Virtual Twin Approach Using Physiologically Based Pharmacokinetic Modeling to Support Precision Dosing of Valproic Acid in Geriatric Patients.

Personalized dosing is particularly important for drugs with narrow therapeutic indices in geriatric patients, who exhibit substantial physiological variability and limited pharmacokinetic (PK) evidence to guide individualized dose selection. Valproic acid (VPA) is an effective treatment option for bipolar disorder in older adults, yet dosing largely relies on therapeutic drug monitoring (TDM), which guides adjustment only after steady state is achieved. This study evaluated a physiologically based pharmacokinetic (PBPK)-guided virtual twin (VT) framework to prospectively predict individual VPA exposure in a real-world geriatric population. A PBPK model for extended-release VPA was developed using the Simcyp Simulator (version 25). Clinical data (n = 74) were collected from elderly patients receiving VPA at the Samsung Medical Center. PBPK model performance was validated against published PK data from healthy adults and independent TDM data from the geriatric patients. The model adequately reproduced observed PK of VPA, with predicted PK parameters ranging from 0.73- to 1.37-fold of observed values. Virtual twins were subsequently generated for each patient by incorporating patient-specific demographic characteristics and routinely available physiological parameters. Individual PK profiles were simulated and compared with measured TDM concentrations. Incorporation of demographic and physiological covariates achieved an absolute average fold error of 1.13 (90% CI: 1.10-1.16), with nearly 90% of predictions falling within the 0.8- to 1.25-fold range of the observed concentrations. These findings suggest that PBPK-guided virtual twins can predict individual VPA exposure using routinely available clinical data and may enable prospective dose optimization, complementing conventional TDM to advance precision dosing in geriatric pharmacotherapy.