Latent Class Regression Model for Assessment of Diagnostic Tests in the Absence of a Gold Standard, with Accommodation for Covariate Information

• Activity Number: 393
• Type: Invited
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Zheyu Wang*+ and Xiao-Hua Andrew Zhou
• Companies: University of Washington and University of Washington
• Keywords: Diagnostic accuracy ; Gold standard ; Latent class models ; Biomarker ; Alzheimer's disease

Abstract

Estimating the accuracy of a diagnostic test requires existence of a gold standard. However, in many research setting, gold standard evaluation may be too costly or unethical to obtain. For example, the definite diagnosis of Alzheimer's disease (AD) on a patient cannot be established until a patient has died and a brain autopsy has been conducted. This issue is becoming more common and pressing with the growing interest and emphasis on preclinical diagnosis and prevention. Moreover, there is a need to include and examine patients' characteristics that may affect disease prevalence or test performance. In this talk, we introduce a new latent class regression model for assessing the accuracy of diagnostic tests in the absence of a gold standard. This new model also allows estimation of covariate-specific diagnostic accuracy of the tests. We also discuss the issue of identifiability in a latent class model. We apply the propose method to a real-world example on assessing the accuracy of the CSF Aß and tau in detecting early AD features without a gold standard.