Patients with severe aplastic anemia were conditioned with cyclophosphamide (200 mg/kg) and given marrow grafts from HLA-identical family members. Among 233 patients transplanted, 225 survived greater than or equal to 14 days and could be evaluated for engraftment. Forty-four of the 225 rejected their graft; 33 of these died and 11 survive. One hundred eighty-one patients had sustained engraftment; of these, 46 died and 135 survived. Binary logistic regression analyses revealed five risk factors for graft rejection: year of transplant, a large number of platelet transfusions, a positive relative response in mixed leukocyte culture, a low marrow cell dose, and omission of donor buffy coat cell infusion for transfused patients. These data show that patients transplanted recently had a lower likelihood of graft rejection than did patients transplanted in earlier years. Conceivably, this was related to changes in transfusion practices, but other factors as yet unidentified are likely to be involved. The data confirm that the largest possible number of marrow cells should be transplanted. Although the difference in the incidence of graft rejection between untransfused and transfused patients was not significant, it should be noted that transfused patients were given buffy coat cells. Because the addition of buffy coat cells results in a higher incidence of chronic graft-v-host disease (GVHD), it is still desirable to transplant patients with marrow alone early in their course before they have been transfused.