Predictive Performance of Bayesian Dosing Software for Vancomycin in Intensive Care Unit Patients

Area of Science:

• Pharmacokinetics and Pharmacodynamics
• Critical Care Medicine
• Computational Pharmacology

Background:

• Updated guidelines recommend Bayesian-derived area under the curve estimation for vancomycin dosing.
• Bayesian dosing software requires thorough evaluation in intensive care unit (ICU) patients.
• This study assesses the predictive performance of commonly used Bayesian software for vancomycin concentrations in ICU settings.

Purpose of the Study:

• To evaluate the performance of three Bayesian software programs in predicting vancomycin concentrations in ICU patients.
• To determine the bias and precision of these software programs for vancomycin dosing.
• To inform the potential use of Bayesian software for personalized vancomycin dosing in critical care.

Main Methods:

• Retrospective analysis of vancomycin concentrations in adult ICU patients.
• Prediction of serum concentrations using a priori and a posteriori Bayesian forecasting.
• Evaluation of predictive performance using relative bias (rBias) and relative root mean squared error.

Main Results:

• Three software programs (SmartDose, Pharmado, PrecisePK) were evaluated using data from 139 patients.
• All programs demonstrated clinically acceptable bias, with one exception in a priori estimation.
• Precision was generally low but significantly improved with a posteriori estimation compared to a priori.

Conclusions:

• Bayesian dosing software holds potential for vancomycin dose optimization in ICU patients.
• Specific Bayesian programs may be better suited for patients with unique physiological or pathological characteristics.
• Further validation is warranted for widespread clinical adoption in critical care settings.