Bayesian forecasting of aminoglycoside dosing requirements in obese patients: influence of subpopulation versus general population pharmacokinetic parameters as the internal estimates

Abstract

An aminoglycoside Bayesian forecaster was evaluated in obese patients. This study assessed the influence of replacing the program-supplied general population parameters (GPP) with obese population parameters (OPP) determined from the study population (n = 26). After entering the required patient information and the first peak and trough levels, patient-specific pharmacokinetic parameters were generated by the Bayesian program based on GPP. These parameters were used to predict peak and trough levels for a second dosage regimen. Next, average OPP determined from a study population were substituted for the GPP, and the peak and trough levels were predicted again. Finally, Bayesian predictions of peak and trough levels were made in a validation population (n = 10), first with GPP, then with OPP. The accuracy of the predictions were evaluated through a prediction error analysis in which mean error indicates bias and mean absolute error and root mean-squared error indicate precision. Means were statistically compared through a Student's t test, with the significance level set at p less than 0.05. For the study and validation populations, peak level predictions based on the OPP had less bias and greater precision than those predicted with GPP. Peaks predicted with GPP were statistically different from the observed peaks as well as the peaks predicted using the OPP. There was no statistical difference between the observed peaks and the predicted peaks using the OPP. The trough level predictions using GPP in the study population had less bias than those predicted using OPP; however, the OPP predictions had less bias in the validation population.(ABSTRACT TRUNCATED AT 250 WORDS)