Chronic graft-versus-host disease (cGVHD) is the leading cause of late treatment-related deaths among recipients of allogeneic bone marrow and blood transplants. However, cGVHD is also associated with fewer relapses. We sought to determine whether severity of cGVHD predicts the magnitude of these effects. One impediment to such an analysis is the current limited/extensive grading system for cGVHD because this classification was designed to identify patients likely to benefit from systemic immune suppression and does not capture the severity of multiorgan involvement. We, therefore, first developed a grading system predictive for survival by using data from 1827 HLA-matched sibling allotransplant recipients reported to the International Bone Marrow Transplant Registry (IBMTR). We found Karnofsky performance score, diarrhea, weight loss, and cutaneous and oral involvement to be independent prognostic variables, from which we generated a grading scheme. We tested this scheme, the limited/extensive classification system, and a classification based on clinical impression of overall cGVHD severity (mild/moderate/severe) in parallel analyses of 1092 HLA-matched sibling transplant recipients from the IBMTR and 553 recipients of unrelated donor marrow from the National Marrow Donor Program. Presence of cGVHD was associated with fewer relapses (relative risk [RR], 0.5-0.6) but more treatment-related mortality (RR, 1.8-2.8) in the 3 analyses. No grading scheme correlated cGVHD severity with relapse rates, but all schemes predicted treatment-related mortality. Survival and disease-free survival of the most favorable cGVHD group in each scheme were similar, or better, than those of patients without cGVHD; these patients may not need aggressive or extended immune suppression.