FLEXIBLE INTERIM ANALYSIS METHOD FOR SAMPLE SIZE RE-ESTIMATION AND EARLY STOPPING: A CONDITIONAL POWER APPROACH

This paper proposes a flexible interim analysis method based on the conditional power approach described in Lan and Trost (1999). Accrued data are analyzed at time t, where 0 Ut) to claim significance. In (3) and (4), the null hypothesis is rejected if final Z-statistic exceeds Zα. This is accomplished by enforcing the appropriate use of conditional power for futility. The boundaries proposed balance the alpha inflation, due both to early stopping to claim significance and sample size re-estimation, with alpha deflation from early stopping for futility. Simulation results will be presented.