Seventy-seven elderly patients (median age 72, range 59-85) with de novo AML were treated with low-dose Ara C (10 mg/m2/12 h over 21 days, for one or two courses). Thirteen (17%) achieved complete remission (CR), 16 (21%) partial remission (PR); 28 (35%) had resistant leukemia, and 20 (26%) early death or death during hypoplasia. Most (86%) of the patients had severe pancytopenia and 58% were hospitalized. Overall median survival was 3 months. Median duration of CR was 9 months. Five CR were longer than 1 year, and two were longer than 4 years. All but one PR were < or = 9 months, and 12/16 were < or = 4 months. Karnofsky index and karyotype (the latter performed for 52 patients) were the only significant prognostic factors of response to treatment (including CR+PR) and survival: poor response rate (8%) and survival (median 0.7 months) were found in patients with Karnofsky index < 60, compared with 44% and 4 months, respectively, in patients with Karnofsky index > or = 60; likewise, patients with rearrangements of chromosome 5 and/or 7 or complex rearrangements had a response rate of 13% and median survival of 1.5 months, compared with 68% and 8 months, respectively, in patients with normal karyotype or single abnormalities (not involving chromosomes 5, 7, or 8). Patients with isolated trisomy 8 had a response rate of 37% but short median survival (2.5 months). Significantly longer survival was seen in responders. Our findings suggest that, overall, low-dose Ara C yields limited results in AML in the elderly. However, it could remain a useful option in elderly patients with AML who are not candidates for intensive chemotherapy (even with the support of growth factors), provided their general condition is not too altered and they do not have an "unfavorable" karyotype (i.e., rearrangements of chromosomes 5 or 7 or complex abnormalities).