Hello everyone,
I am conducting a meta-analysis to evaluate the efficacy of various PTSD treatments (e.g., Prolonged Exposure Therapy, CPT...) using Comprehensive Meta-Analysis (CMA). For each treatment, I have calculated the within-treatment effect size (Hedges' g) based on pre- to post-treatment or pre- to follow-up changes across multiple studies. I also have the corresponding Q-statistics for each treatment group.
Now, I would like to compare these treatments to determine if there are significant differences in their efficacy. Specifically, I am wondering:
- Should I focus on comparing the total Hedges' g scores for each treatment?
- Or should I compare the Q-statistics instead?
- Alternatively, is there a more appropriate method (e.g., subgroup analysis, meta-regression) within CMA to assess differences across treatment types?
I would greatly appreciate any guidance or suggestions on how to approach this comparison in CMA, and also how to do so.
Thank you!
Dear Gelman,
Total Hedges scores for each treatment need to be compared. In addition, for categorical variables, subgroups are required to assess differences between treatment types; for continuous variables, meta-regression is required. Finally, I recommend using R software if possible. Regards.
Dear Gelman,
Since Hedges'g is a standardized effect size you can compare its values between treatment arms, yet it could be challenging when the values of two treatment arms are close and overlapping. Subgroup analysis of each treatment and comparison of their effect size with statistical tests such as the Q test may prove useful in such cases.
To do so in CMA define a comparison column and fill it in with the corresponding treatment for each study. It should look like the figure I've attached (Figure 1). Once you run the analyses it will produce a forest plot of all the studies and you need to conduct a subgroup analysis by the type of comparison and you should get sth similar to Figure 2. Comparison of each treatments effect size for your meta-analysis can be done using method 2 or method 3 explained in the CMA handbook (https://meta-analysis.com/download/Meta-analysis%20Subgroups%20analysis.pdf?srsltid=AfmBOor06LWKrtTPnGxpYtyoiWtshTdVDO54zjpPDkouTKM8H2PF1Hq1). The newer CMA software may have implemented these codes into the software, yet I am not aware of such an update and most likely you will have to do it manually using Excel.
If the studies are multi-armed rather than having a single arm, I would recommend a network meta-analysis as it will allow you to compare and rank the treatments, yet it is not possible with CMA.
- Badges
- Science topic
- Similar topics
- Publishing