[R-meta] Questions about extracting the p-value in subgroup analysis

[Dr. Guido Schwarzer]

Nick,

I am not sure which p-value(s) you are interested in and which website you are referring to (I couldn't find the statement in the current help pages of R package meta). So, the following might not fully answer your question.

As Ishtiaq wrote, the argument test.subgroup.random can be used to print the result of the test for subgroup differences. However, this argument is TRUE by default and, accordingly, the result is part of the R output in your email (see "Test for subgroup differences (random effects model (HK)):").

If you are interested in the p-values for the test of effect in individual subgroups, you can extract this information using

update.meta(m.gen, test.subgroup.random =T , subgroup = ins, tau.common = FALSE)$pval.random.w

or

m.gen.ins <- update.meta(m.gen, test.subgroup.random =T , subgroup = ins, tau.common = FALSE)
m.gen.ins$pval.random.w

These p-values are not part of the printout as this could result in "cherry picking", i.e., identifying subgroups with a significant treatment effect. This point is discussed in the Cochrane Handbook under the topic of "within-subgroup inferences" (see https://training.cochrane.org/handbook/current/chapter-10#section-10-11-3-1).

Best,
Guido