This prospective phase II study evaluated the efficacy of azacitidine (Aza)+erythropoietin (Epo) in transfusion-dependent patients with lower-risk myelodysplastic syndrome (MDS). Patients ineligible for or refractory to full-dose Epo+granulocyte colony stimulation factors for >8 weeks and a transfusion need of greater than or equal to4 units over 8 weeks were included. Aza 75 mg m−2 d−1, 5/28 days, was given for six cycles; non-responding patients received another three cycles combined with Epo 60 000 units per week. Primary end point was transfusion independence (TI). All patients underwent targeted mutational screen for 42 candidate genes. Thirty enrolled patients received greater than or equal toone cycle of Aza. Ten patients discontinued the study early, 7 due to adverse events including 2 deaths. Thirty-eight serious adverse events were reported, the most common being infection. Five patients achieved TI after six cycles and one after Aza+Epo, giving a total response rate of 20%. Mutational screening revealed a high frequency of recurrent mutations. Although no single mutation predicted for response, SF3A1 (n=3) and DNMT3A (n=4) were only observed in non-responders. We conclude that Aza can induce TI in severely anemic MDS patients, but efficacy is limited, toxicity substantial and most responses of short duration. This treatment cannot be generally recommended in lower-risk MDS. Mutational screening revealed a high frequency of mutations.