Section 3 - Part

Summary

Selection designs and futility designs offer investigators a way to screen potential therapies in early phase clinical research with fewer patients than would be required for a traditional phase 3 trial for each candidate. There are some avoidable-pitfalls when planning a futility study. The first is that if the sample size is too small, a rather awkward situation can arise. The last pitfall relates to the use of historical control data in the single-arm design. Selection procedures offer an attractive approach to the problem of screening potentially good treatments. There are many different procedures for general ranking and selection goals such as selection from among more than two treatments, selection of best subsets of treatments, and ranking treatments in order of efficacy. Although selection procedures efficiently achieve their goal of selecting best treatments, the desire to 'test something' with an accompanying statement of statistical significance seems irresistible.