Multiple independent clinical trials are typically conducted at various research or medical centers to test the efficacy of a particular treatment or drug. The results of these clinical trials are then pooled by a somewhat controversial procedure called meta-analysis. According to a series of debate on this statistical issue in the British Medical Journal (Egger et al., 1997), the term meta-analysis was coined in 1976 by the psychologist Gail Glass. Meta-analysis was later revisited by medical researchers for use primarily in randomized clinical trials. A useful definition was given by Huque (1988): “A statistical analysis that combines or integrates the results of several independent clinical trails considered by the analyst to be combinable.”

In a meta-analysis, different statistical methods may be used to combine the data. However, a typical procedure will use a weighted average of the results, with the larger trials having more influence over the smaller ones. The more complicated problems with the use of this type of analysis are, however, the parts that deal with identifying the relevant studies and with expressing the individual results in a standardized format.

Despite its widespread use, meta-analysis continues to be a controversial technique unacceptable to some statisticians. This is because the pooling of results from clinical trials that involved independent but varied study protocols is simply statistically unsound. Epidemiologists continue to take advantage of meta-analysis because a single study often cannot detect or exclude with certainty a modest, yet clinically relevant, difference in two treatment effects.