The Incremental Value of Percent Mammographic Density and Common Single Nucleotide Polymorphism Genotypes for Predicting Absolute Risk of Breast Cancer

• Activity Number: 533
• Type: Invited
• Date/Time: Wednesday, August 6, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Jinbo Chen*+ and Lu Chen and Mitchell Henry Gail
• Companies: University of Pennsylvania Perelman School of Medicine and University of Pennsylvania Perelman School of Medicine and National Cancer Institute

Abstract

The Breast Cancer Risk Assessment Tool (BCRAT) has been widely used to identify women at high risk. But it has only modest discriminatory accuracy as measured by the area under the ROC curve (AUC; 0.58~0.67 across studies). The percent mammographic density has been investigated for its potential to improve the predictive accuracy of the BCRAT. But the increase in the AUC was small. This is disappointing since the percent density is expected to be strongly predictive due to its high attributable risk. Recent work has also evaluated the added value of single nucleotide polymorphism (SNP) genotypes for predicting breast cancer risk, and the results were similarly disappointing. Recently, criteria other than AUC have been developed that are recognized to be more clinically relevant for quantifying the added value of new predictors. In this work, we adapt these criteria for evaluating the accuracy of absolute risk prediction, and use them to fully assess the added value of breast density and SNP genotypes for predicting breast cancer risk. Our results provide benchmarking values of statistical criteria for evaluating breast cancer risk predictors.