Meta-analysis: an update.

Abstract

A fairly new type of research, termed meta-analysis, attempts to analyze and combine the results of previous reports. In 1992 we updated our 1987 survey of 86 meta-analyses of randomized control trial reports in the english language literature with an additional 78. We evaluated the quality of these meta-analyses using a scoring method that lists 23 items in six major areas: study design, combinability, control of bias, statistical analysis, sensitivity analysis, and application of results. Of the 23 individual items, the mean number satisfactorily addressed was 7.63 +/- 2.84 (mean +/- S.D.) for 40 papers published from 1955 through 1982, 6.80 +/- 3.86 for 66 papers published from 1983 through 1986, and 11.91 +/- 4.79 for 58 papers published from 1987 through 1990 (F = 31.3, p < .001). We noted that methodology has definitely improved since our first survey of meta-analyses, but an urgent need still exists for a better search of the literature, quality evaluation of trials, and a synthesis of the results. Recently, meta-analysis has expanded to cover non-randomized studies, including evaluation of diagnostic tests and pooling of epidemiologic studies. There is growing concern for standards, and several methodologic issues remain unresolved.