Multilevel Bayesian models of survival and zero-inflated longitudinal patient-reported outcomes

• 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): Laura A. Hatfield*+ and Mark E. Boye and Michelle Denise Hackshaw and Bradley P. Carlin
• Companies: University of Minnesota and Eli Lilly and Company and Eli Lilly and Company and University of Minnesota
• Keywords: Bayesian ; latent variables ; zero-inflated beta ; survival ; hierarchical models

Abstract

Malignant pleural mesothelioma (MPM) is a fatal pulmonary disease associated with asbestos exposure. Previous studies have established pemetrexed (Alimta(R)) plus cisplatin as effective in prolonging survival. We build hierarchical Bayesian models for longitudinal patient-reported outcomes (PROs) and progression-free survival (PFS). Our methods combine zero-inflated beta regression for the PROs with proportional hazards Weibull regression for the right-censored PFS values using latent random variables. The results indicate significant treatment effects of decreased PRO severity over time and increased PFS. We also observe 1) correlation between the individual probability of non-zero PRO and severity of non-zero PROs, 2) and associations between the PRO latent variables and PFS.