Study Designs for Evaluating Personal Risk Models

• Activity Number: 606
• Type: Invited
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Alice S. Whittemore*+
• Companies: Stanford University
• Keywords: Personal risk models ; two-stage sampling ; ovarian cancer

Abstract

To facilitate targeted prevention strategies and help patients weigh the costs and benefits of preventive interventions, we need statistical models that provide accurate and precise projections of a person's risk of developing a given adverse health outcome. Such models must be evaluated using cohort data on individuals at risk of the outcome. The ideal evaluation would assemble covariates and assign risks to all subjects in a large cohort, and then relate assigned risks to subsequent outcomes. However when the covariates include genetic factors, such covariate assessment may be too costly for feasibility. In this talk I will describe new cost-effective designs for evaluating the added value of covariates, such as personal genotypes. The designs will be illustrated using cohort data from the California Teachers Study (CTS) and population-based case-control data on ovarian cancer.