Cluster randomised trials are usually adjusted for the cluster effect. Is it essential to further adjust the cluster effect during meta-analysis?
If you have the sufficient depth of data to do so for the other studies, you can certainly try. Subsequently, you can present unadjusted and adjusted meta-analysis results and decide what will be your study's main result.
However, it is more likely than not that you will not have the sufficient (depth of) data to do this, and I think it is reasonably fine to combine the adjusted results of the report in question in the generally-unadjusted pool of other studies. Just specify that this is the case in your results, and discuss whether this discrepancy causes a predictable effect in the results (i.e., will this underestimate or overestimate the impact of the intervention/phenomenon, or the direction cannot be ascertained on your end?).
@Solomon, can you please explain the logic behind further adjustment.
Aedrian Abrilla I agree with you. In my understanding, further adjustment of the clustering effect is not mandatory if the primary studies have already adjusted for that. However, as you mentioned, if sufficient data are available, the clustering effect can be further adjusted and both the unadjusted and adjusted meta-analysis can be presented.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis