Draft 4.3.10.2 Quantification of the statistical heterogeneity: I squared

The I square statistic (I²) represents the percentage of the variability in effect estimates that is due to heterogeneity (Deeks et al 2008). I² is the proportion of observed dispersion of results from different studies included in a meta-analysis that is real, rather than spurious (Borenstein et al 2009). The I² index can be interpreted as the percentage of the total variability in a set of effect sizes due to true heterogeneity (between-studies variability) (Huedo-Medina et al 2006). If I² = 0%, this indicates that all variability in effect size estimates is due to sampling error within studies. If I²= 50%, it indicates that half of the total variability among effect sizes is caused not by sampling error, but by true heterogeneity between studies (Huedo-Medina et al 2006).  I² is a percentage and its values lie between 0% and 100% (Higgins et al 2003). A value of 0% indicates no observed heterogeneity, and larger values show increasing heterogeneity (Higgins et al 2003). One proposed suggestion was to consider as low, moderate, and high heterogeneity for I² values of 25%, 50%, and 75% (Higgins et al 2003). Another guide to interpretation was proposed: 0% to 40% might not be important; 30% to 60% may represent moderate heterogeneity; 50% to 90% may represent substantial heterogeneity; 75% to 100% considerable heterogeneity (Deeks et al 2008). Authors of the guide mention that careful interpretation of the value of I² depends on magnitude and direction of effects and strength of evidence for heterogeneity (Deeks et al 2008). With a small number of studies (< 20) and/or average sample size (N <80) the statistical power for I² procedures is less than the usually recommended minimum value of 0.8 (Huedo-Medina et al 2006). With a small number of studies (< 20), both the I² confidence interval and the Q test should be interpreted very cautiously (Huedo-Medina et al 2006).