A review and re-interpretation of a group-sequential approach to sample size re-estimation in two-stage trials.
Bottom Line: In this paper, we review the adaptive design methodology of Li et al. (Biostatistics 3:277-287) for two-stage trials with mid-trial sample size adjustment.We argue that it is closer in principle to a group sequential design, in spite of its obvious adaptive element.Several extensions are proposed that aim to make it even more attractive and transparent alternative to a standard (fixed sample size) trial for funding bodies to consider.
Affiliation: MRC Biostatistics Unit, Cambridge, UK.Show MeSH
Mentions: Fixing C to Zα means that rejection of H0 at stage two via the adaptive design must coincide with a rejection based on a standard analysis using z. The algorithm can be split into the aforementioned steps two and three because equation (7) is independent of n1, and this also makes the numerical optimisation an easier task. The solid line in Figure 4 shows the values of (h, k, , n1) consistent with this strategy. Scales for h and k are shown alongside the p-values for early stopping due to efficacy and futility (Pk and Ph) they imply. For scales and 1- β1, the expected sample size per arm at and n1 are also shown. The red point highlights an interesting and appealing design, where the minimum conditional power equals the unconditional overall power, or β = β1. This occurs at (approximately) h = 1.14, k = 2.24 and n1= 70. This is listed as ‘design 3’ in Table 1. The expected sample size at is approximately 123, which is greater than design 1 in Section 3.4 but is still below the fixed design's sample size.