Pharmacological prevention against relapses in patients with neuromyelitis optica spectrum disorder (NMOSD) is developing rapidly. We aimed to investigate the safety and efficacy of rituximab, an anti-CD20 monoclonal antibody, against relapses in patients with NMOSD.We did a multicentre, randomised, double-blind, placebo-controlled clinical trial at eight hospitals in Japan. Patients aged 16-80 years with NMOSD who were seropositive for aquaporin 4 (AQP4) antibody, were taking 5-30 mg/day oral steroids, and had an Expanded Disability Status Scale (EDSS) score of 7·0 or less were eligible for the study. Individuals taking any other immunosuppressants were excluded. Participants were randomly allocated (1:1) either rituximab or placebo by a computer-aided dynamic random allocation system. The doses of concomitant steroid (converted to equivalent doses of prednisolone) and relapses in previous 2 years were set as stratification factors. Participants and those assessing outcomes were unaware of group assignments. Rituximab (375 mg/m2) was administered intravenously every week for 4 weeks, then 6-month interval dosing was done (1000 mg every 2 weeks, at 24 weeks and 48 weeks after randomisation). A matching placebo was administered intravenously. Concomitant oral prednisolone was gradually reduced to 2-5 mg/day, according to the protocol. The primary outcome was time to first relapse within 72 weeks. Relapses were defined as patient-reported symptoms or any new signs consistent with CNS lesions and attributable objective changes in MRI or visual evoked potential. The primary analysis was done in the full analysis set (all randomly assigned patients) and safety analyses were done in the safety analysis set (all patients who received at least one infusion of assigned treatment). The primary analysis was by intention-to-treat principles. This trial is registered with the UMIN clinical trial registry, UMIN000013453.Between May 10, 2014, and Aug 15, 2017, 38 participants were recruited and randomly allocated either rituximab (n=19) or placebo (n=19). Three (16%) patients assigned rituximab discontinued the study and were analysed as censored cases. Seven (37%) relapses occurred in patients allocated placebo and none were recorded in patients assigned rituximab (group difference 36·8%, 95% CI 12·3-65·5; log-rank p=0·0058). Eight serious adverse events were recorded, four events in three (16%) patients assigned rituximab (lumbar compression fracture and infection around nail of right foot [n=1], diplopia [n=1], and uterine cancer [n=1]) and four events in two (11%) people allocated to placebo (exacerbation of glaucoma and bleeding in the right eye chamber after surgery [n=1], and visual impairment and asymptomatic white matter brain lesion on MRI [n=1]); all patients recovered. No deaths were reported.Rituximab prevented relapses for 72 weeks in patients with NMOSD who were AQP4 antibody-positive. This study is limited by its small sample size and inclusion of participants with mild disease activity. However, our results suggest that rituximab could be useful maintenance therapy for individuals with NMOSD who are AQP4 antibody-positive.Japanese Ministry of Health, Labour and Welfare, Japan Agency for Medical Research and Development, and Zenyaku Kogyo.