This study utilized various machine learning algorithms to predict pharmacokinetic parameters (AUClast and Cmax) based on the results of a Phase 1 clinical trial. This trial assessed the pharmacokinetic profiles of co-administered single formulations ...
This study utilized various machine learning algorithms to predict pharmacokinetic parameters (AUClast and Cmax) based on the results of a Phase 1 clinical trial. This trial assessed the pharmacokinetic profiles of co-administered single formulations versus combined formulations in healthy adult volunteers. By evaluating the predictive performance of AUClast and Cmax using machine learning models, we established that reliable predictions could be achieved with a reduced number of blood sampling points (0 to 6 hours, 11 times) compared to the full sampling schedule (0 to 48 hours, 15 times). The analysis demonstrated that the LASSO algorithm performed optimally for AUClast, while the XGBoost algorithm excelled in predicting Cmax. Both external and internal validations confirmed that the pharmacokinetic parameters predicted using the minimal sampling schedule showed no statistically significant difference compared to those predicted using the full sampling schedule. The predicted AUClast and Cmax exhibited statistical equivalence, affirming the feasibility of the reduced sampling protocol. In conclusion, the results of this study suggest that selecting a minimal blood sampling schedule can reliably estimate pharmacokinetic parameters, thereby reducing the burden of repeated blood draws on clinical trial participants.