Meta-Analysis of One Outcome from Group Sequential Trials with Composite Outcomes: Are Standard Methods Appropriate?

• Activity Number: 618
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: ENAR
• Author(s): Abigail Shoben*+
• Companies: The Ohio State University
• Keywords: meta-analysis ; composite outcome ; sequential monitoring

Abstract

Composite outcomes, in which multiple events are combined into one outcome, are the primary outcome in many clinical trials. Composite outcomes are convenient because they provide a way to collapse over competing risks and provide simple interpretation to physicians and others evaluating effectiveness. However, composite outcomes may not provide clarity if treatment effects differ between the outcomes being combined, and attempts to separate the effects post-hoc either in a single trial or in combined meta-analyses may be problematic. Further complications may arise if clinical trial data were sequentially monitored for safety or efficacy. Such monitoring may be done using the composite outcome or using one or more outcomes separately. Differential monitoring of one or more outcomes in the composite outcome may further complicate post-hoc meta-analyses. We consider as an example recent meta-analyses of data from randomized trials of magnesium sulfate on neurological deficits and death in preterm infants. We illustrate the problems resulting from the separation of the composite outcome into separate outcomes in these meta-analyses and provide guidance for future studies.