Determinants of plasma poly- and perfluoroalkyl substances during pregnancy: The Japan Environment and Children's...

Per- and polyfluoroalkyl substances (PFAS), a group of compounds associated with adverse health outcomes, have been identified widely in humans. This study investigated the determinants of PFAS in plasma collected from pregnant women enrolled in the Japan Environment and Children's Study from 2011 to 2014. Several machine learning approaches were used, and the XGBoost model had the best predictive performance for seven PFAS quantified in more than 50 % of the population (from R² = 0.34 and RMSE = 0.39 ng/mL for perfluorononanoic acid to R² = 0.85 and RMSE = 0.19 ng/mL for perfluoroundecanoic acid). Feature analysis revealed that variation among regional centres was the most critical determinant. While parity was negatively associated with most PFAS concentrations and maternal age was positively associated with it, a positive interaction between parity and age suggests that PFAS accumulation with increasing age may have a greater impact than the reduction in PFAS levels associated with parity. Fish intakes demonstrated strong interactions with regional centres, leading to elevated PFAS levels in specific regions. Moreover, the cumulative effect of multiple weak interactions between determinants may have influenced PFAS concentrations in maternal plasma. Our models correlate more than 200 determinants with maternal PFAS concentrations while accounting for key interactions, providing a novel perspective regarding the effects of physical conditions, behaviours and dietary consumption on PFAS in pregnant women. This study provides valuable insights into the determinants of plasma PFAS concentrations and contributes to risk assessment and management strategies for pregnant populations.