We have performed a retrospective analysis of the clinical, morphologic, and cytogenetic findings in 26 patients diagnosed between January 1969 and September 1991 with acute erythroblastic leukemia de novo (EL or AML-M6). Clonal chromosomal abnormalities were found in 20 (77%) patients (95% confidence interval [CI], 61% to 93%). Loss of all or part of the long arm (q) of chromosomes 5 and/or 7 was observed in 17 (65%) patients (95% CI, 47% to 83%). In addition, the karyotypes were often complex, with multiple abnormalities and subclones. Among the remaining nine patients, six had a normal karyotype and one each had trisomy 8, t(3;3), or t(3;5). The overall frequency of abnormalities of chromosomes 5 and/or 7 observed in our M6 patients is similar to that observed in our patients with therapy-related acute myeloid leukemia (t-AML; 99 of 129 patients, 77%), but substantially higher than that noted in our other patients with AML de novo (French- American-British [FAB] subtypes M1-M5: 52 of 334 patients, 16%). Our M6 patients with abnormalities of chromosomes 5 and/or 7 were older and had a shorter median survival (16 v 77 weeks [P = .005]) than did the M6 patients without these abnormalities. We found no correlation between morphologic features and either cytogenetic abnormalities or clinical outcome. Of note was the finding that the percentage of myeloblasts, which may account for only a small fraction of the total marrow elements when the revised FAB criteria are applied, had no bearing on prognosis. We conclude that acute erythroblastic leukemia, when defined by morphologic criteria, consists of two distinctive subgroups: one group tends to be older, has complex cytogenetic abnormalities, especially of chromosomes 5 and/or 7, and shares biologic and clinical features with t-AML; the other group, with simple or no detectable cytogenetic abnormalities, has a more favorable prognosis when treated with intensive chemotherapy.