On variability of effects for a metaanalysis of rheumatoid arthritis radiographic progression

Variance of an outcome that is the topic of a metaanalysis may be estimated in a variety of ways. Variances are also not used in the same way in every metaanalysis. The metaanalysis may be of a single measure expressed in its original units, or of several such measures simultaneously, or it may be done using unitless effect sizes. We present some observations and practical recommendations on the estimation and use of variances relating to these several approaches to metaanalysis, as they may apply in a metaanalysis of radiographic progression in rheumatoid arthritis.