The preplanned interim analysis of the COMMANDS trial showed greater efficacy of luspatercept than epoetin alfa for treating anaemia in erythropoiesis-stimulating agent (ESA)-naive patients with transfusion-dependent, lower-risk myelodysplastic syndromes. In this Article, we report the results of the primary analysis of the trial. COMMANDS is a phase 3, open-label, randomised, controlled trial conducted at 142 sites in 26 countries. Eligible patients were those aged 18 years or older, with myelodysplastic syndromes of very low risk, low risk, or intermediate risk (as defined by the Revised International Prognostic Scoring System), who were ESA-naive and transfusion dependent, and had a serum erythropoietin concentration of less than 500 U/L. Patients were stratified by baseline red blood cell transfusion burden, serum erythropoietin concentration, and ring sideroblast status, and randomly allocated (1:1) to receive luspatercept (1·0-1·75 mg/kg body weight, subcutaneously, once every 3 weeks) or epoetin alfa (450-1050 IU/kg body weight, subcutaneously, once a week; maximum total dose 80 000 IU) for at least 24 weeks. The primary endpoint was red blood cell transfusion independence lasting at least 12 weeks with a concurrent mean haemoglobin increase of at least 1·5 g/dL (weeks 1-24), evaluated in the intention-to-treat population. The safety population included all patients who received at least one dose of treatment. This trial is registered with ClinicalTrials.gov (NCT03682536; active, not recruiting). Between Jan 2, 2019, and Sept 29, 2022, 363 patients were screened and randomly allocated: 182 (50%) to luspatercept and 181 (50%) to epoetin alfa. Median age was 74 years (IQR 69-80), 162 (45%) patients were female, and 201 (55%) were male. 289 (80%) were White, 44 (12%) were Asian, and two (1%) were Black or African American. 23 (6%) were Hispanic or Latino and 311 (86%) were not Hispanic or Latino. Median follow-up for the primary endpoint was 17·2 months (10·4-27·7) for the luspatercept group and 16·9 months (10·1-26·6) for the epoetin alfa group. A significantly greater proportion of patients in the luspatercept group reached the primary endpoint (110 [60%] vs 63 [35%]; common risk difference on response rate 25·4% [95% CI 15·8-35·0]; p<0·0001). Median follow-up for safety analyses was 21·4 months (IQR 14·2-32·4) for the luspatercept group and 20·3 months (12·7-30·9) for the epoetin alfa group. Common grade 3-4 treatment-emergent adverse events occurring among luspatercept recipients (n=182) were hypertension (19 [10%] patients), anaemia (18 [10%]), pneumonia (ten [5%]), syncope (ten [5%]), neutropenia (nine [5%]), thrombocytopenia (eight [4%]), dyspnoea (eight [4%]), and myelodysplastic syndromes (six [3%]); and among epoetin alfa recipients (n=179) were anaemia (14 [8%]), pneumonia (14 [8%]), neutropenia (11 [6%]), myelodysplastic syndromes (ten [6%]), hypertension (eight [4%]), iron overload (seven [4%]), and COVID-19 pneumonia (six [3%]). The most common serious treatment-emergent adverse events in both groups were pneumonia (nine [5%] luspatercept recipients and 13 [7%] epoetin alfa recipients) and COVID-19 (eight [4%] luspatercept recipients and ten [6%] epoetin alfa recipients). One death (due to acute myeloid leukaemia) considered to be luspatercept-related was reported at the interim analysis. Luspatercept represents a new standard of care for ESA-naive patients with transfusion-dependent, lower-risk myelodysplastic syndromes. Significantly more patients had red blood cell transfusion independence and haematological improvement with luspatercept than with epoetin alfa, with benefits observed across patient subgroups. Celgene and Acceleron Pharma.