The optimal primary endpoint for acute graft-versus-host disease (GVHD) therapeutic trials has not been established. In a retrospective analysis, we examined the response of 864 patients who received prednisone 60 mg/m 2/d for 14 days, followed by an 8-week taper, as initial therapy for acute GVHD from 1990-2007 at the University of Minnesota. Patients received grafts of human leukocyte antigen-matched sibling bone marrow (BM) or peripheral blood (PB; n = 315), partially matched sibling BM or PB (n = 24), unrelated donor BM or PB (n = 313), single (n = 89) or double (n = 123) umbilical cord blood. Day 28 responses were similar to day 56 responses and better than day 14 responses in predicting transplantation-related mortality (TRM). In multiple regression analysis, patients with no response at day 28 were 2.78 times (95% CI, 2.17-3.56 times; P < .001) more likely to experience TRM before 2 years than patients with a response. Other factors associated with significantly worse 2-year TRM include older age, high-risk disease, severe GVHD, and partially matched related BM/PB. No other differences in response by donor source were observed. These data suggest that day 28 is the best early endpoint for acute GVHD therapeutic trials in predicting 2-year TRM.