Objective: This study evaluated if previously published limited sampling models (LSMs) accurately predict midazolam area under the concentration time curve (AUC) during cytochrome P450 (CYP) 3A baseline, inhibition and induction/activation.
Materials and methods: Plasma midazolam concentrations (n = 108) were obtained where intravenous midazolam was co-administered alone or concomitantly with ketoconazole, itraconazole, aprepitant, rifampin, or pleconaril. Observed AUC was calculated using noncompartmental analysis. Predicted AUC was calculated from the LSMs. Bias and precision were determined by percent mean prediction error (%MPE), percent mean absolute error (%MAE), and percent root mean squared error (%RMSE).
Results: Contrasting results were observed for LSMs in predicting CYP3A baseline activity, with the majority of studies resulting in unacceptable bias and precision. During CYP3A inhibition, unacceptable bias and precision were observed from single- and 2-time point LSMs. %MAE and %RMSE values exceeded acceptable limits during CYP3A induction with rifampin. Contrasting results were observed with pleconaril.
Conclusion: The contrasting results during CYP3A baseline and induction/activation, as well as the unacceptable bias and precision during CYP3A inhibition, limits the widespread use of the previously published LSMs.