Identifying Patients with Aggressive Prostate Cancer Using Prostate-Specific Antigen (PSA) Profiles

• Activity Number: 403
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Feng Gao*+ and Chengjie Xiong and Ling Chen
• Companies: Washington University School of Medicine and Washington University School of Medicine and Washington University School of Medicine
• Keywords: Latent class analysis ; joint modeling ; survival data ; longitudinal data ; prostate cancer

Abstract

Mortality for prostate cancer (PCa) continues to be significant in the US with an estimated 3% of American men destined to die of the disease. While the introduction of PSA-based screening has contributed to a significant increase in the number of prostate cancer cases, this also leads to dramatic over-diagnosis and over-treatment for some and ineffective treatment for others. Therefore, the critical question nowadays is not who has PCa, but who is at risk for dying of disease. In this talk we propose a latent class model to uncover subpopulation structure for both PSA trajectories and the intermediate clinical events (adjuvant radiotherapy, hormone therapy, PSA recurrence, etc.) in a longitudinal data with highly unbalanced and irregularly measured PSA. The patterns of PSA trajectories can be viewed as latent classes in a finite mixture where the membership in latent classes is modeled with a multinomial logistic regression. The class-specific PSA trajectories are described by a linear mixed model and the probability of disease outcome (pathological recurrence or death due to PCa) within a latent class is estimated via a proportional hazard model.