In a series of 70 patients diagnosed according to the FAB criteria, 42 clinical and biological disease characteristics were analyzed in order to identify significant prognostic factors by means of univariate and multivariate analysis. The univariate analysis identified ten parameters associated with poor prognosis: Symptoms of anemia, WBC over 10 x 10(9)/l, presence of blast cells, myeloid precursors or erythroblasts in peripheral blood (PB), high bone marrow (BM) cellularity, severe dysthrombopoiesis, percent of blast cells in BM and high serum levels of bilirubin and LDH. The Cox proportional hazards regression method revealed that the combination of high leukocyte counts and BM percentage of blast cells had the strongest predictive relation to survival length (p = 0.002 and p = 0.060 respectively). A new multivariate analysis treating the presence of myeloid and erythroid precursors in PB as a single variable selected only this as the most significant prognostic factor (p = 0.001). Both regression models allowed us to discriminate two significantly different risk groups of patients.