The optimal anesthetic management of patients undergoing thoracotomy for pulmonary resection has not been definitely determined. We evaluated whether general i.v. anesthesia (propofol-fentanyl) provides superior PaO2 during one-lung ventilation (OLV) compared with thoracic epidural anesthesia (TEA) with supplemental local and general anesthetics. We studied 60 patients who had prolonged periods of OLV for elective thoracic surgery for lung cancer and who were prospectively randomized into two groups. In 30 patients (GA group), fentanyl/propofol/rocuronium anesthesia was used. Another 30 patients (TEA group) were anesthetized with propofol/rocuronium/epidural thoracic bupivacaine 0.5%. A double-lumen endotracheal tube was inserted, and mechanical ventilation with 100% oxygen was used during the entire study. Arterial and venous blood gases were recorded before surgery in a lateral position with two-lung ventilation, 15 and 30 min after OLV (OLV + 15 and OLV + 30, respectively) in all patients. We measured PaO2, venous central oxygen tension, arterial and central venous oxygen saturation, venous admixture percentage (Qs/Qt%), and arterial and central venous oxygen content. The mean values for PaO2 during OLV in the GA group after 15 min (175 mm Hg) and 30 min (182 mm Hg) were significantly (P < 0.05) higher compared with the TEA group (120 and 118 mm Hg, respectively). Furthermore, Qs/Qt% was significantly (P < 0.05) increased in the TEA group during OLV. There were no other significant differences. We conclude that using the TEA regimen is associated with a lower PaO2 and a larger intrapulmonary shunt during OLV than with total i.v. anesthesia alone. CONCLUSIONS Sixty patients undergoing elective lung surgery during a prolonged period of intraoperative one-lung ventilation were studied and randomized to receive general i.v. anesthesia or general i.v. anesthesia combined with thoracic epidural anesthesia. The arterial oxygenation in the first group was better than that in the second group during one-lung ventilation.