Prenatal Exposure to Perfluorooctanoate and Risk of Overweight at 20 Years of Age: A Prospective Cohort Study

1Because of their repellent properties and low surface tension, fluorochemicals have been used extensively for more than half a century in various commercial applications, including carpets, textiles, personal care products, leveling and wetting agents and in food-contact materials. Use and production of fluorochemicals has resulted in the environmental presence of perfluoroalkyl acids (PFAAs), which are highly persistent compounds consisting of a fully fluorinated carbon chain (C4–C14) with, for example, a terminal sulfonate or carboxylate functional group. The environmental presence of PFAAs may be a consequence of emissions during manufacturing, use, or disposal of products (direct sources) as well as transformation of other precursors to PFAAs (indirect sources). The two PFAAs most widely detected in humans are the C8-backbone derivatives perfluorooctanoate (PFOA) and perfluoro-octane sulfonate (PFOS), with estimated elimination half-lives of around 4 and 5 years, respectively.

2After reports of global contamination, phaseout of C8-derivatives was initiated in the year 2000 by the major U.S. manufacturers (U.S. Environmental Protection Agency 2010). This has resulted in decreased human exposure compared with peak levels in Europe and the United States. However, recent findings from the Health and Nutrition Examination Survey (NHANES) indicate that PFOA levels in the United states remained relatively stable between 2003 and 2008, suggesting continued direct and indirect sources of human exposures to PFOA.

3PFAAs accumulate primarily via binding to albumin in blood (D’eon et al. 2010; ), and they are readily transported across the placenta. PFAAs are found in highest concentrations in the liver and low-dose human exposures to PFOA and PFOS have been associated with modest increases in liver enzymes and blood lipids in cross-sectional settings. The relevance of these findings is uncertain, however, because prospective studies have not been conducted. Findings from animal studies suggest that PFAAs may act as endocrine disruptors, affecting circulating estrogens levels through pathways such as aromatase induction in the liver or through altering estrogen-expressive genes, although findings with respect to interactions with estrogen receptors have not been consistent.

4Experimental studies have demonstrated that in utero exposures to certain endocrine disruptors, including xenoestrogens, may lead to permanent changes in metabolic pathways that regulate body weight. A recent study using a CD-1 mouse model observed higher body weight and increased insulin and leptin levels in post-pubertal female offspring after low-dose in utero exposures to PFOA. The aim of this study was to explore these findings in environmentally exposed pregnant women with prospective offspring follow-up at 20 years of age.

5The birth cohort. Between April 1988 and January 1989, 965 women with singleton pregnancies were recruited for a birth cohort study in Aarhus, Denmark. This was 80% of a consecutive sample of 1,212 women attending a midwife center in the city, covering a well-defined geographical area. A face-to-face interview was conducted at a routine midwife visit in gestational week 30, covering medical history, anthropometry, diet, lifestyle, and socioeconomic factors. A blood sample was also collected, and it was immediately separated into serum, plasma, and erythrocytes and frozen at –20°C. Further information on maternal health and birth outcomes were extracted from hospital records and the Danish Medical Birth Registry.

6Offspring follow-up. In 2008–2009, study mothers and offspring were contacted, and offspring were asked to fill out a Web-based questionnaire concerning their current health, lifestyle, and dietary habits and their current height and weight. In addition, a tape measure was mailed to all potential participants with instructions on how to measure their waist circumference. Offspring were also invited to participate in a 75–100-min clinical examination, which included standardized anthropometric measures and collection of a fasting blood sample that was immediately centrifuged and frozen at –80°C. The study was approved by the Danish Data Protection Agency and the Danish Council of Ethics (reference no. 20070157), and all participants gave written consent prior to inclusion into the study.

7Mother–offspring pairs available for analysis. A total of 915 offspring of the 965 women with singleton pregnancies recruited in 1988–1999 were located and contacted in 2008–2009, and 692 of the offspring agreed to participate in the follow-up study. Missing values on offspring anthropometry (n = 4) and maternal blood samples (n = 23) left 665 mother–-offspring pairs, or 69% of the original cohort, available for analyses. Of the 665 offspring included in this study, 423 (44% of the original cohort) attended the clinical examination, and 242 completed the Web-based questionnaire only.

8Maternal PFAA samples. Serum samples from gestational week 30 were measured at the Department of Analytical Chemistry at the Norwegian Institute of Public Health in Oslo. The analytical procedure and sources of chemicals used have been described in detail elsewhere. Nineteen PFAAs were analyzed, and 12 were above the limit of quantification (LOQ; 0.05 ng/mL serum for all PFAAs reported in this study) in one or more samples. For quantification of PFOS, the total area of the linear and branched isomers was integrated. The quality of the analytical procedure was monitored by analyzing in-house quality control samples and human serum samples from interlaboratory comparison exercises (Institut national de santé publique Québec 2009). For the interlaboratory comparison exercises, the coefficient of variation for PFOA was between 3% and 9%, depending on the concentration of the reference sample. No traces of PFAAs above LOQ were observed in any of the procedural blanks.

9Offspring biomarkers of adiposity. In short, adiponectin was quantified by a time-resolved immunofluorometric assay based on two antibodies and recombinant human adiponectin (R&D Systems, Abingdon, United Kingdom), as previously described. Leptin was determined by a time-resolved immunofluorometric assay based on commercial reagents (R&D Systems) using recombinant human leptin as the standard; otherwise, quantification was carried out essentially as for adiponectin. Plasma insulin concentrations were determined by the commercial Insulin ELISA kit (Dako, Copenhagen, Denmark).

10Outcome measures. Body mass index (BMI) and waist circumference measures from the clinical examination were used when available (n = 423). Self-reported measures from the Web-based questionnaire were used for the remaining 242 subjects. Although subjects tend to underestimate their weight by self-report, relative ranks appeared to be retained, as the correlation between measured and self-reported BMI was 0.91 for both sexes (Spearman r, n = 423). Use of self-reported measures does, however, result in underestimation of cases when dichotomizing subjects as overweight and obese (BMI ≥ 25 kg/m2). In our data the sensitivity coefficient and specificity coefficients were 0.46 and 0.99, respectively, for identifying overweight and obese female offspring based on self-report. The corresponding numbers were 0.63 and 0.98 for male offspring, respectively.

11Statistical analyses. The mean and SD were used to describe normal variables, whereas median and interquartile range (IQR) were used for skewed variables. Student t-tests or chi-square tests were used to test departures from the null hypothesis for normal outcomes, and the Kruskal–Wallis test was used for skewed outcomes. Linear regression was used for continuous outcomes. For dichotomous outcomes, relative risk (RR) was estimated using log-Poisson regression with robust variance estimation, as implemented in PROC GENMODE in SAS (version 9.2; SAS Institute Inc., Cary, NC, USA). All analyses were performed for males and females separately. When examining association with anthropometric outcomes, maternal PFOA concentrations were divided into quartiles and a test for linear trend was performed by using the ordinal values. For the biomarker analysis, PFOA was entered as a continuous variable.

12When estimating the association between in utero PFOA concentration and offspring anthropometry at 20 years of age, we adjusted for the following potential confounders identified a priori: maternal age (continuous, no missing); maternal education (elementary schooling, high school or technical schooling, university education, higher academic education, other education, 6% missing); maternal smoking (never, < 10 and ≥ 10 cigarettes/day, 6% missing); pre-pregnancy BMI [restricted cubic spline regression, 2% missing]; parity (0, 1, ≥ 2, no missing); infant birth weight (continuous, 0.2% missing); and offspring age at follow-up (continuous, 1% missing). A total of 67 subjects (10%) had missing values for one or more of the covariates included. Maternal age and parity were included because they are known predictors of PFAA exposure. Maternal education and smoking were included to account for potential social and lifestyle confounding. Pre-pregnancy BMI and birth weight were included because they are known predictors of offspring BMI. Missing covariate values were substituted using multiple imputation, as implemented in PROC MI in SAS (SAS Institute, Inc.).

13Because of skewed distributions, offspring biomarkers of adiposity were log-transformed when examining their association with maternal PFOA concentrations, and estimates were adjusted for offspring age and timing of blood sample collection.

14Non-significant differences in maternal PFOA concentration, prepregnancy BMI, and parity were observed between mothers whose offspring did not participate in follow-up, mothers whose offspring completed the Web-based questionnaire only, and mothers whose offspring participated in the clinical examination (Table 1). However, mothers whose offspring did not participate in follow-up were more likely to smoke during pregnancy. Females who participated in clinical examination had significantly lower BMI (≈ 0.8 kg/m2) compared with those who completed the Web-based questionnaire only, whereas non-significant differences were observed for waist circumference and current smoking. Male offspring who participated in the clinical examination were not significantly different from those who completed the Web-based questionnaire only with regard to BMI, waist circumference, and current smoking.

15The median maternal PFOA serum concentration was 3.7 ng/mL (range: 0.1–19.8). Maternal PFOS, perfluorooctane sulfonamide (PFOSA), and perfluorononanoate (PFNA) concentrations increased across quartiles of PFOA concentration (Table 2). Eight other PFAAs were quantified but were not included in analyses because concentrations were low (median: ≤ 0.4 ng/mL) and were within a narrow range (IQR < 0.2 ng/mL). Mothers in the highest quartile of PFOA concentration were less likely to be parous (32% vs. 58%) and were younger (28.7 vs. 29.8 years) than mothers in the lowest quartile of PFOA when enrolled into the original study.

16The prevalence of overweight or obesity (BMI ≥ 25 kg/m2) was similar between male and female offspring (19% and 18%, respectively), but waist circumference above action level II was less common in males (3.5% > 102 cm) than in females (16.2% > 88 cm) (Table 3). All males and 70% of females with waist circumference above action level II were overweight or obese. Females were more likely than males to report that they were currently on a diet (17% vs. 6%).

17Maternal PFOA concentrations were positively associated (p for trend < 0.05) with BMI and waist circumference among female offspring at 20 years of age (n = 345) in both unadjusted and covariate-adjusted analyses (Table 4). In adjusted analyses, female offspring whose mothers were in the highest quartile of PFOA concentration had 1.6 kg/m2 [95% confidence interval (CI): 0.6, 2.6] higher BMI and 4.3 cm (95% CI: 1.4, 7.3) higher waist circumference compared with offspring whose mothers were in the lowest quartile. No statistically significant associations were observed for males (p for trend > 0.05, n = 320). Female offspring of mothers in the highest versus lowest PFOA quartile were also more likely to be overweight [RR = 3.1 (95% CI: 1.4, 6.9)] and to have a waist circumference above action level II at 20 years of age [3.0 (95% CI: 1.3, 6.8)] (Table 5). PFOA was not associated with overweight in male offspring, and RR estimates were consistently null. Only 11 of 320 male offspring had a waist circumference above action level II; consequently, RR estimates were unstable with wide CIs.

18Among female participants who provided a blood sample at clinical examination (n = 252), maternal PFOA concentration was positively associated with insulin, leptin, and the leptin-adiponectin ratio and inversely associated with adiponectin levels (Table 6). The associations corresponded to a 2–7% change in these biomarkers per 1-unit increase in PFOA. Point estimates were similar for male offspring but non-significant (n = 170). Although the precision of the estimated associations between maternal PFOA and the biomarkers was reduced because of the small sample size for this analysis (particularly in males), associations between maternal PFOA and offspring anthropometry were comparable with those for the follow-up participants as a whole, supporting the validity of these estimates. For example, estimated increases in BMI comparing highest with lowest quartile of maternal PFOA concentration were 1.8 kg/m2 (95% CI: 0.6, 3.0) and 0.6 kg/m2 (95% CI: –0.7, 1.9) for female and male offspring who provided a blood sample at clinical examination, compared with 1.6 kg/m2 (95% CI: 0.6, 2.6) and 0.6 kg/m2 (95% CI: –0.3, 1.5) for all female and male participants, respectively.

19Concerning other PFAAs, we observed that PFOS, PFNA, and PFOSA (continuous variables) were in univariate analysis either borderline (p for trend = 0.05 for PFOSA) or significantly (p for trend < 0.05 for PFNA, PFOS) positively associated with female offspring BMI at 20 years. However, after adjusting for PFOA, the regression coefficients became non-significant (p for trend ≥ 0.56 in all cases, data not shown). We also found no evidence that the association between maternal PFOA concentration and offspring anthropometry was influenced by the presence of the other PFAAs quantified. For example, the estimated increase in female offspring BMI with a 1-ng/mL increase in maternal PFOA in a univariate analysis was 0.43 kg/m2 (95% CI: 0.25, 0.60) compared with 0.46 kg/m2 (95% CI: 0.21, 0.71) based on a model that included maternal PFOS, PFOSA, and PFNA as continuous terms.

20Associations with offspring BMI were also stable with respect to maternal parity. For female offspring of nulliparous mothers (n = 191), we estimated a 0.46-kg/m2 (95% CI: 0.27, 0.66) increase in BMI per 1-ng/mL increase in PFOA, compared with 0.36 kg/m2 (95% CI: 0.03, 0.69) for female offspring of parous mothers (interaction p-value 0.65).

21In this cohort of environmentally exposed pregnant women, we observed a positive association between in utero exposure to PFOA and the prevalence of overweight and a high waist circumference among female offspring, but not male offspring, at 20 years of age. In addition, in utero exposure to PFOA was associated with biomarkers of adiposity among male and female offspring, although estimates for males were not significant.

22Animal studies have shown that early--life exposures to xenoestrogens such as diethylstilbestrol and bisphenol-A can lead to permanent disruption in endocrine functions that regulate adiposity. One observation from these studies is that low-dose exposures may result in significant weight changes in the exposed animals only after adult age is reached, whereas a pattern of reduced weight in early infancy followed by rapid catch-up growth is often observed at higher doses. Human studies have provided some support for these findings, including reports of rapid weight gain up to 3 years of age after in utero exposures to polychlorinated biphenyls as well as increased overweight at 6 years of age after in utero exposure to hexachloro-benzene. However, metabolic pathways that might explain these findings are not well characterized, and long-term follow-up studies have not been conducted.

23Indications that environmental exposures to PFAAs may affect cardiometabolic risk factors have, so far, been obtained from cross-sectional studies focusing on end points such as blood lipids and glucose homeostasis, but in these studies, findings with respect to anthro-pometry have been inconsistent. Interpretation of cross-sectional findings on anthropometry is complex, however, as absorption and distribution of PFAAs depend on related factors such as lean body mass and blood volume.

24However, our findings are consistent with a recent experimental study on CD-1 mice that reported increased body weight in postpubertal female offspring after low-dose (0.01–0.3 mg/kg body weight) in utero exposure to PFOA. In that study, increases in weight were most pronounced in mid-life and were accompanied by elevated levels of leptin and insulin in the exposed animals. At higher doses (≥ 1mg/kg), weight gain was not observed. Furthermore, exposing the offspring after weaning as opposed to in utero also had no effect on weight gain. In at least two animal studies, researchers did not observe increased body weight after in utero exposure to PFOA, but one study used higher doses (≥ 1mg/kg) and the other followed animals only until 12 weeks of age.

25Similar to the findings of, we observed significant associations between maternal PFOA and insulin and leptin in addition to adiponectin levels among female offspring. Changes in these biomarkers were modest, however, and corresponded to what could be expected from the observed increase in offspring BMI. The results from our biomarker analysis provide valuable insight, as elevated insulin levels predict central fat in both normal weight and overweight women. Therefore, these results provide added weight to our findings, as BMI and waist circumference are relatively crude measures of body composition.

26The absence of an association between in utero PFOA exposure and weight gain among male offspring is noteworthy but not unexpected. Sex-specific differences with respect to weight gain have been observed for xenoestrogens, where female offspring appear to be more affected. Results for male offspring were not reported in the CD-1 mouse study, and experimental studies have suggested that reproductive development, but not weight gain, may be a more sensitive end point for males with respect to developmental exposures to PFAAs. Despite these observations, results for male offspring in our study should be interpreted cautiously, as many study participants were still in their late adolescence, and relatively few males had high waist circumference compared with females. Because girls gain more fat mass and boys gain more fat-free mass during adolescent age, weight gain in male offspring at a later age when accumulation of fat mass increases cannot be excluded.

27Concerning mechanism, there are at least three potential pathways by which in utero exposure to PFOA might affect offspring weight. First, PFOA may interfere with ovary development in utero, leading to impaired estrogen synthesis among female offspring. This mechanism is supported by the observation that in utero PFOA exposure did not affect weight gain in the previously mentioned study on CD-1 mice when ovariectomy was performed in exposed offspring after delivery. Second, PFAAs may interact with the peroxisome proliferator--activated receptors (PPAR) PPARα and PPARγ, which are involved in lipid metabolism in adipocytes. Whether such an interaction can lead to sex-specific differences in weight gain remains unclear. Third, thyroid hormones might play some role: The results of one cross-sectional study suggested that PFOA exposure may lead to an increased risk of thyroid disease, which is more common in females than males. A recent study from the C8 Health project also observed a modest positive association of PFOA with serum thyroxine and an inverse association with T3 uptake. The overall evidence linking PFAAs to thyroid function is still weak, particularly as prospective studies have not been conducted, and no clear effect on thyroid function has been observed in occupationally exposed individuals.

28A unique strength of our study is the long-term prospective follow-up with outcome assessment at an age where weight changes due to linear growth have leveled off, and it was possible to adjust for several prenatal factors including maternal BMI, smoking, and education. We were also able to support our findings on offspring anthropometry with assessment on serum biomarkers of adiposity in a subset of study participants. Furthermore, quantification of a total of 12 PFAAs made it possible to conclude that PFAAs other than PFOA were unlikely to play an important role in our study population, either because of a low concentration or a lack of association after adjusting for PFOA.

29Concerning limitations, we acknowledge that uncertainties regarding routes of human exposure to PFAAs may have resulted in a failure to identify important confounders. We also acknowledge that maternal PFOA concentrations in late gestation may be influenced by factors such as blood volume expansion, decreased albumin concentration, and fetal uptake, and these factors may predict fetal outcomes such as birth weight. However, with a temporal separation of 20 years, our findings should be less prone to such influence. Furthermore, PFOA concentrations in samples collected at around gestational weeks 6–12 and 24 and in cord blood have been observed to be highly cor-related (r ≈ 0.8–0.9).

30Finally, as with all longitudinal studies, losses during follow-up are inevitable, and this is perhaps the main limitation of our study. Having quantified maternal PFOA concentrations in 874 (or 72%) of 1,212 women who were originally invited to participate, it is reassuring that exposure levels were non-differential with respect to level of participation in the -follow-up study. A higher prevalence of smoking among mothers whose offspring did not participate suggests that losses due to follow-up were related to some extent to maternal lifestyle. There were relatively modest differences in anthropometry between offspring who participated in the clinical examination and those who completed the Web-based questionnaire only. Nonetheless, our secondary analyses of offspring biomarkers of adiposity may be compromised because of the low participation rate, particularly among males.

31In a cohort of environmentally exposed pregnant women, a positive association between in utero exposure to PFOA and the prevalence of overweight and high waist circumference was observed in female offspring at 20 years of age. Our findings are in line with recent experimental findings and provide added support for the hypothesis that early-life exposure to certain endocrine disruptors, even at low concentrations, may play a role in the current obesity epidemic. Given the widespread detection of PFOA in humans and wildlife and observed increase in related perfluorinated compounds of similar biological potential, it is of considerable public health importance to understand and eliminate pathways of human exposures to PFAAs.