across time, correlation plots were created using Kendall’s Tau-B and p-values calculated using the ATS function from the NADA2 R package.²⁸ These analyses were performed only for analytes detected in >50% of samples across both tissues and accounted for censored data.

Given the exploratory nature of this study, associations with a p-value <0.05 were considered “suggestive”. However, to control the type I error rate, we adjusted for multiple testing by determining the number of effective tests (N_eff) based on the correlation structure of the PFAS data using the meff() function from the poolr R package.²⁹ We then used the Bonferroni correction to compute a threshold for associations considered “significant” by dividing 0.05 by the number of effective tests (p < 0.05/N_eff). Associations with a p-value <0.006 or <0.025 were considered “significant” for cord blood and DBS data, respectively.

3. RESULTS

3.1. Participant characteristics

A flow chart depicting sample availability and selection is presented in Figure S1. Sample characteristics are presented in Table 1, and compared with the parent study in Table S3. The cord blood and DBS samples consisted of 66 (45% female) and 50 (54% female) infant participants, respectively. Twenty-six participants overlapped across the two samples. All infants were born at full term (>37 gestational weeks) and mean birth weight ranged from 3.47 kg to 3.60 kg across the two samples. The mean maternal age was approximately 27 years, and over 50% of the sample were residents of the AUA geographic region (Table 1). Characteristics of participants with biospecimens were similar to those without biospecimens (Table S3).

Table 1.

Participant characteristics.

Participant Information from Baseline or Birth
Characteristic	Cord blood n = 66	DBS (4 months) n = 50
Infant sex, n (%)
Female	29 (45%)	27 (54%)
Male	36 (55%)	23 (46%)
Missing	1a	0
Census region, n (%)
AUA	39 (60%)	26 (52%)
NWU	26 (40%)	22 (44%)
ROU	0 (0%)	2 (4.0%)
Missing	1	0
Socioeconomic resources as no. household assetsc, mean (SD)	8.1 (2.8)	8.2 (2.9)
Missing	14	10
Parityb, mean (SD)	1.90 (1.59)	2.07 (1.53)
Missing	25	21
Maternal age in years, mean (SD)	27.33 (5.91)	27.50 (6.37)
Missing	2	0
Maternal BMI in kg/m2, mean (SD)	35.58 (5.80)	33.87 (5.35)
Missing	7	5
Participant Information 4 months post-birth
Characteristic	Cord blood n = 66	DBS (4 months) n = 50
Infant age in days, mean (SD)	126.63 (9.72)	128.98 (16.36)
Missing	7	1
Maternal BMI in kg/m2, mean (SD)	35.01 (6.56)	34.00 (5.77)
Missing	14	9
Infant nutrition source, n (%)
Exclusively breastfed	37 (64%)	35 (74%)
Formula/mixed fed d	21 (36%)	12 (26%)
Missing	8	3

^aLost to follow-up between pre-delivery recruitment and 1-week infant assessment (during which sex was self-reported by parents);

^bParity, 12 nulliparous, 6 primiparous, 23 multiparous in cord blood and 6 nulliparous, 5 primiparous, 18 multiparous in DBS;

^cSocioeconomic resources based on number of items owned from the 15-item material lifestyle score questionnaire;

^dFormula or mixed-fed categories were combined due to small sample size (2 infants were exclusively formula fed); to further characterize the sample, additional participant characteristics are presented in Table S3.

3.2. Description of PFAS concentrations

PFAS concentrations are summarized in Table 2 and depicted graphically in Figure 2. Of the 40 PFAS analytes tested, 19 were detected in cord blood, with 10 detected in >50% of samples (PFBA, PFPeA, PFOA, PFNA, PFDA, PFUnA, PFTrDA, PFHxS, PFOS, and 9Cl-PF3ONS) while 12 were detected in DBS, with 3 detected in >50% of samples (PFBA, PFHxS, and PFOS). Across cord blood and DBS samples, PFHxS stood out as an analyte detected at high concentrations (median 6.61 and 1.84 ng/mL, respectively). For the DBS data, only PFBA exceeded this with a median concentration of 3.44 ng/mL. Across PFAS classes, perfluoroalkyl carboxylic acids were detected at the highest rates, followed by perfluoroalkyl sulfonic acids. The correlation (Kendall’s Tau or Kendall’s Tau-B) between analytes ranged from −0.18 to 0.57 (cord blood) and 0.06 to 0.49 (DBS) (Figures S2 and S3).

Table 2.

Description of PFAS concentrations (ng/mL) in cord blood samples (birth) and dried blood spots (4 months post-birth).

	Cord blood (n = 66)						Dried blood spots (n = 50)
			Percentile (ng/mL)						Percentile (ng/mL)
Analyte	Detection frequency n (%)	Range	25th	50th	75th	95th	Detection frequency n (%)	Range	25th	50th	75th	95th
Perfluoroalkyl carboxylic acids
PFBA*	59 (89.4)	ND – 3.27	1.16	1.35	1.55	2.11	37 (74.0)	ND – 13.13	0.78	3.44	5.79	9.31
PFPeA*	44 (66.7)	ND – 0.83	-	0.38	0.48	0.73	1 (2.0)	ND – 1.52	-	-	-	-
PFHxA	6 (9.1)	ND – 0.19	-	-	-	0.12	0	ND – ND	-	-	-	-
PFHpA	18 (27.3)	ND – 0.24	-	-	0.11	0.14	0	ND – ND	-	-	-	-
PFOA*	64 (97.0)	ND – 4.49	0.43	0.57	0.68	1.13	2 (4.0)	ND – 1.12	-	-	-	-
PFNA*	48 (72.7)	ND – 0.35	-	0.15	0.20	0.28	0	ND – ND	-	-	-	-
PFDA*	53 (80.3)	ND – 0.33	0.1	0.12	0.14	0.18	1 (2.0)	ND – 0.61	-	-	-	-
PFUnA*	64 (97.0)	ND – 0.79	0.18	0.22	0.27	0.37	1 (2.0)	ND – 2.11	-	-	-	-
PFDoA	4 (6.1)	ND – 0.19	-	-	-	0.11	1 (2.0)	ND – 7.41	-	-	-	-
PFTrDA*	37 (56.1)	ND – 0.39	-	0.11	0.14	0.21	5 (10.0)	ND – 3.18	-	-	-	1.57
PFTeDA	4 (6.1)	ND – 0.16	-	-	-	0.11	2 (4.0)	ND – 1.47	-	-	-	-
Perfluoroalkyl sulfonic acids
PFBS	6 (9.1)	13.53	-	-	-	3.69	0	ND – ND	-	-	-	-
PFPeS	0	ND – ND	-	-	-	-	1 (2.0)	ND – 0.70	-	-	-	-
PFHxS*	66 (100)	2.19 – 30.47	4.23	6.61	9.45	14.78	48 (96.0)	ND – 13.89	1.14	1.84	3.43	9.34
PFHpS	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
PFOS*	66 (100)	0.52 – 5.81	0.96	1.14	1.47	2.31	50 (100.0)	0.49 – 15.27	1.00	1.52	3.11	8.40
PFNS	3 (4.5)	ND – 0.34	-	-	-	-	1 (2.0)	ND – 1.47	-	-	-	-
PFDS	1 (1.5)	ND – 0.11	-	-	-	-	0	ND – ND	-	-	-	-
PFDoS	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
Fluorotelomer sulfonic acids
4:2FTS	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
6:2FTS	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
8:2FTS	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
Perfluorooctane sulfonamides
PFOSA	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
NMeFOSA	1 (1.5)	ND – 21.67	-	-	-	-	0	ND – ND	-	-	-	-
NEtFOSA	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
Perfluorooctane sulfonamidoacetic acids
NMeFOSAA	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
NEtFOSAA	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
Perfluorooctane sulfonamide ethanols
NMeFOSE	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
NEtFOSE	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
Per- and Polyfluoroether carboxylic acids
HFPO-DA	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
ADONA	3 (4.5)	ND – 0.58	-	-	-	-	0	ND – ND	-	-	-	-
PFMPA	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
PFMBA	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
NFDHA	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
Ether sulfonic acids
9Cl-PF3ONS*	37 (56.1)	ND – 1.09	-	0.39	0.47	0.61	0	ND – ND	-	-	-	-
11Cl-PF3OUdS	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
PFEESA	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
Fluorotelomer carboxylic acids
3:3FTCA	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
5:3FTCA	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-
7:3FTCA	0	ND – ND	-	-	-	-	0	ND – ND	-	-	-	-

This table is ordered by class and chain length. The * symbol is used to flag analytes detected in >50% of samples for either cord blood or dried blood spots. “ND” and the ‘-’ symbol indicate non-detect values (i.e., < limit of detection values as presented in Table S1).

Figure 2.

Rain cloud plots describing PFAS concentrations.

Figure 2 depicts PFAS concentrations in (A) cord blood collected at birth (n = 66) and (B) dried blood spot collected at 4 months post birth (n = 50). Graphics are presented only for analytes with detectable values in more than one sample, designed for sample descriptive purposes only. Detection frequencies for each PFAS are listed in Table 2. Analytes are colored by class, and the dashed line represents the analyte-specific limit of detection (LOD). Black dots signify observed values, while red dots indicate values below the LOD (depicted graphically as LOD/2). The numbers in the bottom left comer of plot facets indicate the count of values below the LOD. The figure is ordered by class and chain length.

3.3. PFAS concentrations and participant characteristics

To better understand PFAS concentrations by participant characteristics, we examined associations between natural log transformed PFAS concentrations and infant and household level factors of sex, census region of residence, and socioeconomic resources as shown in Table 3. For the cord blood data (Table 3a), we observed a significant association between natural log transformed PFHxS and sex, with higher concentrations in male infants compared with female infants ($\widehat{\beta }=0.376\left[0.131\text{to}0.620\right]$, p = 0.003). While controlling for neonate sex, we also observed a suggestive association between natural log transformed PFPeA and census region of residence, with higher concentrations in infants residing in NWU compared to AUA ($\widehat{\beta }=0.389\left[0.052\text{to}0.727\right]$, p = 0.024). Finally, we observed suggestive associations between greater socioeconomic resources and lower natural log transformed PFUnA ($\widehat{\beta }=\text{−}0.045\left[\text{−}0.081\text{to}\text{−}0.010\right]$, p = 0.012) and 9Cl-PF3ONS ($\widehat{\beta }=\text{−}0.049\left[\text{−}0.085\text{to}\text{−}0.012\right]$, p=0.009) while controlling for sex. For the DBS data (Table 3b), no statistically significant differences were observed between PFAS concentrations and sex while controlling for infant age or socioeconomic resources while controlling for infant age and sex. However, we observed a suggestive association between census region of residence and natural log transformed PFBA ($\widehat{\beta }=\text{−}0.308\left[\text{−}0.596\text{to}\text{−}0.020\right]$, p = 0.036) while controlling for infant age and sex, with lower PFAS concentrations observed in NWU compared with AUA.

Table 3.

Results of regression using maximum likelihood estimation (MLE) examining associations between natural log transformed PFAS concentrations (dependent variable) and infant and household-level factors of sex, census region of residence, and socioeconomic resources.

a. Cord blood (birth)
	Sex, Male vs Female (n = 65)						Census Region, NWU vs AUA (n = 65) (Model adjusted for neonate sex)						Socioeconomic Resources (n = 52) (Model adjusted for neonate sex)
Analyte	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	%Change	p	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	% Change	p	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	% Change	p
PFBA	−0.018	−0.276	0.241	0.982	−1.757	0.893	−0.165	−0.419	0.088	0.848	−15.248	0.201	−0.001	−0.056	0.055	0.999	−0.063	0.982
PFPeA	0.065	−0.236	0.366	1.067	6.717	0.672	0.389	0.052	0.727	1.476	47.613	0.024	0.019	−0.045	0.083	1.019	1.947	0.554
PFOA	−0.150	−0.400	0.099	0.860	−13.971	0.237	0.040	−0.238	0.317	1.040	4.031	0.780	0.010	−0.036	0.056	1.010	0.967	0.682
PFNA	−0.178	−0.466	0.109	0.837	−16.347	0.224	−0.119	−0.352	0.115	0.888	−11.188	0.320	0.003	−0.039	0.046	1.003	0.346	0.872
PFDA	−0.177	−0.452	0.098	0.838	−16.233	0.207	0.009	−0.131	0.148	1.009	0.859	0.904	−0.004	−0.034	0.026	0.996	−0.396	0.796
PFUnA	−0.094	−0.343	0.156	0.911	−8.934	0.462	0.024	−0.157	0.205	1.024	2.435	0.794	−0.045	−0.081	−0.010	0.956	−4.427	0.012
PFTrDA	−0.122	−0.450	0.207	0.886	−11.448	0.468	−0.005	−0.239	0.230	0.995	−0.458	0.969	−0.029	−0.075	0.018	0.972	−2.828	0.224
PFHxS	0.376	0.131	0.620	1.456	45.598	0.003	0.092	−0.157	0.342	1.097	9.676	0.468	−0.003	−0.054	0.048	0.997	−0.327	0.900
PFOS	0.093	−0.151	0.338	1.098	9.759	0.455	−0.032	−0.281	0.218	0.969	−3.135	0.802	0.004	−0.047	0.054	1.004	0.358	0.890
9Cl-PF3ONS	−0.077	−0.406	0.252	0.926	−7.429	0.645	0.069	−0.098	0.236	1.072	7.157	0.417	−0.049	−0.085	−0.012	0.952	−4.778	0.009
b. Dried blood spots (4 months)
	Sex, Male vs Female (n = 49) (Model adjusted for infant age)						Census Region, NWU vs AUA (n = 48) (Model adjusted for infant age and sex)						Socioeconomic Resources (n = 40) (Model adjusted for infant age and sex)
Analyte	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	% Change	p	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	% Change	p	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	% Change	p
PFBA	−0.171	−0.456	0.113	0.843	−15.718	0.238	−0.308	−0.596	−0.020	0.735	−26.508	0.036	0.041	−0.013	0.096	1.042	4.206	0.140
PFHxS	−0.081	−0.363	0.202	0.922	−7.781	0.576	−0.233	−0.520	−0.053	0.792	−20.785	0.110	−0.024	−0.078	0.031	0.976	−2.352	0.390
PFOS	−0.134	−0.417	0.148	0.875	−12.549	0.352	−0.193	−0.479	0.093	0.825	−17.520	0.187	0.010	−0.045	0.064	1.010	0.975	0.726

Table ordered by class and chain length.

$\widehat{\beta }$ = regression estimate for natural log transformed dependent variable (PFAS analyte).

$exp(\widehat{\beta })$ = exponentiated regression estimate for dependent variable (PFAS analyte); for every one unit increase in X (participant characteristic), the value of Y (PFAS analyte) is $exp(\widehat{\beta })$ times the original value while holding covariates constant.

% Change = percent increase or decrease in Y (PFAS analyte level) for every one unit increase in X (infant or household characteristic) while holding covariates constant; calculated as % $\mathrm{Change}=100\ast (exp(\widehat{\beta })\text{−}1)$.

2.5^th = lower bound of 95% confidence interval; 97.5^th = upper bound of 95% confidence interval for $\widehat{\beta }$.

Bolded p-values indicate statistically suggestive associations (p<0.05) while bolded and underlined p-values indicate statistically significant associations (p<0.006 for cord blood data and p<0.025 for dried blood spot data).

NWU, Northwest ‘Upolu; AUA, Apia Urban Area.

Socioeconomic resources based on number of items owned from the 15-item material lifestyle score questionnaire.

Next, we examined associations between natural log transformed PFAS concentrations and maternal parity, age, and BMI as shown in Table 4. For the cord blood data (Table 4a), we observed a statistically significant association between natural log transformed PFOA and parity while controlling for infant sex, with lower PFOA concentrations in children born to women with higher parity ($\widehat{\beta }=\text{−}0.155\left[\text{−}0.259\text{to}\text{−}0.051\right]$, p = 0.003). We also observed a suggestive association between natural log transformed PFDA and maternal age, with higher PFDA concentrations among children born to women with higher age ($\widehat{\beta }=0.013\left[0.001\text{to}0.025\right]$, p = 0.034), as well as suggestive associations between maternal BMI and natural log transformed PFUnA ($\widehat{\beta }=\text{−}0.018\left[\text{−}0.033\text{to}\text{−}0.003\right]$, p = 0.017), PFTrDA ($\widehat{\beta }=\text{−}0.023\left[\text{−}0.045\text{to}\text{−}0.001\right]$, p = 0.043), and 9Cl-PF3ONS ($\widehat{\beta }=\text{−}0.014\left[\text{−}0.027\text{to}\text{−}0.001\right]$, p = 0.033), with lower PFAS concentrations observed in children of women with higher BMI. For the DBS data (Table 4b), we observed statistically significant associations between maternal age and natural log transformed PFBA ($\widehat{\beta }=\text{−}0.047\left[\text{−}0.072\text{to}\text{−}0.022\right]$, p = 0.0002) and PFHxS ($\widehat{\beta }=\text{−}0.044\left[\text{−}0.068\text{to}\text{−}0.019\right]$, p = 0.0005), with lower PFAS concentrations observed in children of women with higher age, as well as a statistically significant association between natural log transformed PFBA and maternal BMI, with higher PFBA concentrations observed in women with greater BMI ($\widehat{\beta }=0.091\left[0.033\text{to}0.089\right]$, p = 0.00002).

Table 4.

Results of regression using maximum likelihood estimation (MLE) examining associations between natural log transformed PFAS concentrations (dependent variable) and maternal factors of parity, age, and BMI.

a. Cord blood (birth).
	Parity (n = 41) (Model adjusted for infant sex)						Maternal age (n = 64) (Model adjusted for infant sex)						Maternal BMI (n = 59) (Model adjusted for infant sex)
Analyte	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	% Change	p	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	% Change	p	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	% Change	p
PFBA	−0.035	−0.138	0.067	0.966	−3.449	0.502	−0.008	−0.032	0.016	0.992	−0.819	0.499	0.006	−0.018	0.029	1.006	0.593	0.622
PFPeA	−0.032	−0.173	0.110	0.969	−3.103	0.663	0.003	−0.027	0.033	1.003	0.278	0.857	0.001	−0.030	0.033	1.001	0.124	0.939
PFOA	−0.155	−0.259	−0.051	0.856	−14.365	0.003	0.021	−0.002	0.044	1.021	2.108	0.073	0.010	−0.013	0.033	1.010	1.013	0.393
PFNA	−0.042	−0.132	0.048	0.959	−4.118	0.361	−0.004	−0.026	0.017	0.996	−0.432	0.689	−0.011	−0.032	0.010	0.989	−1.074	0.313
PFDA	−0.046	−0.096	0.004	0.955	−4.475	0.072	0.013	0.001	0.025	1.013	1.330	0.034	−0.012	−0.025	0.001	0.988	−1.190	0.063
PFUnA	0.008	−0.067	0.082	1.008	0.775	0.839	−0.003	−0.019	0.013	0.997	−0.294	0.712	−0.018	−0.033	−0.003	0.982	−1.776	0.017
PFTrDA	−0.050	−0.138	0.038	0.951	−4.861	0.269	0.006	−0.015	0.027	1.006	0.603	0.580	−0.023	−0.045	−0.001	0.977	−2.269	0.043
PFHxS	−0.087	−0.184	0.011	0.917	−8.304	0.081	0.002	−0.019	0.023	1.002	0.190	0.860	0.014	−0.008	0.037	1.014	1.446	0.205
PFOS	−0.068	−0.165	0.030	0.935	−6.531	0.174	0.007	−0.014	0.028	1.007	0.714	0.510	−0.009	−0.031	0.013	0.991	−0.899	0.426
9Cl-PF3ONS	−0.021	−0.073	0.031	0.979	−2.091	0.425	−0.001	−0.014	0.012	0.999	−0.103	0.878	−0.014	−0.027	−0.001	0.986	−1.377	0.033
b. Dried blood spots (4 months).
	Parity (n = 29)*						Maternal age (n = 49) (Model adjusted for infant age and sex)						Maternal BMI (n = 41) (Model adjusted for infant age and sex)
Analyte	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	% Change	p	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	% Change	p	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	% Change	p
PFBA	-	-	-	-	-	-	−0.047	−0.072	−0.022	0.954	−4.605	0.0002	0.038	0.010	0.066	1.039	3.891	0.007
PFHxS	-	-	-	-	-	-	−0.044	−0.068	−0.019	0.957	−4.268	0.0005	−0.010	−0.038	0.017	0.990	−1.044	0.456
PFOS	-	-	-	-	-	-	−0.023	−0.047	0.002	0.978	−2.230	0.070	−0.012	−0.039	0.016	0.989	−1.150	0.410

Table ordered by class and chain length.

$\widehat{\beta }$ = regression estimate for natural log transformed dependent variable (PFAS analyte).

$exp(\widehat{\beta })$ = exponentiated regression estimate for dependent variable (PFAS analyte); for every one unit increase in X (maternal characteristic), the value of Y (cord blood PFAS concentration) is $exp(\widehat{\beta })$ times the original value while holding covariates constant.

% Change = percent increase or decrease in Y (PFAS analyte level) for every one unit increase in X (maternal characteristic) while holding covariates constant; calculated as % $\mathrm{Change}=100\ast (exp(\widehat{\beta })\text{−}1)$.

2.5^th = lower bound of 95% confidence interval; 97.5^th = upper bound of 95% confidence interval for $\widehat{\beta }$.

Bolded p-values indicate statistically suggestive associations (p<0.05) while bolded and underlined p-values indicate statistically significant associations for cord blood (p<0.006) or dried blood spots (p<0.025).

^*Parity not examined in DBS due to sample size restrictions.

Next, we examined associations between DBS PFAS concentration and nutrition source at four months of life (Table 5, Table S5, and Figure S4). We observed associations between nutrition source and natural log transformed PFBA ($\widehat{\beta }=0.456\left[0.107\text{to}0.804\right]$, p = 0.010) and PFHxS ($\widehat{\beta }=0.434\left[0.087\text{to}0.782\right]$, p = 0.014) while controlling for infant age and sex, with higher PFAS concentrations observed in formula- or mixed-fed infants compared to exclusively breastfed infants.

Table 5.

Results of regression using maximum likelihood estimation (MLE) examining associations between natural log transformed dried blood spot (4 months) PFAS concentrations (dependent variable) and nutrition source (formula-fed or mixed-fed vs. exclusive breastfeeding as reference) while controlling for infant age and sex.

Analyte	$\widehat{\beta }$	2.5th	97.5th	$exp(\widehat{\beta })$	% Change	p
PFBA	0.456	0.107	0.804	1.577	57.716	0.010
PFHxS	0.434	0.087	0.782	1.543	54.342	0.014
PFOS	0.147	−0.201	0.494	1.158	15.778	0.408

n=47. Formula and mixed-fed nutrition source combined due to sample size (2 infants were exclusively formula fed).

$\widehat{\beta }$ = regression estimate for natural log transformed dependent variable (PFAS analyte).

$exp(\widehat{\beta })$ = exponentiated regression estimate for natural log transformed dependent variable (PFAS analyte); in formula or mixed fed infants compared to breastfed infants, the value of y (PFAS analyte) is $exp(\widehat{\beta })$ times the original value while holding infant age and sex constant.

% Change = percent increase or decrease in Y (PFAS analyte level) in formula or mixed fed infants compared to breastfed infants while holding infant age and sex constant; calculated as % Change = 100 * (exp(Est.)−1).

2.5^th = lower bound of 95% confidence interval; 97.5^th = upper bound of 95% confidence interval.

Bolded p-values indicate statistically suggestive associations (p<0.05) while bolded and underlined p-values indicate statistically significant associations (p<0.025).

Finally, we examined correlations between PFAS analytes that were detected in >50% of samples across both tissues (Figure S5; PFBA, PFHxS, and PFOS; n = 26). A statistically significant correlation was observed only for PFHxS (Kendall’s Tau-B = 0.29, p = 0.037).

4. DISCUSSION

4.1. Comparison with existing literature

We present the first data on PFAS contamination in infants from Samoa. Among the detected analytes, PFOS, PFHxS, and PFBA stood out as commonly detected analytes in both cord blood and DBS, which aligns with findings in other studies on PFAS.⁶ Notably, PFOS was present in all samples while PFHxS was observed in 100% of cord blood samples and 96% of DBS. Given the wide application of these chemicals both nationally and globally², and detection of these analytes in blood of 97% of US residents surveyed in 2011-2012³⁰, these observations are not surprising.

In comparing our top results with existing literature (Table S6, cord blood), the median concentration of cord blood PFOS detected in this study (1.14 ng/mL) was lower than those reported in cord blood in New York, US (6.32 ng/mL sampled in 2001-2002)³¹, Shanghai, China (2.51 ng/mL sampled in 2012-2013³² and 3.14 ng/mL sampled in 2013-2016³³), and comparable or higher than that reported in Hamamatsu, Japan (1.2 ng/mL sampled in 2010-2012³⁴), Belgium (0.99 ng/mL sampled in 2008-2014³⁵), and Canada (<LOD sampled in 2008-2011³⁶). Interestingly, much higher median concentrations of PFHxS were detected in this study (6.61 ng/mL in cord blood) compared to those reported in cord blood in existing literature (0.66 ng/mL in New York, US³¹; 0.15 ng/mL in Belgium³⁵; 0.18-0.46 ng/mL in Shanghai, China^32,34; <LOD in Canada³⁶). The high PFHxS concentrations could reflect a replacement transition after the phase-out of PFOS and PFOA since the 2000s.¹⁴ While PFHxS has also been regulated in the US, and some studies have reported serum PFHxS concentrations to be gradually decreasing over the past two decades³⁷, this long chain PFAS continues to be detected at high rates in the Pacific, including Vanuatu¹⁴ and now Samoa. This observation could be a result of continued import of products containing PFHxS from other counties that have no restriction on PFHxS, or a consequence of the long estimated mean biological half-life of PFHxS (2.84-8.5 years).³⁸

PFBA, also used as a replacement of PFOA in the USA³⁹, was detected at comparable or slightly higher concentrations than PFOA in this study, indicating the widespread use of and exposure to PFAS alternatives. PFBA is a short-chain PFAS that has not received a great deal of regulatory scrutiny as it is rapidly excreted, despite its association with thyroid hormone levels in humans⁴⁰ and systemic toxicity in a murine model.³⁹ It is worth mentioning that false detection of PFBA and PFPeA may occur due to analytical interferences.^41,42 Other analytes, including PFPeA, PFOA, PFNA, PFDA, PFUnA, PFTrDA, and 9Cl-PF3ONS, were also detected in over 50% of cord blood samples but at generally lower median concentrations compared to existing literature where comparisons existed.^31,32,43 Many of these are long-chain PFAS with a tendency to bioaccumulate. These findings, paired with evidence of PFAS contamination across the Asia Pacific region^14,44,45 and recent governmental efforts to modernize water infrastructure and reduce PFAS contamination in neighboring American Samoa⁴⁶ highlight the importance of instituting ongoing PFAS monitoring across Samoa.

4.2. Associations with participant factors

In further characterizing PFAS concentrations across participant characteristics, we observed associations with sex, geographic region, and socioeconomic resources, though the specific analytes and directions of effect were not generally consistent across tissues. For example, we observed an association between higher mean PFHxS and male sex in cord blood, but no associations in DBS (Table 3). These findings are not surprising given the generally mixed literature on this topic. For example, in a 2012-2013 study of cord blood PFAS concentrations from Shanghai, no differences in PFAS concentrations by sex were identified.³² Other literature, however, suggests that there are sex differences in specific analytes, including PFHxS, at older ages^47–49 and outcome-impacting sex-specific associations between PFAS and DNA methylation⁵⁰, an epigenetic regulator of gene expression.

In our examination of geographic region, our initial hypothesis was that infants residing in NWU (which is home to ‘Upolu’s two airports) would have higher PFAS concentrations compared to those in AUA. However, in the analysis of cord blood, only one analyte, PFPeA, showed an association with the geographic region in the expected direction. In the DBS data, PFBA demonstrated an association with geographic region, but unexpectedly, lower mean concentrations were observed in NWU compared to AUA (Table 3). While the geographic region measure used here has been associated with a variety of health outcomes in this setting^51,52, these broad categorizations are not ideal for examining environmental exposures. For instance, although both airports are technically located in NWU, the proximity of Fagali’i airport to Apia (situated in AUA as shown in Figure 1) complicates the analysis of associations between PFAS concentrations and geographic region in this study. Factors such as access to fast food or differences in water infrastructure in Apia as Samoa’s single urban center could contribute to this complexity.

In terms of socioeconomic resources, in higher-income countries chronic exposure to toxicants has been historically observed at higher rates in lower income communities that are often located closer to exposure sources⁵³, but the literature suggests that concentrations of PFAS generally increase with socioeconomic status indicators such as income, which is mainly attributed to discordant use of consumer products (e.g., ability to purchase expensive waterproof or stain resistant fabrics) or dietary differences (e.g, higher consumption of fresh marine food).^54,55 Samoa is a lower-middle income country, and our prior work has revealed that there are often unexpected directions of effect, particularly related to social variables.⁵⁶ In fact, we observed lower concentrations of cord blood PFUnA and 9Cl-PF3ONS in infants born to families with greater socioeconomic resources, and no associations in DBS (Table 3). It is possible that this cord blood observation is a result of the ability of families with greater socioeconomic resources to purchase fresh whole foods, compared with lower income families who must rely on cheaper processed and packaged foods. Additional work is needed to better understand potential differences in PFAS bioaccumulation by sex, geographic region, and socioeconomic resources in infants.

We also identified associations between PFAS and several maternal factors. Notably, an inverse association emerged between cord blood PFOA and parity (Table 4). This aligns with established knowledge linking lower PFAS concentrations, especially in the case of PFOA, PFOS, PFNA, and PFHxS, with higher parity in women, possibly due to placental transfer, blood loss during delivery, or lactation.⁵⁷ Additionally, children born to women with higher parity also exhibit lower PFAS concentrations.⁵⁷ However, a significant limitation of our study was the amount of missing parity data.

In the sample data, maternal age had moderate-strong positive correlation with parity (R=0.614, p=0.0004) but had fewer missing data points, leading us to explore it further. Given the correlation, we expected results similar to those with parity when examining associations between PFAS concentrations and maternal age. Surprisingly, while we noted a potential trend between cord blood PFOA and maternal age, the direction of effect differed from that with parity. PFDA ultimately emerged as the only cord blood analyte statistically associated with maternal age, demonstrating higher mean concentrations in older women. In DBS data, different analytes, PFBA and PFHxS, were associated with maternal age but with opposite directions of effect compared to cord blood. These associations contrast with findings from a large systematic review⁵⁷ that reported relationships between maternal age and child PFAS concentrations to be largely null, though the discordance could be attributed to differences in childbearing ages/parity across cultures.

In terms of associations between maternal BMI and child PFAS concentrations, existing literature is quite mixed.⁵⁷ Our findings in both cord blood and DBS revealed associations in this sample, though the directions of effect and specific analytes were, again, inconsistent across tissues. Specifically, we observed inverse associations between maternal BMI and cord blood PFUnA, PFTrDA, and 9Cl-PF3ONS, but positive associations with DBS PFBA. Various factors may contribute to these disparate findings, both across tissues/time points within this study and compared with the broader literature. Potential influences include variations in BMI-related factors such as food choices, socioeconomic status, social support, general population health, or body composition (e.g., fat mass vs lean mass). Future investigations should delve into these factors more comprehensively.

Finally, we observed higher DBS concentrations of two PFAS analytes, PFHxS and PFBA, in infants who were formula- or mixed-fed at 4 months post-birth compared to infants who were exclusively breastfed. These strong associations persisted in post hoc analyses even after controlling for socioeconomic resources (Table S5). These findings are inconsistent with prior work from the Faroe Islands that found, with the exception of PFHxS, PFAS concentrations generally increased at higher rates in breastfed infants than in formula-fed infants.⁵⁸ While the limited exploration of PFAS contamination in formula suggest concentrations are generally low^59,60, more research in this area is needed.⁶¹ In Samoa, where the majority of formula is powder and requires reconstitution with water, the observed higher PFAS concentrations in formula/mixed-fed infants could potentially be linked to water contamination due to the pressing need for improved water infrastructure. Alternatively, it might be associated with formula packaging. Notably, PFAS have been associated with delays and insufficiencies in mammary gland development in mice⁶² and, in other settings, reduced breastfeeding duration in humans.^63,64 In Samoa, however, we believe this is less likely to explain our findings as insufficient breast milk supply is not frequently reported.⁶⁵ Additional work is needed to understand PFAS risk from human milk or formula in this setting as modifiable nutrition-related interventions offers promise to reduce PFAS exposure or bioaccumulation.⁶⁶

4.3. Potential sources of PFAS contamination in Samoa

Sources of PFAS exposure are generally consistent with some variations based on the specific compound, use, environmental behavior, and potential for bioavailability and bioaccumulation. Some of the most common sources of human exposure include proximity to airports or military bases (where PFAS-containing fire-fighting foams have been historically used), consumption of PFAS-contaminated water or food (e.g., fish), and use of products made from PFAS.⁶⁷ Without conducting environmental monitoring studies, however, we cannot say with confidence what the main exposure sources of PFAS are in Samoa. Possible routes of PFAS exposure include contaminated food (e.g., ocean fish, canned foods, meat, dairy) or drinking water^68,69, with added potential transfer to infants through breastfeeding or formula preparation. While water treatment plants and national water standards exist across Samoa, not all families rely on public water sources and those who are dependent on rainfall and surface water may have additional potential exposure routes.⁷⁰ Other routes of potential exposure include soil contact, inhalation (e.g., air, dust), or consumer products via dermal exposure.^68,69 For example, PFAS have been detected in consumer products targeting infants and children (e.g., clothing, bedding, bibs), often lacking explicit disclosure of chemical additives.^71,72 Additional work is needed to survey Samoan individuals regarding types of consumer products that they use frequently, and perform environmental and product sampling.

4.4. Limitations

Despite the study’s strengths, including its longitudinal design and examination of 40 PFAS analytes across two tissues in a unique sample of infants from Samoa, there are limitations that should be acknowledged. First, given the exploratory nature of this study and challenges of collecting biospecimens in a low-resource setting like Samoa, the sample size was small. This was particularly problematic for variables with missing values such as parity and socioeconomic resources, or those with little variability such as nutrition status that forced us to combine groups for formal analyses. While the use of MLE is a strength compared to replacement methods which introduce bias, the sample size was on the threshold of adequacy, particularly for incomplete or sparse variables. To counter this, we implemented a censoring cutoff, requiring a minimum of 50% of the data be above the LOD, but we encourage caution when interpreting results. Ultimately, future investigations in larger sample sizes and supplemental investigations of environmental samples would contribute to a broader and clearer understanding of PFAS exposure in this setting. Finally, while DBS represent an important public health monitoring tool, particularly for low-resource settings, a major challenge of this tissue centers on blood volume uncertainty across punches as a product of discordant card saturation, blood viscosity, or pressure applied during punching. The PFAS concentrations in DBS were generally lower than that detected in cord blood, which were possibly a result of tissue type/blood volume. While we strategically adjusted for blood volume²², future directions include comparing PFAS concentrations across peripheral blood/DBS collected from the same individual, as well as comparing our blood volume normalization approach with other methods such as measuring specific gravity or hemoglobin content of samples.

5. Conclusion

In conclusion, this study presents the first evidence of PFAS contamination in Samoa and sheds light on a global environmental justice issue that impacts communities with limited resources and regulatory frameworks. Ultimately, the inclusion of Samoa’s experiences in the broader conversation on PFAS contamination strengthens the impetus for global collaborative efforts in addressing PFAS-related concerns. Moving forward, expanded investigations of PFAS contamination across environmental compartments and residents of Samoa are needed, including examination of associations of PFAS concentrations and health outcomes, as understanding determinants of PFAS concentrations and their effects is crucial for exposure estimation and implementing effective public health interventions.

Highlights:

• This study provides the first evidence of per- and poly-fluoroalkyl substance (PFAS) contamination in Samoa, shedding light on a previously unexplored issue.

• Concentrations of PFAS in the study sample of infants were generally lower than other populations, with the exception of elevated PFHxS concentrations.

• Further research with larger sample sizes is needed to understand health impacts of PFAS exposure and identify modifiable determinants of PFAS concentrations to guide effective policy measures.

Availability of Data:

The participants included in this study were not consented for data sharing. However, we recognize the importance of data sharing for scientific advancement and to ensure transparency and reproducibility of results and are therefore committed to providing detailed and aggregated results of analyses, or organizing partnerships that allow data sharing with approval from the Samoa Health Research Committee. Researchers interested in discussing this in more detail can contact the corresponding author at law145@pitt.edu for more information.