The “A, B, C” of multiple statistical methods for composite endpoints
Hironori Hara1; David van Klaveren2; Norihiro Kogame3; Ply Chichareon3; Rodrigo Modolo3; Mariusz Tomaniak4; Masafumi Ono1; Hideyuki Kawashima1; Kuniaki Takahashi3; Davide Capodanno5; Yoshinobu Onuma6; Patrick W Serruys7, ;
1. Department of Cardiology, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands; Department of cardiology, National University of Ireland, Galway (NUIG), Galway, Ireland 2. Department of Public Health, Center for Medical Decision Making, Erasmus MC, Rotterdam, The Netherlands; Predictive Analytics and Comparative Effectiveness Center, Institute for Clinical Research and Health Policy Studies, Tufts Medical Center, Bosto 3. Department of Cardiology, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands 4. Department of Cardiology, Erasmus Medical Center, Erasmus University, Rotterdam, The Netherlands; First Department of Cardiology, Medical University of Warsaw, Warsaw, Poland 5. Division of Cardiology, Cardio-Thoraco-Vascular and Transplant Department, CAST, Rodolico Hospital, AOU “Policlinico-Vittorio Emanuele”, University of Catania, Catania, Italy 6. Department of cardiology, National University of Ireland, Galway (NUIG), Galway, Ireland 7. Department of Cardiology, National University of Ireland Galway (NUIG), Galway, Ireland; NHLI, Imperial College London, London, United Kingdom, United Kingdom
As a public service to our readership, this article - peer reviewed by the Editors of EuroIntervention - has been published immediately upon acceptance as it was received. The content of this article is the sole responsibility of the authors, and not that of the journal or its publishers.
Please note that supplementary movies are not available online at this stage. Once a paper is published in its edited and formatted form, it will be accompanied online by any supplementary movies.
To read the full content of this article, please log in to download the PDF.
Composite endpoints are commonly used in clinical trials, and time-to-first-event analysis has been the usual standard. Time-to-first-event analysis treats all components of the composite endpoint as having equal severity and is heavily influenced by short-term components. Over the last decade, novel statistical approaches have been introduced to overcome these limitations. We reviewed win ratio analysis, competing risk regression, negative binomial regression, Andersen-Gill regression, and weighted composite endpoint (WCE) analysis. Each method has both advantages and limitations.The advantage of win ratio and WCE analyses is that they take event severity into account by assigning weights to each component of the composite endpoint. These weights should be pre-specified, because they strongly influence treatment effect estimates. Negative binomial regression and Andersen-Gill analyses consider all events for each patient – rather than only the first event – and tend to have more statistical power than time-to-first-event analysis.Pre-specified novel statistical methods may enhance our understanding of novel therapy when components vary substantially in severity and timing. These methods consider the specific type of patients, drugs, devices, events, and follow-up duration.