External Evaluation of Population Pharmacokinetic Models of Cabotegravir, During Its Oral and Intramuscular Administration in HIV-Infected Patients.

Cabotegravir (CAB), combined with rilpivirine, is the first long-acting injectable therapy approved for HIV-1 maintenance treatment. While adherence and patient satisfaction have been improved, pharmacokinetic (PK) variability remains a concern. Two population PK models have been developed: one based on data from phase I-III trials and the other on routine clinical data. However, neither model has undergone thorough external evaluation. The aim of this study was to evaluate the predictive performance of these models using an independent prospective dataset to evaluate their suitability to support model-informed precision dosing (MIPD). External validation was performed using data from the French observational and multicenter (n = 14) ANRS0255 CARLAPOP study (736 HIV-infected patients, representing 2192 concentrations). Models were implemented using MONOLIX software and evaluated using goodness-of-fit, prediction-based, and simulation-based diagnostics. For Han's model, regarding plasma concentrations following intramuscular administration, Median Prediction Error (MDPE) was -1.2% (PRED) and -4.4% (IPRED); Median Absolute Prediction Error (MDAPE) was 36.6% (PRED) and 17.9% (IPRED). For Thoueille's model, MDPE was -24.2% (PRED) and -9.2% (IPRED); MDAPE was 39.0% (PRED) and 15.3% (IPRED). However, < 70% of predictions were within a 20% error margin with both models. Graphical analyses of Thoueille's model showed systemic bias, particularly in women, nonsmokers, and patients with higher body mass index. Therefore, neither model was considered reliable enough for MIPD application in our population. Although Han et al.'s model demonstrated higher predictive performances, further improvements are required before it can be reliably applied for MIPD in daily routine.