The pharmacokinetics of cyclosporine A (CyA) was studied in 21 uremic patients. The plasma concentrations after an oral dose and a subsequent short-term infusion were analyzed simultaneously by nonlinear regression. Bi- and triexponential disposition models with either zero- or first-order absorption were fitted to the data. A triexponential disposition model with zero-order absorption was generally found to best describe the concentration-time profile. The bioavailability and clearance were estimated to be 0.24 +/- 0.10 and 21 +/- 8 L/hr, respectively. These values differed only marginally from those predicted by the other models. Similar bioavailability estimates were also obtained from a three-compartment model where elimination was assumed saturable, from a deconvolution procedure, and from analyses based on blood concentrations. Markedly higher bioavailabilities (0.34 +/- 0.13) were obtained when a model-independent AUC correction procedure, commonly used to calculate CyA bioavailability, was used. The difference could not be explained by poor description of data in the model-dependent analyses, but rather by overestimation in the model-independent analyses mainly due to errors in the extrapolations used. Thus, by the simultaneous fitting procedure, which is a new approach for estimating CyA bioavailability, drawbacks of the AUC correction procedure could be avoided. Further, future studies of CyA bioavailability could be designed with a markedly shorter and more convenient length of time if analyzed by the proposed method.