Overview
Meta-analyses fail when they pool effect sizes from methodologically incompatible studies, ignore heterogeneity (treating a statistically significant pooled effect from highly heterogeneous studies as a reliable estimate), or fail to address publication bias (the systematic underrepresentation of null results in the published literature that inflates pooled effect size estimates). A pooled effect from 20 studies is only as valid as the consistency of those studies and the completeness of the literature being pooled.
The Meta-Analysis Framework applies systematic inclusion criteria, quantifies heterogeneity before interpreting pooled effects, selects the appropriate statistical model (random vs. fixed effects), and tests for publication bias using funnel plots and Egger's test.