An erroneous assumption is often made in clinical trials that once patients achieve a remission on an early phase of a study, then they can be randomized to various arms of later phases of that trial without stratifying for induction therapy, the contention being that complete remissions imply a homogeneous sample with respect to therapy already received. This short article shows that this is not the case because induction therapies may be equal with respect to the percentage of complete remissions achieved but may differ in overall proportion of patients actually cured. Examples of clinical trials with this flaw are presented. A statistical explanation of the problem, which illustrates specific methodological points, is included. © 1981.