High Dimentional ROC Analysis Can Improve the Correlation Between Histological Grade (HG) and Oncotype-Dx (ONC) Recurrence Risk (RR) in Breast Cancer (BC)

• Activity Number: 425
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Consulting
• Author(s): Yufeng Li*+ and Choo Hyung Lee and Omar Hameed
• Companies: The University of Alabama at Birmingham and The University of Alabama at Birmingham and The University of Alabama at Birmingham
• Keywords: Breast cancer ; ROC analysis ; Recurrence score ; Tumor grade ; three-class diagnostic

Abstract

The ROC analysis has proven remarkably versatile in medical decision making in binary classification settings. It is conceptually appealing and can be visually assessed from a ROC plot. Recently, ROC methodology was extended to three-class diagnostic problems, e.g. histology grade 1, 2 and 3. One may evaluate all pairs of classes using 2-class ROC analyses or an ROC surface or hyper-surface can be constructed. The volume under these surfaces can be used for inference using bootstrap techniques or U-statistics theory. The association between the RS and various clinical and pathologic characteristics has little been tested in the general population of patients who are referred to undergo the test. Using three-class ROC analysis and available data from 458 breast cancer patients at multiple institutions, we attempted to identify the optimal cut point for three RS risk categories.