A randomized two-arm prospective study was planned to assess the role of therapeutic drug monitoring (TDM) coupled with a Bayesian approach in tailoring vancomycin dosages in unstable critically ill patients. Group A (n=16) had their regimen adjusted day-by-day according to TDM and Bayesian forecasting (D(a)); group B (n=16) had their regimen adjusted day-by-day according to Moellering's nomogram (D(M)). Blood samples were collected every 1-2 days to assess the trough and peak plasma concentrations. In group A, the tailored D(a) required for optimizing vancomycin exposure were considerably higher than the D(M) in 7/16 cases, and lower than the D(M) in 1/16 cases. In group B, standard D(M) caused under-treatment in 3/16 cases and over-treatment in 4/16 cases. Most of these patients concomitantly had some conditions that might have altered vancomycin disposition. The TDM-guided Bayesian-based approach should be considered an invaluable tool for clinicians to handle appropriately on real-time vancomycin therapy in critically ill patients.