Acute graft-versus-host disease (GVHD) severity is graded by pattern of organ involvement and clinical performance status using a system introduced by Glucksberg and colleagues 21 years ago. We examined how well Glucksberg grade predicted transplant outcome and constructed a Severity Index not requiring subjective assessment of performance in 2881 adults receiving an HLA-identical sibling T-cell-depleted (n = 752) or non-T-cell-depleted (n = 2129) bone marrow transplant for leukaemia between 1986 and 1992. Relative risks (RR) of relapse, treatment-related mortality (TRM) and treatment failure (TF) (relapse or death) were calculated for patients with (Glucksberg Grade I, II or III/IV acute (GVHD) versus those without acute GVHD and for patients with distinct patterns of organ involvement regardless of Glucksberg grade. Using data for non-T-cell-depleted transplants, a Severity Index was developed grouping patients with patterns of organ involvement associated with similar risks of TRM and TF. Higher Glucksberg grade predicted poorer outcome; however, patients with the same grade but different patterns of skin, liver or gut involvement often had significantly different outcomes. The revised Severity Index groups patients in four categories, A-D. Compared to patients without acute GVHD, RRs (95% confidence interval) of TF were 0.85 (0.69, 1.05) for patients with Index A, 1.21 (1.02, 1.43) with B, 2.19 (1.78, 2.71) with C, and 5.69 (4.57, 7.08) with D. Prognostic utility of the Index was tested in patients receiving T-cell-depleted transplants; similar RRs of TF were observed. An acute GVHD Severity Index is proposed to enhance design and interpretation of clinical trials in the current era of allogeneic blood and bone marrow transplantation.