Sir,

We read with great interest the recent meta-analysis by Chung et al,1 which reached important conclusions about the effect of intravitral bevacizumab in central serous chorioretinopathy; nevertheless, some methodological issues need to be addressed concerning this meta-analysis.1

Specifically, the authors state that ‘The mean difference and SD at the 6-month follow-up were calculated from the data in the included studies.’ This seems an intriguing statement that should be further clarified by the authors to substantiate the validity of the meta-analysis and guarantee the reproducibility of their results. The included studies presented mean±SD at baseline and at the 6-month time point; the SD of the difference (with the latter representing a new measure) was not provided by the included articles.

Given that the variance of a an A–B difference inherently necessitates knowledge about the covariance (A, B), any attempt to estimate the SD of the difference would imply assumptions about the covariance; the latter is not negligible and seems of corollary importance in light of the longitudinal nature of the baseline—6-month comparison. Therefore, the authors should disclose their assumptions regarding the calculation of covariance and provide the relevant formulas with the corresponding statistical references supporting their approach; critical discussion of any limitations potentially stemming from such assumptions would be of interest.

An alternative way would be contact with the authors of each study, asking them to calculate de novo the difference and provide the meta-analysts with the exact SD data. Nevertheless, Chung et al did not provide any statement disclosing contact with the authors of individual studies.

In conclusion, thorough clarification of the methods used by Chung et al1 seems desirable, so as to further solidify the validity of their approach. Reliable calculation of variance often represents a challenging notion in the field of meta-analysis.