Perfluoroalkyl Substances (PFASs) are a group of man-made fluorine-containing compounds that are used broadly in industry and commercial products since the 1950s. PFASs are very persistent in the environment and in living organisms, and thus are detected in wildlife and in humans throughout the globe. PFASs can cross the placental barrier and expose the fetus during the most vulnerable period in development. Animal data suggested that PFASs are neurotoxic and have endocrine disruptive properties. A few cross-sectional studies linked current level of PFASs to hyperactivity disorders in children, but finding from prospective studies are lacking. Additional human studies are urgently needed to examine the potential neuro-developmental impacts of these ubiquitous exposures.
We conducted a case-cohort study within the framework of the Danish National Birth Cohort (DNBC) to investigate whether prenatal exposure to PFASs increases the risk of infantile autism, attention-deficit/hyperactivity disorder (ADHD) or congenital cerebral palsy (CP) in children. Among 83,389 mother-child pairs enrolled in the DNBC during 1996-2002, we identified 890 ADHD cases and 301 autism cases in the Danish National Hospital Register and the Danish Psychiatric Central Registry. We also identified 156 CP cases from the Danish National Cerebral Palsy Register. For cost-efficiency, we randomly selected 220 cases of ADHD and autism each, and all 156 CP cases in the three disease groups. In addition, we randomly selected 550 controls from the cohort frequency matched on child's sex. Sixteen` PFASs were measured in maternal plasma samples from early or mid-pregnancy, but in analyses we focus on the six PFASs that were quantifiable in more than 90% of the samples. PFAS concentrations were analyzed as continuous variables (with natural-log transformation) or categorized into quartiles. We estimated Odds Ratios (OR) and Risks Ratios (RR) for the outcomes.
First, we found that prenatal exposures to PFASs may increase the risk for CP in boys. We observed higher risks of CP in boys with increasing maternal PFAS levels (RR=1.74 (95%CI 1.05-2.88) per one unit (natural-log ng/mL) increase in perfluorooctane sulfonate (PFOS) and RR=1.99 (95%CI 1.15-3.44) per unit increase in perfluorooctanoic acid (PFOA)). We also observed a dose-response pattern of CP risks in boys per PFOS and PFOA quartile (p-trend<0.01). Both spastic unilateral or bilateral CP sub-phenotypes were found to be associated with PFASs, but no association between PFASs and CP was found in girls. For ADHD and autism, we found no consistent patterns to suggest that prenatal PFAS exposures increase the risks of ADHD or autism in children, but some observed positive as well as inverse association in secondary analyses should be further explored.
We also conducted a bias analysis to investigate whether conditioning on live-born status may induce sufficiently large bias to explain some of the unexpected inverse associations seen between prenatal PFAS exposures and ADHD in our and previous studies. We employed directed acyclic graphs to present our causal assumptions and the hypothesized scenarios, and used Monte-Carlo techniques to simulate a realistic pregnancy cohort based on the distributions in the DNBC and prior research knowledge. We found that the protective associations of PFAS on ADHD observed in the live-born cohort could possibly be explained by selection bias if PFAS reduces fetal survivals, while PFAS has a true null effect on ADHD. The magnitude of this selection bias due to fetal death are generally small, unless the exposure is a strong determinant of fetal loss and there are one or more strong risk factors of the outcome that also reduce chances of fetal survival. One way to reduce or even eliminate the bias due to conditioning on fetal survival is to adjust for the common causes of the outcome and fetal losses in analyses.
In conclusion, we found that prenatal exposure to PFASs may increases the risks for CP in boys, and no consistent associations were found between maternal PFAS levels and ADHD or autism in children. In additional bias analysis, we quantitatively showed that conditioning on life-born status in observational study can yield a negative bias if an exposure of interests reduces conception or fetal survival among the high risks fetuses for the targeted outcome.