The objectives of this study were to develop transplantation regimens for patients with advanced myelodysplastic syndrome (MDS) that would be associated with low transplantation-related mortality and improved relapse-free survival. Sixty patients with advanced MDS or acute myeloid leukemia evolving from MDS (sAML), 12 to 62 years old (median, 40 years), were conditioned with busulfan (7 mg/kg) and TBI (6 x 200 cGy) (BU/TBI) and received transplants from related (n = 20) or unrelated donors (n = 40). By French-American-British (FAB) criteria, 21 patients had refractory anemia with excess blasts (RAEB), 16 had RAEB in transformation (RAEB-T), 15 had sAML, and 8 had chronic myelomonocytic leukemia (CMML). By International Prognostic Scoring System (IPSS) criteria, 1 patient had low, 10 had intermediate-1, 13 had intermediate-2, and 31 had high-risk MDS (5 patients had proliferative CMML). All evaluable patients achieved sustained engraftment. The cumulative incidence (CI) of acute GVHD grades II to IV was 83% with unrelated donors and 85% with related donors. The CI of relapse was 25% at 3 years. The incidence of nonrelapse mortality (NRM) at 100 days was 38%. The Kaplan-Meier estimate of survival was 26% at 3 years. Major causes of death were relapse, organ failure, GVHD, and infection. In multivariate analysis, improved relapse-free survival was associated with good cytogenetic risk (P = .002) and shorter disease duration (P = .004). NRM was increased with longer disease duration (P = .0002), positive cytomegalovirus serology (P = .02), and male sex (P = .02). Relapse was associated with poor cytogenetic risk (P = .0004). Thus, BU/TBI conditioning as used in this trial was associated with relapse rates comparable to those observed with a previously used more intensive regimen combining BU/TBI with cyclophosphamide. However, despite the omission of cyclophosphamide, transplantation-related morbidity and mortality were considerable, particularly with transplants from unrelated donors. Future trials should explore the efficacy and tolerability of reduced-intensity conditioning regimens.