Incorporation of Survival Information in Diagnostic Test Evaluation

• Activity Number: 167
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Chang Xu*+ and Dean Billheimer and Bonnie Lafleur
• Companies: Ventana Medical Systems and University of Arizona and Ventana Medical Systems
• Keywords: Bayesian approach ; diagnostic tests ; latent class models ; sensitivity ; specificity

Abstract

Typical measures of diagnostic accuracy for a new test, such as sensitivity and specificity, are non-identifiable in the absence of a gold standard. Hui and Walter (1980), Walter (1984), and Walter and Irwig (1988) described methods for estimating accuracy using latent class models. Johnson, Gastwirth, and Pearson (2001) described methods to overcome non-identifiability by use of two tests, and two populations with different underlying prevalences. We extend these ideas using Bayesian estimation methods with one population and three tests, where one of the tests is defined by a survival function based on therapeutic response. We demonstrate that diagnostic accuracy can be estimated using a combination of diagnostic tests and survival information, and that the Bayesian solutions are robust to prior specification.