Exact sequential tests for single samples of discrete responses using spending functions

Sequential tests are increasingly used to reduce the expected sample size o f trials in medical research. The majority of such methods are based on the assumption of normality for test statistics. In clinical trials yielding a single sample of discrete data, that assumption is often poorly satisfied. In this paper we show how a novel application of the spending function app roach of Lan and DeMets can be used together with exact calculation methods to design sequential procedures for a single sample of discrete random var iables without the assumption of normality. A special case is that of binom ial data and the paper is illustrated by the design of a cytogenetic study which motivated this work. Copyright (C) 2000 John Wiley & Sons, Ltd.