Hello,
I am struggling with some questions. I would like to know if it is possible to use before-after trials and randomized controlled trials in the same meta-analysis? Someone recommended me to separate the effect in between and within, is that correct?
And how could I deal with different kinds of outcomes? sometimes I have means and SD, but sometimes I just have percentages : before 5% of the people did this and after intervention 80%? Is there a way to convert percentages in another kind of measure? (odd ratios but just in one-arm interventions ?)
In a meta-analysis, you need to fix what is your main objective.
If you want to see any association, then you need to calculate the odds ratio, hence to calculate odds ratio you need a 2x2 table. So, you can include studies, those allow the comparison of the intervention group of a study relative to the comparison or placebo group.
Yes, You can include data from before-after trials and randomized controlled trials in the same meta-analysis and perform subgroup analysis.
Look for CI and p-value instead of means and SD. You can calculate the number of the population using CI and p-value easily.
If you have the number of the total population, then, of course, you can calculate the population size using the percentage.
I think you have combined two different questions
i) How do I combine two different types of outcomes (frequency vs mean difference)?
In very simple terms you can't because they are different measures. Technically probably not possible but certainly entirely incorrect.
ii) Can you combine results from before and after studies with those of randomised controlled trials?
Here I think the answer is you can (if you say had a trial and a before and after trial both reporting mean differences) but you shouldn't. There are two reasons for this - one is because the relative strength of the methods is so vastly different it is potentially very misleading to combine the results. the second is the although the outcomes appear to be the same, they are not - one is the difference between a treated ground and an untreated group (for example). The other is the mean change in a group that are treated. The assumption that these two are equivalent is unlikely to be justified.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis