I am conducting meta-analysis on complication outcomes related to a rather new surgical technique. Hence, published studies contain rather small study samples. Is there a minimum to the number of participants in a study for it to be eligible for inclusion in meta-analysis?
Thank you
i trhink you will find an answer for your question in Chapter 16: Special topics in statistics(chocrane handbook for interventions systematic review)
Cochrane Handbook for Systematic Reviews of Interventions
http://handbook.cochrane.org/
Hey Christian!
Short answer: You should include studies with small sample size, if they fullfill your defined requirements (if somebody later says that you don't need these studies for whatever reason you can still exclude them).
Long answer: There is one reason why studies with small samples used to be exluded: The higher risk for publication bias (small study bias) when conducting a meta-analysis. Smaller studies are easier to conduct and hence there will be more studies with interesting result. And studies with interesting results have a higher chance to get published. Accordingly, you have to expect more small studies showing favourable effects. With modern statistical methods you can detect small study bias and there is no need to exclude these studies anymore.
In my opinion very large studies are sometimes more problematic than small studies, because in the former you will have significant results even when they effct is practically irrelevant (this is the reason why you should never interpret the p value alone).
Maybe you are interested in this short presentation:
http://skin.cochrane.org/sites/skin.cochrane.org/files/uploads/CSG-COUSIN_March%202015_M%20Grainge.pdf
Hi Christian,
Your analysis should account for small sample sizes as well. You're probably going to want to conduct a random-effects analysis, and thus, the weights will be determined by he individual study variance/SE. I've seen many meta-analyses published with small sample sizes, so I don't think this will be a problem for you. Goodluck!
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Applied Statistics
- Survey Design