Using Incident Cases to Evaluate Prognostic Markers with Time-Varying Performance

• Activity Number: 359
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Aasthaa Bansal*+ and Patrick Heagerty
• Companies: University of Washington and University of Washington
• Keywords: Prognosis ; Biomarkers ; Medical Decision Making ; Hazard Ratios ; Receiver Operating Characteristic Curve

Abstract

Prognostic markers use an individual's characteristics at a given time to predict future disease events, with the goal of guiding medical decision-making. If an accurate prediction can be made, then a prognostic marker could be used to identify subjects at greatest risk for subsequent serious adverse events, and appropriate for targeted therapeutic intervention. Often a marker is measured at a single baseline time point and used to guide decisions at multiple subsequent time points; however, the performance of the marker may vary over time as an individual's underlying clinical status changes. We provide an overview of modern statistical methods for evaluating the time-varying accuracy of a prognostic marker measured at baseline. We compare approaches that consider cumulative versus incident events and we find that an incident event approach is more appropriate for evaluating trends over time. We also compare the common approach of using hazard ratios from Cox proportional hazards regression to more recently developed approaches using time-dependent receiver operating characteristic (ROC) curves. The summaries are illustrated using a multiple myeloma study of candidate biomarkers.