[R-meta] Clarifications about pooling of continous outcomes when geometric means are present

[Viechtbauer, Wolfgang (NP)]

Dear Alberto,

No, as far as I understand your question, CVLN and SDLN are not measures that are relevant for your meta-analysis. One would have to use the equations given in the article you referenced.

Best,
Wolfgang

> -----Original Message-----
> From: R-sig-meta-analysis <r-sig-meta-analysis-bounces using r-project.org> On Behalf
> Of Alberto Enrico Maraolo via R-sig-meta-analysis
> Sent: Tuesday, December 12, 2023 13:19
> To: r-sig-meta-analysis using r-project.org
> Cc: Alberto Enrico Maraolo <albertomaraolo using mail.com>
> Subject: [R-meta] Clarifications about pooling of continous outcomes when
> geometric means are present
>
> Dear all,
>
> I would like to have a help if possible on the pooling of continuous data,
> specifically Pk/Pd outcomes in the presence of means, medians, and geometric
> means.
> The goal is to have a (standardized) mean difference between patients with two
> different conditions.
> Medians under certain assumption can be converted to means and the conv.fivenum
> functions serves the purpose.
>
> The problem is the pooling of means with geometric means, since the latter
> cannot be re-converted to arithmetic means without raw data.
>
> So, as suggested by the Cochrane Handbook, “a meta-analysis may be then
> performed on the scale of the log-transformed data”; the Handbook describes how
> to derive the natural logs from the geometric means and how to compute standard
> deviations of the log-transformed data from confidence intervals.
> Since “log-transformed and untransformed data cannot be mixed in a meta-
> analysis”, the issue is to go from raw to transformed data for standard means,
> and the suggestion is to follow the formulas from the paper of Higgins 2008.
>
> My question is: can I follow the chunks suggested in this page:
>
> https://wviechtb.github.io/metafor/reference/escalc.html?
>
> I mean the section 3a (“Measures for Quantitative Variables”), by resorting to
> "CVLN" for the log transformed coefficient of variation and to "SDLN" for the
> log transformed standard deviation in each group, in order to have log-
> transformed data to be pooled with the natural logs from the geometric means.
>
> If there is some line of code also for faster handling of geometric means in
> order to derive natural logs it would be nice as well.
>
> Warm regards,
> Alberto
>
> Alberto Enrico Maraolo, MD, MSc (Antimicrobial Stewardship, Evidence Synthesis),
> FESCMID
>
> Infectious Diseases Specialist, Member of the Steering Committee of SIMIT (ID
> Italian Society)
>
> Cotugno Hospital, AORN dei Colli, Naples, Italy
> mail: mailto:albertomaraolo using mail.com
>
> Alberto Enrico Maraolo, MD, MSc (Antimicrobial Stewardship, Evidence Synthesis),
> FESCMID
>
> Specialista in Malattie Infettive, Consigliere Nazionale Direttivo SIMIT
> (Società Italiana di Malattie Infettive e Tropicali)
>
> Dirigente Medico, AORN dei Colli - Ospedale Cotugno, Napoli