This chapter’s fundamental tenet is that the conventional methodology used to evaluate new drugs is incongruent with clinical reality; in particular, the former frequently underestimates the complexity of the latter. I will point out that this situation is by no means inevitable. Indeed, a large number of papers in statistical literature have pointed out flaws in conventional designs and have proposed alternatives; however, this literature has largely been ignored. By conventional methodology, I refer to: (i) a phase 1 trial whose sole endpoint is toxicity that is quantified using standard criteria and evaluated using a “3+3 design” in order to determine a dose for a subsequent phase 2 trial; (ii) a single-arm phase 2 trial formally concerned only with a single measure of efficacy, followed by (iii) a large randomized phase 3 trial intended to compare the new drug with standard treatment. Although the chapter’s focus is acute myeloid leukemia (AML), I believe readers can generalize its points.