|
Abstract:
|
To improve the communication of scientific discoveries, the USA FDA-NIH Biomarker Working Group has drafted the “Biomarker, Endpoints, and other Tools (BEST) Resource”. This document contains specific definitions for various types of biomarkers and advocates for adequate validation and context of use. Indeed, the document defines Predictive Biomarkers (PBs) and provides real-world examples; however, the BEST Resource does not specify how to design studies to validate PBs, nor it discusses the type of data or analyses. Consequently, this lack of specification opens a space for statistical thinking to fill in.
On the technical side of the statistical thinking and PBs, Abad Van der Elst and Molenberghs (AVM) used causal inference to develop a set of statistical methods to assess and validated candidate PBs and have applied these to personalize medicine. In this talk, we will discuss why the testing and validation of candidates PBs are integral part of vaccine development and how these are not often treated appropriately. Also, we will argue the AVM’s methods are useful in vaccine trials and apply these in a clinical trial that failed to meet its primary endpoints.
|
