After failure of erythropoiesis-stimulating agents (ESAs), lenalidomide (LEN) yields red blood cell (RBC) transfusion independence (TI) in 20-30% of lower-risk non-del5q myelodysplastic syndrome (MDS). Several observations suggest an additive effect of ESA and LEN in this situation. We performed a randomized phase III study in 131 RBC transfusion-dependent (TD, median transfusion requirement six RBC units per 8 weeks) lower-risk ESA-refractory non-del5q MDS. Patients received LEN alone, 10 mg per day, 21 days per 4 weeks (L arm) or LEN (same schedule) + erythropoietin (EPO) beta, 60,000 U per week (LE arm). In an intent-to-treat (ITT) analysis, erythroid response (HI-E, IWG 2006 criteria) after four treatment cycles (primary end point) was 23.1% (95% CI 13.5-35.2) in the L arm and 39.4% (95% CI 27.6-52.2) in the LE arm (P=0.044), while RBC-TI was reached in 13.8 and 24.2% of the patients in the L and LE arms, respectively (P=0.13). Median response duration was 18.1 and 15.1 months in the L and LE arms, respectively (P=0.47). Side effects were moderate and similar in the two arms. Low baseline serum EPO level and a G polymorphism of CRBN gene predicted HI-E. Combining LEN and EPO significantly improves erythroid response over LEN alone in lower-risk non-del5q MDS patients with anemia resistant to ESA.