UC Irvine

Previous studies produced the C8 exposure assessment which included PFOA release assessment, integrated fate and transport modeling, and dose reconstruction to predict the serum PFOA concentration of each individual in the C8 Health Project population from 1951 to 2008. The serum concentration predictions were used in various C8 Science Panel epidemiological studies to evaluate whether there is a ‘probable link’ between PFOA exposure and health effects. One such study analyzed the association with preeclampsia among the participants and found a moderate association (Savitz et al., 2012a). For this study, uncertainties in spatiotemporal predictions of PFOA water/air concentrations and in individual-level variables used in the dose-reconstruction and pharmacokinetic models likely resulted in some exposure measurement error, potentially affecting the validity of the epidemiological findings.

The main objective of this dissertation was to analyze the impacts of different sources of input parameter uncertainty/variability in the C8 exposure assessment and study the impacts on the serum PFOA concentration predictions and the epidemiological association with preeclampsia. Monte Carlo (MC) simulation techniques were used to conduct these uncertainty analyses in this exposure assessment-environmental epidemiology model system.

I found that autocorrelated and shared uncertainty in the PFOA water concentration estimates produced a highly variable set of plausible serum PFOA concentrations for study participants; however, it had less impact on the AOR of preeclampsia occurrence. Together, inter-individual variability and epistemic uncertainty in independent individual-level exposure parameters including water ingestion rates, the serum PFOA half-life, and the volume of distribution for PFOA impacted the serum PFOA concentration predictions and their association with preeclampsia moderately, with a 25% bias towards the null in the AOR of preeclampisa occurrence. Geocoding based uncertainty in residential addresses and work history together moderately impacted the rank exposure among the participants and caused a 41% bias away from the null in the AOR of preeclampsia occurrence.

Future studies with complex exposure scenarios and multiple sources of variability/uncertainty might benefit from our approach of separating out the different kinds of uncertainty to better understand their individual impacts on the validity of epidemiological associations