Is There Heterogeneity?
Very few trials are conducted in exactly the same way. It would be unethical to repeat exactly the same trial on the same patients as you should know the result. The usual assumption in a meta-analysis is that the included trials are all similar.
Heterogeneity is the amount of variation in the trial results in a systematic review. It is important to look for this as the trials may not be very different. Heterogeneity is only measured on results (outcome). Population and intervention differences are not usually assessed in a meta-analysis.
There are three ways of measuring heterogeneity in meta-analysis:
Firstly (and easiest), look to see if most of the box and whisker icons are showing similar treatment effects. If they are all similar, this is reassuring that there is little heterogeneity.
Cochrane’s Q is a statistical test that examines heterogeneity. A p value of less than 0.05 indicates that there is significant heterogeneity.
The I 2 test is a more informative test.
Values less than 25% mean low heterogeneity, less than 50% moderate heterogeneity and greater than 75% high heterogeneity.
High values mean more heterogeneity and a less valid result.