Preclinical data suggest that retinoids, eg, all-trans retinoic acid (ATRA), lower concentrations of antiapoptotic proteins such as bcl-2, possibly thereby improving the outcome of anti-acute myeloid leukemia (AML) chemotherapy. Granulocyte colony-stimulating factor (G-CSF) has been considered to be potentially synergistic with ATRA in this regard. Accordingly, we randomized 215 patients with newly diagnosed AML (153 patients) or high-risk myelodysplastic syndrome (MDS) (refractory anemia with excess blasts [RAEB] or RAEB-t, 62 patients) to receive fludarabine + ara-C + idarubicin (FAI) alone, FAI + ATRA, FAI + G-CSF, or FAI + ATRA + G-CSF. Eligibility required one of the following: age over 71 years, a history of abnormal blood counts before M.D. Anderson (MDA) presentation, secondary AML/MDS, failure to respond to one prior course of chemotherapy given outside MDA, or abnormal renal or hepatic function. For the two treatment arms containing ATRA, ATRA was given 2 days (day-2) before beginning and continued for 3 days after completion of FAI. For the two treatment arms including G-CSF, G-CSF began on day-1 and continued until neutrophil recovery. Patients with white blood cell (WBC) counts >50,000/microL began ATRA on day 1 and G-CSF on day 2. Events (death, failure to achieve complete remission [CR], or relapse from CR) have occurred in 77% of the 215 patients. Reflecting the poor prognosis of the patients entered, the CR rate was only 51%, median event-free survival (EFS) time once in CR was 36 weeks, and median survival time was 28 weeks. A Cox regression analysis indicated that, after accounting for patient prognostic variables, none of the three adjuvant treatment combinations (FAI + ATRA, FAI + G, FAI + ATRA + G) affected survival, EFS, or EFS once in CR compared with FAI. Similarly, there were no significant effects of either ATRA ignoring G-CSF, or of G-CSF ignoring ATRA. As previously found, a diagnosis of RAEB or RAEB-t rather than AML was insignificant. There were no indications that the effect of ATRA differed according to cytogenetic group, diagnosis (AML or MDS), or treatment schedule. Logistic regression analysis indicated that, after accounting for prognosis, addition of G-CSF +/- ATRA to FAI improved CR rate versus either FAI or FAI + ATRA, but G-CSF had no effect on the other outcomes. We conclude that addition of ATRA +/- G-CSF to FAI had no effect on CR rate, survival, EFS, or EFS in CR in poor prognosis, newly diagnosed AML or high-risk MDS.