This study compared epidural and intravenous fentanyl infusions for pain relief for the first 20 h after thoracotomy, in order to examine whether an thoracic epidural fentanyl infusion offers clinical advantage over an intravenous infusion. Forty patients were assigned randomly to receive either fentanyl epidurally and saline intravenously or fentanyl intravenously and saline epidurally in a double-blind fashion. For each patient the fentanyl infusion was titrated to a rate required for pain relief (pain score less than 3, maximum 10). Patients reported similar median pain scores, but in the epidural group the required mean fentanyl infusion rate was less (0.95 +/- 0.23 vs. 1.67 +/- 0.46 micrograms.kg-1.h-1, P = 0.0001) and plasma fentanyl concentrations were less at 4 and 18 h (4 h: 0.81 +/- 0.27 vs. 1.38 +/- 0.36 ng.ml-1, P = 0.0001; 18 h: 0.94 +/- 0.32 vs. 1.54 +/- 0.65 ng.ml-1, P = 0.0007) than those in the intravenous group. Respiratory function was better preserved and the incidence of nausea and sedation was less in the epidural group than in the intravenous group. In conclusion there appears to be a clinical advantage to the epidural infusion over the intravenous infusion of fentanyl for analgesia after thoracotomy.