This randomized phase II study was designed to compare the efficacy and tolerability of dacarbazine (DTIC) and bleomycin, vincristine, lomustine and DTIC (BOLD) combined with natural interferon-alpha (nIFN-alpha) or recombinant interferon-alpha2b (rIFN-alpha2b) in patients with advanced melanoma. The treatment arms were: A, DTIC plus nIFN-alpha; B, BOLD plus nIFN-alpha; C, DTIC plus rIFN-alpha2b; D, BOLD plus rIFN-alpha2b. One hundred and eight patients were randomized, of whom 106 were eligible to be analysed for efficacy. Overall, 56% of patients had abdominal visceral and/or bone involvement. The response rates were 8% (2/25) in arm A, 13% (4/31) in arm B, 12% (3/25) in arm C and 24% (6/25) in arm D. The differences were not statistically significant by the usual chi-squared test. However, when analysed using the Cochran-Armitage trend test, the one-sided P values were close to significant (0.085 and 0.033). All of the eight complete responses occurred in patients with soft tissue and/or lung metastases and the BOLD regimens produced six of them. There were no significant differences in survival (arm A, 11.1 months; arm B, 9.8 months; arm C, 9.1 months; arm D, 7.5 months; P=0.62). BOLD was more toxic than DTIC. With the present sample size, there were no statistically significant differences in efficacy between the arms, but there was a trend towards a higher response rate with BOLD plus rIFN-alpha2b. Patients with soft tissue or lung metastases may achieve more complete responses with BOLD regimens.