When does combining markers improve classification performance and what are implications for practice?

• Activity Number: 190
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Aasthaa Bansal*+ and Margaret Sullivan Pepe
• Companies: University of Washington and Fred Hutchinson Cancer Research Center
• Keywords: two-phase design ; biomarker combination ; diagnosis ; ROC ; classification

Abstract

When an existing standard marker does not have sufficient classification accuracy on its own, new markers are sought with the goal of yielding a combination with better performance. The primary criterion for selecting new markers is that they have good performance on their own and preferably be uncorrelated with the standard. We investigate the increment in performance that is possible by combining a novel marker with a moderately performing standard marker under a variety of biologically motivated models for their joint distribution. We find that an uncorrelated continuous marker with moderate performance on its own usually yields only a minimal improvement and identify other settings that lead to large improvements. Our results suggest changing current strategies for identifying markers for possible combination. We find that given studies with large numbers of subjects, a broadened strategy can be more fruitful than the standard strategy that selects panels of markers as candidates based on their marginal performance