Prediction of biliary excretion in rats and humans using molecular weight and quantitative structure-pharmacokinetic relationships

The aims were (1) to evaluate the molecular weight (MW) dependence of biliary excretion and (2) to develop quantitative structure-pharmacokinetic relationships (QSPKR) to predict biliary clearance (CL_b) and percentage of administered dose excreted in bile as parent drug (PD_b) in rats and humans. CL_b and PD_b data were collected from the literature for rats and humans. Receiver operating characteristic curve analysis was utilized to determine whether a MW threshold exists for PD _b. Stepwise multiple linear regression (MLR) was used to derive QSPKR models. The predictive performance of the models was evaluated by internal validation using the leave-one-out method and external test groups. A MW threshold of 400 Da was determined for PD_b for anions in rats, while 475 Da was the cutoff for anions in humans. MW thresholds were not present for cations or cations/neutral compounds in either rats or humans. The QSPKR model for human CL_b showed a significant correlation (R ²=0.819) with good prediction performance (Q ²=0.722). The model was further assessed using a test group, yielding a geometric mean fold-error of 2.68. QSPKR models with significant correlation and good predictability were also developed for CL_b in rats and PD_b data for anions or cation/neutral compounds in rats and humans. Both CL_b and PD_b data were further evaluated for subsets of MRP2 or P-glycoprotein substrates, and significant relationships were derived. QSPKR models were successfully developed for biliary excretion of non-congeneric compounds in rats and humans, providing a quantitative prediction of biliary clearance of compounds.