Design and Analysis of Noninferiority Condom Failure Studies when Control Event Rates are Expected to Approach Zero

• Activity Number: 90
• Type: Topic Contributed
• Date/Time: Monday, August 8, 2005 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Douglas Taylor*+
• Companies: Family Health International
• Keywords: non-inferiority ; bernoulli

Abstract

Condom breakage and slippage studies often are conducted to compare the functionality of an experimental condom type to that of a standard latex type. In the absence of pregnancy or sexually transmitted disease outcome data, studies that examine breakage and slippage may suffice when seeking regulatory approval to market a new condom type. The typical functionality study is a crossover trial in which enrolled couples use one or more condoms of each type, recording incidents of condom failure after each act of intercourse. On the surface, this problem lends itself to a simple noninferiority type of trial design, using an analysis method that accounts for correlated Bernoulli outcome data (e.g., GEE). In practice, however, the design, analysis, and interpretation of condom functionality studies is made challenging due to complicating factors such as low events rates in certain population strata, the difficulty in choosing a meaningful noninferiority margin based on functionality measures, informative cluster sizes, and learning effects. We consider the implications of various study design and analysis methods.