Sample size calculation in randomised phase II selection trials using a margin of practical equivalence

BACKGROUND: In rare cancers or subtypes of common cancers, a comparison of multiple promising treatments may be required. The selected treatment can then be assessed against the standard of care (if it exists) or used as a backbone for combinations with new, possibly targeted, agents. There could be different experimental therapies or different doses of the same therapy, and either can be done in combination with standard treatments. A ’pick-the-winner’ design is often used, which focuses on efficacy to select the most promising treatment. However, a treatment with a slightly lower efficacy compared to another treatment may actually be preferred if it has a better toxicity or quality of life profile, is easier to administer, or cheaper. METHODS: By pre-defining a margin of practical equivalence in order to calculate the sample size, a more flexible assessment can be made of whether the treatments have very different effects or are sufficiently close so that other factors can be used to choose between them. Using exact binomial probabilities, we calculated the sample size for two- and three-arm randomised selection trials including a margin of practical equivalence with a variety of input parameters. RESULTS: We explain conceptually the margin of practical equivalence in this paper, and provide a free user-friendly web application to calculate the required sample size for a variety of input parameters. CONCLUSION: The web application should help promote the randomised selection design with a margin of practical equivalence, which provides greater flexibility than the ’pick-the-winner’ approach in assessing the results of selection trials.