We hypothesized that intrathecal fentanyl infusion would provide excellent analgesia, require lower doses than necessary for the epidural or intravenous route of administration, and reduce the incidence and/or severity of side effects. Accordingly, we studied 12 patients during 48 h after thoracotomy (three pneumonectomies, six lobectomies, and three multiple resections of metastases or pleural surgery). The mean dose of fentanyl infused intrathecally was 0.81 +/- 0.26 microgram.kg-1 x h-1, and plasma fentanyl concentrations ranged between 0.49 +/- 0.19 and 0.72 +/- 0.34 ng/ml. Four patients needed a supplementary bolus of intrathecal fentanyl. Pain scores decreased below 30/100 within 1 h when measured at rest but required 24 h to decrease to the same level during coughing. Pulmonary function tests returned to approximately 50% of preoperative values within 1 h of fentanyl infusion. Mean respiratory rates averaged 19 +/- 4, and no episode of apnea was detected. Pruritus, nausea, and headache occurred, respectively, in four, one, and zero patients. Excessive pressure in the infusion system occurred frequently, limiting fentanyl infusion in two patients. All catheters were removed intact; however, one broke outside of the patient's back. This study demonstrates that intrathecal fentanyl infusion can safely provide rapid and intense analgesia but that current 32-gauge intrathecal catheters are not well suited for prolonged postoperative use.