Evidence suggests that the salvage therapy utilized for relapsed and refractory acute myelogenous leukemia (AML) should differ based on the duration of a patient's complete remission (CR), the principal predictor of outcome. While standard regimens have produced higher CR rates than investigational regimens, these rates have not translated into improved survival in patients with initial remission durations of <1 year. Accordingly, there is no need to give standard regimens to these patients who rather should receive investigational therapy once relapse is discovered. In contrast, in patients with initial remission durations of 1–2 years, standard regimens do increase survival compared to investigational regimens. A somewhat artificial distinction has been placed between phase i and phase ii studies. The agents to be studied in phase ii trials are many, but the patients are limited, so we need to be more innovative in our trial designs. One such proposal, utilizing a bayesian selection design which calls for randomizing a small number of patients among several investigational treatments, will be discussed.
