Issues in Determining Noninferiority Limits in Context with Retrospective Comparisons to Placebo

• Activity Number: 564
• Type: Topic Contributed
• Date/Time: Wednesday, August 3, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Steven Julious*+
• Companies: University of Sheffield
• Keywords: non-inferiority ; non-inferiority margin ; fda ; chmp

Abstract

To assess the efficacy of a new investigative treatment a non-inferiority study is undertaken when it is no longer ethical to have a placebo control. Instead an active controlled trial is undertaken. The objective is to show that the new treatment is no worse than the active control A major issue in designing a non-inferiority study is the setting of the non-inferiority limit. The FDA discuss setting a limit so it would be possible to demonstrate superiority over placebo. This comparison would need to be done indirectly as placebo is not given concurrently. There is an issue with indirect comparisons if they are done retrospectively as the effect over placebo may not be as great today as when a placebo controlled trial was last undertaken The CHMP highlight the possibility of using a survey of experts to help quantify the margin. This could overcome some problems of retrospective comparisons - but can the actual empirical evidence be ignored? An approach that uses both the objective observed data and subjective opinion to set a non-inferiority margin would useful. This presentation explores setting a non-inferiority limit and discusses how simple Bayesian approaches may be useful