I am working on a meta-analysis and have a question about data analysis in RevMan. Some outcomes, such as VAS show a negative result when the value increases, while others, such as pressure pain threshold (PPT), show a positive result when the value increases. Should I reverse the direction of experimental and control group labels in the forest plot? For example, in VAS, the right side represents the control group, while in PPT, the right side represents the experimental group.
Reversing the experimental and control group labels, I think, wouldn't change anything in RevMan (please correct me, Eman Mansour and future responders to this thread, if this statement is inaccurate). What matters, in my view, is you get around RevMan's reflex to treat higher values of any measure as more "adverse" or "negative."
Assuming that your desired outcome measure presentation is appropriate, consider affixing a negative sign (-) to all the central tendency (mean or SMD or median, etc.) estimates that you have for the outcomes that "behave" like the PPT you mentioned and check if it fixes the issue. If the original estimate has a negative sign, adding another negative sign will lead to its cancellation and, thus, it will have a positive sign. As a shortcut, multiply all central tendency values (not the dispersion measures such as standard deviation) that you have by -1.
I you have time, you can check my recent publication where I did a similar decision (for HDL-C):
Article Metformin extended-release versus metformin immediate-releas...
Data Supplementary material | Metformin extended-release versus m...
Good luck!
I agree with Aedrian Abrilla here, changing the direction or sides of the outcome won't help and might confuse the readers. Properly address the value of every parameter you have. I presume in your case, it is more viable to properly explain the forest plot.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis