Leveraging a Previously Published Population Pharmacokinetic Model to Predict Rivaroxaban Exposure in Real-World Patients

Population pharmacokinetic models built from clinical trial populations may not generalize well to real-world patients, raising questions about whether labeled rivaroxaban dosing achieves target exposure outside trial settings. This study used a previously published rivaroxaban population pharmacokinetic model to predict drug exposure in 230 real-world patients via three different methods, comparing results across three software packages (NONMEM, Phoenix NLME, and Monolix). While average labeled dosing was likely to achieve the target AUC range across the population, individual patient AUC values frequently fell outside this reference range, and post hoc clearance estimates varied by up to 50% between individual patients despite good agreement between software packages on average. The study demonstrates both the utility and the individual-level limitations of applying a prior population pharmacokinetic model to predict real-world rivaroxaban exposure.