The aim of this study was to find a procedure allowing estimation of individual pharmacokinetic parameters for adriamycin with minimal cost and disturbance for the patient. Twenty-five patients with breast cancer were treated by short infusion of adriamycin at a dose of 12 mg m-2 week-1 (41 courses). Population characteristics were determined on 15 randomly chosen courses (10 patients, group I) in order to define two optimal sampling times (26 min and 24 h) and to perform Bayesian estimation on the remaining 26 courses (17 patients, group II). For patients of group II, Bayesian estimation (BE) associated with a reduced sub-optimal sampling protocol (20 min and 24 h) was compared with maximum likelihood estimation (MLE), the classical procedure. Regression analysis of clearance values obtained after BE versus MLE indicated a high correlation coefficient (r = 0.969) with the slope (a = 0.991 +/- 0.085) and the intercept (b = 2.271 +/- 4.810) close to 1 and 0 respectively. This original method is thus valid to measure accurately adriamycin clearance; it improves patient comfort and can be used routinely.