Plasma per- and polyfluoroalkyl substance mixtures during pregnancy and duration of breastfeeding in the New Hampshire Birth Cohort Study

Abstract

Background:

Prior studies suggest that prenatal per- and polyfluoroalkyl substances (PFAS) exposures are associated with shorter breastfeeding duration. Studies assessing PFAS mixtures and populations in North America are sparse.

Methods:

We quantified PFAS concentrations in maternal plasma collected during pregnancy in the New Hampshire Birth Cohort Study (2010–2017). Participants completed standardized breastfeeding surveys at regular intervals until weaning (n=813). We estimated associations between mixtures of 5 PFAS and risk of stopping exclusive breastfeeding before 6 months or any breastfeeding before 12 months using probit Bayesian kernel machine regression. For individual PFAS, we calculated the relative risk and hazard ratio (HR) of stopping breastfeeding using modified Poisson regression and accelerated failure time models respectively.

Results:

PFAS mixtures were associated with stopping exclusive breastfeeding before 6 months, primarily driven by perfluorooctanoate (PFOA). We observed statistically significant trends in the association of perfluorohexane sulfonate (PFHxS), PFOA, and perfluorononanoate (PFNA) (p-trends≤0.02) with stopping exclusive breastfeeding. Participants in the highest PFOA quartile had a 28% higher risk of stopping exclusive breastfeeding before 6 months compared to those in the lowest quartile (95% Confidence Interval: 1.04, 1.56). Similar trends were observed for PFHxS and PFNA with exclusive breastfeeding (p-trends≤0.05). PFAS were not associated with stopping any breastfeeding before 12 months.

Conclusions:

In this cohort, we observed that participants with greater overall plasma PFAS concentrations had greater risk of stopping exclusive breastfeeding before 6 months and associations were driven largely by PFOA. These findings further support the growing literature indicating that PFAS may be associated with shorter duration of breastfeeding.

Introduction

Breastfeeding is associated with well-known benefits including reduced risk of asthma, obesity, and early-life infections for children and lower risk of breast and ovarian cancers, Type 2 diabetes, and high pressure for the lactating parent (Centers for Disease Control and Prevention, 2023a; Meek and Noble, 2022). Up until 2022, exclusive breastfeeding was recommended until at least six months of age, and breastfeeding along with solid foods up to one year of age or beyond. This recommendation is now two years or longer per the World Health Organization Guidelines, American Academy of Pediatrics (AAP), and other agencies (Meek and Noble, 2022). Yet, only about one-third of mothers in the United States continue breastfeeding until 12 months (National Immunization Survey 2022). This is likely because of several known barriers, including sociocultural, medical, nutritional, and infant-level factors, ranging from issues with breastmilk supply and latching to competing responsibilities for work and family (Teich et al., 2014).

Aside from these known barriers, endocrine-active chemicals can interfere with hormonal mechanisms that regulate both the initiation of breastfeeding and sustained milk production (Abbott, 2009; Criswell et al., 2020). These chemicals include certain per- and polyfluoroalkyl substances (PFAS), which in experimental studies disrupt mammary gland differentiation (Tucker et al., 2015; White et al., 2007; Yang et al., 2009), interfere with placental hormones, or alter expression of genes encoding milk proteins and, by extension, milk composition (Criswell et al., 2020; Suh et al., 2011; White et al., 2007). Prior studies suggest that PFAS may decrease breastfeeding duration (Fei et al., 2010; Nielsen et al., 2022; Rokoff et al., 2023; Romano et al., 2016; Timmermann et al., 2022). For instance, greater maternal blood serum or plasma concentrations of perfluorooctanoate (PFOA) and perfluorooctane sulfonate (PFOS) during pregnancy have been associated with shorter duration of breastfeeding in multiple studies (Fei et al., 2010; Rokoff et al., 2023; Romano et al., 2016; Timmermann et al., 2017; Timmermann et al., 2022); an additional study reported similar associations when using residential address before delivery as a proxy for exposure to PFAS-contaminated water (Nielsen et al., 2022). In contrast, a large study from Norway reported a decreased risk of early cessation of breastfeeding with certain PFAS (perfluorononanoate (PFNA), perfluorodecanoate (PFDA), perfluoroundecanoate (PFUnDA)), and no association with PFOA and PFOS, which were observed at lower concentrations than most prior studies (Rosen et al., 2018). Similarly lower concentrations observed in a later U.S. study observed no association of PFAS with breastfeeding duration among primipara (Friedman et al., 2023). Few prior studies have examined influence of PFAS mixtures on breastfeeding duration (Friedman et al., 2023; Rokoff et al., 2023), and the majority of studies have been from Europe (Fei et al., 2010; Nielsen et al., 2022; Rosen et al., 2018; Timmermann et al., 2017; Timmermann et al., 2022) with fewer studies from North America (Friedman et al., 2023; Rokoff et al., 2023; Romano et al., 2016). To expand and strengthen these areas of research, we examined the potential effects of PFAS mixtures on duration of breastfeeding in a population of rural pregnant people recruited from prenatal clinics in New Hampshire, USA.

Materials and Methods

Study Population

The New Hampshire Birth Cohort Study (NHBCS) is an on-going prospective study that began in 2009. During prenatal care in a study clinic, eligible pregnant people are enrolled, as described previously (Gilbert-Diamond et al., 2016). Briefly, eligibility criteria included having a singleton pregnancy, English literacy, and use of a private water system (e.g.,well). Questionnaires and medical record reviews collect extensive information on lifestyle, diet and demographic factors as well as pregnancy and birth outcomes. Biological samples are also collected for analysis, including blood samples during gestation as described below. Participants in the current study were enrolled between 2010 (when implementation of the breastfeeding questionnaire began) and 2017. Participants were interviewed 4, 8, and 12 months following delivery regarding breastfeeding practices as well as the introduction of water, juice or other liquids, formula, non-human milk, and solid foods into the infant diet. Participants were asked to report either the date or the child’s age in weeks/months at the last feeding with human milk. Participants were asked about their breastfeeding history prior to the study pregnancy, including number of children previously breastfed and the duration of any breastfeeding for each previously breastfed child. Previous breastfeeding is a route of maternal PFAS excretion (Barbarossa et al., 2013; Criswell et al., 2023; Mondal et al., 2014) and previous breastfeeding history is a strong predictor of future breastfeeding success (Nagy et al., 2001; Whalen and Cramton, 2010) and therefore an important confounder when investigating the effect of PFAS on breastfeeding duration (Timmermann et al., 2023). The study was approved by the Committee for the Protection of Human Subjects at Dartmouth College and all participants provided informed consent. The Centers for Disease Control and Prevention (CDC) laboratory’s involvement did not constitute engagement in human-subjects research.

Exposure Assessment

Maternal blood samples were collected at approximately 24 to 28 weeks’ gestation and processed into aliquots of plasma within 24 hours after collection. Plasma aliquots were stored at −80 °C until overnight shipment on dry ice to CDC’s National Center for Environmental Health laboratory. Plasma samples were analyzed for PFAS concentrations using on-line solid phase extraction-liquid chromatography-isotope dilution tandem mass spectrometry (Kato et al., 2018). The PFAS measured were: perfluorohexane sulfonate (PFHxS), linear PFOS (n-PFOS), perfluoromethylheptane sulfonate isomers (Sm-PFOS), linear PFOA (n-PFOA), branched PFOA isomers (Sb-PFOA), PFNA, PFDA, PFUnDA, and 2-(N-methylperfluorooctane sulfonamido) acetate (MeFOSAA). Total PFOA concentrations (PFOA) were calculated by adding the concentrations of n-PFOA and Sb-PFOA. Similarly, total PFOS concentrations (PFOS) were determined by summing concentrations of n-PFOS and Sm-PFOS (Centers for Disease Control and Prevention, 2022). Blanks and spiked quality control serum materials were included for quality assurance in each analytical batch. The limit of detection (LOD) was 0.1 ng/mL for each PFAS. As previously described, for statistical analyses, concentrations with a signal recorded to be >0.05 ng/mL but <0.1 ng/mL were assigned 0.1/√2. Concentrations with a signal recorded as <0.05 ng/mL were assigned 0.05/√2 (Romano et al., 2022).

Outcome Assessment

Per the AAP guideline, we defined exclusive breastfeeding as feeding with human milk without supplementation (including water, juice, non-human milk, formula, or solid foods) with the exception of vitamins, minerals, and medications (AAP, 2012). Any breastfeeding was defined as the duration in months that the participant reported any amount of breastfeeding, irrespective of supplementation with formula, liquids, or solid foods. We specifically examined reaching the benchmarks of 6 months of exclusive breastfeeding and 12 months of any breastfeeding, as these were the durations recommended by the AAP (AAP, 2012) until the 2022 update as mentioned (Meek and Noble, 2022). However, starting solid foods anywhere between 4 and 6 months is generally accepted medical practice in the U.S. (Borowitz, 2021). As a sensitivity analysis, we explored exclusive breastfeeding to the earlier time point of 4 months.

Statistical Analysis

Among participants in the cohort with plasma analyzed for PFAS, 982 completed postpartum surveys that included questions about breastfeeding. We excluded participants from the analysis with preterm deliveries (<37 weeks), gestational diabetes or preeclampsia, leaving normal, low-risk pregnancies (n=892). After excluding participants who had not breastfed (n=7) or were missing answers to breastfeeding questions (n=72), there were 813 participants in the final analytic sample. Plasma PFAS concentrations were similar for participants included and excluded from the analytic population (Supplemental Material; Table S1).

First, we examined univariate statistics of PFAS concentrations, duration of any and exclusive breastfeeding, and potential covariates. We compared the observed PFAS concentrations in our cohort to those observed among females in the National Health and Nutrition Examination Survey (NHANES) from 2009 – 2010 to 2017 – 2018 (Centers for Disease Control and Prevention, 2023b) and prior studies of blood serum or plasma PFAS concentrations and breastfeeding (Fei et al., 2010; Romano et al., 2016; Rosen et al., 2018; Timmermann et al., 2017; Timmermann et al., 2022). We calculated Spearman correlation coefficients to understand the correlation between different PFAS. Potentially confounding variables were identified a priori based on known associations with PFAS and breastfeeding duration. To account for normal plasma volume expansion in pregnancy, we also adjusted for gestational week of blood sample at collection (Costantine, 2014; Kato et al., 2014). The final multivariable models were adjusted for: maternal age, race, pre-pregnancy body mass index (BMI), smoking status, marital status, parity, gestational week at blood sampling, and total duration of prior breastfeeding. Our primary analysis assessed the influence of PFAS mixtures on cessation of exclusive breastfeeding before 6 months or cessation of any breastfeeding before 12 months. To contextualize our findings within the prior literature, we performed secondary analyses that evaluated the association of individual PFAS with cessation of exclusive breastfeeding before 6 months, cessation of any breastfeeding before 12 months and time to cessation of breastfeeding.

We used probit Bayesian Kernel Machine Regression (BKMR) to assess mixture effects for the five frequently detected PFAS (>75%: PFHxS, PFOS, PFOA, PFNA and PFDA) via the bkmr package in R (Bobb et al., 2018; Bobb et al., 2015). PFAS concentrations were log2-transformed to account for skewness, mean centered and univariate scaled. Continuous covariates were also centered and scaled. Using the Markov Chain Monte Carlo (MCMC) sampler, models were fit by running 50,000 iterations. We evaluated binary outcomes as cessation of exclusive breastfeeding before 6 months and cessation of any breastfeeding before 12 months using the probit implementation of BKMR to estimate the mean difference in the probability of cessation of breastfeeding for each PFAS while controlling for correlations among PFAS within the mixture (Bobb et al., 2018). Posterior inclusion probabilities identify the PFAS most strongly associated with breastfeeding outcomes, and 95% credible intervals were used to determine the uncertainty in the associations and assess the precision of the estimates.. Plots of exposure-response functions were created to visualize and interpret the shape of the dose-response function. We evaluated the overall effect of the mixture by comparing the response when PFAS were at a sequence of percentiles as compared to when the five PFAS are at their 50th percentile. We generated univariate dose-response curves for each individual PFAS while all other PFAS were fixed at their median concentration. Bivariate dose-response relations were assessed by generating the exposure-response function of a single PFAS while a second PFAS was fixed at a number of quantiles. We also summarized the contribution of single PFAS to the response by comparing the 25^th and 75^th percentile when all other PFAS are fixed at a given percentile (25^th, 50^th, 75^th).

Secondary analyses applied multivariable modified Poisson regression models with robust standard errors to estimate the relative risk (RR) (Zou, 2004) of ending exclusive breastfeeding before 6 months and cessation of any breastfeeding before 12 months. To assess time to stopping exclusive or any breastfeeding according to individual maternal plasma PFAS concentrations, we used covariate-adjusted multivariable Weibull accelerated failure time (AFT) models to calculate hazard ratios using the survival and SurvRegCensCov packages (Hubeaux and Rufibach, 2015; Therneau and Grambsch, 2000). PFHxS, PFOS, PFOA, and PFNA were divided into equally distributed quartiles. PFDA, PFUnDA, and MeFOSAA were turned into binary variables to account for heavily right-skewed distributions of the plasma concentrations. For PFDA, the median was calculated (0.2 ng/mL) and utilized as the binary cut-point for the exposure. PFUnDA and MeFOSAA were dichotomized based on the LOD (0.1 ng/mL). We used a censor time of 6 months for exclusive breastfeeding and 12 months for the any breastfeeding models.

Analyses were completed in R (v4.0.2) or SAS (v9.4).

Results

Participants in our sample of people who breast fed were 31.6 years of age on average (std=4.7) at enrollment, non-Hispanic white (97.7%), and married (86.9%) (Table 1). The majority had never smoked (87.6%) and were divided among 42.7% as nulliparous, 37.5% with one prior birth, and the remainder two or more. Among participants who had breastfed prior to the study pregnancy, the mean cumulative duration of prior breastfeeding was 15.1 months (std=13.7). The pregnancies, the mean duration of any breastfeeding of their infants was 9.2 months (std=7.4) with 28.5% of infants still receiving some human milk at 12 months. Among participants with sufficient information available to determine durations of exclusive breastfeeding (n=608), the mean duration of exclusive breastfeeding was 3.9 months (std=2.4) and 29% were exclusively breastfeeding for 6 months or more.

Table 1.

Select Characteristics of New Hampshire Birth Cohort Study Participants (2010–2017) who Breastfed (n=813)

	n(%); mean(std)
Maternal age	31.6 (4.7)
Non-Hispanic white	794 (97.7%)
Pre-pregnancy BMI	25.8 (5.7)
Smoking
Never	686 (87.6%)
Current	45 (5.8%)
Former	52 (6.6%)
Parity
0	346 (42.7%)
1	304 (37.5%)
2	115 (14.2%)
3+	46 (5.7%)
Marital Status
Married	681 (86.9%)
Single, Widowed, Divorced	103 (13.1%)
Gestational Week at Blood Sample	28.1 (2.5)
Prior Breastfeeding Duration (months)*	15.1 (13.7)
Any Breastfeeding at 12 months
Yes	232 (28.5%)
No	581 (71.5%)
Duration (months)	9.2 (7.4)
Exclusive Breastfeeding at 6 months
Yes	176 (29.0%)
No	432 (71.1%)
Duration (months)	3.9 (2.4)

^*Among pregnant people who previously breastfed a child (n=426).

Missing on maternal BMI (n=12), smoking (n=30), parity (n=2), marital status (n=29), gestational week at blood sample (n=1), prior breastfeeding duration (n=23) and exclusive breastfeeding (n=205).

PFHxS, PFOS, PFOA and PFNA were frequently detected (>94%), while PFDA (76%), PFUnDA (41%), and MeFOSAA (37%) were less commonly detected (Table 2). Overall, median plasma PFAS concentrations in the NHBCS were similar to serum concentrations reported for the U.S. general population of females in NHANES (Centers for Disease Control and Prevention, 2023b). The PFOS concentrations were generally lower than those observed in other studies of breastfeeding duration (Fei et al., 2010; Rokoff et al., 2023; Romano et al., 2016; Timmermann et al., 2023; Timmermann et al., 2017; Timmermann et al., 2022) (Supplemental Material; Table S2). The correlation among PFAS, as determined by Spearman rank correlation coefficients, ranged from 0.31to 0.67 and were all statistically significant (Supplemental Material; Table S3).

Table 2:

Distribution of maternal plasma PFAS concentrations (ng/mL) measured during pregnancy in the New Hampshire Birth Cohort Study (2010–2017) and females in the National Health and Nutrition Examination Survey (NHANES) (2009–2018)

NHBCS			NHANES Females
PFAS	% >LODa	Median (25th-75th percentile) (ng/mL)	09–10	11–12	13–14	15–16	17–18
			Median (ng/mL)b
PFHxS	99%	0.70 (0.50, 1.10)	1.30	0.97	1.00	0.90	0.80
PFOS	100%	3.50 (2.40, 5.20)	7.80	5.27	4.10	3.60	3.30
PFOA	100%	1.24 (0.74, 1.74)	2.70	1.78	1.67	1.37	1.27
PFNA	94%	0.50 (0.40, 0.70)	1.15	0.77	0.60	0.50	0.40
PFDA	76%	0.20 (0.10, 0.20)	0.30	0.19	0.20	0.10	0.20
PFUnDA	41%	<LOD (<LOD, 0.20)	0.20	0.12	<LOD	<LOD	0.10
MeFOSAA	37%	<LOD (<LOD, 0.10)	0.20	0.10	<LOD	<LOD	0.10

PFHxS =perfluorohexane sulfonate; PFOS=perfluorooctane sulfonate; PFOA=perfluorooctanoate; PFNA= perfluorononanoate; PFDA=perfluorodecanoate; PFUnDA=perfluoroundecanoate; MeFOSAA=2-(N-Methyl-perfluorooctane sulfonamido) acetate

^aThe limit of detection (LOD) was 0.1 ng/mL for all PFAS in the NHBCS.

^bFor NHANES, LOD was 0.1 ng/mL for all PFAS except 0.2 ng/mL for PFOS (in 2009–2010 and 2011–2012).

In our primary BKMR analysis, we observed an overall effect of the mixture of PFHxS, PFOS, PFOA, PFNA and PFDA on cessation of exclusive breastfeeding before 6 months (Figure 1a) driven largely by PFOA concentrations [PIPS: 0.36 (PFHxS), 0.37 (PFOS), 0.89 (PFOA), 0.30 (PFNA) and 0.44 (PFDA)]. Higher quantiles of the mixture of these five PFAS corresponded to increases in the estimated probability of exclusive breastfeeding cessation before 6 months, as compared to when all 5 PFAS were at their median (Figure 1a). We also observed a strong upward trend in the risk of exclusive breastfeeding cessation before 6 months with greater PFOA plasma concentration when all remaining PFAS are fixed at their median (Figure 1b). Greater plasma PFNA and PFHxS corresponded to increased risk of exclusive breastfeeding cessation before 6 months, whereas small decreases in this risk were observed for PFOS and PFDA (Figure 1b). We did not observe strong evidence of interactions among PFAS (Figure 1c), though there was some suggestion of an interaction between PFOA and PFOS, in which greater plasma PFOA was more strongly associated with increased probability of exclusive breastfeeding cessation before 6 months when PFOS plasma concentrations were relatively low (25^th percentile) versus relatively high (75^th percentile) (Figure 1c). Given the potential for infants to start solid foods between 4 and 6 months, we defined exclusive breastfeeding through 4 months and repeated the BKMR analysis. We did not find the results to be materially different to 6 months (Supplemental Material; Figure S1).

Figure 1.

BKMR results for per- and polyfluoroalkyl substance (PFAS) mixtures on risk of stopping exclusive breastfeeding before 6 months

(a) Overall effect of PFAS mixture on the probability of stopping exclusive breastfeeding before 6 months, when simultaneously setting all PFAS plasma concentrations (ng/mL) to different deciles compared with their median values, (b) Effect of individual PFAS plasma concentration (ng/mL) on the probability of stopping exclusive breastfeeding before 6 months, (c) Effect of a single PFAS plasma concentration (ng/mL) the probability of stopping exclusive breastfeeding by 6 months when setting a second PFAS fixed at specified quantiles [0.25, 0.5, 0.75], (d) Effect of single PFAS plasma concentration (ng/mL) on the probability of stopping exclusive breastfeeding before 6 months, where a single PFAS is at the 75th percentile as compared to when that PFAS is at its 25th percentile. Remaining PFAS are fixed at specified quantiles [0.25, 0.5, 0.75].

PFHxS =perfluorohexane sulfonate; PFOS=perfluorooctane sulfonate; PFOA=perfluorooctanoate; PFNA= perfluorononanoate; PFDA=perfluorodecanoate

The overall effect of the five PFAS mixture on risk of any breastfeeding cessation before 12 months was null (Figure 2a) with contrasting effects observed for individual PFAS and cessation of any breastfeeding (Figure 2b). This included a decrease in risk of any breastfeeding cessation before 12 months with greater plasma PFNA, slight increase in risk for PFOS, slight inverted U-shape for PFHxS, and U-shape for PFDA (Figure 2b and 2c). No individual PFAS appeared to dominate over others to drive the association with cessation of any breastfeeding [PIPs: 0.38 (PFHxS), 0.37 (PFOS), 0.26 (PFOA), 0.46 (PFNA) and 0.32 (PFDA)]. We did not observe interactions among the PFAS (Figure 2c; Figure 2d) in the analysis of any breastfeeding cessation by 12 months.

Figure 2.

BKMR results for per- and polyfluoroalkyl substance (PFAS) mixtures on risk of stopping breastfeeding before 12 months

(a) Overall effect of PFAS mixture on the probability of stopping breastfeeding before 12 months, when simultaneously setting all PFAS plasma concentrations (ng/mL) to different deciles compared with their median values, (b) Effect of individual PFAS plasma concentration (ng/mL) on the probability of stopping any breastfeeding before 12 months, (c) Effect of a single PFAS plasma concentration (ng/mL) on the probability of stopping breastfeeding before 12 months with setting a second PFAS fixed at specified quantiles [0.25, 0.5, 0.75], (d) Effect of single PFAS plasma concentration (ng/mL) on the probability of stopping breastfeeding before 12 months, where single PFAS is at the 75th percentile as compared to when that PFAS is at its 25th percentile. Remaining PFAS fixed at specified quantiles [0.25, 0.5, 0.75].

PFHxS =perfluorohexane sulfonate; PFOS=perfluorooctane sulfonate; PFOA=perfluorooctanoate; PFNA= perfluorononanoate; PFDA=perfluorodecanoate

We observed parallel findings for the five PFAS mixture when assessing PFAS individually using traditional methods in secondary analyses. We observed slight increases in RR for stopping exclusive breastfeeding before 6 months for PFHxS (p-trend=0.02), PFOA (p-trend=0.01), and PFNA (p-trend=0.01). No strong effects of any PFAS on the risk of any breastfeeding cessation before 12 months were identified (Table 3). Similarly, accelerated failure time models indicated that increasing quartiles of PFHxS (p-trend=0.05) and PFNA plasma concentrations (p-trend=0.01) corresponded to a greater hazard of exclusive breastfeeding cessation before 6 months, and we also observed a suggestive increase for PFOA (p-trend=0.10) (Table 3). Individuals with PFOS in the second, third and fourth quartile had greater hazard of stopping exclusive breastfeeding before 6 months versus the first quartile. No clear associations between individual PFAS and time to stopping any breastfeeding by 12 months were observed.

Table 3.

Adjusted Relative Risks (RR) and Hazard Ratios (HR) for stopping exclusive breastfeeding before 6 months and any breastfeeding before 12 months by quartiles of per- and polyfluoroalkyl substances (PFAS) plasma concentrations (ng/mL) during pregnancy

		EXCLUSIVE BREASTFEEDING		ANY BREASTFEEDING
PFAS	Quartile	RR (95%CI)	HR (95% CI)	RR (95%CI)	HR (95% CI)
PFHxS	<0.50	Ref	Ref	Ref	Ref
	0.50 to <0.70	1.10 (0.92, 1.32)	1.34 (0.99,1.82)	1.01 (0.88, 1.17)	1.03 (0.80,1.34)
	0.70 to <1.10	1.13 (0.97, 1.33)	1.28 (0.94,1.73)	1.03 (0.91, 1.18)	1.04 (0.81,1.35)
	≥1.10	1.21 (1.03, 1.42)	1.48 (1.08,2.02)	1.09 (0.95, 1.25)	1.08 (0.83,1.41)
	p for trend	0.02	0.05	0.15	0.52
PFOS	<2.30	Ref	Ref	Ref	Ref
	2.30 to <3.40	1.17 (1.00, 1.37)	1.26 (0.94,1.69)	0.97 (0.85, 1.11)	1.00 (0.78,1.30)
	3.40 to <5.10	1.14 (0.97, 1.34)	1.39 (1.03,1.87)	0.98 (0.87, 1.12)	1.09 (0.85,1.40)
	≥5.10	1.10 (0.94, 1.30)	1.22 (0.91,1.65)	1.01 (0.88, 1.15)	1.03 (0.79,1.33)
	p for trend	0.62	0.46	0.72	0.84
PFOA	<0.74	Ref	Ref	Ref	Ref
	0.74 to <1.24	1.15 (0.95, 1.40)	1.30 (0.96,1.76)	1.10 (0.94, 1.29)	1.15 (0.89,1.5)
	1.24 to <1.74	1.15 (0.94, 1.40)	1.32 (0.96,1.82)	1.06 (0.91, 1.25)	0.98 (0.75,1.3)
	≥1.74	1.28 (1.04, 1.56)	1.36 (0.98,1.89)	1.00 (0.85, 1.19)	0.90 (0.67,1.2)
	p for trend	0.01	0.10	0.34	0.11
PFNA	<0.40	Ref	Ref	Ref	Ref
	0.40 to < 0.50	0.98 (0.81, 1.19)	1.22 (0.91,1.64)	0.89 (0.77, 1.03)	0.78 (0.61,0.99)
	0.50 to <0.70	1.07 (0.91, 1.25)	1.24 (0.92,1.66)	0.89 (0.79, 0.99)	0.86 (0.67,1.10)
	≥0.70	1.17 (1.01, 1.36)	1.38 (1.02,1.86)	0.88 (0.79, 0.99)	0.79 (0.61,1.02)
	p for trend	0.01	0.01	0.07	0.16
PFDA	< Median (0.20)	Ref	Ref	Ref	Ref
	≥ Median	0.98 (0.86, 1.11)	1.19 (0.97,1.45)	1.00 (0.90, 1.11)	0.90 (0.76,1.08)
PFUnDA	<LOD (0.10)	Ref	Ref	Ref	Ref
	≥LOD	1.00 (0.89, 1.12)	1.00 (0.82,1.22)	0.96 (0.87, 1.07)	0.95 (0.79,1.14)
MeFOSAA	<LOD	Ref	Ref	Ref	Ref
	>LOD	0.93 (0.83, 1.05)	1.04 (0.84,1.27)	0.93 (0.85, 1.03)	1.14 (0.96,1.36)

All models adjusted for maternal age, race, pre-pregnancy BMI, smoking, marital status, parity, blood sampling week and prior breastfeeding duration. P-value for trend calculated by assigning the median value in each quartile and modeling as a continuous variable.

Bold p for trend are p<0.05

PFHxS =perfluorohexane sulfonate; PFOS=perfluorooctane sulfonate; PFOA=perfluorooctanoate; PFNA= perfluorononanoate; PFDA=perfluorodecanoate; PFUnDA=perfluoroundecanoate; MeFOSAA=2-(N-Methyl-perfluorooctane sulfonamido) acetate; LOD= limit of detection (was 0.1 ng/mL for all PFAS)

Discussion

In a general population cohort from New Hampshire, we found relatively low plasma PFAS concentrations during pregnancy, and observed that participants with greater overall plasma concentration of five PFAS (PFHxS, PFOS, PFOA, PFNA, PFDA) were at increased risk of exclusive breastfeeding cessation by 6 months. This association appeared to be primarily driven by PFOA, though suggestive associations of PFHxS, PFOS, and PFNA with exclusive breastfeeding cessation were observed when considering PFAS individually. We did not detect associations between PFAS and duration of any breastfeeding in this cohort.

Our results are consistent with the majority of prior research observing shorter duration of exclusive breastfeeding among individuals with greater PFOA plasma/serum concentrations, but extend this work through the investigation of PFAS mixtures. To date, the most consistent associations have been observed for PFOA and PFOS in regards to shorter duration of both exclusive and any breastfeeding (Timmermann et al., 2023). Two prior studies observed associations between PFOA and PFOS and shorter duration of exclusive breastfeeding (Fei et al., 2010; Timmermann et al., 2017) though two other studies, including one prior U.S. study, observed no association of PFAS with exclusive breastfeeding (Nielsen et al., 2022; Rokoff et al., 2023; Romano et al., 2016). Although the present study did not observe associations of PFAS with any breastfeeding cessation, four prior studies have suggested that PFOS and PFOA are associated with decreased duration of any breastfeeding (Fei et al., 2010; Romano et al., 2016; Timmermann et al., 2017; Timmermann et al., 2022). A study from Sweden also observed that women with greater exposure to a mixture of PFAS from contaminated drinking water had a ~30% greater hazard of terminating breastfeeding before 3 months (Nielsen et al., 2022). Discrepancies in findings among studies may partially result from the specific mixtures of PFAS evaluated or differences in PFAS concentrations across populations. Prior studies have largely not assessed the influence of PFAS mixtures on breastfeeding. In the present study, PFOA appears to be the primary driver of the association of the five PFAS mixture with shorter duration of exclusive breastfeeding. Nevertheless, the observed effect of the five PFAS mixture on stopping breastfeeding underscores the relevance of carefully considering PFAS mixtures as compared to individual PFAS in future epidemiologic studies of breastfeeding.

Hormonal mechanisms regulating lactation are susceptible to interference by endocrine disruptors (Abbott et al., 2009; Criswell et al., 2020), but the biological mechanism underlying the observed associations between greater PFAS and shorter duration of breastfeeding across multiple epidemiologic studies (Fei et al., 2010; Nielsen et al., 2022; Romano et al., 2016; Timmermann et al., 2017; Timmermann et al., 2022) is not well understood. Criswell et al. reviewed several possibilities indicated by the toxicological evidence (2020), including that PFAS suppress prolactin and placental lactogen signaling (Suh et al., 2011), adversely influence mammary gland differentiation and impair lactogenesis (White et al., 2007), interfere with mammary gland development (Macon et al., 2011; Tucker et al., 2015; White et al., 2007; White et al., 2011), and may alter the expression of milk protein genes (White et al., 2007) with possible implications for both the quantity and quality of milk produced (Bhat et al., 2017). Initial studies in humans have not supported a role of prolactin (Timmermann et al., 2022) but the mechanisms by which PFAS may interfere with human lactation remain unknown (Criswell and Romano, 2022; Timmermann et al., 2023).

Potentially modifiable factors that influence lactation are of public health interest, as breastfeeding is associated with short- and long-term health benefits for both mother and child. As in our cohort, <37% of U.S. mothers continue any breastfeeding until 12 months (National Immunization Survey 2022) and fewer still will meet the newly recommended 24 months (Meek and Noble, 2022). Similarly, < 25% of U.S. mothers exclusively breastfed (no solids, water or other liquids) for 6 months (National Immunization Survey 2022). The benefits of human milk for children are well described, with health benefits ranging from fewer infections and improved cognitive development in childhood to reduced risk of diabetes extending well into adulthood (Ip et al., 2007; Tozzi et al., 2012). Recent meta-analyses support that continuation of lactation for 12 months or more also confers protection against common adverse maternal cardiometabolic health outcomes, including a 30% risk reduction for diabetes and a 13% risk reduction for hypertension (Rameez et al., 2019). Known barriers to breastfeeding include sociocultural, medical, nutritional, and infant-level factors (Teich et al., 2014), and range from insufficient milk supply and latch to responsibilities for work/school and family. Endocrine-disrupting chemicals may be under-appreciated as potential risk factors for shorter duration of breastfeeding (Criswell et al., 2020; Karmaus et al., 2005; Rogan et al., 1987), as evidenced by observed associations of PFAS with breastfeeding in the present study.

Our study had important limitations as well as notable strengths. Sample size was somewhat modest for the mixture analysis, but ours is also among the largest studies of PFAS and breastfeeding in a U.S. population to date. We did not have information about reasons for stopping breastfeeding in the NHBCS. In prior research, the reason for stopping was not associated with maternal PFOA concentrations (Romano et al., 2016), providing some reassurance that any observed associations between plasma PFAS concentrations and a shorter duration of breastfeeding are not strongly confounded by factors such as caregiver’s return to work. The outcome was reliant on self-reported information and could potentially have resulted in some misclassification, though we expect that this would be non-differential with respect to PFAS exposure. Breastfeeding is an important route of maternal PFAS excretion (Barbarossa et al., 2013; Criswell et al., 2023; Mondal et al., 2014), and previous breastfeeding history is a major predictor of future breastfeeding success (Nagy et al., 2001; Whalen and Cramton, 2010). A strength of the current study is the ability to directly account for confounding by prior breastfeeding history in the analyses.

Conclusions

In this rural cohort from New Hampshire, we observed a greater risk of stopping exclusive breastfeeding prior to 6 months among individuals with greater plasma concentration of a mixture of PFHxS, PFOS, PFOA, PFNA, and PFDA during pregnancy, though this association was largely driven by PFOA. These findings further support prior evidence that PFOA can contribute to shorter duration of exclusive breastfeeding. Our findings also underscore the importance of evaluating the effect of PFAS mixtures on breastfeeding.

Abbreviations

PFAS

per- and polyfluoroalkyl substances

PFHxS

perfluorohexane sulfonate

PFOA

perfluorooctanoate

PFOS

perfluorooctane sulfonate

PFNA

perfluorononanoate

PFDA

perfluorodecanoate

PFUnDA

perfluoroundecanoate

MeFOSAA

2-(N-methylperfluorooctane sulfonamido) acetate

AAP

American Academy of Pediatrics

NHBCS

New Hampshire Birth Cohort Study

CDC

Centers for Disease Control and Prevention

LOD

limit of detection

NHANES

National Health and Nutrition Examination Survey

BKMR

Bayesian kernel machine regression