One Sample Group Sequential Trial Design for Correlated Binary Data

• Activity Number: 44
• Type: Contributed
• Date/Time: Sunday, August 9, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Ming Zhou* and Yang Zhao and Larry Shen
• Companies: Bristol-Myers Squibb Company and University of Tennessee Health Science Center and Pharmapace, Inc.
• Keywords: Correlated Binary Data ; Group Sequential Trial ; One Sample ; Independent Increment ; Canonical Distribution

Abstract

Group sequential experiments where the primary outcome is binary are common statistical practice in many different fields, such as bio-medical science, quality control, etc. The case when the binary data are independent has been well studied in the literature, where designs can be constructed using the general group sequential design theory based on the canonical joint distribution. When the binary responses are observed in clusters, the responses within a cluster are usually correlated and the intra-cluster correlation needs to be considered. In this paper, we study the problem of one sample group sequential test of binomial proportion for correlated binary data. In particular, we discuss under what conditions the canonical joint distribution still holds. Further, if the canonical joint distribution is not satisfied, we give an upper bound for the type I and type II errors, which can be of important practical use. Numerical examples are given for illustration purpose.