We retrospectively assessed whether normalized bone marrow WT1 levels could be used for risk stratification in a consecutive series of 584 acute myeloid leukemia (AML) patients. A cutoff value of 5065 copies at diagnosis identified two prognostic groups (overall survival (OS): 44 ± 3 vs 36 ± 3%, P=0.023; leukemia-free survival (LFS): 47 ± 3 vs 36 ± 4%, P=0.038; and cumulative incidence of relapse (CIR): 37 ± 3 vs 47 ± 4%, P=:0.043). Three groups were identified on the basis of WT1 levels post-induction: Group 0 (WT1 between 0 and 17.5 copies, 134 patients, OS: 59 ± 4%, LFS:59 ± 4% and CIR: 26 ± 4%); Group 1 (WT1 between 17.6 and 170.5 copies, 160 patients, OS: 48 ± 5%, LFS:41 ± 4% and CIR: 45 ± 4%); and Group 2 (WT1 >170.5 copies, 71 patients, OS: 23 ± 6%, LFS: 19 ± 7% and CIR: 68 ± 8%) (P<0.001). Post-intensification samples distinguished three groups: patients with WT1 >100 copies (47 patients, 16%); an intermediate group of patients with WT1 between 10 and 100 copies (148 patients, 52%); and a third group with WT1 <10 copies (92 patients, 32%). Outcomes differed significantly in terms of OS (30 ± 7%, 59 ± 4%, 72 ± 5%), LFS (24 ± 7%, 46 ± 4%, 65 ± 5%) and relapse probability (CIR 72 ± 7%, 45 ± 4%, 25 ± 5%), all P<0.001. WT1 levels in bone marrow assayed using the standardized ELN method provide relevant prognostic information in de novo AML.