A reassessment of the interest of therapeutic drug monitoring (TDM) is needed for a better definition of its applications. It rests on the hypothesis of the existence of an interindividual variability of the dose-effect relationship, this variability being influenced by pharmacokinetic variability. Different types of endpoints were used in validation studies (pharmacokinetic, indirect, economic) and few studies have used clinical endpoints. The question should be: does TDM allow for a better dose adjustment than one based only on criteria other than drug concentration? Retrospective studies have a 'diagnostic test' approach and only prospective clinical trials will provide a real validation. Factors which limit the setting up of such studies are: the number of subjects to include, the often imprecise measurement of drug exposure and poor knowledge of the pharmacokinetic-pharmacodynamic relationship and its variability.