ABSTRACT
Despite a great e¤ort, we often face a high degree of uncertainty about parameters when designing a trial or justifying the sample-size at the design stage. This could involve the initial estimates of within-or between-patient variation, a control group event rate for a binary outcome, the treatment e¤ect desired to be detected, the recruiting pattern, or patient compliance, all of which a¤ect the ability of the trial to address its primary objective (Shih, 2001). This uncertainty can also include the correlation between the measures (if a repeated measure model is used) or among di¤erent variables (multiple endpoints, covariates). If a small uncertainty of prior information exists, a classic design can be applied. However, when the uncertainty is greater, a classic design with a …xed sample-size is inappropriate. Instead, it is desirable to have a trial design that allows for re-estimation of samplesize in the middle of the trial based on unblinded data. Several di¤erent algorithms have been proposed for sample-size re-estimation, including the conditional power approach and Cui-Hung-Wang’s approach based on the ratio of observed e¤ect size versus the expected e¤ect size. In this chapter, we will evaluate the performance of di¤erent sample-
size modi…cation methods. Operationally, it is a concern that sample-size re-estimation will release the unblinded e¢ cacy data to the general public prematurely. Using a discrete function for sample-size re-estimation is suggested, such that the exact e¤ect size would not be revealed. We will study the impact on e¢ ciency of this information-mask approach. The adjusted p-value, the point estimate and con…dence interval calculation will also be discussed. It is important to di¤erentiate the two di¤erent properties: those prop-
erties at the design stage and those at the interim analyses. For example, power is an interesting property at the design stage, but at the time of in-
of concern. From a statistical point of view, most adaptive design methods do not require a prespeci…cation of sample-size adjustment rules at design stage. How to adjust the sample size can be determined right after the interim analysis. In later of the chapter, we’ll give two examples using SSR: a myocardial
infarction prevention trial and a non-inferior adaptive trial with FarringtonManning margin. Summaries and discussions will be presented in the last section.
