Why it matters when you are writing
The output most readers recognize is the forest plot: one row per study showing its effect and interval, with a diamond at the bottom for the pooled estimate. Larger and more precise studies pull the diamond further, which is the point of weighting. Reading the plot rather than the summary sentence tells you whether the pooled figure represents a consistent picture or averages across studies that disagree.
Heterogeneity is the question that decides whether pooling was legitimate at all. Statistics such as I² quantify how much variation exceeds chance, but the substantive judgment comes first: do these studies measure the same thing in comparable populations? Pooling a school-based intervention with a clinical one produces a number that describes nothing real, however tidy the confidence interval looks.
Publication bias distorts the input. Studies that find nothing are published less often and later, so the literature available to pool is skewed toward positive results. Funnel plots and tests for asymmetry give a partial check, and searching trial registries and gray literature gives a better one. Neither fully solves it, which is why a careful meta-analysis discusses the risk rather than declaring it absent.
