Group sequential designs using a family of type I error probability spending functions

Abstract

Performing planned or unplanned interim analyses on accumulating data in clinical trials is a frequent practice. In this paper, we propose a general one-parameter family of type I error probability spending functions to construct customized group sequential boundaries with unequal increments in information time. This proposed family generalizes the spending functions of Lan and DeMets¹ and Kim and DeMets.² We give an example to illustrate the use of this family.