Noninferiority Trial Designs for Odds Ratios and Risk Differences

• Activity Number: 386
• Type: Invited
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Statisticians in the Pharmaceutical Industry
• Author(s): Joan Hilton*+
• Companies: University of California, San Francisco
• Keywords: active-controlled trial ; allocation ratio ; ancillary parameter

Abstract

We show that for noninferiority trials in which response rates are equal under the alternative hypothesis, the common response rate is a design parameter whose value lies between the Experimental and Control rates. Thus setting it equal to the Control rate underestimates the minimum overall sample size, N*. Given the Control rate, the margin, and type I and II errors, our algorithm finds the design value of the common response rate and the corresponding N* and optimal allocation ratio, gamma*. We find that gamma* is increasingly imbalanced as the margin increases, the direction of the imbalance depends on the scale of the margin (for odds ratio VS risk difference), and multiple values of gamma* map to N*. The last feature allows trade-offs between an optimal gamma* at N* and a preferred gamma at a larger N. Power is lost if the scale of the margin is changed between design and analysis.