Hi, I'm hoping someone can point me in the right direction. I'm using CMA v2.2 and am conducting a meta-analysis of clinical trials (drug vs. placebo), using post-treatment means to calculate Hedges's g between conditions.
A reviewer wrote: "you don't control the post-treatment data for baseline. This can be done easily in CMA (see the book by Borenstein et al 2009)."
I have combed through Borenstein et al.'s "Introduction to Meta-Analysis" (2009) and for the life of me I cannot figure out how to control for pre-treatment (baseline) scores. Any advice you can give would be much appreciated.
My solutions are: 1) to check whether the original study assesses the baseline balance of pre-treatment scores, and 2) put the pre-treatment data in the RevMan to see whether there is significant difference between two groups. Hopefully it will help you.
I would compute the standardized mean difference with pretest scores and I would test if the effect size reach statistical significance.
Best!
Testing for statistical significance for the differences in baseline means between the two groups is not appropriate, because statistical testing is an inferential process which goes beyond the data and confounding (one condition of which is between-group imbalance in the baseline score) is the property of the data. Also, the result of statistical tests depends on a sample size. If a trial is too large, even negligible difference in baseline score between the groups will be statistically significant. My suggestion would be to calculate mean differences, i.e., mean change (between baseline and follow-up values: post treatment minus pre-treatment mean values) for each study arm (drug and placebo) and compare them in your meta-analysis. So, you would input in CMA mean change from baseline (post munus pre) instead of post-treatment mean values for both groups. This would control for your baseline between-group imbalance in the score. Hope this helps. Best luck.
lots of helpful thoughts above! I agree, if you've got access to differences of differences and/or baseline and final values, there are a few nice effect size options in CMA which will allow you to go a little further than working with only the final means here that should make your reviewer happy.
I have a similar question, but using R (metafor ). Any suggestions as to my options in R if I'd like to include baseline data when comparing treatments vs. controls?
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis