How to Select a PK/PD Model

Abstract

For drug administration in anesthesia practice, the goal is to achieve a desired and stable clinical effect, which requires accurate control of blood and brain drug concentrations. For intravenous drugs, commercial medical devices are unavailable to measure the drug concentrations. Instead, pharmacokinetic and pharmacodynamic models are used to estimate plasma and effect-site concentrations of intravenous drugs. In the case that two or more models are available, one model has to be selected. The model applicability should be considered with their internal and external validity. For the evaluation of the pharmacokinetic model performance, metrics derived from percentage performance error are used such as “median performance error” indicating the bias, “median absolute performance error” indicating the inaccuracy, “divergence” showing the shift of performance error or absolute performance error against time, and “wobble” showing the intraindividual variability of the prediction error. Visual inspection of the goodness of fit plots is also applied for the model evaluation. In the prediction of drug concentration in a patient, the background of the model and applied patient such as patient characteristics influences the predictive performance of the model. For the model application for a patient, a model will be selected from available models in the commercial devices or from all published models by the recommendation of someone or the user of the model. For the model selection, variable information such as the results of the model evaluation, the background of the model, the background of the patient and dosing plan, and the data set are used for the external validation.