Heterogeneity
What is heterogeneity?
Looking at the forest plot, one can see that the results for the individual studies are all different. This difference is called ‘heterogeneity’. Heterogeneity may be due to differences in the study protocols, such as differences in doses of treatments, or study populations. However, heterogeneity may exist purely by chance too. Heterogeneity is an important factor to consider in all meta-analyses, since too much heterogeneity can invalidate the pooled results.
Heterogeneity is usually assessed either visually using a forest plot or more formally using statistical tests, such as the Cochran’s Q test. However, more recently a statistical test called the 'I2 test' has been used. The I2 test gives a measure of how much variation between the studies is due to heterogeneity. A numerical value is presented for I2 which can range from 0%, indicating no heterogeneity, up to 100% indicating that all of the variation between the studies is due to heterogeneity. If an I2 value of 85% or more is seen, then the meta-analysis should not be performed since the studies are too different to combine together. In our example, the I2 value indicates that 45% of the variation is due to heterogeneity.