Primary resistant acute myeloid leukemia (AML) has a very poor prognosis. Etoposide-mitoxantrone-cytarabine (EMA) timed sequential chemotherapy including a first sequence combining mitoxantrone (12 mg/m(2) per day over 3 days) with cytarabine (500 mg/m(2) per day over the same period), and a second sequence consisting in etoposide (200 mg/m(2) per day for 3 days) and cytarabine as in the first sequence, has been proposed as a salvage regimen. Over a 10-year period, 66 primary resistant AML patients have been treated by EMA salvage chemotherapy. All patients displayed intermediate- or high-risk karyotypic abnormalities. Of the 66 patients, 24 [36%, 95% confidence interval (CI): 25-49%] achieved complete remission (CR). Thirty-eight patients were resistant to EMA chemotherapy and four patients died from toxicity during aplasia. After CR achievement, 18 patients received consolidation therapy. Five patients with an HLA-identical sibling donor underwent allogeneic stem cell transplantation (SCT), one patient received autologous SCT, two patients received a second course of EMA chemotherapy, and ten were scheduled for 6-monthly maintenance courses (mini-EMA). Median follow-up was 7.3 years. At the time of analysis, 21 of the 24 patients (87%) who achieved CR have relapsed. Median disease-free survival (DFS) was 5 months (95% CI: 4.3-7.7 months). Median overall survival (OS) was 5 months (95% CI: 3.8-6.7 months). There were only two long-term remitters (3%). In the univariate analysis, CR achievement was mainly related to white blood cell (WBC) count at the time of starting salvage therapy with poorer outcome for patients with more aggressive leukemia (WBC count > or =10 x 10(9)/l) (CR rates: 50% vs 10%, p<0.001). Overall survival was also influence by WBC count (median OS: 7.2 months vs 2.8 months, respectively, for WBC <and > or =10 x 10(9)/l, p<0.0001). Initial karyotype was not a significant prognostic factor either for CR achievement or for DFS or OS when comparing patients with normal karyotype and those with chromosomal abnormality. In multivariate analysis, WBC count less than 10 x 10(9)/l with the absence of circulating blasts at the time of starting salvage therapy appeared to be of favorable prognostic value for CR achievement ( p=0.002), while WBC count less than 10 x 10(9)/l appeared to be of favorable prognostic value for survival ( p<0.0001). Using these two objective parameters of proven significance, we devised a prognostic system of immediate clinical utility for prognostic stratification and risk-adapted therapeutic choices. Patients with both factors (WBC count <10 x 10(9)/l and no circulating blasts) or with at least one at the time of starting salvage therapy had a CR rate of 50% and were therefore candidates for intensified post-remission therapy. All other patients displayed a very poor outcome and must be oriented after failure of first-line therapy to alternate therapeutic programs based on investigational drugs.