Implementing type I & type II error spending for two-sided group sequential designs

Group sequential designs have become the mainstay for addressing efficacy and ethical issues when monitoring clinical trials. Several different procedures of defining stopping rules have been developed for the formulation of a sequential design, one of these being direct specification of type I and type II error spending. There are also different methods that have been proposed to fit a two-sided design for a given error spending function. Two methods that differ on when type II error begins to be spent are the flexible implementation of the unified family by Kittelson and Emerson and the method of Chang, Hwang, and Shih. Trial designs formulated by the latter are unable to mimic the boundaries of the unified family, which includes the two-sided symmetric designs of Emerson and Fleming, the two-sided designs of Pampallona and Tsiatis, and the double triangular designs of Whitehead and Stratton. Design operating characteristics of these two methods are compared over a wide range of commonly used size, power and error spending function combinations.