Abstract Aim To guide the efficient implementation of Model‐Informed Precision Dosing (MIPD) by comparing the predictive performance of published tacrolimus population pharmacokinetic (popPK) models with an in‐house model in Uruguayan kidney transplant recipients, using comprehensive external evaluation methodology. Methods Twenty‐five adult tacrolimus popPK models were identified through systematic review and externally evaluated using retrospective data from 39 patients (396 steady‐state concentrations). In parallel, an in‐house model was developed with prospective and retrospective data from 49 patients (727 concentrations). Predictive performance was assessed across different data integration levels: a priori predictions and sequential Bayesian forecasting with progressive incorporation of individual observations. Results All published models and the in‐house model showed poor a priori predictive performance, with even the best performing models yielding MAPE values near 50% and fewer than 30% of predictions within 30% of observed concentrations. Bayesian forecasting progressively improved bias, accuracy and dispersion of prediction errors. Initial incorporation of three trough concentrations produced limited benefit, but substantial improvements were observed upon integration of a pharmacokinetic curve composed of at least 3 post‐dose concentrations, at which point several published models and the in‐house model achieved high precision, with mean bias below 10% and MAPE below 30%. Two published models exhibited comparable performance to the in‐house model, which achieved marginally better results as Bayesian forecasting incorporated successive individual observations, without providing a meaningful overall advantage. Conclusion Published tacrolimus popPK models can achieve clinically reliable predictions when combined with Bayesian updating, performing comparably to in‐house developed models. These findings support a resource‐efficient strategy for implementing tacrolimus MIPD without the need for local model development, provided that a pharmacokinetic curve is obtained early post‐transplantation.
Umpiérrez et al. (Wed,) studied this question.