Modern Approaches to Clinical Trials Using SAS: Classical, Adaptive, and Bayesian Methods

3.5 Summary and Conclusions

Several different methods for blinded and unblinded sample size re-estimations were presented in this chapter for normally distributed data. Blinded methods reviewed included the one-sample variance method and the adjusted one-sample variance method. Unblinded SSR methods included Fisher’s combination test, the inverse normal method, the CHW procedure, the promising zone design and conditional type I error preserving procedures (Proschan and Hunsberger, Müller and Schäfer). With the exception of the promising zone and the Müller and Schäfer procedure, all the other unblinded methods can be viewed as a special form of a combination test.

These procedures can be extended to the case of a binary endpoint based on the ...