Multivariate B-Value: A Tool for Monitoring Data with Multiple Co-Primary Endpoints

• Activity Number: 513
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yansong Cheng*+ and Surajit Ray and Ying Zhu
• Companies: Boston University and Boston University and Biogen Idec
• Keywords: conditional power ; multivariate B-value ; multiple co-primary endpoints ; reverse multiplicity ; interim analysis

Abstract

Many clinical trials are required to declare significant efficacy on two or more primary endpoints simultaneously. Multiple co-primary endpoints problem is also called reverse multiplicity problem. Intersection-Union Test (IUT) is the standard test for such problem. How to monitor the clinical trial with multiple co-primary endpoints is lack of consideration. This paper extends the B-value to multiple dimensions that makes the B-value tool be applicable for the clinical trial study with considering multiple co-primary endpoints. The correlation among the co-primary endpoints are taken into consideration. The overall conditional power (CP) is defined as the probability of declaring all endpoints are significant at the end of study, conditioning on the observed information at interim analysis. The overall CP of the IUT is no greater than the smallest CP of all sub-hypothesis and it falls between the two extreme cases: one is all the endpoints are independent and the other is all the endpoints are perfectly correlated. The example of comparing two samples with two normal co-primary endpoints is shown.