The 6-MP versus placebo clinical trial in acute leukemia

Abstract

Background This article gives the status of clinical cancer research in the 1950's–1960's and tells the story of the development and conduct of the 6-mercaptopurine (6-MP) versus placebo clinical trial in acute leukemia through the initiation, design, conduct and analysis stages, with emphasis on the ethical aspects of randomizing patients to 6-MP or placebo when in remission. Purpose The specific objective was to compare the lengths of remission for patients receiving 6-MP or placebo after achieving complete or partial remission from steroid treatment. Methods A randomized, double-blind, placebo controlled sequential study was conducted in which patients were paired by remission status at each of the eleven institutions participating in the study, and randomized to 6-MP or placebo within each pair of patients. A preference for 6-MP or placebo was recorded depending on which patient in the pair had the longer remission. The preferences were plotted according to a restricted sequential procedure devised by Peter Armitage and, depending on which boundary of the design was crossed, a statistically significant difference could be declared favoring 6-MP, placebo or no preference. Conclusions The trial established the efficacy of 6-MP for maintaining longer remissions in acute leukemia and led to the concept of 'adjuvant chemotherapy', namely that patients with minimal disease have a substantially better response to chemotherapy than patients with advanced disease, a concept that has been followed in many other forms of cancer. Statistically, the fact that many patients were still in remission when the study was stopped (i.e. the length of remission data for these patients was 'right censored') led to the development of a generalized Wilcoxon test and was an important influence on Cox's development of the proportional hazards model. The trial had an innovative design in the early 1960's and has been an important influence on subsequent clinical research in cancer and statistical research in survival analysis.