Univariate Versus Multivariate Surrogates in the Single-Trial Setting

Abstract-In spite of medical and methodological advances, the identification of good surrogate endpoints has remained a challenging endeavor. This may, at least partially, be attributable to the fact that most researchers have only focused on univariate surrogates endpoints. In the present work, we...

Full description

Saved in:
Bibliographic Details
Published in:Statistics in biopharmaceutical research 2019-07, Vol.11 (3), p.301-310
Main Authors: Van der Elst, Wim, Alonso, Ariel Abad, Geys, Helena, Meyvisch, Paul, Bijnens, Luc, Sengupta, Rudradev, Molenberghs, Geert
Format: Article
Language:English
Subjects:
Causal inference
Information theory
Multivariate surrogate endpoints
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Abstract-In spite of medical and methodological advances, the identification of good surrogate endpoints has remained a challenging endeavor. This may, at least partially, be attributable to the fact that most researchers have only focused on univariate surrogates endpoints. In the present work, we argue in favor of using multivariate surrogates and introduce two new complementary metrics to assess their validity. The first one, the so-called individual causal association, quantifies the association between the individual causal treatment effects on the multivariate surrogate and true endpoints, while the second one quantifies the treatment-corrected association between the multivariate surrogate and the true endpoint outcomes. The newly proposed methodology is implemented in the R package Surrogate and a Web Appendix, detailing how the analysis can be conducted in practice, is provided. Supplementary materials for this article are available online.
ISSN:1946-6315
1946-6315
DOI:10.1080/19466315.2019.1575276