Evaluation of Several Multiple Comparison Procedures for Noninferiority Trials with Two Doses of a Treatment and a Control with a Binary Success Rate Endpoint

• Activity Number: 249
• Type: Contributed
• Date/Time: Tuesday, August 8, 2006 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Hongling Zhou*+ and Mohammad Huque
• Companies: U.S. Food and Drug Administration and U.S. Food and Drug Administration
• Keywords: non-inferiority ; multiplicity ; clinical trial ; hypothesis testing

Abstract

Some non-inferiority trials use 2 doses of the test drug for claim of non-inferiority for each dose to the active control, which should be adjusted for multiplicity. As the traditional Bonferroni and Sidak adjustments are conservative, several other approaches could be taken using the closed testing principle to increase the power of the test. These procedures first test for the intersection null hypothesis H0 (intersection of H01 and H02) for passing the gate that clears the assertion that at least one of the two doses is non-inferior to the active control without being specific. After H0 is rejected at level a, then H01 and H02, which are inferiority null hypotheses for dose 1 and 2 respectively, are then each tested at the same level a. We will address several tests in this regard and will show their asymptotic properties and power performance.