Administration of large doses of fentanyl is a popular method to provide postoperative analgesia after thoracotomy. It is however unclear whether epidural lumbar (L) or epidural thoracic (T) administration of fentanyl confers any major advantage over intravenous (iv) infusion. Using a randomized prospective study design, we compared the potential benefits of L, T, and iv fentanyl administration after thoracotomy in 50 patients. Epidural catheters were not injected during surgery. Postoperatively a fentanyl infusion (5 micrograms/ml) was started at 1 microgram.kg-1.h-1 after a bolus of 1 microgram/kg and adjusted to maintain a score < or = 30/100 at rest using a visual analog scale (VAS) for pain. Data were prospectively collected before surgery, at fixed intervals during the 48 h of fentanyl infusions, and the day of discharge. There was no difference between the groups in overall quality of analgesia at rest and after coughing, quantity of fentanyl delivered (L = 1.15 +/- 0.38, T = 1.22 +/- 0.23, iv = 1.27 +/- 0.3 micrograms.kg-1.h-1), incidence of pruritus needing treatment (L = 2, T = 1, iv = 0 patients), need to decrease fentanyl infusion rate because of side effects (L = 2, T = 2, iv = 4 patients), importance of pulmonary infiltrates, or arterial blood gas values. One patient (L group) needed naloxone (0.04 mg iv). Intravenous patients were more frequently nauseated (P = .009) and needed boluses of fentanyl more often (L = 3 +/- 9, iv = 6 +/- 12, T = 4 +/- 8; P = .04).(ABSTRACT TRUNCATED AT 250 WORDS)