Clinical trial data are often reviewed periodically by Data Monitoring Committees. For many test statistics used for the comparison of two treatments, the stochastic process under review behaves like a discrete Brownian motion when the sample size is large. Under this framework, control of Type I error rate is straightforward and many user-friendly software are available on the market.

In this talk, we will discuss how to discretize continuous boundaries and use alpha-spending functions to design sequential clinical trials. Potential problems for the evaluation of relationship between power and sample size will also be presented.