Estimating required information size and quantifying diversity in a meta-analysis.
Jesper Brok, Copenhagen Trial Unit
Christian Gluud, Copenhagen Trial Unit
Kristian Thorlund, Copenhagen trial
*Jørn Wetterslev, Copenhagen Trial Unit
Keywords: meta-analysis, heterogeneity, sample size
There is increased awareness that meta-analyses need to attain an information size to detect or reject an anticipated intervention effect. Information size calculations need to consider the total variability in a meta-analysis to preserve the desired error rates. We derive a correction factor that readily yields the expression for the required information size under any meta-analysis model. We also obtain a central measure of diversity in a meta-analysis, defined as the percentage of total variance reduction when the model of pooling is changed from a random-effects model to a fixed-effect model. We show that this measure is an unbiased quantification of the percentage of variation due to inter-trial variability rather than sampling error in a meta-analysis.We prove that diversity is greater than heterogeneity in all random-effects models. The properties and capabilities of diversity and the traditional measure for heterogeneity are illustrated in 7 meta-analysis