Bayesian and least-squares methods for vancomycin dosing.

The authors assessed the performance of a Bayesian and a least squares method for predicting individual pharmacokinetic parameters for vancomycin. For clearance, the best performance of both methods was an absolute error of approximately 5%. This level of accuracy required 4 serum vancomycin concentrations with the least squares method but could be achieved with a peak and trough concentration with the Bayesian method. For volume of distribution, the best performance occurred with 3 or 4 levels with both methods and amounted to an error of about 15%. In conclusion, both methods of estimating vancomycin pharmacokinetics perform comparably, but the Bayesian method appears to require fewer data.