Thirty patients undergoing closed mitral valvotomy were prospectively randomised to receive either thoracic or lumbar epidural catheter. General anaesthesia consisted of morphine sulphate 0.15 mg/kg (single dose given before skin incision), thiopentone sodium 4-6 mg/kg, vecuronium and halothane titrated to stable haemodynamics. In the immediate postoperative period, pain was assessed by VAS (visual analogue scale) and VRS (verbal ranking score) and an epidural fentanyl bolus of 1.5 microg/kg was given followed by an infusion of 0.4 microg/kg/hr. Pain was assessed after 30 min and if pain relief was still inadequate, another fentanyl bolus of 1 microg/kg was administered, followed by an increase in infusion rate to 0.6 microg/kg/hr. If two consecutive pain scores were satisfactory (VAS <4, VRS <1) maintenance dose of fentanyl was decreased by 0.2 microg/kg/hr. Thoracic group received significantly less total dose of fentanyl in 24 hrs period (446.7 +/- 101.70 microg) compared with the lumbar group (705.33 +/- 181.03 microg) (p<0.01). The mean infusion rate was also significantly less in the thoracic group as compared with the lumbar group (0.44 +/- 0.08 microg/kg/hr vs 0.61 +/- 11 microg/kg/hr, p<0.001). The side effects were comparable between both the groups and none of the patients had significant respiratory depression. The data suggest that thoracic epidural fentanyl infusion is superior to lumbar infusion for post thoracotomy pain relief because of smaller dose requirement.