Bayesian Modeling of Correlated Binary Data from the Cryotherapy for Retinopathy of Prematurity (CRYO-ROP) Study

• Activity Number: 30
• Type: Contributed
• Date/Time: Sunday, August 6, 2006 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Claudia Pedroza*+ and Betty Tung
• Companies: The University of Texas School of Public Health and The University of Texas School of Public Health
• Keywords: Bayesian multilevel modeling ; correlated binary data ; multicenter clinical trials ; random effects models ; mixed effects

Abstract

Data arising from multicenter clinical trials can be complex to analyze due to the possibility of correlation between observations within a clinical center. If the correlation is ignored, variance estimates could potentially be incorrect. In this talk, we present a Bayesian multilevel model which accounts for correlation within a study center when estimating the incidence of ROP. We show that this methodology allows for great flexibility both in the choices of distributions and also in the structure of the model, i.e. random effects for the centers and/or patients. Methods presented will be applied to a data set from the CRYO-ROP study. This data set consists of a binary outcome (ROP) and baseline variables (e.g., gestational age, race, birth weight, gender). We will also discuss the ramifications that a random effects study can have on a multicenter clinical trial.