The latter reveals the proportion of variability in a meta-analysis that is the result of between-study variation and not error within studies due to random sampling.
Advantages of meta-analysis include an increase in precision due to increased effect size, control for between-study variation and overcoming bias of individual studies [ 14, 15].
I is the proportion of total variation provided by between-study variation, and I values of 0, 250 50, and 75% represent no, low, moderate, and high heterogeneity, respectively (32).
We aggregated outcomes data at the study level and performed statistical calculations with Review Manager (RevMan) 5.0 (2009; The Cochrane Collaboration, Oxford, UK) using random-effects models [ 23], which incorporate both within-study and between-study variation and generally provide more conservative effect estimates when heterogeneity is present.
I is the proportion of total variation contributed by between-study variation (Higgins and Thompson, 2002), and values of 25%, 50%, and 75% have been regarded as representing low, moderate, and high heterogeneity, respectively (Higgins et al, 2003).
Study-specific estimates were combined using the DerSimonian and Laird random-effects model, which incorporates both within- and between-study variation (DerSimonian and Laird, 1986).
To take within and between study variation into account, we constructed HSROC and forest plot for NT-proBNP, which derived an area under the curve (AUC) of 0.87 (95% CI 0.83 to 0.89) and a summary OR of 21.6 (95% CI 8.33 to 55.97), respectively (figures 3 and 4).
Random-effects model was used to combine study-specific RR (95%CI), which considers both within-study and between-study variation (DerSimonian and Laird, 1986).
42 A major limitation of this meta-analysis is the between study variation in pooled sensitivities and specificities, which are evident in figure 3.
