Pregnancy Exposure to Perfluoroalkyl Substances and Associations With Prolactin Concentrations and Breastfeeding in the Odense Child Cohort

Abstract

Context

Human exposure to perfluoroalkyl substances (PFAS) has been associated with reduced duration of breastfeeding, although not consistently so, and mechanisms by which PFAS might affect breastfeeding are unknown.

Objective

To examine the association between early pregnancy serum-PFAS concentrations and breastfeeding termination and to elucidate the potential role of serum-prolactin concentrations in pregnancy.

Materials and Methods

Pregnant women from the Odense Child Cohort provided blood samples for analysis of 5 major PFAS (n = 1300) and prolactin concentrations (n = 924). They subsequently provided information about the duration of breastfeeding in questionnaires at 3 and 18 months postpartum, and a subgroup also provided breastfeeding information via weekly cell phone text messages. Associations between serum-PFAS concentrations and breastfeeding termination were analyzed using Cox regressions, while linear regression was used to assess associations between serum-PFAS and prolactin concentrations.

Results

Increased serum concentrations of perfluorooctane sulfonic acid, perfluorooctanoic acid, perfluorononanoic acid, and ∑PFAS were associated with a 16% (95% CI: 4%-30%), 14% (95% CI: 2%-26%), 14% (95% CI: 3%-27%), and 20% (95% CI: 6%-36%), respectively, increased risk of terminating breastfeeding at any given time after childbirth. Serum-PFAS concentrations were not associated with serum-prolactin concentrations.

Conclusions

These findings are of public health importance due to the global exposures to PFAS. Because breastfeeding is crucial to promote both child health and maternal health, adverse PFAS effects on the ability to breastfeed may have long-term health consequences.

Perfluoroalkyl substances (PFAS) are a group of persistent environmental chemicals considered to cause diverse adverse health effects in humans (1). Due to the widespread use in products such as furniture, carpets, clothing, food packaging, and firefighting foams (2,3), PFAS are ubiquitous in the environment (4). Humans are mainly exposed to PFAS through contaminated food and water as well as dust from treated textiles and other materials. However, in infants, breast milk may be the most important source of exposure (5,6).

Although PFAS and other environmental chemicals are transferred to the infant through breastfeeding (6,7), breastfeeding is also known to be beneficial to the health of both mother and child (8). The World Health Organization recommends that infants are breastfed exclusively for the first 6 months of life and partially up to 2 years or beyond (9). However, in the United States, 16% of the children born in 2016 were never breastfed, 57% were still being breastfed at 6 months of age, and 25% were being breastfed exclusively at 6 months of age (10). In Denmark, only 17% were breastfed exclusively for 6 months (11). Of note, the Danish health and food authorities recommend introducing food to infants between 4 and 6 months of age (12,13), and the definition of exclusive breastfeeding allows up to 1 infant formula meal per week.

Societal, social, and psychological factors as well as individual choice affect the duration of breastfeeding (14,15), but physiology also plays an important role (16). Many women experience early undesired weaning (17) and frequently report inadequate milk supply as a leading cause (18-20). The reasons for inadequate milk supply are poorly understood (16), but 3 epidemiological studies have linked PFAS exposures to early breastfeeding termination (21-23), although 1 of the studies ascribed the findings to confounding factors from previous breastfeeding (23), and a newer study did not confirm the findings (24). In the mouse, perfluorooctane sulfonic acid (PFOS) exposure resulted in reduced prolactin family hormone concentrations in serum (25), thus suggesting that PFAS might affect the ability to lactate via an endocrine pathway. Still, this aspect has so far not been elucidated in human studies.

The aim of the present study was to examine the association between maternal exposure to PFAS and termination of breastfeeding among Danish women and to explore the role of serum concentrations of prolactin.

Materials and Methods

Pregnant women living in Odense municipality, Denmark were recruited to the Odense Child Cohort (OCC) between January 2010 and December 2012 (26), with a total of 2875 women recruited among 6707 eligible. The women who consented were older, had lower body mass index (BMI), were less often smokers and had lower parity than nonparticipants (26). Following enrollment in late first trimester or beginning of second trimester, the women were asked to donate a blood sample, which was analyzed for PFAS and prolactin. Additional blood was collected for prolactin assessment at approximate gestational age (GA) 28 weeks. Questionnaires were sent to the women twice during pregnancy (26), and information about maternal education, smoking during pregnancy, previous breastfeeding, and breastfeeding expectations were obtained from the questionnaires. Any missing information about maternal education was retrieved from hospital charts. Information about maternal prepregnancy BMI, parity, maternal age at delivery, GA at delivery, child sex, and birth weight was obtained from hospital charts. Maternal national origin was defined based on her mother’s country of birth, as recorded by Odense municipality. From the 2875 women included in the OCC, 2657 live-born singleton children and their mothers were eligible for this study (Fig. 1). Among these parturitions, 106 mothers were included with 2 children, and 1 was included with 3 children.

PFAS Exposure Assessment

A total of 1699 women donated a blood sample for PFAS analysis at a median GA of 12 weeks (5th-95th percentiles, 9-19 weeks). Five PFAS [perfluorohexanesulfonic acid (PFHxS), PFOS, perfluorooctanoic acid (PFOA), perfluorononanoic acid (PFNA), and perfluorodecanoic acid (PFDA)] were quantified in serum at University of Southern Denmark using online solid phase extraction followed by liquid chromatography and triple quadropole mass spectrometry as described by Haug et al (27). As previously described, the first 649 samples were analyzed between September 2011 and September 2013, and the remaining samples were analyzed in January 2019 (28). The between-batch imprecision was below 10.5%, and the limit of detection (LOD) was 0.03 ng/mL (28). Values below the LOD were replaced with LOD/2. The accuracy of the PFAS method is continuously secured by participation in the German Quality Assessment program (G-EQUAS) organized by the German Society of Occupational Medicine. Total PFAS exposure was calculated as the sum of the 5 PFAS.

Prolactin Assessment

The first blood sample was available for prolactin analysis from 1056 pregnancies at a median GA of 12 weeks (5th-95th percentiles, 10-16 weeks). The second blood sample for prolactin analysis was available from 1524 pregnancies at median GA 29 weeks (5th-95th percentile: GA 28-30 weeks).

As previously described (29), stored biobank samples were analyzed batchwise for prolactin in sera using an electro-chemiluminescence immunoassay. The Cobas prolactin II assay used was calibrated against the World Health Organization Third International Standard for Prolactin (84/500). Interassay coefficients of variance for prolactin batch analysis were determined using Seronorm Immunoassay Liq-1/2 (Sero) and was <4.4% (quality control level 1; 210 mIU/L) and <3.3% (quality control level 2; 903 mIU/L) (29).

Breastfeeding

At approximately 3 and 18 months postpartum, the mothers answered questions about duration of exclusive breastfeeding, any breastfeeding, and the use of formula feeding. Furthermore, a subgroup of 1100 mothers giving birth from April 2012 until October 2013 received a weekly cell phone text message asking them to indicate extent of current breastfeeding and use of supplementary formula, starting at 3 days postpartum (30). Information from text messages was considered more reliable than the questionnaires, but the latter was relied upon when text message data had not been collected.

In Denmark, all newborns receive home visits by a nurse (ie, a health visitor with special training in neonatal and infant care) to assess and advise on growth and general health and well-being of the child. Routine visits are initiated within the first 2 weeks after birth and repeated about age 2, 5, and 9 months. In addition, parents can consult the health visitor by phone or request additional visits if needed. Information from health visitor contacts was acquired for all women who had PFAS measured during pregnancy and who terminated breastfeeding within 16 weeks postpartum. Reasons for terminating breastfeeding was categorized as follows: (1) insufficient lactation and/or reduced child growth, (2) possibly insufficient lactation (eg, child is uneasy and seeks the breast frequently/takes long to feel satisfied), (3) unrelated to insufficient lactation (eg, maternal use of medication), or (4) reason not provided. The latter was assumed to reflect early termination of breastfeeding as a choice made by the mother rather than caused by breastfeeding problems. For 26 mother-child dyads, the registered duration of exclusive and/or any breastfeeding based on the text messages and questionnaire data was not in accordance with the information from the health visitor’s report. In these cases, the health visitor’s report was assumed to be correct.

Statistical Analyses

Breastfeeding duration distributions were examined in relation to maternal and birth characteristics. Associations between PFAS exposure and termination of exclusive and any breastfeeding were analyzed in Cox regression models with time since childbirth (in weeks) as the underlying time. Exit time for women who never initiated breastfeeding or quit before 1 week was set to 0.01 weeks to include these women in the analyses. However, we also performed sensitivity analyses in which these women were excluded. Potential confounding variables were identified using a directed acyclic graph based on a priori knowledge (Fig. 2). Thus, the analyses were adjusted for parity (1, 2, >2), education (high school or less, high school + 1-4 years, high school + >4 years), prepregnancy BMI (<25, 25-29.9, ≥30), smoking (any, none), national origin (Danish, Western, non-Western), expectations of giving formula (before 1 week, after 1 week but before 1 month, after 1 month but before 6 months, no formula within 6 months), duration of previous breastfeeding (0, ≤4, >4 months, no previous births), and lack of milk with previous child(ren) (no, yes, no previous birth). Although preterm birth is associated with shorter duration of breastfeeding, it was not considered a potential confounder, as we would not expect it to be associated with pregnancy PFAS exposure (Fig. 2). Information about the potential confounding variables was missing for 27% of the women; crude analyses were therefore performed with and without individuals with missing data. Adjusted analyses were performed with and without censuring of women for whom reasons for early termination of breastfeeding was either given as not related to insufficient lactation or not given in the health visitor’s report. PFAS concentrations were log10 transformed, and beta estimates were back-transformed to express the hazard ratio for terminating breastfeeding with each doubling in PFAS concentrations by using e^[β * log₁₀(2)]. Schoenfeld residuals revealed nonproportional hazards for parity, smoking, duration of previous breastfeeding, and previous inadequate lactation, and thus, the adjusted Cox models were stratified by these covariates (equal coefficients across strata but with a baseline hazard distinct for each stratum). No significant (P < 0.05) deviations from proportional hazards were found for any of the PFAS. Differences between primiparous and multiparous women were tested by including an interaction term between parity (1, ≥2) and PFAS exposure in the regression models. Furthermore, sensitivity analyses were performed with additional adjustment for pregnancy prolactin concentrations (log10 transformed) along with the exact GA at the time of serum sampling to consider the possible impact of differences in prolactin concentrations.

Associations between log10-transformed pregnancy prolactin concentrations and termination of exclusive and any breastfeeding were analyzed in Cox regression models in the same way as previously described for PFAS. The Cox models were stratified by parity and smoking, and they were adjusted for timing of prolactin measurement and BMI. The analyses were performed with and without censuring of women for whom reasons for early termination of breastfeeding was either given as not related to insufficient lactation or not given in the health visitor’s report.

The associations between log10-transformed PFAS exposure and log10-transformed prolactin concentrations were analyzed by linear regression, and the beta estimates were subsequently back transformed to express the percentage difference in prolactin concentrations for a doubling in PFAS concentrations by using (2^β − 1)*100. Likewise, the associations between log10-transformed PFAS exposure and change in prolactin concentrations from GA week 10 to 28 were analyzed by linear regression, and the beta estimates were subsequently back-transformed to express the difference in prolactin concentration change for a doubling in PFAS concentrations by using β * log10(2). The analyses were adjusted for the exact GA at serum sampling, parity (1, 2, >2), BMI (<25, 25-29.9, ≥30), and smoking (any, none) as according to the directed acyclic graph in Figure 2. The assumptions of homoscedasticity and normal distribution of the residuals were accepted based on visual inspection of plots showing residuals against fitted values and quantile-normal plots of the standardized residuals, respectively. In the OCC, elevated PFAS concentrations have previously been shown to be associated with higher fasting glucose and insulin among women with risk factors for gestational diabetes mellitus (GDM) (31), and prolactin concentrations have been inversely associated with risk of GDM (29). Thus, in a subgroup of women with available information on predefined risk factors for GDM [standard Danish screening criteria (32); BMI ≥ 27 kg/m², family history of diabetes mellitus, present multiple pregnancy, glucosuria during pregnancy, previous GDM, or previous delivery of a macrosomic infant (birth weight ≥ 4500 g)], the prolactin analyses were tested for interactions between PFAS and having risk factors for GDM.

All regression analyses were performed with cluster-robust standard errors to account for dependence between mothers being included more than once. The assumption of PFAS log-linearity was assessed by including log10(PFAS) squared along with log10(PFAS) in the model, and no significant (P < 0.05) deviations were observed. All analyses were conducted in Stata version 16.1.

Ethics

Written informed consent was obtained from all women included in the OCC. The study was carried out in accordance with the Helsinki Declaration II and has been approved by the Regional Scientific Ethical Committees for Southern Denmark and the Danish Data Protection Agency.

Results

Among the 2675 mother-child-dyads in OCC, information about duration of exclusive and any breastfeeding was obtained for 2184 and 1967 mother-child dyads, respectively, of which 47% was obtained from weekly text messages. PFAS measurements and information about either exclusive or any breastfeeding were available for 1300 mothers (Fig. 1). PFAS were found in median concentrations between 0.29 ng/mL and 7.57 ng/mL, and the sum of all 5 PFAS was 10.86 ng/mL (Table 1). Six mothers had PFHxS concentrations below the LOD, while the other PFAS were detected in all samples. The median durations of any and exclusive breastfeeding were 33 weeks and 11 weeks, respectively, and women tended to breastfeed longer if they were older and more educated, had lower BMI, nonsmokers, were of non-Western origin, gave birth vaginally and at term, expected to breastfeed exclusively for at least 6 months, had given birth before, had breastfed previously for more than 4 months, and had not experienced inadequate lactation or problems in general with previous breastfeeding (Table 2).

After confounder adjustment, each doubling in serum concentrations of PFOS, PFOA, and PFNA increased the risk of terminating breastfeeding at any given time after birth by 16% (95% CI: 4%-30%), 14% (95% CI: 2%-26%), and 14% (95% CI: 3%-27%), respectively, while each doubling in the sum of the 5 PFAS increased the risk by 20% (95% CI: 6%-36%) (Table 3). Censuring of women for whom reasons for early termination of breastfeeding was either given as not related to insufficient lactation or not provided by the health visitor, slightly strengthened these associations. Notably, all hazard ratio estimates in the censured analyses were above 1, suggesting that PFAS exposure increases the risk of terminating breastfeeding, although the estimate for PFDA was close to 1, and CIs were wide. In the crude analysis, a doubling in serum-PFOA concentrations was associated with an increased risk of terminating exclusive breastfeeding (hazard ratio: 1.13, 95% CI: 1.06-1.20), but the association vanished after confounder adjustment. After confounder adjustment and censuring of women for whom reasons for early termination of breastfeeding was either given as not related to insufficient lactation or not provided by the health visitor, a doubling in serum-PFHxS concentrations was associated with 8% (95% CI: 2%-14%) lower risk of terminating exclusive breastfeeding. Serum concentrations of PFOA, PFNA, and PFDA were not associated with termination of exclusive breastfeeding (Table 3). Exclusion of women who did not initiate breastfeeding or terminated breastfeeding within the first week did not substantially change the results [Supplementary Table 1 (33)], and no significant differences between primiparous and multiparous women were found [Supplementary Table 2 (33)].

No significant association was found between pregnancy prolactin and hazards of terminating breastfeeding (Table 4). In crude analyses, increased serum-PFAS concentrations were associated with increased pregnancy prolactin concentrations both at GA weeks 10 and 28 as well as increased change in prolactin concentration from GA week 10 to 28, but after confounder adjustment, the associations were attenuated or even reversed, and no clear trends were observed across the analyses (Table 5). When testing interactions between PFAS and GDM risk factors, a doubling in serum-PFDA concentrations was associated with 16.3% (95% CI: 5.8%-27.8%) higher prolactin concentrations at GA 10 weeks among women with GDM risk factors, whereas it was associated with a slight decrease in prolactin concentration among women with no GDM risk factors. No other significant interactions were found [Supplementary Table 3 (33)]. Additional adjustment for prolactin concentrations had little or no effect on the estimates in the analyses of association between serum-PFAS concentrations and risk of terminating breastfeeding [Supplementary Table 4 (33)].

Discussion

In this large prospective study, higher serum concentrations of PFAS were associated with increased risk of terminating breastfeeding. When censuring women who terminated breastfeeding early for reasons not related to insufficient lactation, the associations became slightly stronger, thus supporting the hypothesis that the association between PFAS and reduced breastfeeding duration was limited to cases with insufficient lactation. This result is in accordance with findings from the Health Outcomes and Measures of the Environment (HOME) study, where PFOA and, to some degree, PFOS concentrations were associated with breastfeeding termination at 3 and 6 months among American women (21). Furthermore, PFOS, PFOA, PFNA, and, to some degree, PFHxS and PFDA were associated with shorter duration of breastfeeding among Faroese women in the Children’s Health and the Environment in the Faroe Islands (CHEF) cohorts (22). Similar associations were also observed among women in the Danish National Birth Cohort although only among multiparous women (23), thus suggesting that the association could be caused by confounding from previous breastfeeding. However, in the present study, we took previous breastfeeding into account, and differences between primiparous and multiparous women were marginal, which confirms the findings from the HOME study (21) and the CHEF cohorts (22).

Exclusive Breastfeeding

In the present study, we found no consistent trends in the associations between PFAS exposure and exclusive breastfeeding. Higher serum-PFHxS concentrations were associated with a decreased risk of terminating exclusive breastfeeding. With no plausible biological explanation, we ascribe this 1 significant finding to chance, and the lack of consistency between findings relating to any and exclusive breastfeeding could be due to imprecision of the exclusive breastfeeding information based on questionnaires (30). In the Norwegian Mother and Child (MoBa) cohort study, increased concentrations of PFNA, PFDA, and PFUnDA were also associated with decreased risk of terminating breastfeeding, and they suggested that divergencies across cohorts could be due to differences in exposure levels and profiles (24). Serum-PFAS concentrations in the present study were lower than those found in the Danish National Birth Cohort (1996-2002) (23) and in the HOME Study (2003-2006) (21). Furthermore, concentrations of some PFAS were lower than seen in the MoBa study (1999-2008) (24) and the older CHEF cohort (1997-2000) (22), whereas concentrations were similar to those found in the younger CHEF cohort (2007-2009) (22). Thus, these differences alone are unlikely to explain the reverse associations found in the MoBa study and for exclusive breastfeeding in the present study.

The Possible Role of Prolactin

To our knowledge, the present study is the first to examine the association between PFAS and pregnancy prolactin concentrations in humans, but a Japanese study has reported that decreased concentrations of prolactin in female infants were associated with higher maternal pregnancy PFOS concentrations (34). In our study, serum-PFAS concentrations generally did not seem to affect pregnancy prolactin concentration at GA week 10 or 28 or the change from week 10 to 28. However, among women with GDM risk factors, higher serum-PFDA concentrations were associated with increased prolactin concentrations. Higher third trimester concentrations of prolactin have previously been associated with reduced glucose tolerance based on data from a 2-h oral glucose tolerance test (35). Therefore, in a metabolically vulnerable group of pregnant women, PFAS-mediated changes in prolactin concentrations may hypothetically be a contributing environmental risk factor for insulin resistance, supporting previously published findings (31). In our study we measured prolactin only during pregnancy, and prolactin concentrations in GA weeks 10 and 28 were not associated with hazards of terminating breastfeeding. We cannot exclude the possibility that PFAS could affect prolactin at other time points than those measured in our study—in particular, postpartum. Still, evidence available did not suggest that prolactin concentrations during pregnancy were associated with PFAS exposure.

Although the mechanism by which PFAS might affect breastfeeding duration remains unknown, several studies in mice have shown that PFOA can affect mammary gland development (36-38), at present the most sensitive developmental outcome of PFOA exposure (1). Furthermore, PFAS can activate the peroxisome proliferator-activated receptor α (39), which has been shown to impair mammary lobuloalveolar development in mice when activated during pregnancy (40).

Strengths and Weaknesses

Strengths of the present study include the study size and access to information about both previous breastfeeding duration and apparent reasons for terminating breastfeeding. Also, we had data on repeated measurement of prolactin in early and late pregnancy. Further, we previously tested for macroprolactin (higher molecular weight complexes of prolactin and immunoglobulins), using polyethylene glycol precipitation in a subgroup of 101 of the cohort women and found that total prolactin measurements during pregnancy were not affected by macroprolactinemia (29). However, we were unable to discriminate between prolactin of placental origin and prolactin secreted from the pituitary gland (29), which could have affected our ability to detect associations. In addition, prolactin assessment during pregnancy rather than postpartum may have affected the ability to detect an association, as could any imprecision of the exclusive breastfeeding measure. However, both the health visitor and the women were unaware of their PFAS exposure when reporting breast feeding durations and reasons for breastfeeding termination, thereby likely causing a nondifferential misclassification and subsequent dilution of the results. Finally, we measured PFAS exposure during pregnancy, while experimental studies suggest that mammary gland development is more likely affected by exposure incurred prenatally (37,41), during early postnatal life (36), or during puberty (38,42). Accordingly, unmeasured early-life exposure to PFAS may have affected the ability of the cohort mothers to breastfeed, without this association being detected.

Conclusions

In this study among Danish women, we found that higher serum-PFAS concentrations were associated with shorter duration of breastfeeding. These findings are of public health importance, since breastfeeding is one of the most effective ways to promote child and maternal health. Thus, if PFAS exposure negatively affects the ability to breastfeed, this may have a long-term health impact on both the mother and child. Future studies should preferably include more detailed questions about reasons for breastfeeding termination and study the mechanism by which PFAS affect lactation.

Acknowledgments

The families in the Odense Child Cohort are acknowledged for their participation and commitment to the study. The healthcare professionals at the Hans Christian Andersen Children’s Hospital, and the technicians at the Department of Environmental Medicine are acknowledged for their careful examination of the children and analysis of perfluoroalkyl substances, respectively. The authors thank the home nurses/health visitors for their recruitment assistance.

Financial Support: This work was supported by Odense University Hospital, Region of Southern Denmark, municipality of Odense, the Mental Health Service of the Region of Southern Denmark, Odense Patient data Explorative Network (OPEN), Danish Council for Independent Research, Medical Sciences (grant no 8020-00123B_FSS), and the Novo Nordisk Foundation (Grant nos. NNF17OC0029404 and NNF19OC0058266). Further, P.G. is supported by the National Institute of Environmental Health Sciences (NIEHS) of the NIH (ES021477).

Additional Information

Disclosures: P.G. has provided paid expert assistance in legal cases involving perfluoroalkyl substance–exposed populations. Otherwise, the authors have no actual or potential competing financial interests.

Data Availability

Restrictions apply to the availability of some or all data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.

References