[R-meta] Three-level meta-analysis using netmeta

[Viechtbauer, Wolfgang (SP)]

Dear Natan,

Gerta, Guido, and colleagues (i.e., the netmeta authors) should really be the ones to address your question about netmeta. But I am once again procrastinating with my own work by answering questions on this mailing list and once I get going, I am hard to stop, so let me say just a few words about this.

netmeta is indeed a very nice choice for a network meta-analysis as it simplifies some of the intricacies involved, compared to the more manual process required when fitting such models with metafor. However, you can run network meta-analyses also with metafor, if you know how to set things up (and are willing to do a bit of 'post-processing' of the results). So that is the disadvantage of using metafor for these purposes. The advantage is that it will allow you to expand on the underlying models in essentially arbitrary ways, including the addition of additional levels and/or random effects to capture further dependencies. Also, one can run meta-regression analyses in this context, which I believe is not (yet) possible with netmeta (although this is something that may be in the works?).

If the more complex data structure in the present context require the specification of corresponding models, then I do not see why using metafor is actually a problem; quite the opposite.

Best,
Wolfgang

>-----Original Message-----
>From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces using r-project.org] On
>Behalf Of Natan Gosmann
>Sent: Wednesday, 17 March, 2021 9:25
>To: r-sig-meta-analysis using r-project.org
>Subject: [R-meta] Three-level meta-analysis using netmeta
>
>Dear all,
>
>Our group recently submitted for publication a three-level network
>meta-analysis that was sent to peer review and one aspect of concern that
>was raised during the review process is the fact that we performed our data
>analysis using metafor (through methods described by Wolfgang elsewhere).
>Reviewers stated that moving to netmeta might be preferable, given that it
>is specifically designed for network meta-analyses. Although we recognize
>the great value of netmeta for network meta-analyses, we are not sure that
>currently the package is able to perform three-level meta-analyses (even
>after carefully reading its detailed manual). Can we perform our data
>analysis (using the three-level/network approach) using netmeta? If yes,
>any suggestion or reference with any guidance on how to do it would be of
>great help.
>
>Best,
>Natan Gosmann