Hello everyone,
I am conducting a meta-analysis to find if x has superior effect than y on the outcome. I used the SMD method and I got a significant moderate heterogeneity of a value I2: 52%
should I do the subgroup analysis to find out where the heterogneity came from?
I have only 16 studies and if i want to subgroup by intervention types (EG), I dont have an equal number of studies for each group. also when I wanted to subgroup by health status, I had 6 studies that used persons with MS, 1 study that used persons with stroke, 1 study used healthy population, 3 studies with MCI etc... so the number of studies is not equal among the subgroups. can I still do that?
An I² value of 52% suggests moderate heterogeneity among the studies. While there isn't a strict threshold for what constitutes acceptable heterogeneity, values around 50% are relatively common in meta-analyses.
you can conduct subgroup analysis for one group of studies based on major characteristics while not subgrouping the other studies.
No minimum number but generally a meta-analysis with < 10 studies will not be robust to the findings of a new study.
As is generally the case, the more data availibile the greater confidence in the results.
https://www.researchgate.net/post/What_is_the_minimum_number_of_studies_in_each_category_required_for_a_multiple_subgroup_analysis_to_be_viable_in_a_meta-analysis
A similar question to yours has been posted to ResearchGate & answered in 2021.
In summary, there is no unanimous consensus on the minimum number required for subgrouping. It would help if you tried addressing the heterogeneity by using a random effects model (instead of a fixed effects model) and using a leave-one-out test.
If those two failed, try your luck with a subgroup analysis.
- Badges
- Science topic
- Similar topics
- Correction