BACKGROUND Antileukemic chemotherapy has been used for two decades to treat high-risk myelodysplastic syndrome (refractory anemia with excess of blasts [RAEB] and RAEB in transformation into acute leukemia [RAEB-t]) patients. Because the results of standard regimens have been disappointing, high-dose chemotherapeutic regimens were investigated recently. In the absence of randomized trials, the relative merits of various treatment regimens are unknown. METHODS The authors analyzed the outcome for 394 newly diagnosed patients treated between 1991 and 1999 with five regimens consisting of intermediate- or high-dose cytosine arabinoside (A) in combination with idarubicin (I), and introduced cyclophosphamide (C) and the new agents fludarabine (F) and topotecan (T) into new combinations with A. In addition to defining the role of high-intensity chemotherapy in the overall outcome for patients with RAEB-t and RAEB, the authors determined the relative merits of the five regimens (IA, FA, FAI, TA, and CAT), accounting for the nonrandom distribution of the prognostic covariates. RESULTS The overall complete response (CR) rate of 58% was significantly associated with karyotype, performance status (PS), treatment in the laminar air flow room, duration of antecedent hematologic disorder and age, but not French-American-British or International Prognostic Scoring System risk categories. Multivariate analysis did not identify statistically significant differences in CR rates obtained with each regimen. Induction death rates increased with age with all but the TA regimen; they were lowest with TA (5.4%) and highest with FAI (20.7%), and these differences were significant in patients older than 65 years. The trend for time to death was the same as for time to recurrence in all groups. Multivariate analysis of time to death identified treatment regimen (FA, FAI, and CAT), cytogenetic status (-5/-7), increasing age, and PS greater than 2 as significant independent unfavorable prognostic factors. After prognostic variables were accounted for, survival with IA treatment remained superior to that of FA and FAI but comparable to TA, and CR duration was only marginally shorter with FA. Landmark analysis showed the overall survival of responders to be superior to that of nonresponders, the difference remaining significant after adjustment for prognostic covariates. CONCLUSIONS Although the newer regimens did not improve outcome, TA and CAT produced results comparable to those of IA and may be considered treatment alternatives. The TA regimen was particularly effective in RAEB patients and could be delivered safely, with low induction mortality. Our results indicated that although CR seemed associated with survival advantage, innovative post-remission managements represent a challenge because improvement in outcome is not likely to come from intensified therapy.