The requirements for supplementary 3 ml epidural injections of bupivacaine 0.5% (top-ups) were used in a randomized double-blind study to compare the effects of five types of thoracic epidural infusions given at 2.5 ml h-1 for the first 24 h after major surgery to the upper abdomen in 99 patients and the lower abdomen in 72. The infusions were: bupivacaine 0.167% alone; diamorphine 0.167 mg ml-1 (0.417 mg h-1) in bupivacaine 0.167%; sufentanil 2 micrograms ml-1 (5 micrograms h-1) in 0.167% bupivacaine; sufentanil 4 micrograms ml-1 (10 micrograms h-1) in 0.167% bupivacaine; and sufentanil 4 micrograms ml-1 (10 micrograms h-1) in normal saline. The patients who had upper abdominal surgery were on average older than those having lower abdominal surgery and a larger proportion of them were female. They received on average fewer top-ups. After both upper and lower abdominal surgery, epidural infusions of bupivacaine alone required the most frequent supplementation (inter-quartile range 6-14 top-ups in 24 h) and the two sufentanil-bupivacaine mixtures required the fewest (interquartile range 0-12 top-ups in 24 h). The infusions of sufentanil without bupivacaine were significantly less effective than the sufentanil-bupivacaine mixtures after upper (but not lower) abdominal surgery. Although the two sufentanil-bupivacaine mixtures were indistinguishable in analgesic effectiveness after either upper or lower abdominal surgery, the lower (5 micrograms h-1) dose rate of sufentanil gave a significantly higher average breathing rate and lower average PaCO2 for the first 24 h after lower (but not upper) abdominal surgery. Blood samples were taken (as an afterthought) from 11 patients receiving sufentanil 10 micrograms h-1, just before the epidural infusion was stopped. The concentrations were mostly above the range for systemic analgesia, but below the values that would have been expected if a steady state had been achieved.