An Application of Joint Hypothesis Testing in a Randomized Clinical Trial

• Activity Number: 440
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Teaching of Statistics in the Health Sciences
• Author(s): Nicole Zimmerman* and Edward Mascha and Negmeldeen Mamoun
• Companies: Cleveland Clinic and Cleveland Clinic and Cleveland Clinic
• Keywords: Joint hypothesis testing ; noninferiority ; superiority

Abstract

Clinical trials often use multiple primary outcomes to assess efficacy of an intervention. Joint hypothesis testing provides straightforward interpretation of such studies while preserving type I error. Investigators might claim an intervention better than others based on a priori defined rules such as superiority (SUP) on all outcomes, noninferiority (NI) on all outcomes, or SUP on at least one.

We assessed effects of intravenous acetaminophen (IVA) on pain management (PM) after cardiac surgery measured by opioid consumption and pain, requiring NI on both outcomes and SUP on at least one to reject the null hypothesis. We performed one-sided tests at the overall 0.05 significance level for NI and SUP, using Bonferroni correction only for SUP since NI was required for both outcomes. The estimated 90% confidence intervals (CI) for ratio of means of opioids (0.8, 1.1) and mean difference for pain (-1.3, -0.5) were below NI deltas of 1.15 and 1. The estimated upper limits of 95% CIs for SUP were below the criterion of 0 for pain (-1.4, 0.4) but above the criterion of 1 for opioids (0.7, 1.1), rejecting the null and demonstrating that IVA is more effective than placebo on PM.