Bayesian analysis of physiologically based pharmacokinetic model and exposure reconstruction for perchloroethylene
MetadataShow full item record
Perchloroethylene (PCE) is a pollutant distributed widely in the environment and the primary chemical used in dry cleaning. Liver cancer induced by PCE has been observed in mice, and central nervous system (CNS) effects have been observed in dry-cleaning workers. The objectives of this study were to 1) derive population distributions of physiologically based pharmacokinetic (PBPK) model parameters, which will subsequently be used in PCE exposure reconstruction, 2) predict the trend of percentages of PCE metabolized in the liver under thdifferent exposure conditions and 95 upper percentile for fraction PCE metabolized at a concentration of 1ppm with posteriors, 3) determine relationship between brain concentration of PCE and effect on visual evoked potentials, 4) perform sensitivity analysis of PBPK model parameters of PBPK model for PCE to model outputs to identify sensitive parameters to outputs and to ascertain effects of parameter transformation on sensitivity analysis results, and 5) reconstruct occupational exposure profiles to PCE with PBPK model based on sparse thbiomonitoring data. The 95 percentile for fraction PCE metabolized at a concentration of 1ppm was estimated to be 1.89%. Ventilation perfusion ratio (VPR) and blood/air partition coefficients (PB) in either original or transformed form were shown to be sensitive to variability in blood and alveolar air concentrations of PCE; variability in parameters clearance (ClC) in either form is most sensitive to model-predicted blood concentrations of trichloroacetic acid (TCA) or urinary excretion of TCA. Atmospheric PCE levels in the working environment and background levels of PCE had high correlations with the biomarkers, blood and alveolar PCE concentrations. Lastly, posterior distributions of PBPK model parameters and exposure parameters were used to perform Monte Carlo (MC) simulations. Bootstrap sampling of MC sample with respect to likelihood of model outputs was used to construct distributions of exposure profiles.