We have documented that adrenaline 2.0 micro g.ml- 1 markedly improves relief of dynamic pain when added to a thoracic epidural analgesic infusion of bupivacaine 1 mg.ml- 1 and fentanyl 2 micro g.ml- 1. Concern about possible adverse effects on spinal cord blood flow, expressed by others, prompted us to find the lowest concentration of adrenaline needed to produce effective and reliable pain relief after major surgery. A prospective, randomized, double-blind, parallel group study was carried out in 36 patients after major thoracic or upper abdominal surgery. Patients with only mild pain when coughing during titrated thoracic epidural infusion of approximately 9 ml per hour of bupivacaine 1 mg.ml- 1, fentanyl 2 micro g.ml- 1, and adrenaline 2.0 micro g.ml- 1 were included. The study was conducted as a dose-finding study comparing three different adrenaline concentrations in the epidural mixture (0.5, 1.0, and 1.5 micro g.ml- 1) with each other and with adrenaline 2.0 micro g.ml- 1 in our standard epidural mixture. On the 1st postoperative day, the patients were randomly allocated into three equal groups of 12 patients each, and given a double-blind epidural infusion at the same rate, but with different adrenaline concentrations (0.5, 1.0, or 1.5 micro g.ml- 1). The effects were observed for 4 h or until pain when coughing became unacceptable in spite of rescue analgesia. Rescue analgesia consisted of up to two patient-controlled epidural bolus injections per hour (4 ml) and subsequent i.v. morphine, if necessary. All patients received rectal paracetamol 1 g, every 6th hour. Main outcome measures were pain intensity at rest and when coughing, evaluated by a visual analogue scale and an overall quality of pain relief score. The extent of sensory blockade was evaluated by determining dermatomal hypaesthesia to cold. Pain intensity when coughing increased (P < 0.001) and the number of hypaesthetic dermatomal segments decreased (P < 0.002) when the concentration of adrenaline was reduced below 1.5 micro g.ml- 1 in the triple epidural mixture. This change started within two hours after reducing the concentration of adrenaline below 1.5 micro g.ml- 1. The differences in pain intensities at rest were less pronounced. After 4 h with adrenaline 0.5 or 1.0 micro g.ml- 1 pain intensity when coughing was unacceptable in spite of rescue analgesia. After restarting the standard epidural mixture with adrenaline 2.0 micro g.ml- 1, pain intensity was again reduced to mild pain when coughing and the sensory blockade was restored. Occurrence of pruritus increased with a decreasing adrenaline concentration. Adrenaline in a dose-related manner improves the pain-relieving effect and sensory blockade and decreases the occurrence of pruritus of a low-concentration thoracic epidural analgesic infusion of bupivacaine 1 mg. ml- 1 and fentanyl 2 micro g.ml- 1 after major thoracic or upper abdominal surgery. The minimally effective concentration of adrenaline, when added to bupivacaine 1 mg.ml- 1 and fentanyl 2 micro g.ml- 1, to maintain relief of dynamic pain is approximately 1.5 micro g.ml- 1. The data clearly document that dynamic, cough-provoked pain is a more sensitive outcome measure for postoperative pain relief than pain at rest.