Subgroup analysis - minimum requirements and conclusions to draw from them

Question

In a metaanalysis, what are the minimum requirements a subgroup analysis and what may be explained by them?

1. Is it possible to explain heterogeneity by them?
2. Do you need a certain number of studies with a minimum size for them?

Answer 1

If you divide the dataset into two or more groups and fit a model to each separately then there are disadvantages as you do not get a direct comparison between the groups and if you are using random effects models you get less precise estimates of $\tau^2$. See this post and the answer from Wolfgang and his article which he cites there

Is stratified meta-analysis more or less objective than meta-regression?

Fitting a meta-regression model would seem superior unless there are very different phenomena going on in each group but even then I would do the meta-regression first to establish that fact before I stratified.