I am performing a Meta-Analysis about a mental disease, comparing intervention vs. control groups and try to extract outcome results from published study papers.

In some clinical randomised trials the results for the main outcome, continuous in this example, are given only as "adjusted difference in mean scores" e.g. "-8,9" as well as the Confidence Interval 95% start and end and the hint that the results come from "repeated measure analysis of covariance, adjusting for baseline and clinic center with random effect model. Comparing an intervention with a control group.

  1. Is this the effect size of the intervention?
  2. What information can be extracted from the change scores?
  3. Is it possible to calculate the Mean Difference (MD) and Standard Deviation (SD) for BOTH groups (intervention & control) from this given "change score"?

(In my case the answer to the 3. question part is needed to produce forest plots in the software RevMan.)


Here's my answers to your questions:

1- I think so, but cannot be sure as I have not read the article. This look like a unstandardized difference of mean. If the authors do not explain how they calculated this change score, I would try to contact them.

2- From what you are saying of your meta-analysis, I understand that you are trying to calculate a standardized difference of means. Like Cohen's d:

$$ d = \frac{m_1 - m_2}{ s_\text{pooled}} $$

$m$ being the mean of group 1 and 2. So you have the top part. You are missing the bottom and you would need the standard deviation of each group to calculate it.

3-Point 2 pretty much answer this question. My strategy would be to contact the others to try to obtain means per group and any variance measure that you could transform into SD.

The meta-analyst work often require to contact the authors and you will see that most of them are nice people.

Another strategy would be to use another effect size type that could fit your data, but I don't know anything about that.

| cite | improve this answer | |
  • 1
    $\begingroup$ I edited your answer. Notice that CV supports $\LaTeX$ formatting and this is better way of posting formulas then as images. $\endgroup$ – Tim Apr 8 '15 at 13:50
  • $\begingroup$ Thank you for your answer. I was anticipating that I need to contact the authors and will do that when I am able to understand more of the statistics behind the Meta-Analysis. $\endgroup$ – Sebastian Apr 9 '15 at 7:15
  • $\begingroup$ I learned meta-analysis by myself last year...don't worry, you'll get there! $\endgroup$ – Emilie Apr 9 '15 at 11:24

Change scores entail a host of hidden assumptions and inefficiencies. I go into this in detail in Chapter 13 of http://biostat.mc.vanderbilt.edu/wiki/pub/Main/ClinStat/bbr.pdf .

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.