Sample-size calculations for trials that inform individual treatment decisions: a ‘true-choice’ approach

Abstract

Sample size decisions for clinical trials should be taken in such a way as to maximize informed choice by reducing scientific uncertainty about the consequences of an intervention. Recent approaches to trial design have focused on the potential decision impact of the trial when deciding whether the trial should be undertaken, and how large it ought to be. For the most part these approaches are concerned with the impact of trials either on clinical opinion or on collective reimbursement recommendations. Our purpose is to model the contribution of clinical trials to patient-level decision-making and to propose a way of assessing this contribution at the design stage. The model is developed within the framework of Bayesian decision theory. It is presumed that some patients make choices that they would not have made in the presence of perfect information about the likely consequences. These 'false' choices would be reversed in response to a fully informative (ie, very large) trial of the competing interventions. By contrast, choices that would not change in response to a fully informative trial are termed 'true' choices since they accurately reflect patient preferences. An impact plot is proposed which maps how the expected numbers of 'true' and 'false' choices change in response to a trial of any given size. The approach is illustrated with reference to the choice of delivery mode for term breech presentation, using data obtained before the recent term breech trial. Applications in other contexts are indicated. No account is taken of the magnitude of expressed patient preferences for one treatment over another. The upside is that the need for detailed utility-elicitation is obviated. The approach is a pragmatic aid to trial design in settings where patient preference drives the choice between alternative treatments.