In a subgroup analysis, you can combine or leave out study data in all sorts of ways to explore how your treatment works in different subpopulations.
You might do a subgroup analysis to compare:
Men versus women
Age categories (old vs young, or break down further)
Severity of disease (Early or late stages)
Patients with other serious conditions (your disease of interest as well as diabetes, hypertension or obesity)
Patients also taking another medication (or receiving another type of treatment)
Study design features (blinded or not, different levels of dosing, etc)
However, this type of analysis is only possible if studies report their information by these different categories.
Next, take a look to see if any one study is having a dramatic effect on your average result.