OBJECTIVE To study the predictive performance of a target-controlled infusion (TCI) system of propofol in patients undergoing coronary bypass graft (CABG) surgery, using a referenced pharmacokinetic set derived from healthy patients. Also, to determine the propofol concentrations required for clinically acceptable induction and maintenance of anesthesia when combined with midazolam as premedication and a continuous alfentanil infusion and to study the hemodynamic stability of this technique. METHODS Prospective noncomparative study analysis. METHODS Operating room at a university hospital. METHODS Twenty-on patients with good left ventricular function undergoing coronary artery surgery. METHODS Patients were anesthetized using a continuous infusion of alfentanil (mean infusion rate: 1 microgram/kg/min) and propofol administered by TCI. RESULTS The predictive performance of the TCI system (212 arterial samples) was measured at specified time points before, during, and after bypass. The TCI system underestimated the measured blood propofol concentrations with a bias of +21.2% and +9.6% during the prebypass and the bypass periods, respectively. The predictive inaccuracy, expressed by the median absolute prediction error, was 23% and 18.5%, respectively. Mean target propofol concentrations required to induce and maintain anesthesia before bypass were 0.92 microgram/mL and 3.64 micrograms/mL, respectively. In the period during and after bypass, the mean target concentrations required to maintain anesthesia was 2.22 micrograms/mL. The administration of propofol by TCI was still associated with some short episodes of hemodynamic instability that were easily controlled by adjusting the target concentration in the majority of the patients. Therefore, the overall quality and ease of control of anesthesia were considered as being good or adequate. CONCLUSIONS In this group of patients undergoing CABG surgery, the TCI system used underestimated the measured propofol concentrations. However, the predictive performance of the selected mean pharmacokinetic parameters derived from healthy patients was acceptable during the whole surgical procedure.