A Bayesian Model for Misclassified Binary Outcomes and Correlated Survival Data, with Applications to Breast Cancer

• Activity Number: 163
• Type: Topic Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Min Yi*+ and Sheng Luo and Xuelin Huang and Kelly K. Hunt
• Companies: The University of Texas Health Science Center at Houston and The University of Texas School of Public Health and MD Anderson Cancer Center and MD Anderson Cancer Center
• Keywords: Bayesian ; Logistic regression ; Misclassi cation ; MCMC ; Breast tumor relapse ; WinBUGS

Abstract

Breast cancer patients after breast conservation therapy may experience breast tumor relapse. The relapse is classified into two distinct statuses by method subject to misclassification errors. Additionally, some observed survival information (e.g., time to relapse and time from relapse to death) is strongly correlated with the relapse status. This article presents a Bayesian approach to modeling the misclassified binary breast tumor relapse status and the correlated survival information. Our methods are motivated by, and applied to, the dataset of breast carcinoma. There are two important innovations in our model: (1) to utilize the additional survival times correlated with the relapse status to improve the parameter estimates; (2) to provide tools to address the correlation between two diagnostic methods conditional on the true relapse status.