We examined the efficacy of epidural butorphanol to either prevent or relieve pruritus associated with epidural morphine infusion in children. Forty-six children were randomized to receive either epidural morphine (M) or epidural M with butorphanol (B) for postoperative analgesia. They received bupivacaine and either M 50 microg.kg-1 or the same dose of M plus B 10 microg.kg-1. Following surgery, a continuous infusion of 0.1% bupivacaine with either M 20 microg.ml-1 or M 20 microg.ml-1 + B 4 microg.ml-1 was given at a rate of 0.3 ml.kg-1.h-1. Pain scores and pruritus scores were recorded every 4 h during epidural infusion. Subjects with a pruritus score=2 received diphenhydramine 0.5 mg.kg-1 i.v. and were switched to an alternate epidural infusion; subjects receiving M (group M) were switched to M+B while subjects receiving M+B (group B) were switched to hydromorphone (H) 4 microg.ml-1. There was no difference in the initial incidence of pruritus (group M 11/18; group B 13/28). No subject in group M required a second change of epidural infusion because of continued pruritus after being switched to M+B; five of 13 subjects in group B continued to experience pruritus after being switched to H and required a second change of epidural infusion or an alternate analgesic modality (P=0.038). The median pruritus score in the first 24 h after changing epidural infusions was 0 in subjects in group MDelta (changed from M to M+B) and 1 in subjects in group BDelta (changed from M+B to H; P=0.012). While the median sedation score in the first 24 h was 1 in both groups, there was a greater incidence of sedation scores of 2 in group B than group M (28% vs 12.3%; P=0.021). B 10 microg.kg-1 was not effective in preventing pruritus associated with bolus epidural administration of M 50 microg.kg-1 in children. B 1.2 microg.kg-1. h-1 was effective in relieving pruritus associated with continuous epidural infusion of M 6 microg.kg-1.h-1.