Survival times in chronic myeloid leukaemia (CML) may vary widely depending on the risk profiles of patients. This fact is frequently not, or not sufficiently, considered in evaluating survival in CML, and some studies do not report risk profiles. Therefore we analysed the relative impact of risk profile and therapy on survival using the median survival times of therapy groups and of risk groups of the three-arm randomized German CML Study I (interferon alpha v hydroxyurea v busulphan; median survival times 65 v 56 v 45 months, n = 490, median observation time 70.4 months). The impact of risk profile (Sokal) on survival as determined by the survival difference between high and low risk patients (40 months) was twice the maximum survival difference between treatment groups (20 months). A similar ratio was obtained after stratification for therapy and for risk profile. Since Sokal's index has been reported to prognostically discriminate IFN-treated patients less well than chemotherapy-treated patients, a new score with better discrimination of IFN-treated patients was also used. The results were similar for both scores. We conclude that the risk profile at diagnosis is still more important for survival of CML patients than therapy. Therefore patients should be stratified according to risk profile for comparisons of survival times between studies and treatment arms.