Correct Unequal Tail Probabilities When Comparing Two Independent Proportions

• Activity Number: 628
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Jin Xu*+ and Frank Liu
• Companies: Merck and Merck
• Keywords: Unequal allocation ; Tail probability ; Safety evaluation ; Proportion

Abstract

When safety is not their primary focus, clinical trials often compare safety profiles by 95% confidence interval for treatment differences in terms of proportion of subjects reporting various adverse experiences. If the two treatments truly have the same risk profile, we should expect the same chance finding of any statistical significance against either one of the two groups, that is, the Type I errors at both tails are the same. However, in this presentation we show that an unequal sample size for the two groups can lead to wildly unequal tail probabilities when the endpoints being compared are discrete and the event rates are low, which leads to a false appearance that one group has a worse safety profile. This unequal tail probability problem may cause unnecessary concerns when comparing rare serious adverse events in clinical trials with unequal sample sizes. We show how to correct this unequal tail probabilities for the popular Miettinen and Nurminen's method in both stratified and unstratified analyses and demonstrate the advantages of this correction.