Fludarabine in addition to cytosine-arabinoside (ARA-C) increases the accumulation of ARA-C-5'-triphosphate (ARA-CTP), which is responsible for the cytotoxic effect in leukemic blasts. In a randomized phase 3 trial, patients with high-risk myelodysplastic syndrome (MDS) (n = 91) or elderly patients with acute myeloid leukemia (AML) (n = 43) were randomized to receive 2 induction courses consisting of ARA-C (2 g/m2 days 1 through 5) and granulocyte colony-stimulating factor (G-CSF) (filgrastim, 5 microg/kg) during and after chemotherapy with or without fludarabine (25 mg/m2, days 1 through 5) (FLAG versus AG). Consolidation consisted of daunorubicin (45 mg/m2, days 1 through 3) and ARA-C (200 mg/m2, days 1 through 7). Complete remission (CR) rate following AG was 65% versus 71% with FLAG (P =.49). Overall survival (OS) at 24 months was 24% for AG treatment and 39% for FLAG (P =.32). Event-free survival (EFS) at 2 years was 10% and 19% (P =.31) for the AG and FLAG treatments, respectively. Platelet and granulocyte recovery times after the second cycle were prolonged in the FLAG treatment group. Grades 3 to 4 neurotoxicities were more often reported in the FLAG arm (14% versus 3%, P =.03), whereas no significant differences in other toxicities were observed. In a cohort of patients, the in vivo accumulation of ARA-CTP in leukemic cells was determined. Although ARA-CTP accumulation in leukemic cells after FLAG was enhanced, clinical outcome in terms of CR rate, OS, EFS, and disease-free survival (DFS) was not significantly improved by combining fludarabine with ARA-C.