The Bayesian approach to Meta-analysis: a Case-study

I. Verdinelli, K. Andrews, K. Detre and P. Peduzzi


We consider the meta-analysis study of Yusuf et al. (1994) that used data collected by the Coronary Artery Bypass Graft Surgery Trialist Collaboration group from seven randomized clinical trials to compare coronary artery bypass graft surgery versus medical treatment.

Among other characteristics of the studies involved, Yusuf et al. examined a particular aspect, regarding the relation among mortality rates and gravity of the baseline risk characteristics of the patients. As an alternative, we propose here the use of Bayesian meta-analysis for the same classification of patients that they considered. The method of stratification, denoted ``Step-wise Risk Score'', includes clinical and angiographic baseline characteristics of patients gif and was developed in Yusuf et al. (1994).

In this paper, we are mainly concerned in a detailed revision of the Bayesian methodology for meta-analysis studies. We will show the flexibility of the Bayesian method, examine the basic assumptions on which the method relies, and illustrate on the availability of the relevant software while examining this special case study.

Here is the full postscript text for this technical report. It is 285745 bytes long.