Amikacin Bayesian forecasting in critically ill patients with sepsis and cirrhosis.

This study was designed to determine the population pharmacokinetic parameters of amikacin in two subpopulations of intensive care unit patients with sepsis and cirrhosis and sepsis without cirrhosis. The authors evaluated the usefulness of the obtained parameters to forecast the serum amikacin concentrations in a validation group of patients with sepsis and cirrhosis when used as a priori distribution in a Bayesian forecaster. The population parameters were estimated by a nonparametric expectation maximization algorithm (NPEM), and the accuracy of the predictions were evaluated through a prediction error analysis. Significant differences (p < 0.05) were found in Vd (0.668 versus 0.470 l/kg) and K (0.0701 versus 0.161 h-1) between subpopulations of patients with and without cirrhosis. The model derived for patients with cirrhosis used as a priori distribution, with and without feedback, was superior to the model derived for patients with sepsis in forecasting amikacin serum concentrations. The results show the relevance of using the specific model for the subgroup with cirrhosis as a priori distribution in a Bayesian forecaster to obtain a nonbiased prediction with an acceptable precision.