As curative bone marrow transplantation is available only to a minority of patients with chronic myelogenous leukemia (CML), drug therapy remains of central interest. Several nonrandomized studies have suggested that interferon-alpha (IFN) may prolong survival in CML. In a randomized multicenter study the influence of IFN versus busulfan or hydroxyurea (HU) on survival of Philadelphia-positive (Ph+) CML was examined. A total of 513 Ph+ patients were randomized for treatment as follows: 133 for IFN, 186 for busulfan, and 194 for HU. IFN-treated CML patients have a significant survival advantage over busulfan-treated (P = .008), but not over HU-treated patients (P = .44). The longer survival is due to slower progression to blast crisis. Median survival of IFN-treated patients is 5.5 years [5-year survival, 59%; 95% confidence interval (CI), 48%-70%], of busulfan-treated patients, 3.8 years (5-year survival, 32%; CI, 24%-40%), and of HU-treated patients, 4.7 years (5-year survival, 44%; CI, 36%-53%). Patients who continue on IFN survive longer than those in whom IFN is discontinued before blast crisis (P = .007). Complete hematologic IFN-responders have a survival advantage over partial responders or nonresponders (P = .02). Cytogenetic IFN-responders have no significant survival advantage over nonresponders (P = .2). Patients who attain white blood cell (WBC) counts of 10 x 10(9)/L or less have a survival advantage in the IFN (P = .007) and HU (P = .05) groups. Whereas toxicity in the IFN group was considerably higher than in the busulfan or HU groups, long-lasting cytopenias necessitating discontinuation of therapy as observed with busulfan have not been seen with IFN or HU. The problems of conventional prognostic scores (Sokal's score, Score 1) that we observed in IFN-treated patients support the idea that IFN changes the natural course of CML. We conclude that, with regard to survival of CML in the chronic phase, IFN is superior to busulfan and as effective as HU. Whether and to what extent IFN is superior to HU appears to depend, at least in part, on the degree of WBC suppression by HU-therapy and on the risk profile of the patients.