The underlap coefficient as measure of a biomarker’s discriminatory ability in a three-class disease setting

Vanda Inacio, School of Mathematics

Co-authors: Zhaoxi Zhang, University of Edinburgh; Miguel de Carvalho, University of Edinburgh

Abstract: Existing summary measures of a biomarker’s discriminatory ability in a three-class disease setting rely on a restrictive stochastic ordering assumption, which may lead to incorrect conclusions. To address this, we propose the underlap coefficient, a novel summary index for assessing a biomarker’s ability to distinguish simultaneously between three disease groups. The utility of our new index is illustrated in evaluating how a potential Alzheimer’s disease (AD) biomarker differentiates individuals with normal cognition, mild impairment, and severe dementia or AD, while also assessing the influence of age and gender on its discriminatory power