Meta-analysis: science or religion?

Meta-analysis of clinical trials has assumed increasing importance in recent years as a means of reaching conclusions regarding the safety and efficacy of treatments evaluated in individual trials. This article discusses impediments to meta-analyses and enumerates needs that should be satisfied to enhance their usefulness. Key among the needs are improved means of identifying trials prior to publication via prospective registration of them and vocabulary and nomenclature for identifying reports of meta-analyses in the published literature.