Heterogeneity refers to the variability in results or outcomes across studies included in a meta-analysis. This variability can arise from differences in patient characteristics, treatment protocols, or study design, affecting the conclusions that can be drawn from the meta-analysis.
Heterogeneity can be measured using the I2 (I squared) value, which describes the percentage of variation across studies that is due to heterogeneity rather than chance. Interpreting I2 values is subjective, and there are different guidelines that can be followed. As a rule of thumb Higgins 2003 uses the following:
- Low heterogeneity: I2 = 25%
- Moderate heterogeneity: I2 = 50%
- High heterogeneity: I2 = 75%
Imagine we have a collection of studies evaluating the effectiveness of PRP in improving live birth rates in IVF patients. These studies vary in their findings—some show a significant improvement in live birth rates with PRP, while others show little to no effect. An I2 value of 50% in this context indicates that half of the variation in the live birth rates across these studies is due to methodological differences between the studies. This suggests that this treatment might work better in certain situations or for certain patients, and that more research may be needed.
A chi-squared test is used to assess the statistical significance of heterogeneity; a p-value less than 0.1 suggests significant heterogeneity (http://facpub.stjohns.edu/~sees/Stats/heterogeneity.pdf).