Please see attached. More events are a good result. The intervention - FGDM - has proportionally more events than the control, but it appears to be the opposite in the forest plot?
It is only one study included in the second subgroup. Meta-analysis is performed if you have two or more combinable studies.
Dear Tony,
If you check the bottom of the forest plot, you will see 'Favors FGDM' on the left, and 'Favors control' on the right. This means that, yes, the result supports that FGDM is favored compared to the control. I guess the reason that you see it in the opposite direction to what you'd expect is that generally speaking, more events are unfavorable. Think about medicine and particularly death, more events are a bad outcome.
Thanks Robert. Yes, in this case more events (more family reunifications) are a good thing. The forest plot is suggesting more is bad. I can't see how to sort this in Revman. the terms used at the bottom of the plot are misleading, i can swap them around, but the data above goes with them. Is there no "more is better" option (like there is in Covidence for example)
Hi! How did you go about this conundrum, sir? I am facing the same issue in my meta-analysis.
I think to get around the positive effect showing up as negative, you must "rephrase" the values in terms of the opposite (i.e., proportion of families that DID NOT reunite instead of proportion of families that reunited). It sounds bad reading the forest plot that way, but I think in those terms, FGDM in your example will be favored.
Another issue altogether is in terms of continuous outcomes wherein the larger value is "better." Thanks!
An easy fix for this that has worked for me is to go into the settings for the analysis and under "graph" where you would do things like change the scale for your forest plot you can just change the left and right graph labels (ie change which side favours experimental and which favours control)
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis