The potential benefits of unrelated donor marrow transplantation are offset by the immunologic complications of graft-versus-host disease (GVHD) and infection. Therefore, we used cryopreserved umbilical cord blood (UCB) as a strategy to reduce the risks of GVHD and treatment-related mortality (TRM) and improve survival. Data on 102 patients (median age 7.4 years) who received transplants between 1994 and 2001 for the treatment of malignant (n = 65; 68% were high-risk patients) and nonmalignant (n = 37) diseases were evaluated. Log-rank tests and Cox regression analyses were used to determine the effects of various demographic, graft-related, and treatment factors on engraftment, GVHD, TRM, relapse, and survival. As of October 15, 2001, the median follow-up was 2.7 years (range, 0.3-7.2). Incidences of neutrophil and platelet engraftment were 0.88 (CI, 0.81-0.95) and 0.65 (CI, 0.53-0.77), respectively. Notably, incidences of severe acute and chronic GVHD were 0.11 (CI, 0.05-0.17) and 0.10 (CI, 0.04-0.16), respectively. At 1 year after transplantation, proportions of TRM and survival were 0.30 (CI, 0.21-0.39) and 0.58 (CI, 0.48-0.68), respectively. In Cox regression analyses, CD34 cell dose was the one factor consistently identified as significantly associated with rate of engraftment, TRM, and survival. Despite the low incidence of GVHD, the proportion of patients with leukemia relapse at 2 years was 0.17 (CI, 0.00-0.38) and 0.45 (CI, 0.28-0.61) for patients with standard and high-risk disease, respectively. There is a high probability of survival in recipients of UCB grafts that are disparate in no more than 2 human leukocyte antigens (HLAs) when the grafts contain at least 1.7 × 10 5 CD34 + cells per kilogram of recipient's body weight. Therefore, graft selection should be based principally on CD34 cell dose when multiple UCB units exist with an HLA disparity of 2 or less.