Versions Compared
Key
 This line was added.
 This line was removed.
 Formatting was changed.
HideElements  


The standard chisquared test (Cochran Q test) for statistical heterogeneity tests the statistical hypothesis that the true treatment effects (the effect size parameters) are the same in all the primary studies included in metaanalysis (Sutton et al 2000). This statistical test uses a test statistic Q that has a chisquared distribution on k1 degrees of freedom (k represents the number of studies) under the statistical hypothesis; the corresponding pvalue for the test statistic is examined (Sutton et al 2000). The statistical power of the test is in most cases very low due to the small number of studies; heterogeneity may be present even if the Q statistic is not statistically significant at conventional levels of significance such as 0.05. A cutoff significance level of 0.10 rather than the usual 0.05 has been advocated (Sutton et al 2000). If results of the test are statistically significant (p<0.05) the statistical hypothesis that the true treatments effects (the effect size parameters) are the same in all the primary studies included in metaanalysis (the hypothesis of homogeneity) is rejected, therefore, it is considered that there is statistical heterogeneity. With a small number of studies (< 20), the Q test should be interpreted very cautiously (HuedoMedina et al 2006). It is not appropriate to decide the metaanalysis model (fixed or random effects model) based on the results of the Chi squared statistical test (Q test) for heterogeneity.