Thirty patients were prospectively randomized to receive either thoracic or lumbar epidural fentanyl infusion for postthoracotomy pain. Epidural catheters were inserted, and placement was confirmed with local anesthetic testing before operation. General anesthesia consisted of nitrous oxide, oxygen, isoflurane, intravenous fentanyl citrate (5 micrograms/kg), and vecuronium bromide. Pain was measured by a visual analogue scale (0 = no pain to 10 = worst pain ever). Postoperatively, patients received epidural fentanyl in titrated doses every 15 minutes until the visual analogue scale score was less than 4 or until a maximum fentanyl dose of 150 micrograms by bolus and an infusion rate of 150 micrograms/h was reached. The visual analogue scale score of patients who received thoracic infusion decreased from 8.8 +/- 0.5 to 5.5 +/- 0.7 (p < or = 0.05) by 15 minutes and to 3.5 +/- 0.4 (p < or = 0.05) by 45 minutes. The corresponding values in the lumbar group were 8.8 +/- 0.6 to 7.8 +/- 0.7 at 15 minutes and 5.3 +/- 0.9 at 45 minutes (p < or = 0.05). The infusion rate needed to maintain a visual analogue scale score of less than 4 was lower in the thoracic group (1.55 +/- 0.13 micrograms.kg-1 x h-1) than in the lumbar group (2.06 +/- 0.19 microgram.kg-1 x h-1) during the first 4 hours after operation (p < or = 0.05). The epidural fentanyl infusion rates could be reduced at 4, 24, and 48 hours after operation without compromising pain relief. Four patients in the lumbar group required naloxone hydrochloride intravenously.(ABSTRACT TRUNCATED AT 250 WORDS)