I do not understand what you mean by "P value analysis between the treatment modalities for each main outcome". Are you talking about doing subgroup meta-analyses or the old-fashioned, improper vote counting procedure (i.e.., counting the numbers of studies with significant and nonsignificant findings)?

That was the question asked by one of the Reviewers. I do not think he meant to perform a subgroup meta-analysis or a vote counting procedure.

The situation of my meta-analysis is more or less the following:

- papers 1, 2, 3 proposed treatment X, and provided technical success and mortality rates

- papers 4, 5, 6 proposed treatment Y, and provided technical success and mortality

- papers 7, 8, 9 proposed treatment Z, and provided technical success and mortality

So, I simply synthesized outcomes in my one-group meta-analysis:

- technical success and mortality rates (estimated proportions) for treatment X

- technical success and mortality rates (estimated proportions) for treatment Y

- technical success and mortality rates (estimated proportions) for treatment Z

Now, by saying "I suggest to perform a P value analysis between the treatment modalities for the technical success", I think the Reviewer is asking me to "statistically" compare treatment X with treatment Y and Z (and provide a P). However, X, Y and Z have never been compared within a single paper (no comparative studies available in literature).

Do you think it is possible (and appropriate!) to perform such analysis?

I hope it is clearer, now.

Thank you

Yes, I think the three treatments (assuming they are treating the same condition) can be compared in your analysis. That is the purpose of a subgroup analysis. For example, see Table 1 of our attached paper in which we compare the effectiveness of three drugs (i.e., diclofenac vs. Ibuprofen vs. naproxen) even though none of the the 3 drugs were compared in a given study. Groups of studies are compared against other groups of studies. The p values in the last column of our Table 1 are analogous to those obtained when running a one-way ANOVA; they tell you whether there is a statistical difference among your groups of studies. If there is a significant overall p-value, you can then do pairwise comparisons to determine which groups of studies are different from another.