2002 Volume 50 Issue 6 Pages 363-370
In developing a vancomycin (VCM) dosage regimen, we assessed the prediction accuracy of VCM serum concentration using population pharmacokinetic parameters in 30 patients and by the Bayesian method in 12 patients (April, 1997-July, 2000). The use of population pharmacokinetic parameters appropriately estimated the serum VCM concentration as judged by the mean prediction error (ME;-0.75μg/mL), mean absolute prediction error (MAE; 3.21μg/mL), and root mean squared prediction error (RMSE; 3.97μg/mL) at troughs (n=30), and ME 2.71μg/mL, MAE 4.59μg/mL, and RMSE 5.24μg/mL at peaks (n=26). Prediction accuracy worsened, however, after about 9 days of VCM treatment due to a decrease in VCM clearance during therapy. Predictive performance by the Bayesian method was ME-3.30μg/mL, MAE 3.90μg/mL, and RMSE 4.93μg/mL at troughs (n=12), and ME 1.67μg/mL, MAE 5.73μg/mL, and RMSE 7.48μg/mL at peaks (n=11). The adequacy of the dosage determined by the Bayesian method was good for patients whose VCM half-life estimated from observed concentrations was less than 1.25 times the mean population half-life. For patients whose estimated half-life was more than 1.45 times of the mean population pharmacokinetic parameter, the prediction error was high. The dosage regimen for patients in this group thus requires careful adjustment.