Twelve patients with acute nonlymphoblastic leukemia (ANL, n = 6) or acute lymphoblastic leukemia (ALL, n = 6) in first complete remission received cyclophosphamide 120 mg/kg and total body irradiation (TBI) 12 Gy followed by HLA-identical sibling marrow that had been depleted of T cells by incubation with anti-CD2 (with or without anti-CD8) monoclonal antibody and rabbit complement. These 12 patients were compared historically to 25 patients with ANL (n = 15) or ALL (n = 10) in first remission given cyclophosphamide 120 mg/kg and TBI 12 Gy followed by non-T cell depleted HLA-identical sibling marrow for parameters of relapse and survival. Thirty-six of the 37 patients received cyclosporin as post transplant prophylaxis for graft-versus-host disease (GVHD). All surviving patients have been followed for a minimum of one year from transplant. The actuarial rate of leukemic relapse in the T-depleted group was 62% compared to 37% in the non-depleted group (p less than 0.02). Additionally, relapse occurred significantly earlier post transplant in the T-depleted recipients (p = 0.012). Actuarial survival at two years post transplant was 24% for the T-depleted recipients and 41% at six years post transplant for the non-depleted recipients (not significant, p = 0.37). We have previously shown that GVHD is less severe in patients given T cell depleted transplants. Taken together, these findings suggest that (under the protocol conditions used) a graft-versus-leukemia effect is not separable from a GVHD effect in man.(ABSTRACT TRUNCATED AT 250 WORDS)