One hundred ninety-five adult patients with refractory or first relapse acute myelogenous leukemia (AML) were randomly assigned to receive high-dose cytarabine (HiDAC), 3 g/m2 as a three-hour intravenous (IV) infusion every 12 hours for four doses, followed by 6,000 IU/m2 asparaginase (ASNase) administered at hour 42, or HiDAC without ASNase. Treatment was repeated on day 8. The median patient age was 52 years. There was an overall superior complete remission (CR) rate for HiDAC/ASNase (40%) v HiDAC (24%), P = .02. Subset analysis according to prior response and age showed the following CR rates: 54% from HiDAC/ASNase treatment of refractory AML in patients less than 60 years, and 31% in patients greater than 60 years; CR from HiDAC in the same refractory groups were 18% (less than 60) and 0% (greater than 60); 37% from HiDAC/ASNase treatment of relapsed AML in patients less than 60 years, and 43% in patients greater than 60 years; CRs from HiDAC in the same relapsed groups were 33% (less than 60) and 21% (greater than 60). Toxicity in the two treatment arms was comparable and consisted primarily of leukopenia, thrombocytopenia, mild hepatic dysfunction, diarrhea, conjunctivitis and serositis, and hyperglycemia. There was only one case of transient cerebellar toxicity and no cutaneous toxicity. Median time to full hematologic recovery was 5 weeks. There was an overall survival benefit for patients treated with HiDAC/ASNase (19.6 weeks) compared with HiDAC (15.9 weeks), P = .046, primarily attributable to effects in refractory patients. Median time to failure for refractory patients who achieved CR was 38.5 weeks with HiDAC/ASNase, and 13.3 weeks for those treated with HiDAC. For relapsed patients in CR from HiDAC/ASNase the median time to failure was 17.7 weeks and 18.3 weeks for HiDAC. The overall 42% CR rate from HiDAC/ASNase v 12% from HiDAC in patients with refractory AML indicates that HiDAC/ASNase is not cross-resistant with standard-dose cytarabine (SDAC) and anthracyclines. We conclude that HiDAC/ASNase has substantial activity in poor-prognosis AML and that this combination warrants further trials in earlier stage disease.