Activity Number:
|
249
- Contributed Poster Presentations: International Statistical Institute
|
Type:
|
Contributed
|
Date/Time:
|
Monday, July 31, 2017 : 2:00 PM to 3:50 PM
|
Sponsor:
|
International Statistical Institute
|
Abstract #324076
|
|
Title:
|
A Parametric Test for Comparing ROC Curves from Biomarkers with a Limit of Detection
|
Author(s):
|
Maria Pardo* and Alba MarĂa Franco Pereira and Christos T. Nakas
|
Companies:
|
Complutense University of Madrid and Complutense University of Madrid and University of Thessaly
|
Keywords:
|
Area under the ROC curve ;
Biomarkers ;
Limit of detection ;
ROC curve
|
Abstract:
|
The use of biomarkers is of growing importance in clinical prognosis or diagnosis of disease. Receiver Operating Characteristic (ROC) curves are often used to evaluate the prognostic performance of a continuous biomarker. However, biomarker measurements frequently have undetectable levels below a detection limit; usually denoted as the limit of detection (LOD). Ignoring observations below the LOD or substituting them by some replacement value as a method of correction will lead to biased ROC estimators. In this work we propose a parametric test for comparing two correlated normally distributed biomarkers subject to the limit of detection. The test procedure is based on asymptotic normality. Simulation results show substantial gains of efficiency by using the proposed approach when the two ROC curves cross. An illustrative example is presented from a study on Colorectal Cancer biomarkers.
|
Authors who are presenting talks have a * after their name.