Abstract We used flow cytometry to quantify minimal residual disease (MRD) in 56 patients with acute myeloid leukemia (AML) expressing a leukemia-associated phenotype. Thirty-four patients aged 18 to 60 years were entered into the AML-10 protocol (induction, consolidation, and autologous stem-cell transplantation [ASCT]), whereas 22 patients older than 60 years received the AML-13 protocol (induction, consolidation, and consolidation II). After induction, the level of MRD that was best associated with treatment outcome was 4.5 × 10−4 residual leukemic cells. However, the outcome in patients with at least 4.5 × 10−4 cells (n = 26) was not significantly different from that in patients with fewer leukemic cells (n = 30); there were 15 (58%) relapses in the first group and 12 (40%) relapses in the second. After consolidation, the most predictive MRD cutoff value was 3.5 × 10−4cells: 22 patients had an MRD level of 3.5 × 10−4 cells or higher and 17 (77%) of these patients had relapse, compared with 5 of 29 patients (17%) with lower MRD levels (P < .001). An MRD level of 3.5 × 10−4 cells or higher after consolidation was significantly correlated with poor or intermediate-risk cytogenetic findings, a multidrug resistance 1 (MDR1) phenotype, short duration of overall survival, and short duration of relapse-free survival (P = .014, .031, .00022, and .00014, respectively). In multivariate analysis, this MRD status was significantly associated with a high frequency of relapse (P < .001) and a short duration of overall (P = .025) and relapse-free survival (P = .007). ASCT did not alter the prognostic effect of high MRD levels after consolidation: the relapse rate after transplantation was 70%. Thus, we found that an MRD level of 3.5 × 10−4 cells or higher at the end of consolidation strongly predicts relapse and is significantly associated with an MDR1 phenotype and intermediate or unfavorable cytogenetic findings.