An analysis was conducted of clinical and laboratory variables associated with response to remission induction therapy and remission duration in 440 patients with acute myelogenous leukemia treated between 1975 and 1983. The complete remission rate was 259/440 (59%). Specific cytogenetic abnormalities such as t (8; 21), t (15; 17), and inv16 were found to be favorable for response to therapy and/or remission duration, whereas those with a normal (diploid) karyotype had an intermediate prognosis. All other karyotypic abnormalities were associated with lower response rates and short complete remission durations. The karyotypes were classified as favorable, intermediate, and unfavorable groups for response and remission duration after allowing for all the other observed clinical and laboratory values related to prognosis. The cytogenetic classification was prospectively validated in an independent test group of 130 patients treated between 1983 and 1986 and showed a consistent relationship to response and remission duration. Logistic regression and proportional hazard models developed from the initial 440 patients were prospectively evaluated in the test group of 130 patients. Clear stratifications of patients into good, intermediate, and poor risk groups were obtained in the prospective tests. The karyotype of the leukemia cells is an independent prognostic variable for response and remission duration in acute myelogenous leukemia.