In severe aplastic anemia, approximately one-third of responders to standard horse antithymocyte globulin (h-ATG) plus cyclosporine (CsA) will relapse. Anecdotal experience has suggested that a gradual CsA taper might avoid relapse, but this practice has not been rigorously assessed prospectively. In 2003, we adopted a strategy to taper CsA beyond 6 months, with the intention to reduce hematologic relapse compared with our extensive historical experience. In total, 102 patients received h-ATG/CsA for 6 months in two sequential clinical protocols: 67 patients (66%) responded and all had the CsA dose tapered per protocol over the subsequent 18 months (total of 2 years). The rate of relapse at 5 years was 33% (95% CI 27-44%), which did not differ from our large historical relapse experience (patients treated before 2003) of 30-40%, in protocols in which CsA was simply discontinued at 6 months. However, time to relapse was prolonged by about 1 year with the longer CsA course. The rates of clonal evolution and overall survival did not differ between the two cohorts. We infer from this large prospective study that CsA taper as implemented delayed but did not prevent relapse. The kinetics of relapse with long course CsA does suggest that a lower long-term dose might be adequate to maintain patients in remission.