Due to the homogeneity of the studies included in my review I plan to have a narrative synthesis of the finding's, as opposed to a producing a meta-analysis. Though, I do hope to include effect sizes (+ confidence intervals) where possible and to produce a forest plot (without the summary measure) to give the readers a sense of the size of the effect(s). Is it acceptable to produce a forest plot with only those studies who provide the necessary information (e.g. 15 out of the 29 studies included) or do I need to have effect sizes (+CIs) for ALL studies included in order to include a forest plot?
Yes. It is just a graphical way of representing the results of those studies. As long as you don't carry out the metanalysis portion due to the heterogeneity of the studies, it should be acceptable. Also, in SR with metanalysis, if some of the studies are missing the required data to include them in the MA part, it is OK to exclude them for that part of the paper.
Just make sure that the data you put in the forest plot is the same units (don't mix OR with RR, or something like that; if necessary convert).
A forest plot is the primary end-product of a meta-analysis. A systematic review always precedes a meta-analysis. A systematic review's objective is to identify all studies that meet your research question's inclusion and exclusion criteria. These studies must be included in your meta-analysis unless there is insufficient data to calculate an effect size. In those situations, you need to document your efforts to locate the missing data. If I was a manuscript reviewer for a systematic review/meta-analysis paper and it was obvious that some studies had been intentionally left out, I would recommend rejecting the manuscript. And that decision would be unaffected about whether there was no heterogeneity or a lot of heterogeneity.
Thanks Gordon. Yes, I have made every attempt to calculate ES/CIs for those who do not report them in the paper (nor sufficient information to allow me to calculate them) through contacting authors - something which I will document in the report.
Samantha Booth Look at the Comprehensive Meta-Analysis (Professional version) software. It is capable of many different data entry formats. Go to meta-analysis.com and download the trial version.
How to calculate the Confidence Interval(95%) in forest plot for Systematic Review and Meta Analyses?
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series