Bayesian forecasting of plasma vancomycin concentration using time-dependent pharmacokinetics

A new prediction method of vancomycin (VCM) pharmacokinetics has been developed using the modified Bayesian forecasting method involved in time-dependent pharmacokinetics in clearance. We investigated to evaluate the usefulness of this new prediction method compared with that of the ordinary Bayesian forecasting method. Serum samples, obtained from 4 patients at least 3 different days during the period for the VCM treatment were assayed by fluorescence polarization immunoassay. VCM pharmacokinetic parameters and predicted serum VCM concentrations were calculated using this new method and the ordinary one according to the one-compartment model. The precision of the predicted serum VC-M concentrations by these two methods at the third experimental day were evaluated with the mean prediction error (ME), mean absolute prediction error (MAE) and root mean squared error (RMSE). The most precise and least-bias prediction of serum VCM concentrations were observed using this new prediction method (ME: -0.36+/-1.40, MAE: 1.13+/-0.82 and RMSE: 1.37). The time-dependent decrease of VCM clearance was observed in all patients. Therefore, the fitting of the actual serum VCM concentrations obtained using the ordinary method produced less precise results than that using this new method. These results suggest the usefulness of this new prediction method considering time-dependent changes in VCM clearance.