Evaluation of transplacental clearance prediction methods in maternal-fetal PBPK modeling: Implications for...

Accurate prediction of fetal drug exposure is essential for optimizing pharmacotherapy but remains challenging due to limited clinical data. Maternal-fetal physiologically based pharmacokinetic (m-f PBPK) modeling has emerged as a key tool for these predictions. Transplacental clearance (CL_PD) is a critical parameter in m-f PBPK modeling, but various estimation methods exist, and their predictive performance has not been systematically evaluated.

Model drugs were selected based on available data for CL_PD estimation and m-f PBPK modeling. PBPK models for non-pregnant adults were developed in Simcyp and extended to pregnancy. Four CL_PD estimation methods were evaluated: ex vivo human placental perfusion (exHPP), quantitative structure-property relationship (QSPR), Caco-2 permeability-based in vitro-in vivo correlation (Caco-2 IVIVC), and midazolam-based calibration with Caco-2 scaling (MDZ-Caco2 Cal). Predictive performance was assessed using predictive success rate (PSR), defined as the percentage of drugs with >50% of observed umbilical cord blood concentrations within simulated 90% prediction intervals.

Nine drugs with diverse properties were tested. AUC and C_max predictions for non-pregnant adults were within 0.5-2-fold of observed values. Maternal plasma concentrations were well predicted, with most observations falling within the 90% prediction intervals. However, fetal exposure predictions were limited (PSR: 14.3-33.3%). The exHPP approach had the highest performance (33.3%), followed by MDZ-Caco2 Cal (28.6%) and QSPR (22.2%), while Caco-2 IVIVC performed the worst (14.3%).

CL_PD prediction methods show variable performance and limited accuracy for fetal exposure, posing a significant uncertainty in m-f PBPK modeling. Further refinement is needed for more reliable fetal exposure predictions.