Busulfan pharmacokinetic parameters are useful in predicting the outcome of allogeneic bone marrow transplantation (BMT). Standard pharmacokinetic measurements require multiple blood samples. Various limited sampling models (LSM) have been proposed for reducing the sample number required for these measurements, essentially for patients with malignant disorders undergoing BMT. This study was undertaken to evaluate the existing LSM for busulfan pharmacokinetics to find out the most suitable method for patients with thalassaemia major undergoing BMT. Busulfan levels in plasma samples were analysed by HPLC. The AUC calculated by non-compartmental analysis using the program 'TOPFIT' was compared with previously published LSMs. Our seven sample pharmacokinetic data for AUC calculation was compared with the published LSMs. The three sample models suggested by Chattergoon et al and Schuler et al showed significant agreement with AUC TOPFIT (R(2) = 0.98 and 0.94, respectively) in our clinical context. Other models resulted in significant over or under representation of observed values (Vassal's model R(2) = 0.61; Chattergoon's two sample model R(2) = 0.84; four sample model R(2) = 0.83; Schuler's two sample model R(2) = 0.79). By these data the three sample LSM proposed by Chattergoon et al and Schuler et al are suitable for calculation of the AUC in patients with thalassaemia major undergoing BMT conditioned with oral busulfan.