A Closed Testing Procedure Based on Ordered Alternatives in Dose-Response Studies

• Activity Number: 514
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Girish Aras*+ and Juan Li
• Companies: Amgen Inc. and Amgen Inc.
• Keywords: Closed testing procedure ; Dose-response study ; Ordered alternatives

Abstract

In dose-finding trials, clinicians may believe a priori that the ordering of the responses is known, so the efficacy is assumed to increase monotonically with dose. Under this assuption, the ratio test have far greater power than omnibus procedures, and in addition, they provide protection against the possibility of specifying an incorrect functional form. To establish a dose-response profile, not only we need to test doses against the placebo group, but also doses among themselves. With multiple hypotheses to be tested, protecting family-wise error rate becomes an issue. As per the closed testing principle, we need to consider all possible intersections of the null hypothesis of interest. Order restricted inference combined with a closed testing procedure offers a neat solution to establish a dose-response profile that protects family-wise error rate in a strong sense.