You are here:
Replication of PBTK Models for Bisphenol A with High-Throughput TK Models
Citation:
Pearce, R., R. Sayre, Woodrow Setzer, Chris Grulke, R. Thomas, AND J. Wambaugh. Replication of PBTK Models for Bisphenol A with High-Throughput TK Models. Presented at Society of Toxicology Annual Meeting, San Antonio, Texas, March 11 - 15, 2018. https://doi.org/10.23645/epacomptox.7017293
Impact/Purpose:
This research evaluates the difference between using generic, high-throughput PBTK models and tailored, chemical-specific PBTK model. Using Bisphenol A as a case study, we conclude that the errors are only slightly larger when using high throughput PBTK.
Description:
Physiologically-based toxicokinetic (PBTK) models that are tailored to the chemicals of interest are generally considered to be more accurate than simpler toxicokinetic (TK) models parameterized with minimal in vitro and in silico data. However, within the context of high-throughput risk evaluation, a slightly lower level of accuracy may perform equally well in estimating relative risk. In this study, we compared fitted human PBTK models for bisphenol A (BPA) to the R package “httk,” which includes a 1-compartment TK model and a generic PBTK model. We also tested a 2-compartment model, consisting of clearance and rest-of-body compartments with physiologically based volumes, connected by the hepatic blood flow. The models from httk assumed a first-order, non-restrictive clearance and were parameterized with logP, pKa, and in vitro intrinsic clearance, protein binding and blood to plasma partitioning. The fitted PBTK model from Yang et al. (2015) and the httk models yielded similar time-plasma concentration curves in the prediction of human in vivo data from Thayer et al. (2015). The 2-compartment and generic PBTK models made very similar predictions, with root-mean-square error (RMSE) (the square root of the average squared difference in measured and predicted concentrations) of 2.9 and 5.1 nM and average-fold error (AFE) (the average quotient of the measured and predicted concentrations when the dividend is larger than the divisor) of 2.6 and 3.3, respectively. The fitted model (Yang et al., 2015) performed the best (RMSE of 0.33 nM and AFE of 1.4), and the 1-compartment model performed the worst (RMSE of 43 nM and AFE of 23). In this case study, the 2-compartment model performed best among the simplified models in predicting BPA TK, serving as a proof of concept for further investigation of TK models based on high throughput data. This abstract does not necessarily reflect U.S. EPA policy. Thayer, K. A., et al. "Pharmacokinetics of bisphenol A in humans following a single oral administration." Environ. Int. 83 (2015): 107-115. Yang, X., et al. "Development of a physiologically based pharmacokinetic model for assessment of human exposure to bisphenol A." TAAP 289.3 (2015): 442-456.
URLs/Downloads:
DOI: Replication of PBTK Models for Bisphenol A with High-Throughput TK Models
PEARCE_SOT2018.PDF (PDF, NA pp, 936.265 KB, about PDF)