Validating Surrogate Endpoints from an Information Theory Perspective

• Activity Number: 113
• Type: Topic Contributed
• Date/Time: Monday, August 2, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Ariel Alonso Abad*+
• Companies: Interuniversity Institute for Biostatistics and Statistical Bioinformatics
• Keywords: Meta-analysis ; Surrogate Markers ; Information Theory

Abstract

The very mention of surrogate endpoints has always been very controversial due to a number of well-known unfortunate historical events. However, while many would like to avoid surrogate endpoints altogether, sometimes surrogates will be the only reasonable alternative, especially when the true endpoint is rare and/or distant in time. It is then best to use {\em validated\/} surrogates, but one clearly needs to reflect on the very meaning of validation. Like in most clinical decisions, statistical arguments will play a major role, but ought to be considered in conjunction with clinical and biological evidence. Recently, using information theory (IT), a new unified approach to the validation problem has been proposed. This approach implies a new definition of surrogacy that has an appealing and simple interpretation.