Limited Predictive Performance of Existing Amisulpride PopPK Models: External Validation and Proposal of Model-Based Remedial Regimens for Non-Adherence

This work sought to assess how well previously published population pharmacokinetic (PopPK) models of amisulpride perform in individuals with schizophrenia using an independent dataset, and to design dosing adjustments suitable for patients who miss or delay their amisulpride doses. A comprehensive literature search of PubMed, Embase, and Web of Science was used to collect PopPK models for comparison. A total of 390 serum samples from 361 hospitalized Chinese adults diagnosed with schizophrenia were used for external evaluation. Predictive accuracy was examined using both prediction-based and simulation-based metrics. When the published models showed limited predictive quality, a revised PopPK model tailored to our cohort was built. Monte Carlo simulations were then applied to explore various non-adherence patterns and assess appropriate corrective dosing strategies.
Among the five models reviewed, four relied on trough data from schizophrenia cohorts, and one incorporated single-dose data from healthy elderly subjects together with trough levels from older individuals with Alzheimer’s disease. Population and individual prediction errors spanned −92.89% to 27.02% and −24.82% to 4.04%, respectively. Simulation-based diagnostics, including NPDE, revealed systematic bias in all evaluated models. Consequently, a refined one-compartment model integrating estimated creatinine clearance (eCLcr) as a covariate influencing apparent clearance (CL/F) was derived. For delayed dosing, when the delay is ≤12 hours, half of the missed dose should be taken immediately before returning to the original schedule; if the delay approaches 24 hours, the regular dosage regimen should be resumed without compensation. Existing models do not show sufficient predictive robustness for use across different clinical centers. Prospective research will be required to substantiate the utility of our revised model before clinical implementation. Model-based simulations offered a structured method to design rational management approaches for missed or postponed doses.