The treatment of relapsed acute myeloid leukaemia (AML) remains unsatisfactory. We conducted a phase II randomized trial where patients received intermediate-dose cytarabine for 4 d followed by gemtuzumab ozogamicin on day 5 (Arm A), or combined with liposomal daunorubicin for 3 d (Arm B), or cytarabine given for 5 d combined with cyclophosphamide for 3 d and topotecan by continuous infusion for 5 d (Arm C). Eligible patients had primary refractory AML, a first relapse after a remission of <1 year, or a second or greater relapse. The primary objective of this trial was attainment of a conventional complete remission (CR) or a CR without platelet recovery (CRp) in at least 40% of patients. The CR/CRp rates for the 82 eligible patients were 3/26 (12%) in Arm A, 2/29 (7%) in Arm B, and 1/27 (4%) in Arm C. No patients who had relapsed within 6 months of initial CR or who had suffered multiple relapses responded. More than 95% of patients subsequently died of AML. No unexpected toxicities were encountered. We conclude that none of these three regimens were effective enough in the treatment of high-risk relapsed or refractory AML to warrant further study. This trial was registered at http://www.clinicaltrials.gov as #NCT00005962.