Acute graft-versus-host disease (aGVHD) has been classified according to the Seattle criteria as grades 0, I, II, III, and IV for 20 years. The predictive value of such detailed grading is a matter of debate; publications usually report GVHD as present or absent or as absent, moderate, or severe. The Working Party Chronic Leukemia of the European Group for Bone Marrow Transplantation analyzed data of 1,294 patients transplanted from an allogeneic donor for chronic myelogenous leukemia (CML) in first chronic phase and tested the predictive value of aGVHD grading for the following end-points: day 100 mortality (D100M), transplant-related mortality (TRM), relapse incidence (RI), leukemia-free survival (LFS), and survival (SURV). aGVHD was absent in 462 patients (35.7%), grade I occurred in 335 (25.8%), grade II in 264 (20.5%), grade III in 110 (8.5%), and grade IV in 123 patients (9.5%). A total of 297 patients (23%) died within 100 days, 495 patients (38%) died of any TRM, and 100 patients (8%) died of relapse. D100M according to grades 0, I, II, III, and IV was 17%, 13%, 19%, 38%, and 70%, respectively, with significant difference between 0-II versus III-IV. TRM was 28%, 27%, 43%, 68%, and 92%, respectively, with a distinct separation between 0-I versus II-IV. RI showed a continuous decrease of 37%, 30%, 23%, 18%, and 8%, respectively, with increasing aGVHD. LFS was 45%, 51%, 44%, 26%, and 7%, respectively, and was best for patients with grade I aGVHD. This finding was also reflected in a better overall survival (60%, 64%, 53%, 30%, and 8%, respectively). The better LFS for grade I aGVHD patients compared with patients with grade 0 or II aGVHD was confirmed (P = .05) in a multivariate analysis. These data document the value of the present 5-point grading of aGVHD, ie, different outcome is observed depending on endpoint analyzed. Restricting information about aGVHD to presence or absence is not warranted.