Optimizing Predictive Performance of Bayesian Forecasting for Vancomycin Concentration in Intensive Care Patients.

Area of Science:

• Pharmacokinetics and Pharmacodynamics
• Computational Biology
• Clinical Pharmacy

Background:

• Therapeutic drug monitoring (TDM) is essential for optimizing vancomycin dosing in intensive care units (ICUs).
• Bayesian forecasting, particularly using maximum a posteriori (MAP) estimation, is a key tool for model-based vancomycin dose optimization.
• Evaluating and improving the performance of these forecasting methods is critical for patient outcomes.

Purpose of the Study:

• To evaluate the predictive performance of standard maximum a posteriori (MAP) Bayesian forecasting.
• To compare standard MAP with two novel approaches: adaptive MAP and weighted MAP.
• To assess the impact of historical therapeutic drug monitoring (TDM) data on forecasting accuracy.

Main Methods:

• Utilized a vancomycin TDM dataset from 408 ICU patients.
• Compared standard MAP, adaptive MAP (iterative data handling), and weighted MAP (likelihood weighting) methods.
• Evaluated percentage error (PE) across scenarios using historical TDM data from 1 to 7 days prior.

Main Results:

• Adaptive MAP demonstrated the lowest mean median percentage error (-4.5%) compared to standard MAP (-7.7%) and weighted MAP (-6.7%).
• Adaptive MAP also exhibited the narrowest inter-quartile range of percentage error, indicating greater consistency.
• Prediction errors increased with the inclusion of historical TDM data further back in time, irrespective of the MAP method used.

Conclusions:

• The adaptive MAP method offers superior predictive performance over standard MAP for vancomycin TDM.
• Adaptive MAP is a promising approach for enhancing model-based vancomycin dose optimization in ICUs.
• Limiting historical data inclusion to one day (standard/weighted MAP) or two days (adaptive MAP) is recommended to maintain predictive accuracy.