Applications of Bayesian algorithm in TDM practice

Abstract

The Bayesian forecasting techniqe has recently been introduced as a new area of clinical pharmacokinetics, and is now becoming more powerful and useful tool to assist dosage adjustment for individual patients in therapeutic drug monitoring practice (TDM). This newer, and a sophisticated method is based on the principles of Bayes' theory and Maximum Likelihood Estimation, and utilizes prior information on the distribution of population pharmacokinetic parameters, means and variances, as well as drug serum concentrations. With the aid of a small computers, Bayesian weighted least square fitting method can work with at least one observation, whereas standard weighted least square fitting methods need more than four or five observations for accuracy and precision. The Bayesian approach in clinical pharmacokinetics involves the prediction of pharmacokinetic parameters, dosage regimens, and serum drug concentrations. The application of Bayesian procedure has been studied for several drugs such as digoxin, theophylline, aminoglycosides, phenytoin, and lidocaine. In this paper, I present a review of the theoretical background of Bayesian procedure, the applications in therapeutic drug monitoring practice, and the remaining problems to improve the predictive performance.