The prognosis for patients with acute myeloid leukemia (AML) is heterogeneous. A minority of patients have clinical and biologic features associated with a very high risk of relapse. For the remaining patients, no clear prognostic factors can be identified at diagnosis. The degree of treatment response is likely to be an informative predictor of outcome for these patients. Modern assays to detect AML cells that are undetectable by conventional morphologic techniques, ie, minimal residual disease (MRD), can potentially improve measurements of treatment response. It is plausible that modifications to treatment based on the results of these assays will improve clinical management and ultimately increase cure rates. Established MRD assays for AML are based on either polymerase chain reaction amplification of genetic abnormalities or flow cytometric detection of abnormal immunophenotypes. Residual disease and treatment response can be measured by these assays in a manner that is much more sensitive and objective than that afforded by conventional morphologic examination. The expanding use of MRD testing is beginning to change the definitions of treatment response and of remission. Other clinically informative uses of MRD testing include the detection of early relapse and the evaluation of the efficacy of new antileukemic agents.