Bayesian Sample-Size Determination for Studies with Censored Cost-Effectiveness

• Activity Number: 374
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Daniel Beavers*+ and James D. Stamey
• Companies: Wake Forest School of Medicine and Baylor University
• Keywords: cost effectiveness ; Bayesian ; sample size ; INMB ; bivariate normal ; Weibull-gamma

Abstract

Cost effectiveness estimates for outcomes from survival based data can be biased when outcomes are right-censored. Censored observations have both missing effectiveness data and missing costs that are likely correlated, and the estimation of study sample size and power must be able to account for both components as well as their dependence. We propose a method for sample size and power determination in the presence of right-censored cost effectiveness, focusing specifically on the incremental net monetary benefit of treatment. The method utilizes a flexible simulation-based approach for estimation of costs and effectiveness, and we allow for two different types of data. First, we propose a model that assumes both costs and effectiveness are bivariate normally distributed. Next, we utilize a Weibull-gamma model for effectiveness and costs, respectively. We evaluate the performance of each model and compare to naïve estimates.