A phase III clinical trial was developed to test whether the addition of etoposide to a high-dose cytosine arabinoside regimen would improve the remission rate, duration of remission, and survival in relapsed and refractory patients with acute myelogenous leukemia. One hundred and thirty-one patients stratified by age, performance status, percentage of marrow blasts, platelet count, bilirubin and presence or absence of clinical infection, refractory or relapsed (+/- 9 months) were randomized to receive high-dose cytosine arabinoside, 3 g/m2 every 12 h for 6 days with or without three doses of etoposide, 100 mg/m2 days 7-9. Of 67 patients randomized to cytosine arabinoside alone, 31% obtained a complete remission with a median remission duration of 11.9 months. Of 66 patients randomized to the combination regimen, 38% obtained a complete remission with a median duration of 25 months. None of these differences were statistically significant. Significantly (p = 0.036) longer survival was seen in patients on the combination regimen under the age of 50. There was no difference in overall survival. Six and 8%, respectively, of patients were free of disease at 5 years. The addition of etoposide to a high-dose cytosine arabinoside regimen had at best a marginal effect at the expense of some increase in toxicity.