External Evaluation of Population Pharmacokinetic Models for Factor VIII in Chinese Patients with Hemophilia A

Although numerous population pharmacokinetic (PopPK) models related to factor VIII (FVIII) replacement therapy have been published, the vast majority have not undergone external validation. This study aims to externally validate existing PopPK models of FVIII, derived from both domestic and international sources using independent datasets. PopPK models for FVIII were identified and reconstructed using their respective control files. Data were collected from Chinese patients with hemophilia A who received FVIII therapy. The dataset included essential information such as dosing regimens, blood sampling times, plasma concentrations, and blood group, which were used to generate model input files. The predictive accuracy of each model was assessed through prediction-based diagnostics. A visual predictive check (VPC) was conducted to evaluate the agreement between model predictions and observed concentrations. Model validity was further tested using normalized prediction distribution errors (NPDE). In addition, a Bayesian forecasting approach was employed to explore whether incorporating prior concentration points could enhance the predictive performance of the models. A total of 258 samples were collected from 45 patients to constitute the external validation dataset. Following screening, six models were successfully reconstructed for external validation. Prediction-based diagnostics analysis indicated that Model D showed relatively smaller median prediction error. However, VPC and NPDE analyses both revealed clear deviations from model assumptions, with NPDE indicating systematic model misspecification. Bayesian forecasting analysis further indicated that the model's predictive performance could be improved by incorporating 1-3 prior concentration values.