High-dose cyclophosphamide (Cy) has been promoted as curative therapy for severe aplastic anemia (SAA). However, our randomized trial comparing antithymocyte globulin (ATG) and Cy was terminated early because of excess morbidity/early mortality in the Cy arm. We now report analysis of secondary endpoints at a median of 38 months. Relapse occurred in 6 (46%) of 13 responders in the ATG arm versus 2 (25%) of 8 in the Cy arm (P =.38). Five (31%) of 16 patients in the ATG arm and 4 (27%) of 15 patients in the Cy arm had evidence of paroxysmal nocturnal hemoglobinuria (PNH) at diagnosis, with no substantial change in the overall percentage of glycophosphatidyl inositol (GPI)-anchored protein-deficient neutrophils over extended follow-up in individual patients in either arm. Bone marrow cytogenetic abnormalities have been observed among surviving patients in both arms (2 of 14 ATG versus 1 of 12 Cy, P =.70). High-dose Cy does not prevent relapse or clonal evolution in SAA.