2026 Volume 49 Issue 4 Pages 683-690
Teicoplanin is an important antibiotic for methicillin-resistant Staphylococcus aureus infections. To enhance its clinical effectiveness while preventing adverse effects, therapeutic drug monitoring (TDM) of teicoplanin trough concentration is recommended. Given the importance of the early attainment of therapeutic concentrations for treatment success, initial dosing regimens, including loading and maintenance doses, are deliberately designed based on patient information. However, initial dose planning for teicoplanin strongly relies on clinician expertise. This study aimed to use a machine learning (ML) approach to integrate clinicians’ knowledge into a predictive model for initial teicoplanin dose planning. First, we confirmed that dose planning by pharmacists specialized in TDM (TDM pharmacists) significantly improved early therapeutic target attainment for patients who were not admitted to intensive or high care units. Subsequently, we used a dataset of initial teicoplanin dose plans created by TDM pharmacists to train the model that emulates their dosing decision-making process. Although the prediction accuracies of the ML model were modest (45.8 and 66.7% for the loading and maintenance doses, respectively), the model successfully learned the basic policy of dose planning, suggesting that ML approaches have potential utility in supporting appropriate initial teicoplanin treatment.
