Evaluating Incomplete Multiple Imperfect Diagnostic Tests with a Probit Latent Class Model

• Activity Number: 658
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Yi Zhang*+ and Donglin Zeng and Haitao Chu
• Companies: The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill and University of Minnesota
• Keywords: EM algorithm ; probit latent class model ; missing at random ; missing not at random ; diagnostic test

Abstract

Accurate diagnosis of a molecularly defined subtype of cancer is important toward its effective prevention and treatment. Since a gold standard may be unavailable, tumor sub-type status is commonly measured by multiple imperfect diagnostic markers. Furthermore, some subjects are only measured by a subset of diagnostic tests due to cost or compliance. In this paper, we propose a Probit latent class (PLC) model to model latent values for diagnostic tests within diseased and non-diseased groups. Unstructured correlations are used to model dependence among tests. EM algorithm is used to estimate diagnostic accuracy parameters, prevalence, and correlations. The proposed method is applied to analyze data from the NCI Colon Cancer Family Registry (C-CFR) on diagnosing microsatellite instability (MSI) for hereditary nonpolyposis colorectal cancer (HNPCC) with eleven biomarker tests. Simulations are conducted to evaluate the small-sample performance of our method.