Meta-analysis of clinical trials based on censored end-points: simplified theory and implementation of the statistical algorithms on a microcomputer.

Meta-analysis is increasingly being used in clinical research. While the meta-analytical algorithms for pooling the results of studies using non-censored end-points are now sufficiently standardized, the management of trials based on censored end-points is still controversial, and specific algorithms of meta-analysis are still needed (for example, censored end-points are commonly used in survival studies of cancer patients). In the present article we review the various algorithms that have thus far been utilized to perform a meta-analysis based on censored end-points, and we propose a new and original approach that combines two of the existing algorithms. Our approach is designed to formulate any meta-analysis of survival studies in terms of the calculation of an odds ratio. The meta-analytical odds ratio proposed here is called the 'log-rank' odds ratio of meta-analysis. A microcomputer program is described that implements the various methods of meta-analysis that can handle clinical trials based on censored end-points. Some examples are presented to illustrate the use of the program.