Perfluoroalkyl acid levels in first-time mothers in relation to offspring weight gain and growth

(1)

Contents lists available atScienceDirect

Environment International

journal homepage:www.elsevier.com/locate/envint

Per fluoroalkyl acid levels in first-time mothers in relation to offspring weight gain and growth

Irina Gyllenhammar

^a,⁎

, Barbro Diderholm

^b

, Jan Gustafsson

^b

, Urs Berger

^c,d

, Peter Ridefelt

^e

, Jonathan P. Benskin

^c

, Sanna Lignell

^a

, Erik Lampa

^f

, Anders Glynn

^a

aNational Food Agency, P.O. Box 622, 751 26 Uppsala, Sweden

bDepartment of Women's and Children's Health, Uppsala University, 751 85 Uppsala, Sweden

cDepartment of Environmental Science and Analytical Chemistry (ACES), Stockholm University, 106 91 Stockholm, Sweden

dDepartment Analytical Chemistry, Helmholtz Centre for Environmental Research– UFZ, 04318 Leipzig, Germany

eDepartment of Medical Sciences, Clinical Chemistry, Uppsala University, 751 85 Uppsala, Sweden

fUCR Uppsala Clinical Research Center, Uppsala University, Uppsala, Sweden

A B S T R A C T

We investigated if maternal body burdens of perﬂuoroalkyl acids (PFAAs) at the time of delivery are associated with birth outcome and if early life exposure (in utero/nursing) is associated with early childhood growth and weight gain. Maternal PFAA body burdens were estimated by analysis of serum samples from mothers living in Uppsala County, Sweden (POPUP), sampled three weeks after delivery between 1996 and 2011. Data on child length and weight were collected from medical records and converted into standard deviation scores (SDS).

Multiple linear regression models with appropriate covariates were used to analyze associations between maternal PFAA levels and birth outcomes (n = 381). After birth Generalized Least Squares models were used to analyze associations between maternal PFAA and child growth (n = 200). Inverse associations were found between maternal levels of perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA), and per- fluoroundecanoic acid (PFUnDA), and birth weight SDS with a change of −0.10 to −0.18 weight SDS for an inter-quartile range (IQR) increase in ng/g PFAA. After birth, weight and length SDS were not significantly associated with maternal PFAA. However, BMI SDS was significantly associated with PFOA, PFNA, and PFHxS at 3 and 4 years of age, and with PFOS at 4 and 5 years of age. If causal, these associations suggest that PFAA affects fetal and childhood body development in different directions.

1. Introduction

Owing to their stability and combined water and oil repelling prop- erties, perﬂuoroalkyl moieties have been incorporated into numerous commercial products for over 50 years. Over 3000 per- and poly- ﬂuoroalkyl substances (PFASs) are known to exist on the global market.

Among these, the perﬂuoroalkyl acids (PFAAs) have garnered consider- able international attention due to their environmental persistence and global occurrence in humans and wildlife. Pathways of human exposure are numerous and include food, drinking water, dust and air. In addition to direct exposure to PFAAs via any one of the aforementioned routes, exposure can also occur via precursors which may transform to PFAAs following exposure (D'Hollander et al., 2010; Vestergren et al., 2012;

Gebbink et al., 2015; Gyllenhammar et al., 2015).

Studies investigating associations between maternal PFAA concentrations in serum/plasma and birth weight in humans have reported

conflicting results (Johnson et al., 2014; Bach et al., 2015). Most studies have focused on perfluorooctane sulfonate (PFOS) and per- fluorooctanoate (PFOA) although humans are exposed to numerous other PFASs, only a fraction of which are known.

Studies of associations between maternal PFAA levels and childhood weight gain are scarce. Previous work has pointed out negative associations between prenatal PFOA and PFOS exposure and child weight at 5 and 12 months of age (Andersen et al., 2010) and positive associations between maternal PFOS levels and weight at 20 months (Maisonet et al., 2012), and PFOA and risk of overweight at 20 years of age (Halldorsson et al., 2012). Increased prenatal PFAA exposure has also been associated with increased adiposity in 3–20-year-olds (Halldorsson et al., 2012; Braun et al., 2016; Mora et al., 2017).

However, results from studies of prenatal PFAA exposure and childhood weight gain and adiposity are not consistent as some of them show no signiﬁcant associations (Andersen et al., 2013; Barry et al., 2014).

https://doi.org/10.1016/j.envint.2017.12.002

Received 2 June 2017; Received in revised form 27 November 2017; Accepted 1 December 2017

⁎Corresponding author.

E-mail address:Irina.gyllenhammar@slv.se(I. Gyllenhammar).

Available online 20 December 2017

T

(2)

Intra uterine growth restriction (IUGR) and/or being born small for gestational age (SGA) have been related to increased risk of impaired glucose tolerance, increased adiposity, and increased blood pressure later in life, especially among individuals with an early and rapid catch- up in weight after birth (Ibanez et al., 2006; Kerkhof et al., 2012; Gishti et al., 2014). Therefore, it is of interest to investigate fetal/neonatal PFAA exposure in relation to fetal, infant and child weight development and growth.

The aim of the present study was to investigate if maternal body burdens of PFAA at delivery are associated with birth outcomes and offspring weight gain and growth. Based on the present knowledge, we hypothesize that fetal PFAA exposure could lead to a decreased birth weight and thereby possibly also influence offspring weight development.

2. Materials and methods

2.1. Participants

Serum samples were collected 1996–2011 from ﬁrst-time mothers from Uppsala County, Sweden, within the POPUP study (Persistent Organic Pollutants in Uppsala Primiparas). The study was restricted to singleton births with no birth defects. All women were Swedish born and almost all women had a full-term pregnancy (37–42 weeks), except in a few cases with the shortest gestational length being 35 weeks and the longest 43 weeks. The women in the POPUP cohort are homogenous regarding ethnic background, as indicated by the fact that only about 4% of the Swedish population had one or two foreign-born parents in 2006 (Socialstyrelsen, 2009). Samples were taken three weeks after delivery and thereafter biobanked at the Swedish National Food Agency. For details about recruitment, blood sampling, and collection of personal characteristics data seeGlynn et al. (2007)andLignell et al.

(2009). In total 381 mothers who delivered theirﬁrst child were analyzed for PFAAs (Table 1). During the ﬁrst years of the study (1996–1999) women were also sampled in the 3rd trimester of pregnancy. The study was approved by the local ethics committee in Up- psala, Sweden, and the participating women gave informed consent for themselves and for their children.

2.2. Chemical analyses

Serum levels of 7 PFAAs, including perfluorooctanoic acid (PFOA), perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA), per- fluoroundecanoic acid (PFUnDA), perfluorobutane sulfonate (PFBS), perfluorohexane sulfonate (PFHxS), and perfluorooctane sulfonate (PFOS), were determined in mothers three weeks after delivery. In a subset of the participating women (n = 20), the aforementioned PFAAs were also analyzed in 3rd trimester serum in order to investigate correlations in levels between late pregnancy and 3 weeks after delivery.

The analytical procedure has been described previously (Glynn, 2012;

Vestergren et al., 2012). Briefly, serum was spiked with isotopically- labeled internal standards, extracted with acetonitrile, and then sub- jected to clean-up with graphitized carbon, prior to instrumental analysis by ultra performance liquid chromatography-tandem mass spec- trometry (UPLC-MS/MS). Method detection limits (MDLs) were defined based on the quantified background contamination signals. In the ab- sence of procedural blank contamination, MDLs were defined as the lowest concentration in a serum sample giving a signal-to-noise ratio of 3. All batches included a procedural blank (water) and a control sample (replicate pooled human serum, n = 21). In addition, NIST SRM 1957 was analyzed to evaluate the accuracy of the method. CVs for control samples were typically between 15 and 22% with the exception of PFBS which showed higher variability, owing to concentrations closer to MDLs (Supplementary data, Table A1). Analysis of NIST reference material revealed that measured concentrations were consistent with reference values for all targets (Supplementary data, Table A2).

2.3. Infants/children

Weight, length and head circumference of the newborn infants (n = 381) were collected from the Swedish Medical Birth Register.

During infancy and childhood, measurements of weight, length/height and head circumference at 3, 6, 12 and 18 months and 3, 4 and 5 years of age (n = 200) were collected from child health records in Uppsala County. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared at ages 3, 4, and 5 years. Standard deviation scores (SDS) were calculated from growth standards according toKarlberg et al. (2001),Wikland et al. (2002), andNiklasson and Albertsson-Wikland (2008).

2.4. Glomerularﬁltration rate (GFR)

In order to calculate estimated maternal GFR (eGFR), levels of creatinine and cystatin C were analyzed in serum samples from the mothers three weeks after delivery. In 20 women creatinine and cystatin C were also measured in late pregnancy, in order to investigate if the eGFR at this time point corresponded to that estimated after delivery.

Serum samples were analyzed on Abbott Architect ci8200 and ci16200 instruments (Abbott Park, IL, USA). Reagents (8L24-02) for the enzymatic IDMS-traceable creatinine method were from Abbott.

Reagents for the immunoturbidimetric cystatin C method (Cystatin C reagent kit, no 1101) were from Gentian (Moss, Norway), and the re- cently described IFCC-certiﬁed international calibration procedure (Grubb et al., 2014) was used. The total analytical imprecision of Table 1

Characteristic of the mothers and children participating in the study.

Characteristics N Mean (range) %

Mothers

Age (years) 381 29 (20–41)

Pre-pregnancy BMI^a(kg/m²) 380 23 (16–40) Weight gain during pregnancy (%)^b 379 22 (−4.1–54) Weight loss from delivery to sampling (%)^c 343 14 (4.3–28)

Fish consumption (g/day) 360 30 (0−220)

Years of education 372

≤4 years of high school 123 33

1–3 years of higher education 81 22

> 3 years of higher education 168 45

Smoking during pregnancy 380

Smoker 44 12

Non-smoker 234 62

Former smoker 102 27

Children

Boys 217 57

Girls 164 43

Gestational length (days) 381 280 (244–302)

Birth weight (g) 381 3575 (2159–5420)

Birth weight (SDS) 381 −0.15 (−3.1–3.1)

Birth length (SDS) 381 −0.057 (−3.9–3.0)

Head circumference 376 −0.37 (−3.6–3.3)

SGAWd 14 3.6

SGALd 21 5.5

GFRcce(mL/min/1.73 m²) 305 103 (48–176)

GFRcreatf(mL/min/1.73 m²) 305 92 (60–166)

Total breastfeeding (months) 192 6.5 (0– > 12)

Breastfeeding at 3 months 184 96

Breastfeeding at 6 months 161 84

aBody mass index.

bWeight gain was calculated as weight of the mother just before delivery (Wd) compared to weight before pregnancy (Wbp): [(Wd− Wbp) / Wbp] × 100.

cWeight of the mother just before delivery (Wd), weight of the mother at the time of blood sampling (Ws), and the birth weight of the child (Wb) were used to calculated this weight loss: [(Wd− Ws− Wb) / Wd] × 100.

dSmall for gestational age, weight or length <−2SDS.

eEstimated glomerularﬁltration rate, calculated using cystatin C-levels.

fEstimated glomerularﬁltration rate, calculated using creatinine-levels.

(3)

cystatin C measurements were 3.1% and 2.0% at 12 and 38μmol/L, respectively. Correspondingﬁgures for creatinine were 3.4% and 2.3%

at 1.7 and 3.2 g/L. eGFR was calculated using either cystatin C (GFRcc) or creatinine level (GFRcreat). GFRccwas calculated according toGrubb et al. (2014)and GFRcreataccording toNyman et al. (2014)in the unit mL/min/1.73 m²BSA (body surface area).

2.5. Statistical analyses

Statistical analyses were performed using R version 3.3.1 (R Core Team, 2016) with the rms (Harrell, 2017) and nlme (Pinheiro et al., 2016) packages. PFBS, PFDA, and PFUnDA were the only targets investigated in this work with concentrations below MDLs. Due to the reduced number of detects for PFBS (67%), samples were grouped based on mothers with PFBS concentrations < MDL and mothers with PFBS concentrations≥MDL. In contrast, PFDA and PFUnDA displayed a much higher frequency of detects (96 and 90% of samples, respectively). For these targets, concentrations below the MDL were imputed by loge-transforming the data≥MDL for a given year and then ﬁtting these data to a cumulative normal distribution using a Generalized Reduced Gradient algorithm available through the Solver add-in for Microsoft Excel (John, 1998). Prior studies have demonstrated that imputation-based approaches such as this oﬀer a better estimate of non- detectable concentrations with smaller bias and fewer errors compared to replacement techniques (e.g. 0.5 × MDL or MDL/2^1/2) (Croghan and Egeghy, 2003).

The correlations between PFAA serum levels in 3rd trimester and 3 weeks after delivery were tested with Spearman's rank correlation test.

Multiple linear regression models were used to analyze associations between maternal PFAA levels and birth outcomes. Generalized Least Squares models (Pinheiro and Bates, 2000) were used to analyze associations between PFAA levels and child weight gain and growth. For birth outcomes the covariates sampling year, maternal age, pre-pregnancy BMI, years of education, smoking during pregnancy, maternal weight gain during pregnancy, maternal weight loss after delivery, and total ﬁsh consumption were included as covariates in the regression models. We have registered medical data on 278 of the women in the study and 2 women had gestational diabetes and 9 women had gestational hypertension. It was not considered as confounders in this size of population. For child weight gain and growth the covariates sampling year, maternal age, pre-pregnancy BMI, maternal weight gain during pregnancy, maternal weight loss after delivery, whether or not the child was breastfeeding at three and/or six months and duration of breastfeeding and years of education were included. These variables have been shown to be signiﬁcantly associated with serum PFAA levels (Bjermo et al., 2013; Brantsaeter et al., 2013; Gyllenhammar et al., 2015) or with birth weight and child growth (Preedy, 2012). To investigate possible confounders, a DAG was performed and the covariates that are used in the models have been evaluated.

Generalized Least Squares is an extension of ordinary least squares to handle repeated measurements within an individual but circumvents the problem mixed models with random eﬀects have of requiring

complex approximations of distributions of test statistics (Harrell, 2001). What is required is the speciﬁcation of the within individual correlation structure. The correlation structure used in this study was an autoregressive-moving average lag 1, which assumes that the correlation between two time points depends on the distance in time between them.

Initially, all continuous variables were modeled using restricted cubic splines with four interior knots. Knowing that these models may overﬁt the data, degrees of freedom were re-allocated based on each variable's importance in each model as judged by each variable'sχ² statistic minus the degrees of freedom. The rationale behind this strategy is that it is probably more important to model the important variables correctly, i.e. allow for non-linear/non-monotonic de- pendencies, than it is for less important variables. Associations were allowed to diﬀer between boys and girls by including interaction terms between sex and the PFAAs. The interaction terms were then all kept or removed based on their importance in the models. Sensitivity analyses were done with eGFR as an additional adjustment factor.

Estimates in tables andﬁgures are presented as the predicted change in the outcome variable for an inter-quartile range (IQR) increase in each PFAA, i.e. when the PFAA is increased from the 25th to the 75th percentile.

3. Results

3.1. PFAA levels in maternal serum

Maternal PFAA levels are shown inTable 2. In the Uppsala mothers, PFOS was present at the highest concentrations, followed by PFHxS and PFOA. Levels of PFBS above the MDL were found in 57% of the mothers. Serum levels of each homologue (PFOA, PFNA, PFDA, PFUnDA, PFHxS, and PFOS) were signiﬁcantly correlated between late pregnancy and three weeks after delivery (Spearman's correlation coeﬃcient:

0.64–0.94, n = 20, p < 0.001–0.002) in a subset of the participating women (Table 3).

3.2. eGFR

eGFR diﬀered slightly between calculations based on creatinine and cystatin C (Table 1). Simple regression analyses showed that maternal Table 2

PFAA levels (ng/g) in mothers 3 weeks after delivery.

PFAA n Mean (SE) Min 25th 50th 75th Max < MDL (%)

PFOA 381 2.4 (0.06) 0.20 1.6 2.3 3.0 13 0

PFNA 381 0.46 (0.01) 0.062 0.31 0.41 0.57 2.4 0

PFDA 381 0.23 (0.007) < MDL 0.14 0.20 0.28 1.1 3

PFUnDA 381 0.19 (0.006) < MDL 0.11 0.17 0.26 0.91 15

PFBS 351^a 0.03 (0.003) < MDL < MDL 0.012 0.037 0.80 43

PFHxS 379^b 3.6 (0.20) 0.32 1.4 2.4 4.0 26 0

PFOS 377^b 14 (0.44) 2.1 7.4 13 19 46 0

a30 samples were not quantiﬁable due to elevated blank levels.

bThe internal standard signal was too low for quantiﬁcation in 2–4 samples.

Table 3

Spearman's correlation between PFAA levels in 3rd trimester and 3 weeks after delivery, n = 20.

PFAA Correlation coeﬃcient p

PFOA 0.94 < 0.001

PFNA 0.80 < 0.001

PFDA 0.64 0.002

PFUnDA 0.67 0.001

PFHxS 0.83 < 0.001

PFOS 0.82 < 0.001

(4)

GFRcc was signiﬁcantly associated with maternal levels of PFOA (β = 0.004 ± 0.002, p = 0.022). GFRccwas also inversely associated with gestational length (β = −0.07 ± 0.03, p = 0.025) but not with birth SDS weight or SDS length. GFRccwas slightly lower in late pregnancy than three weeks after delivery, estimated to 95 ± 20 and 107 ± 14 mL/min/1.73 m², respectively (Mann Whitney test, p = 0.066). GFRcreat, was higher in late pregnancy than three weeks after delivery, 117 ± 10 and 97 ± 11 mL/min/1.73 m², respectively (Mann Whitney test, p < 0.0001). No statistically signiﬁcant correlations were seen between late pregnancy and post-delivery GFRcc or GFRcreat(data not shown).

3.3. Birth outcomes

At birth there were inverse associations between birth weight SDS and maternal levels of long-chained PFCAs (Fig. 1), being significant for PFDA and PFUnDA and borderline so for PFNA (p = 0.054). The results for associations between PFAAs and birth length SDS followed the same pattern as for birth weight SDS but were not significant. Maternal PFOS concentrations had a borderline significant association with gestational length (p = 0.056) (Fig. 1). All variables were modeled linearly in all models with the exception of maternal BMI and weight gain during pregnancy which were modeled using restricted cubic splines with three knots. When the associations were adjusted for eGFR the

statistical significance for birth weight remained for PFDA and PFUnDA (Supplementary data, Figs. A1 and A2). Since a high maternal weight gain during pregnancy could be a consequence of a fetus gaining a lot of weight, adjusting for it may bias the association between PFAA and birthweight. The statistical analyses for birth outcomes were thus performed without maternal weight gain during pregnancy as sensitivity analyses but the results showed no significant differences from those with adjustment for maternal weight gain (Fig. 1, Supplementary data, Fig. A3).

No associations were found between PFAA levels and head circumference (SDS) (Fig. 1) or between PFBS and birth outcomes (data not shown).

3.4. Growth and weight development in early childhood

We observed no signiﬁcant associations between maternal PFAA levels and childhood weight SDS during the period from 3 months to 5 years of age, although there was a tendency of positive associations for PFOA, PFNA, PFHxS, and PFOS already at 3 months of age (Fig. 2).

No associations were observed between length SDS and maternal PFAA levels (Fig. 3). For BMI SDS, only analyzed at 3, 4, and 5 years of age, PFOA, PFNA and PFHxS had a similar pattern with positive associations between maternal levels and BMI SDS in children at age 3 and 4 although becoming less strong and not signiﬁcant at age 5 (Fig. 4).

Contrast

PFOA PFNA PFDA PFUnDA PFHxS PFOS

−4−3−2−101

Gestational length [Days]

PFOA PFNA PFDA PFUnDA PFHxS PFOS

−0.4−0.20.00.20.4

Birth weight [SDS]

−0.4−0.20.00.20.4

PFOA PFNA PFDA PFUnDA PFHxS PFOS Birth length [SDS]

−0.20.00.20.4

PFOA PFNA PFDA PFUnDA PFHxS PFOS Head circumference [SDS]

Fig. 1. Associations between maternal PFAA concentrations and birth outcomes (n = 377–381). Multiple linear regression models were used with the covariates sampling year, maternal age, pre pregnancy BMI, maternal weight gain during pregnancy, maternal weight loss after delivery, years of education, smoking during pregnancy, and totalﬁsh consumption. Results are shown as adjusted mean (95% conﬁdence interval) change in gestational length or SDS for an interquartile range increase in maternal PFAA.

(5)

Maternal PFOS levels were also positively associated with BMI SDS but on the contrary the association became stronger with increasing age and was signiﬁcant at 4 and 5 years of age (Fig. 4).

Maternal levels of PFOA and PFHxS were positively associated with head circumference SDS and significant at 18 months for PFOA and 12 and 18 months for PFHxS (Fig. 5). No other significant associations were found. No evidence of different associations for boys or girls were found as p-values for the interaction terms between sex and the PFAAs in the models varied between 0.20 and 0.79 for birth outcomes and between 0.19 and 0.99 for the longitudinal outcomes.

All variables were modeled linearly in the models for growth and weight development with the exception of attained age which was modeled using a restricted cubic spline with four knots. Attained age was modeled with a second degree polynomial for the outcomes BMI and Head circumference since there were not enough time points to generate the spline basis functions.

4. Discussion

In the present study increased maternal levels of PFNA, PFDA and PFUnDA were associated with reduced birth weight. We observed no consistently signiﬁcant associations for PFBS, PFHxS, PFOS, and PFOA.

Associations between birth weight and exposure to PFOA and PFOS have been examined in several previous studies. However, investiga- tions of other PFAAs and birth outcomes are scarce. In Taiwanese girls, similar to our results, prenatal levels of PFNA, PFDA, PFUnDA, and perﬂuorododecanoic acid (PFDoDA) were inversely associated with birth weight (Wang et al., 2016). In contrast to our observations, in a study of Danish mothers and children 2008–2013, estimated mean birth weight was reduced in groups with PFHxS exposure above the lowest exposure quartile. This was also observed for perﬂuoroheptane sulfonate (PFHpS), which was not studied by us (Bach et al., 2016). In

another study, from South Korea, on cord serum levels of nine PFAAs (including PFOA and PFOS) and birth weight no signiﬁcant associations were found (Lee et al., 2016). Our results are consistent with previous studies showing associations between increased PFAA exposure and reduced birth weight, however the relation of diﬀerent homologues remains unclear.

Three recent reviews of PFOA and PFOS associations with birth weight have concluded that there is suﬃcient human and animal evidence that developmental exposure to PFOA and PFOS reduces fetal growth/birth weight (Johnson et al., 2014; Lam et al., 2014; Bach et al., 2015). Although PFOA and PFOS concentrations were in the same range in the present study we did not observe associations between increased maternal PFOS and PFOA and decreased birth weight in the statistical analyses. This could for instance be due to diﬀerence in study design, covariates included in the regression models, and the relatively limited sample size in our study.

At least three potential pathways have been suggested by which in utero exposure to the long-chain perfluorinated carboxylic acids (PFCAs) might affect offspring weight (Halldorsson et al., 2012): Firstly, they may interfere with ovary development in utero, leading to impaired estrogen synthesis; secondly, PFAAs may interact with the per- oxisome proliferator-activated receptors (PPAR) PPARα and PPARγ, which are involved in lipid metabolism in adipocytes; and thirdly, PFAA may interact with thyroid hormones (Halldorsson et al., 2012).

It has been suggested that fetal size, maternal GFR, and maternal PFAA concentrations are all correlated, and that assessment of the fetal size (or birth weight) associations with maternal PFAA should be adjusted for maternal GFR, especially if maternal PFAA is determined around delivery(Verner et al., 2015). Urinary excretion of PFCAs de- creases with increased chain-length of the homologues up to C9-C10, and among the PFAAs studied by us PFOA appears to be the homologue with the highest degree of urinary excretion (Zhang et al., 2013). In our Age in months

Weight (SDS)

−0.2 0.0 0.2 0.4

PFOA PFNA PFDA

3 6 12 18 36 48 60

PFUnDA

3 6 12 18 36 48 60

PFHxS

3 6 12 18 36 48 60

−0.2 0.0 0.2 0.4 PFOS

Fig. 2. Associations between maternal PFAA concentrations and weight SDS over time in children at 3 to 60 months (n = 182–193). Generalized Least Squares models were used with the covariates sampling year, maternal age, pre pregnancy BMI, maternal weight gain during pregnancy, maternal weight loss after delivery, years of education, and total time of breastfeeding. Results are shown as adjusted mean (95% conﬁdence interval) change in SDS for an interquartile range increase in maternal PFAA.

(6)

study, PFOA was the only PFAA for which maternal serum levels were significantly associated with maternal eGFR measured 3 weeks after birth, in this case a positive association. This is in contrast to data in- dicating that higher GFR would lead to lower PFOA levels (Verner et al., 2015). Other studies have shown associations between low GFR and reduced birth weight (Morken et al., 2014), however that was not shown in the present study when eGFR was used. Maternal eGFR adjustment of the associations between birth weight and maternal PFAAs had no marked influence on the results in the present study (Supple- mentary data, Figs. A1 and A2). All women in this study were healthy with normal renal function, and it may be possible that the inter-individual variation in GFR was not large enough to confound associations between long-chain PFCAs and birth weight due to low urinary excretion of these PFAA homologues. GFR during the 3rd trimester did not correlate with that analyzed 3 weeks after delivery demonstrating that the intra-individual variation also was large in this cohort with healthy pregnant women. This could be due to the changes in renal function that take place during normal pregnancy and thefirst weeks postpartum (Hussein and Lafayette, 2014). Blood samples for eGFR estimations taken at other times before or during pregnancy, or in the first weeks after delivery may give other results. There are also un- certainties using eGFR, since all equations estimating GFR from creatinine or cystatin C are approximations. However, the gold standard for measuring GFR is using plasma or urinary clearance of an exogenous injected filtration marker, e.g. iohexol clearance (Levey and Inker, 2016). Such methods are invasive and not feasible in the present set- ting.

There is an association between intrauterine growth restriction and risk of impaired glucose tolerance and increased blood pressure later in life possibly leading to metabolic syndrome in adulthood (Ibanez et al., 2006). Especially individuals with an early and rapid catch-up weight gain and growth are at risk (Kerkhof et al., 2012). Animal models have

shown possible mechanisms via reduced number of nephrons in the kidney and pancreatic beta-cell hypoplasia underlying these conditions (Woods et al., 2004; Green et al., 2010). Early catch-up weight gain is also associated with obesity later in life (Monteiro and Victora, 2005).

Since we found that maternal PFDA and PFUnDA, and to some extent also PFNA, were negatively associated with birth weight SDS, the question is whether higher maternal levels of these PFAAs at delivery are associated with faster weight gain and growth during childhood.

We observed a tendency of positive associations between maternal PFOA, PFNA, PFHxS and PFOS and child weight already at 3 months of age. This result did however not give strong support for the hypothesis that reduced fetal growth due to in utero exposure to PFAA results in early-life catch-up growth, since we did only observe borderline sig- niﬁcant decreased birth weight SDS with increased maternal PFNA, and no signiﬁcant associations for PFOA, PFHxS, and PFOS. Higher maternal PFOA, PFNA, and PFHxS levels were associated with increased BMI SDS at 3 and 4 years of age, and higher maternal PFOS with higher BMI SDS at 4 and 5 years of age. Rapid growth during early childhood and high childhood BMI has been associated with obesity later in life (Brisbois et al., 2012).

Studies on associations between PFAS exposure and childhood weight gain and growth are few and not consistent. In study on British girls high in utero exposure to PFOS was associated with a lower birth weight and a higher weight at 20 months, compared to those with lower exposure (Maisonet et al., 2012). Danish women had a higher risk of overweight at 20 years of age after prenatal exposure to PFOA (Halldorsson et al., 2012). In a Danish study maternal PFOS and PFOA levels during pregnancy were inversely related to child weight up to 7 years of age, although not statistically signiﬁcant (Andersen et al., 2010, 2013). In this Danish cohort negative associations between PFOS and PFOA levels and birth weight have previously been reported (Fei et al., 2007, 2008). Increased prenatal PFAA exposure has also been Age in months

Length (SDS)

−0.2 0.0 0.2

PFOA PFNA PFDA

3 6 12 18 36 48 60

PFUnDA

3 6 12 18 36 48 60

PFHxS

3 6 12 18 36 48 60

−0.2 0.0 0.2 PFOS

Fig. 3. Associations between maternal PFAA concentrations and length SDS over time in children at 3 to 60 months (n = 182–193). Generalized Least Squares models were used with the covariates sampling year, maternal age, pre pregnancy BMI, maternal weight gain during pregnancy, maternal weight loss after delivery, years of education, and total time of breastfeeding. Results are shown as adjusted mean (95% conﬁdence interval) change in SDS for an interquartile range increase in maternal PFAA.

(7)

associated with increased adiposity in children and adolescents (Halldorsson et al., 2012; Braun et al., 2016; Mora et al., 2017).

However, other studies of prenatal PFAA exposure and childhood weight and adiposity show no signiﬁcant associations (Andersen et al., 2013; Barry et al., 2014; Timmermann et al., 2014). None of these studies investigated child weight gain or linear growth.

Some studies suggest diﬀerent associations of PFAAs and birth outcomes by infant sex.Andersen et al. (2010)studied 1010 children and found that increased maternal levels of PFOS and PFOA were associated with lower weight and BMI at 5 and 12 months of age in boys.

Washino et al. (2009)found negative associations between increased maternal levels of PFOS and birth weight in girls.Robledo et al. (2015) studied children to 234 couples and association between a several contaminants representing different classes and found an inverse association between maternal perfluorooctanesulfonamide (PFOSA) and birth weight in boys. Our data do not support different associations by sex for any of the studied outcomes. Tests for interaction in regression models require a larger sample size than tests for main effects and have limited power in small data sets. Our sample size is much smaller than inAndersen et al. (2010)but is on par with the sample size inWashino et al. (2009)and larger than the sample used byRobledo et al. (2015) so sample size alone cannot explain the differences. One explanation could be the different outcome measurements used. While the above mentioned studies studied birthweight in grams, we used sex and age standardized measures so the effects of sex on the outcomes are already accounted for. Differences in covariates included in the regression models and study design may also explain the lack of clear PFAA by sex interactions in our study.

4.1. Strengths and limitations

In the present study seven diﬀerent PFAA were evaluated. Most

other studies in this ﬁeld have only presented results for PFOS and PFOA. The use of standard deviation scores have the advantage of di- rectly correcting for gestational age as well as postnatal age, thus compensating for the variations in actual time for the individual measurements. In the present study maternal levels of PFAA were analyzed in serum sampled three weeks after delivery, which makes the results harder to interpret. The ideal study design would have included esti- mation of maternal PFAA before pregnancy or at least in early pregnancy, in order to avoid the suggested confounding with maternal GFR around delivery. Serum levels of each homologue (PFOA, PFNA, PFDA, PFUnDA, PFHxS, and PFOS) in POPUP women were strongly correlated between late pregnancy and three weeks after delivery and in an earlier study strong correlations have also been seen between PFAA concentrations after delivery and in both early and late pregnancy (Glynn et al., 2012). Method of delivery could be suggested to be a confounder in this type of study. We have not found any association with PFAA concentrations or birth outcome and method of delivery (data not shown). A study from Sweden found no diﬀerence in mean estimated blood loss between planned vaginal delivery and planned caesarean section studying 451 women (Larsson et al., 2011). This demonstrates that maternal PFAA levels measured 3 weeks after delivery provide good estimates of levels in early and late pregnancy, although residual confounding still cannot be excluded. We used birth outcomes data from the Swedish Medical Birth register and body weight and length from child health records, instead of self-reported data.

5. Conclusion

The present study (including PFOA, PFNA, PFDA, PFUnDA, PFHxS and PFOS), on one hand indicates that early-life exposure to long- chained PFCAs is associated with decreased fetal growth, and on the other hand an increased childhood BMI with increasing maternal PFOA, Age in months

BMI (SDS)

−0.2 0.0 0.2 0.4

PFOA PFNA PFDA

36 48 60

PFUnDA

36 48 60

PFHxS

36 48 60

−0.2 0.0 0.2 0.4 PFOS

Fig. 4. Associations between maternal PFAA concentrations and BMI SDS over time in children at 36 to 60 months (n = 182–193). Generalized Least Squares models were used with the covariates sampling year, maternal age, pre pregnancy BMI, maternal weight gain during pregnancy, maternal weight loss after delivery, years of education, and total time of breastfeeding. Results are shown as adjusted mean (95% conﬁdence interval) change in SDS for an interquartile range increase in maternal PFAA.

(8)

PFNA, PFHxS, and PFOS. Further follow-up of associations between early-life PFAA exposure and growth at later ages, including puberty, is warranted.

Acknowledgements

Funding was provided by the Swedish Environmental Protection Agency (219 1202), the Swedish Research Council Formas (2010- 1300), and the Swedish Civil Contingencies Agency (dnr 1637/2012).

Appreciation is expressed to the participating women and to the mid- wives who assisted in recruitment, interviewing, and sample collection.

We also thank Ingalill Gadhasson, Anne-Soﬁe Kärsrud, Thomas Alsberg, and Oskar Sandblom for excellent laboratory support.

Appendix A. Supplementary data

Supplementary data to this article can be found online athttps://

doi.org/10.1016/j.envint.2017.12.002.

References

Andersen, C.S., Fei, C., Gamborg, M., Nohr, E.A., Sorensen, T.I., Olsen, J., 2010. Prenatal exposures to perﬂuorinated chemicals and anthropometric measures in infancy. Am.

J. Epidemiol. 172 (11), 1230–1237.

Andersen, C.S., Fei, C., Gamborg, M., Nohr, E.A., Sorensen, T.I., Olsen, J., 2013. Prenatal exposures to perﬂuorinated chemicals and anthropometry at 7 years of age. Am. J.

Epidemiol. 178, 921–927.

Bach, C.C., Bech, B.H., Brix, N., Nohr, E.A., Bonde, J.P., Henriksen, T.B., 2015.

Perﬂuoroalkyl and polyﬂuoroalkyl substances and human fetal growth: a systematic review. Crit. Rev. Toxicol. 45, 53–67.

Bach, C.C., Bech, B.H., Nohr, E.A., Olsen, J., Matthiesen, N.B., Bonefeld-Jorgensen, E.C., Bossi, R., Henriksen, T.B., 2016. Perﬂuoroalkyl acids in maternal serum and indices of fetal growth: the Aarhus Birth Cohort. Environ. Health Perspect. 124, 848–854.

Barry, V., Darrow, L.A., Klein, M., Winquist, A., Steenland, K., 2014. Early life per- ﬂuorooctanoic acid (PFOA) exposure and overweight and obesity risk in adulthood in

a community with elevated exposure. Environ. Res. 132, 62–69.

Bjermo, H., Darnerud, P.O., Pearson, M., Barbieri, H.E., Lindroos, A.K., Nalsen, C., Lindh, C.H., Jonsson, B.A., Glynn, A., 2013. Serum concentrations of perﬂuorinated alkyl acids and their associations with diet and personal characteristics among Swedish adults. Mol. Nutr. Food Res. 57, 2206–2215.

Brantsaeter, A.L., Whitworth, K.W., Ydersbond, T.A., Haug, L.S., Haugen, M., Knutsen, H.K., Thomsen, C., Meltzer, H.M., Becher, G., Sabaredzovic, A., Hoppin, J.A., Eggesbo, M., Longnecker, M.P., 2013. Determinants of plasma concentrations of perﬂuoroalkyl substances in pregnant Norwegian women. Environ. Int. 54, 74–84.

Braun, J.M., Chen, A., Romano, M.E., Calafat, A.M., Webster, G.M., Yolton, K., Lanphear, B.P., 2016. Prenatal perﬂuoroalkyl substance exposure and child adiposity at 8 years of age: the HOME study. Obesity (Silver Spring) 24, 231–237.

Brisbois, T.D., Farmer, A.P., McCargar, L.J., 2012. Early markers of adult obesity: a review. Obes. Rev. 13, 347–367.

Croghan, C., Egeghy, P., 2003. Methods of Dealing With Values Below the Limit of Detection Using SAS. Southern SAS User Group, pp. 22–24.

D'Hollander, W., de Voogt, P., De Coen, W., Bervoets, L., 2010. Perﬂuorinated substances in human food and other sources of human exposure. Rev. Environ. Contam. Toxicol.

208, 179–215.

Fei, C., McLaughlin, J.K., Tarone, R.E., Olsen, J., 2007. Perﬂuorinated chemicals and fetal growth: a study within the Danish National Birth Cohort. Environ. Health Perspect.

115, 1677–1682.

Fei, C., McLaughlin, J.K., Tarone, R.E., Olsen, J., 2008. Fetal growth indicators and perﬂuorinated chemicals: a study in the Danish National Birth Cohort. Am. J.

Epidemiol. 168, 66–72.

Gebbink, W.A., Glynn, A., Darnerud, P.O., Berger, U., 2015. Perﬂuoroalkyl acids and their precursors in Swedish food: the relative importance of direct and indirect dietary exposure. Environ. Pollut. 198, 108–115.

Gishti, O., Gaillard, R., Manniesing, R., Abrahamse-Berkeveld, M., van der Beek, E.M., Heppe, D.H., Steegers, E.A., Hofman, A., Duijts, L., Durmus, B., Jaddoe, V.W., 2014.

Fetal and infant growth patterns associated with total and abdominal fat distribution in school-age children. J. Clin. Endocrinol. Metab. 99, 2557–2566.

Glynn, A., 2012. Perﬂuorerade alkylsyror (PFAA) i Uppsala dricksvatten. In: R. B. A.

Division (Ed.), Riskvärdering. Swedish National Food Agency.

Glynn, A., Aune, M., Darnerud, P.O., Cnattingius, S., Bjerselius, R., Becker, W., Lignell, S., 2007. Determinants of serum concentrations of organochlorine compounds in Swedish pregnant women: a cross-sectional study. Environ. Health 6, 2.

Glynn, A., Berger, U., Bignert, A., Ullah, S., Aune, M., Lignell, S., Darnerud, P.O., 2012.

Perﬂuorinated alkyl acids in blood serum from primiparous women in Sweden: serial sampling during pregnancy and nursing, and temporal trends 1996–2010. Environ.

Sci. Technol. 46, 9071–9079.

Green, A.S., Rozance, P.J., Limesand, S.W., 2010. Consequences of a compromised Age in months

Head Circumference (SDS)

−0.2 0.0 0.2 0.4

PFOA PFNA PFDA

3 6 12 18

PFUnDA

3 6 12 18

PFHxS

3 6 12 18

−0.2 0.0 0.2 0.4 PFOS

Fig. 5. Associations between maternal PFAA concentrations and head circumference SDS over time in children at 3 to 18 months (n = 152–177). Generalized Least Squares models were used with the covariates sampling year, maternal age, pre pregnancy BMI, maternal weight gain during pregnancy, maternal weight loss after delivery, years of education, and total time of breastfeeding. Results are shown as adjusted mean (95% conﬁdence interval) change in SDS for an interquartile range increase in maternal PFAA.

(9)

intrauterine environment on islet function. J. Endocrinol. 205, 211–224.

Grubb, A., Horio, M., Hansson, L.O., Bjork, J., Nyman, U., Flodin, M., Larsson, A., Bokenkamp, A., Yasuda, Y., Blufpand, H., Lindstrom, V., Zegers, I., Althaus, H., Blirup-Jensen, S., Itoh, Y., Sjostrom, P., Nordin, G., Christensson, A., Klima, H., Sunde, K., Hjort-Christensen, P., Armbruster, D., Ferrero, C., 2014. Generation of a new cystatin C-based estimating equation for glomerularﬁltration rate by use of 7 assays standardized to the international calibrator. Clin. Chem. 60, 974–986.

Gyllenhammar, I., Berger, U., Sundstrom, M., McCleaf, P., Euren, K., Eriksson, S., Ahlgren, S., Lignell, S., Aune, M., Kotova, N., Glynn, A., 2015. Inﬂuence of con- taminated drinking water on perﬂuoroalkyl acid levels in human serum—a case study from Uppsala, Sweden. Environ. Res. 140, 673–683.

Halldorsson, T.I., Rytter, D., Haug, L.S., Bech, B.H., Danielsen, I., Becher, G., Henriksen, T.B., Olsen, S.F., 2012. Prenatal exposure to perﬂuorooctanoate and risk of overweight at 20 years of age: a prospective cohort study. Environ. Health Perspect. 120, 668–673.

Harrell, F.E., 2001. Regression Modeling Strategies : With Applications to Linear Models, Logistic Regression, and Survival Analysis. Springer, New York.

Harrell, F.E.J., 2017. rms: Regression Modeling Strategies. R Package Version 5.1-1.

Hussein, W., Lafayette, R.A., 2014. Renal function in normal and disordered pregnancy.

Curr. Opin. Nephrol. Hypertens. 23, 46–53.

Ibanez, L., Ong, K., Dunger, D.B., de Zegher, F., 2006. Early development of adiposity and insulin resistance after catch-up weight gain in small-for-gestational-age children. J.

Clin. Endocrinol. Metab. 91, 2153–2158.

John, E., 1998. Simpliﬁed curve ﬁtting using spreadsheet add-ins. Int. J. Eng. Educ. 14, 375–380.

Johnson, P.I., Sutton, P., Atchley, D.S., Koustas, E., Lam, J., Sen, S., Robinson, K.A., Axelrad, D.A., Woodruﬀ, T.J., 2014. The Navigation Guide - evidence-based medicine meets environmental health: systematic review of human evidence for PFOA eﬀects on fetal growth. Environ. Health Perspect. 122, 1028–1039.

Karlberg, J., Luo, Z.C., Albertsson-Wikland, K., 2001. Body mass index reference values (mean and SD) for Swedish children. Acta Paediatr. 90, 1427–1434.

Kerkhof, G.F., Willemsen, R.H., Leunissen, R.W., Breukhoven, P.E., Hokken-Koelega, A.C., 2012. Health proﬁle of young adults born preterm: negative eﬀects of rapid weight gain in early life. J. Clin. Endocrinol. Metab. 97, 4498–4506.

Lam, J., Koustas, E., Sutton, P., Johnson, P.I., Atchley, D.S., Sen, S., Robinson, K.A., Axelrad, D.A., Woodruﬀ, T.J., 2014. The Navigation Guide - evidence-based medicine meets environmental health: integration of animal and human evidence for PFOA eﬀects on fetal growth. Environ. Health Perspect. 122, 1040–1051.

Larsson, C., Saltvedt, S., Wiklund, I., Andolf, E., 2011. Planned vaginal delivery versus planned caesarean section: short-term medical outcome analyzed according to in- tended mode of delivery. J. Obstet. Gynaecol. Can. 33, 796–802.

Lee, E.S., Han, S., Oh, J.E., 2016. Association between perﬂuorinated compound concentrations in cord serum and birth weight using multiple regression models. Reprod.

Toxicol. 59, 53–59.

Levey, A.S., Inker, L.A., 2016. GFR as the“Gold Standard”: estimated, measured, and true.

Am. J. Kidney Dis. 67, 9–12.

Lignell, S., Aune, M., Darnerud, P.O., Cnattingius, S., Glynn, A., 2009. Persistent organochlorine and organobromine compounds in mother's milk from Sweden 1996–2006: compound-speciﬁc temporal trends. Environ. Res. 109, 760–767.

Maisonet, M., Terrell, M.L., McGeehin, M.A., Christensen, K.Y., Holmes, A., Calafat, A.M., Marcus, M., 2012. Maternal concentrations of polyﬂuoroalkyl compounds during pregnancy and fetal and postnatal growth in British girls. Environ. Health Perspect.

120, 1432–1437.

Monteiro, P.O., Victora, C.G., 2005. Rapid growth in infancy and childhood and obesity in

later life—a systematic review. Obes. Rev. 6, 143–154.

Mora, A.M., Oken, E., Rifas-Shiman, S.L., Webster, T.F., Gillman, M.W., Calafat, A.M., Ye, X., Sagiv, S.K., 2017. Prenatal exposure to perﬂuoroalkyl substances and adiposity in early and mid-childhood. Environ. Health Perspect. 125, 467–473.

Morken, N.H., Travlos, G.S., Wilson, R.E., Eggesbo, M., Longnecker, M.P., 2014. Maternal glomerularﬁltration rate in pregnancy and fetal size. PLoS One 9, e101897.

Niklasson, A., Albertsson-Wikland, K., 2008. Continuous growth reference from 24th week of gestation to 24 months by gender. BMC Pediatr. 8, 8.

Nyman, U., Grubb, A., Larsson, A., Hansson, L.O., Flodin, M., Nordin, G., Lindstrom, V., Bjork, J., 2014. The revised Lund-Malmo GFR estimating equation outperforms MDRD and CKD-EPI across GFR, age and BMI intervals in a large Swedish population.

Clin. Chem. Lab. Med. 52, 815–824.

Pinheiro, J.C., Bates, D.M., 2000. Mixed-eﬀects Models in S and S-PLUS. Springer, New York.

Pinheiro, J., Bates, D., DebRoy, S., D, S., 2016. nlme: Linear and Nonlinear Mixed Eﬀects Models_. R Package Version 3.1-128. R Core Team.

Preedy, V.R., 2012. Handbook of Growth and Growth Monitoring in Health and Disease.

Springer, New York.

R Core Team, 2016. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.

Robledo, C.A., Yeung, E., Mendola, P., Sundaram, R., Maisog, J., Sweeney, A.M., Barr, D.B., Louis, G.M., 2015. Preconception maternal and paternal exposure to persistent organic pollutants and birth size: the LIFE study. Environ. Health Perspect. 123, 88–94.

Socialstyrelsen, 2009. 374 Folkhälsorapport.

Timmermann, C.A., Rossing, L.I., Grontved, A., Ried-Larsen, M., Dalgard, C., Andersen, L.B., Grandjean, P., Nielsen, F., Svendsen, K.D., Scheike, T., Jensen, T.K., 2014.

Adiposity and glycemic control in children exposed to perﬂuorinated compounds. J.

Clin. Endocrinol. Metab. 99, E608–614.

Verner, M.A., Loccisano, A.E., Morken, N.H., Yoon, M., Wu, H., McDougall, R., Maisonet, M., Marcus, M., Kishi, R., Miyashita, C., Chen, M.H., Hsieh, W.S., Andersen, M.E., Clewell 3rd, H.J., Longnecker, M.P., 2015. Associations of perﬂuoroalkyl substances (PFAS) with lower birth weight: an evaluation of potential confounding by glo- merularﬁltration rate using a physiologically based pharmacokinetic model (PBPK).

Environ. Health Perspect. 123, 1317–1324.

Vestergren, R., Berger, U., Glynn, A., Cousins, I.T., 2012. Dietary exposure to per- ﬂuoroalkyl acids for the Swedish population in 1999, 2005 and 2010. Environ. Int.

49, 120–127.

Wang, Y., Adgent, M., Su, P.H., Chen, H.Y., Chen, P.C., Hsiung, C.A., Wang, S.L., 2016.

Prenatal exposure to perﬂuorocarboxylic acids (PFCAs) and fetal and postnatal growth in the Taiwan Maternal and Infant Cohort Study. Environ. Health Perspect.

124, 1794–1800.

Washino, N., Saijo, Y., Sasaki, S., Kato, S., Ban, S., Konishi, K., Ito, R., Nakata, A., Iwasaki, Y., Saito, K., Nakazawa, H., Kishi, R., 2009. Correlations between prenatal exposure to perﬂuorinated chemicals and reduced fetal growth. Environ. Health Perspect. 117, 660–667.

Wikland, K.A., Luo, Z.C., Niklasson, A., Karlberg, J., 2002. Swedish population-based longitudinal reference values from birth to 18 years of age for height, weight and head circumference. Acta Paediatr. 91, 739–754.

Woods, L.L., Weeks, D.A., Rasch, R., 2004. Programming of adult blood pressure by maternal protein restriction: role of nephrogenesis. Kidney Int. 65, 1339–1348.

Zhang, Y., Beesoon, S., Zhu, L., Martin, JW., 2013. Biomonitoring of perﬂuoroalkyl acids in human urine and estimates of biological half-life. Environ. Sci. Technol. 47 (18), 10619–10627.