We analyzed prognostic factors of response, response duration, and possible impact on survival of epoetin alpha, epoetin beta, or darbepoetin alpha (DAR) with or without granulocyte colony-stimulating factor in 403 myelodysplastic syndrome (MDS) patients. Sixty-two percent (40% major and 22% minor) and 50% erythroid responses were seen, and median response duration was 20 and 24 months according to IWG 2000 and 2006 criteria, respectively. Significantly higher response rates were observed with less than 10% blasts, low and int-1 International Prognostic Scoring System (IPSS), red blood cell transfusion independence, serum EPO level less than 200 IU/L, and, with IWG 2006 criteria only, shorter interval between diagnosis and treatment. Significantly longer response duration was associated with major response (IWG 2000 criteria), IPSS low to INT-1, blasts less than 5%, and absence of multilineage dysplasia. Minor responses according to IWG 2000 were reclassified as "nonresponders" or "responders" according to IWG 2006 criteria. However, among those IWG 2000 minor responders, response duration did not differ between IWG 2006 responders and nonresponders. Multivariate adjusted comparisons of survival between our cohort and the untreated MDS cohort used to design IPSS showed similar rate of progression to acute myeloid leukemia in both cohorts, but significantly better overall survival in our cohort, suggesting that epoetin or DAR treatment may have a favorable survival impact in MDS.