METHODS: BELIEVE was a phase 3, randomised, double-blind, placebo-controlled study performed at 65 sites in 15 countries. The trial included adults with transfusion-dependent β-thalassaemia or haemoglobin E/β-thalassaemia and Eastern Cooperative Oncology Group score of 0-1. Patients were randomly assigned (2:1) using integrated response technology stratified by region to luspatercept (1·0-1·25 mg/kg) or placebo administered subcutaneously once every 21 days. After study unblinding, patients could receive luspatercept in the open-label extension phase (crossover allowed). The primary endpoint results (proportion of patients with reduction in transfusion burden of ≥33% and ≥2 RBC units during weeks 13-24) are described elsewhere; herein we present an update to the primary endpoint analysis consequent to late-reported transfusion events. We also report long-term efficacy (intention-to-treat population) and safety data (safety population) for patients followed up for approximately 3 years. This trial is registered on ClinicalTrials.gov (NCT02604433) and is completed.

FINDINGS: Between May 2, 2016, and May 16, 2017, 336 patients were randomly assigned to luspatercept (n=224) or placebo (n=112). The median age of patients was 30 years (IQR 23-40); 195 (58%) were female and 141 (42%) male. As of Jan 5, 2021, the median duration of treatment in the luspatercept group was 153·6 weeks (IQR 81·0-171·0) and median study follow-up was 163·1 weeks (140·5-176·2). Due to the difference in treatment duration between the luspatercept and placebo groups, no comparative analyses between the two groups were performed after week 96. Patients in the luspatercept group showed a sustained reduction in RBC transfusion burden from baseline through week 192, with mean decreases of 6·2 RBC units (SD 5·7) during weeks 97-144 and 6·4 RBC units (4·3) during weeks 145-192. In the luspatercept group, a 33% or greater reduction in transfusion burden from baseline was observed in 173 (77%) patients over any 12-week interval and in 116 (52%) patients over any 24-week interval. The median total duration of 33% or greater transfusion burden reduction response during any period of at least 12 weeks was 586·0 days (IQR 264·0-1010·0). The most common grade 3 or worse treatment-emergent adverse events (TEAEs) among all patients who received luspatercept (n=315, including 92 patients who crossed over after study unblinding) were anaemia (nine [3%]), increased liver iron concentration (seven [2%]), and bone pain (seven [2%]); serious TEAEs occurred in 71 (23%) patients. No treatment-related deaths occurred in any group during the study.

INTERPRETATION: These long-term results affirm luspatercept's efficacy in addressing key unmet needs of patients with transfusion-dependent β-thalassaemia with a manageable safety profile.