The utility of a Bayesian parameter estimation program in the interpretation of cyclosporin concentrations was investigated in a group of 32 patients following renal transplantation. The program was evaluated by comparing concentrations predicted from individual estimates of pharmacokinetic parameters with measured concentrations. A one-compartment model incorporating an exponential, time-related change in clearance was used for data collected in the first 4 weeks after transplantation, and predictions of concentrations measured during weeks 5–14 were made using three schemes: a changing clearance model using all data from week 1 onward; a changing clearance model using data from week 4 onward; and a nonchanging clearance model using data from week 4 onward. Results demonstrated that predictions made by the Bayesian program were unreliable during the first 4 weeks of therapy, but that there was a progressive improvement as time after transplantation increased. The changing clearance model was superior to the constant clearance model and its performance was not compromised by including data from the first 4 weeks of therapy. Although the Bayesian approach may help with the interpretation of blood cyclosporin concentrations during maintenance therapy, the large variability in the pharmacokinetics of orally administered cyclosporin limits the usefulness of the approach in the early weeks following transplantation.