Advantages and Challenges in Using a Composite Endpoint in Cardiovascular Outcome Trials
View Presentation View Presentation
Allison L Barnes, GlaxoSmithKline  Rebekkah S Brown, GlaxoSmithKline  Greg Cicconetti, GlaxoSmithKline  Richard Y Davies, GlaxoSmithKline  *Indrani Nandy, GlaxoSmithKline  Jennifer B Shannon, GlaxoSmithKline 

Keywords: composite endpoint, cardiovascular outcome

GlaxoSmithKline has recently concluded two large Phase III cardiovascular (CV) outcome trials using the drug darapladib. Many endpoints in these studies including the primary one have been composite endpoints. This talk will discuss some of the challenges in the planning, conduct and analysis of these trials related to the composite endpoints.

In CV outcome trials the primary endpoint is often a composite one – time to first occurrence of specific CV conditions. The choice of these conditions depends on the anticipated clinical mechanism of action of the experimental drug. The composite nature of the endpoints avoids multiplicity issue, mitigates the issue of competing risk, and helps keep the cost of the trial down through the smaller sample size for a given level of power.

The calculation of sample size depends on the hazard ratio (HR) assumed for each component, the relative frequency of each component within the composite endpoint, the perceived event rate, the rate of study withdrawal, the expected duration of recruitment and the desired length of the treatment period. Additionally, the inclusion of a component on which the experimental drug has no or minimal effect could dilute the treatment effect. Further, in the case when the composite primary endpoint is statistically significant, but the significance is driven mostly by one component, the regulatory approved label may not contain the composite endpoint. Therefore, the choice of the components and estimation of their relative effects need careful clinical versus statistical consideration while designing the study.

In the collection of CV events data, the adjudication process is typically used to confirm the occurrence of endpoint events; this practice standardizes the definition of events submitted as endpoints by the investigators. The added value of this practice, given the costs and burden to the trial, has become a topic of debate. The concordance of investigator report and adjudicated report of endpoints may be compromised if the adjudication referral process includes too many events, while a more restrictive referral philosophy has an increased risk of missing true events. Additional challenges include determining event dates (since dates could come from different sources, e.g. procedure date, symptom onset date, etc.) and the necessity of coordinating data collection and analysis strategies on which date to use. In the analysis of recurrent events, how many events are counted could pose an issue – e.g. when a fatal myocardial infarction occurs which leads to CV death a few days later, should this be counted as 2 events or 1 event for the subject? For a well-defined composite endpoint it may sometimes be challenging which component analysis to present – mutually exclusive components versus clinically meaningful components; the clinical interpretation of these components plays a critical role.