Removing studies from subgroup analysis

Question

Im currently working on a meta-analysis of randomized controlled trials, which examine the use of a certain graft in certain surgical intervention. In my pre-specified protocol, I planned to perform a subgroup analysis based on the thickness of the graft.

So the subgroup analysis was planned to be having two groups only:

Group A: Intervention with full-thickness graft

Group B: Intervention with thinned or half-thickness graft

However, I noted that there are 5 out of 17 studies that failed to provide if the graft was thinned or not.

My question is:

Should I

1. Assume that the graft was not thinned and include them in Group A
2. Simply remove these studies from the subgroup analysis
3. Add a third subgroup labelled as "unknown thickness"

Which one makes the analysis properly conducted in this case?

Answer 1

Subgroup analyses are for hypothesis generation. In this case you are doing an indirect comparison of full vs. thinned/ half-thickness grafts to determine if there is evidence of a difference in outcome measures between full and half-thickness grafts. Since you don't know the thickness in five of the studies then you don't include them in the analysis. You could do sensitivity analyses to see if the results would be different if you assumed they were all full-thickness or all half-thickness.