Community-facing toxicokinetic models to estimate PFAS serum levels based on life history and drinking water exposures

Abstract

BackgroundPer- and polyfluoroalkyl substances (PFAS) are drinking water contaminants. Tools to assess the potential body burden associated with drinking PFAS-contaminated water may be helpful for public health assessment of exposed communities. MethodsWe implemented a suite of one-compartment toxicokinetic models using extensively calibrated toxicokinetic parameters (half-life and volume of distribution). We implemented the models both in the R programming language for research purposes, and as a web estimator for the general public (built in typescript.js). These models simulate exposure to PFAS water concentrations for individuals with varying characteristics such as age, sex, weight, and breastfeeding history. The models account for variability and uncertainty in parameter inputs to produce Monte Carlo-based estimates of serum concentration. For children, the models additionally account for gestational exposure, lactational exposure, and potential exposure through formula feeding. For adults who have borne children, the models account for clearance through birth and breastfeeding. We ran simulations of individuals with known PFAS water and serum concentrations to evaluate the model. We then compared the predicted serum PFAS concentrations to measured data. ResultsThe models accurately estimate individual-level serum levels for each PFAS for most adults within ½ order of magnitude. We found that the models somewhat overestimated serum concentrations for children in the tested locations, and that these overestimates are generally within an order of magnitude. DiscussionThis paper presents scientifically robust models that allow users to estimate serum PFAS concentrations based on known PFAS water concentrations and physiologic information. However, accuracy in historical water concentration inputs, exposure from non-drinking water sources, and life-history characteristics of individuals present a complex problem for individual estimation. Additional refinements to the model suite to improve the prediction of individual results may consist of including duration of exposure and additional life-history characteristics.