Most recent cyclosporine (CsA) pharmacokinetic (PK) studies have focused on noncompartmental analysis. Because CsA undergoes significant first-pass elimination after oral dosing, the most appropriate compartment model may need to take this process into account for the construction of a valid population PK model for Sandimmune (SAN) and Neoral (NEO) formulations. Twenty patients with cardiac transplants were stabilized for at least 4 weeks on a certain dose of SAN, then changed to the same daily dose of NEO. Blood samples were obtained at times 0, 1, 2, 3, 4, 6 and 12 hours after dosing at steady state. Pharmacokinetic modeling was performed using ADAPT II. Quality of fit was assessed by visual graph inspections, R2 values, and Akaike criterion test. Eight pharmacokinetic models were constructed and evaluated. These included one- and two-compartment with and without a first-pass effect and a time-lag. Neoral and SAN data were consistently best fitted using a two-compartment or the two-compartment first-pass model. However, a time-lag process was found to be necessary for SAN. The use of a two-compartment first-pass with (SAN) or without (NEO) a time-lag process appears to fit CsA concentrations at least as well as a two-compartment model. This first-pass model may be very useful for population pharmacokinetics and Bayesian control analysis.