Pirarubicin (4'-O-tetrahydropyranyldoxorubicin, THP-Adriamycin) is a new anthracycline antibiotic that has recently been developed because its reduced cardiac toxicity is associated with an antitumour efficacy similar to that of doxorubicin. Pirarubicin is characterised by strong haematological toxicity, which has been shown to be correlated with pharmacokinetic parameters, especially the area under the time-concentration curve. To obtain routine pharmacokinetic evaluations of pirarubicin for dose monitoring we developed a limited sampling strategy relying on three blood samples taken at the end of the infusion and at 12 and 24 h post-infusion. The characteristics of interindividual variability were assessed on the first courses of treatment performed in 18 patients; the model was then validated on 10 independent first courses of treatment performed in 10 other patients. The main pharmacokinetic parameters (half-lives, total volume of distribution, total plasma clearance) were estimated in the test group by maximum-likelihood estimation using all samples and by Bayesian estimation using three samples. The concordance between the two estimates was correct (the bias and precision for clearance were 2.3% and 12.1%, respectively), which shows that this limited sampling strategy can be used in routine drug monitoring.