ABSTRACT
In this chapter, we introduce assurance, a concept close in principle to average power, but more general, as it does not restrict itself to the positive outcome of a significance test to define the success of a clinical trial. We show how studies can be sample sized to achieve a particular level of assurance. In the same way as average power, assurance is also bounded by the prior probability that the treatment effect is positive and this impacts on our ability to sample size a trial to achieve a given assurance. To overcome this restriction, we show how to sample size a study based on normalised assurance, in which the normalisation is with respect to the prior bound. We broaden the use of assurance by considering a series of trials and show how conditional assurance, for example, the assurance of a phase 3 trial, conditional on the positive outcome of a phase 2b study can be calculated and used. We show how the same idea can be applied to studies with interim analyses. Finally, we extend the use of assurance to non-inferiority trials including a fixed margin approach, the so-called synthesis method and Bayesian approaches.
