We retrospectively assessed seventy-four consecutive patients with AML over 65 years of age (median 71; range 65-88) treated with an individualized approach in two specialized cancer centers. Patients were managed according to their performance status (PS) and associated diseases in both institutions. The proportion of patients with poor PS (3-4) was higher in center 1 (37%) than in center 2 (10%) and in center 1 palliative treatment was given more frequently (16/32 patients) than in center 2 (7/42 patients). Fifty-one patients received intensive combination chemotherapy including an anthracycline and ara-C or VP16 (2 patients) and 36 patients received a second intensive course as reinduction or as consolidation treatment after complete remission. Patients not eligible for myelosuppressive chemotherapy were treated with palliative measures (23 patients). With intensive chemotherapy, complete remission (CR) was achieved in 29 of 51 patients (57%), after first (20 patients) or second course (9 patients) and the rate of deaths during induction was 14% (7 patients). The CR rate was lower for patients with performance status >or= 2 (48%) as compared to patients with performance status <or= 1 (78%) and for patients with secondary AML (46%) as compared to patients with de novo AML (60%). Median survival was 9.1 months for patients receiving intensive treatment and 1.2 months for patients receiving palliative treatment (P=0.001). In the Cox model for overall survival, treatment with curative intent was associated with longer survival (hazard ratio: 0.33, 95% confidence interval: 0.18-0.58) and poor performance status (PS >or= 2) was associated with reduced survival (hazard ratio: 3.29, 95% confidence interval: 1.75-6.17). Overall 2-years and 5-years survival were 20% and 11% for the patients treated intensively. From this study it appears that an individualized approach of treatment with intensive chemotherapy for selected patients offers a substantial CR rate and an improvement in survival. This analysis also suggests that differences in outcome between single institutions can be explained mainly by referral and selection biases