Of the 1826 women with singleton, live births, and gestational age >34 weeks, 69 were missing GWG data. Of the remaining 1757 women with GWG data who met the study inclusion criteria, 34 participants were missing first trimester PFAS data and 456 were missing cord blood PFAS data. The analytical samples thus comprised 1723 participants with first trimester maternal exposure data and 1301 participants with cord blood exposure data.
Study population characteristics are shown in Table 1
. The majority of women were of normal pre-pregnancy BMI, gained in excess of the U.S. IOM gestational weight gain guidelines, had a household income greater than $50,000, and were non-smokers. Descriptive statistics for key demographic characteristics were compared between the analytic sample and the whole MIREC cohort. The rate of current smokers was slightly lower (5.1%) in the analytical sample than the overall cohort (5.5%). There were no notable differences in maternal age at delivery or pre-pregnancy BMI between the analytical sample and the overall cohort.
Median maternal plasma PFAS concentrations were higher than those in cord blood (Table 2
). The percentage of maternal plasma samples with detectable values of PFOA, PFOS, and PFHxS was 99.8%, 99.8%, and 95.7%, respectively. The percentage of cord blood plasma samples with detectable values of PFOA, PFOS, and PFHxS was 89.1%, 48.2%, and 38.0%, respectively. The Spearman correlation coefficient among chemicals varied from 0.34 (between cord blood PFOA and maternal PFHxS) to 0.68 (between maternal and cord blood levels of PFOA) (all p
-values < 0.05).
This association between 1st trimester PFAS concentrations and total GWG did not vary according to pre-pregnancy BMI (all p
-values for the BMI x
PFAS product terms >0.1) (Table 3
). In order to present estimates independent of pre-pregnancy BMI, all results are presented by strata of BMI. Among underweight/normal and obese women, each PFAS was positively associated with modest increases in GWG though the association was only statistically significant in the PFOS-GWG model among underweight/normal and normal women. Doubling PFOS concentrations was associated with a 0.39 (95% CI: 0.02, 0.75) kg increase in GWG in the underweight/normal pre-pregnancy BMI subgroup. Among overweight women, the association was positive for PFOA and inverse for PFOS and PFHxS though the null value was included in all confidence intervals.
In the analysis using continuous GWG as the independent variable, GWG was significantly associated with elevated odds of high cord blood PFOA and PFOS concentrations (Table 4
). An interquartile increase in GWG (7 kg) was associated with 33% increase in odds of high cord blood PFOA levels (95% CI: 1.13, 1.56). Associations were similar among male (OR = 1.40 95% CI: 1.12, 1.75) and female infants (OR = 1.26 95% CI: 1.00, 1.59) (p
-value interaction term = 0.76). An IQR increase in GWG exposure was also associated with significantly increased odds of high PFOS exposure (OR = 1.20 95% CI: 1.03, 1.40) with similar results among males and females.
In this cohort of Canadian women and their newborns, a doubling of PFOS concentrations was associated with modest, statistically significant increases in GWG among women in the underweight or normal pre-pregnancy BMI category but not the overweight or obese pre-pregnancy BMI category. It is possible that the lack of statistical significance in the overweight and obese categories was influenced by the lower sample size. For example, the magnitude of association between PFOA and GWG is the same among underweight and obese women yet the confidence interval is notably wider among the smaller subgroup of obese women. We also observed that 1 kg and IQR (7 kg) increases in GWG were associated with statistically significantly elevated concentrations of cord blood PFOA and PFOS. There was no observed effect modification by infant sex in the relationship between GWG and cord blood PFAS concentrations. Exposure concentrations in the MIREC study were comparable to a representative sample of the Canadian population. Median PFAS concentrations in women ages 20–39 in the Canadian Health Measures Study (median PFOS = 6.4 ug/L, PFOA = 2.1 ug/L, PFHxS = 1.2 ug/L) were comparable to MIREC study participants (median PFOS = 4.6 ug/L, PFOA = 1.7 ug/L, PFHxS = 1.0 ug/L) [6
A limited number of epidemiological and experimental studies provide some insight into the physiological relations between PFAS exposure and metabolic and hormonal pathways. One birth cohort study reported that in utero
PFOA exposure was associated with obesity among the offspring at age 20 [10
]. Low-dose in utero
PFOA exposure has been reportedly associated with mid-life increases in leptin and insulin in an animal model [8
]. A mouse cell study reported that all three PFASs measured in this study were associated with increased expression of genes, such as the leptin gene, involved in lipid metabolism and adipocyte differentiation [11
]. These findings provide some biological plausibility for a positive association between PFAS exposure and GWG. Among adults, a limited number of studies have examined the relations between PFAS concentrations and metabolic function. For example, a cross-sectional analysis of the Canadian Health Measures Study reported that blood PFHxS concentrations, but not PFOS or PFOA, were positively associated with total cholesterol levels in adults [29
]. Another study among pregnant women reported that maternal PFAS concentrations were associated with higher levels of thyroid hormones [30
However, PFAS exposure has also been shown to induce peroxisome proliferator-activated receptor (PPAR) pathways [8
]. PPAR pathway activation may minimize obesity related inflammation and have an anti-obesogenic effect [31
]. This potential mechanism provides biological plausibility for the previously observed inverse association between PFOA exposure and birth weight [12
]. In sum, PFAS exposure may operate through multiple physiological pathways involved in hormonal and metabolic homeostasis. Moreover, these multiple pathways may operate in opposing manners. It is not possible, based on our data, to determine which, if any, of these mechanisms underlie the observed results. Further experimental work that builds upon current understanding of the adiposity related effects of PFAS [32
] and that attempts to determine how placental transfer of PFAS varies according to GWG and BMI is necessary to define biological mechanisms underlying the observed results.
Due to the ubiquity of exposure and multiple pathways for contact with PFASs, it is also difficult to determine the relative contribution of different exposure sources to an individual’s body burden [33
]. Diet, particularly ingestion of animal fats, meat, and snack foods (e.g., microwave popcorn), is an established and primary source of PFAS exposure [33
]. Ingestion of contaminated food may result in simultaneous increases in caloric intake, leading to gestational weight gain and its association with maternal PFOA levels. We attempted to preserve a temporal relation between maternal exposure and GWG by using a first trimester measure of exposure. However, due to their long half-life, first trimester measures are correlated with levels throughout pregnancy [26
]. Thus, we cannot definitively state whether maternal PFOA concentrations are a determinant of GWG or a consequence of the ingestion of contaminated food that may elevate both PFOA levels and GWG. Considering that specific sources of exposure cannot be ascertained based on maternal and fetal blood concentrations, it is difficult to determine whether maternal ingestion patterns (consumption of PFAS contaminated food via food packaging, cookware, or processing) contributed to the observed relationships. The observed associations between GWG and neonatal PFOA exposure may be driven by the high correlation (r
= 0.68) between maternal and cord blood measures. If women with excess GWG have high PFOA levels, the relation between GWG and cord blood PFOA levels may be explained by the distribution of maternal contaminants into the fetal compartment.
Our study has several strengths, notably the relatively large sample size and availability of PFAS data in both mothers and cord blood. We were able to control for key covariates and had a study population recruited from 10 different cities across Canada. Interpretation of our results warrants consideration of three primary limitations. First, as previously noted, is the inability to disentangle the potential influence of an unmeasured confounder, namely dietary patterns, on the relation between PFOA and GWG. Second, it is possible that our results were confounded by maternal changes in plasma volume. As plasma volume increases throughout pregnancy [36
], levels of contaminants may be diluted. We anticipate this potential influence to be minimal because maternal plasma PFAS concentrations were measured during the first trimester, prior to the time when maximal changes in plasma volume occur [36
]. Any potential influence of plasma volume dilution would have likely resulted in negative confounding and an underestimate of the true associations between maternal PFAS concentrations and GWG. Cord blood PFASs concentrations are largely insulated from plasma volume changes [36
], and, therefore, also unlikely to be influenced by plasma dilution. Third, we did not have the capacity to adjust for other physiological characteristics, namely glomerular filtration rate [38
], that may confound the observed associations. In addition, due to the exploratory nature of this research, our objective was to examine the associations with individual chemicals. We did not attempt to examine potential synergy among chemicals or cumulative exposure. Fourth, the lack of statistical significance in the overweight and obese categories may have been influenced by the low sample size in these subgroups. Replication in a study with a larger sample size would be necessary to rule out type 2 error in the null association between maternal PFAS concentrations and gestational weight gain.