On 2017 Apr 03, Jeruza Neyeloff commented:

The general suggestion raised by Holman is appropriate: when dealing with different data sets and adjusting the spreadsheets provided, we recommend seeking help from a statistician or someone experienced in meta-analyses.

Particularly, on the cited steps, the additional file indeed does not place a lower bound on I², and depends on the researcher to know that negative values are not possible. The paper should also have more clearly described to not use a recalculated I² statistic, since it should not change according to the model chosen (fixed or random).

As stated in the article, one of the main advantages of conducting the analysis with a step-by-step approach is a better understanding of the complete analysis process and of the formulas used. However, specific-purpose software or statistical packages will yield more stable results and usually require less data manipulation by the end user. In my research group, we tend to use the spreadsheet as a teaching tool (or to generate forest plots for presentations), and to use the R Meta and Metafor packages for more complex analyses.

We hope the article, by providing a view of how the calculations involved in these analyses work, continues to aid students and researchers interested in meta-analyses.

On 2015 Dec 29, Rebecca Holman commented:

A clinical researcher asked me for help in using the Additional file 1: Meta-analyses and forest plots in MS Excel provided by the authors of this paper. The file had produced a strange result for this researcher's meta analysis. My concern is around step 8 of "Steps in analyzing data and producing a forest plot". The authors definition of I² is slightly different to that presented previously (see for example Higgins JP, 2003). The authors of the current manuscript do not place a lower bound of 0% on the value of I² . Hence, in some meta-analysis data sets, the additional file can result in a value of I² of less than 0%. In addition, the authors have not placed a lower bound of 0 on the value of Q-(k-1) when calculating v in step 9B "Random effects model". In some meta-analysis data sets, this can lead to negative values of v (cell M16 in the additional file). This can lead to incorrect results for a random effects based meta-analysis. When viewed in conjunction with previous comments on this paper, I feel that researchers should exercise caution when using the formulas in or additional Excel file to this paper to perform calculations or obtain figures for a meta-analysis.

On 2015 Dec 29, Rebecca Holman commented:

A clinical researcher asked me for help in using the Additional file 1: Meta-analyses and forest plots in MS Excel provided by the authors of this paper. The file had produced a strange result for this researcher's meta analysis. My concern is around step 8 of "Steps in analyzing data and producing a forest plot". The authors definition of I² is slightly different to that presented previously (see for example Higgins JP, 2003). The authors of the current manuscript do not place a lower bound of 0% on the value of I² . Hence, in some meta-analysis data sets, the additional file can result in a value of I² of less than 0%. In addition, the authors have not placed a lower bound of 0 on the value of Q-(k-1) when calculating v in step 9B "Random effects model". In some meta-analysis data sets, this can lead to negative values of v (cell M16 in the additional file). This can lead to incorrect results for a random effects based meta-analysis. When viewed in conjunction with previous comments on this paper, I feel that researchers should exercise caution when using the formulas in or additional Excel file to this paper to perform calculations or obtain figures for a meta-analysis.