Format

Send to

Choose Destination
Can J Cardiol. 2016 Nov;32(11):1356.e21-1356.e28. doi: 10.1016/j.cjca.2016.02.067. Epub 2016 Mar 3.

Composite End Points in Acute Heart Failure Research: Data Simulations Illustrate the Limitations.

Author information

1
Department of Medicine, University of Alberta, Edmonton, Alberta, Canada.
2
Duke Clinical Research Institute, Durham, North Carolina, USA.
3
Canadian VIGOUR Centre, Edmonton, Alberta, Canada. Electronic address: jae2@ualberta.ca.

Abstract

Composite end points are frequently used in clinical trials of investigational treatments for acute heart failure, eg, to boost statistical power and reduce the overall sample size. By incorporating multiple and varying types of clinical outcomes they provide a test for the overall efficacy of the treatment. Our objective is to compare the performance of popular composite end points in terms of statistical power and describe the uncertainty in these power estimates and issues concerning implementation. We consider several composites that incorporate outcomes of varying types (eg, time to event, categorical, and continuous). Data are simulated for 5 outcomes, and the composites are derived and compared. Power is evaluated graphically while varying the size of the treatment effects, thus describing the sensitivity of power to varying circumstances and eventualities such as opposing effects. The average z score offered the most power, although caution should be exercised when opposing effects are anticipated. Results emphasize the importance of an a priori assessment of power and scientific basis for construction, including the weighting of individual outcomes deduced from data simulations. The interpretation of a composite should be made alongside results from the individual components. The average z score offers the most power, but this should be considered in the research context and is not without its limitations.

PMID:
27499377
DOI:
10.1016/j.cjca.2016.02.067
[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Elsevier Science
Loading ...
Support Center