Observed Trough-Based Vancomycin Dosing Decisions Versus Model-Predicted Bayesian AUC-Guided Recommendations

Background: Vancomycin monitoring has traditionally relied on trough concentrations; however, recent guidelines recommend AUC-guided monitoring to optimize exposure while minimizing nephrotoxicity risk. This study compared trough-based clinical dosing decisions with predicted Bayesian AUC-guided recommendations. Methods: A retrospective study was conducted in adult patients receiving intravenous vancomycin between January and December 2025. Observed trough-based dose adjustments were compared with model-predicted Bayesian AUC-guided recommendations derived from steady-state trough concentrations. Agreement was evaluated using Cohen's kappa and Bowker's test. Results: Fifty-two patients were included. Agreement between trough-based and AUC-guided recommendations occurred in 26 cases (50%) with fair concordance (κ = 0.24). Among patients whose dose was maintained based on trough monitoring (n = 29), model-predicted AUC-guided recommendations indicated dose reduction in 48.3%. Among patients with acute kidney injury (n = 12), model-predicted AUC-guided recommendations suggested dose reduction in 58.3%, compared with 16.7% using trough-based decisions. Conclusions: Trough-based dosing showed limited agreement with model-predicted Bayesian AUC-guided recommendations, with frequent differences in the classification of vancomycin exposure and corresponding dosing recommendations.