I'm performing meta-analyses of a continuous measure based on clinical studies. For one subgroup of interest, there are only two studies available: one with n=5 patients and one with n=1 patient. Ideally, I would like to pool them together.
I am weighting the meta-analyses by SD/SQRT(n). I made an assumption that the standard deviation of the n=1 study would be the same as the average SD across the other studies, and then performed a fixed effects meta-analysis.
But I'm questioning whether this is a reasonable approach, or would it be better to leave these estimates unpooled?