Biomarker Evaluation Under Tree/Umbrella Ordering with Application to Lung Cancer

• Activity Number: 25 - Medical Devices and Diagnostics Speed Session
• Type: Contributed
• Date/Time: Sunday, August 8, 2021 : 1:30 PM to 3:20 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Yi Gao* and Lili Tian
• Companies: State University of New York at Buffalo and State University of New York at Buffalo
• Keywords: Biomarker evaluation; ROC curve; Sensitivity; Generalized inference; MOVER; Copulas

Abstract

This talk targets on biomarker evaluation under tree/umbrella ordering which is quite common in diagnostic studies(or biomarker study),involving distinguishing a main class(health class) from the other main class(diseased class) which consists of several diseased subclasses.e.g. differentiation between normal and non-small lung cancer tissues consisting of three unordered subtypes(i.e. adenocarcinoma, large cell carcinoma and squamous-cell carcinoma) falls in tree/umbrella ordering. While pooling is a common practice for such scenario, statistical researchers have pointed out such pooling practice can lead to invalid results,and defining biomarker evaluation under tree/umbrella ordering needs special handling.We focus on a specific biomarker evaluation metric of particular practical importance: sensitivity given specificity under tree/umbrella ordering, and explore extensively parametric and non-parametric methods for confidence interval estimation of sensitivity at a given specificity of single biomarker and of the difference between sensitivities of two correlated biomarkers at a given specificity. All methods are evaluated. A published microarray data for lung cancer is analyzed