Meta‑analyses Can Mislead

When a field is full of low‑quality or selectively reported studies, combining them can produce a more confident but more biased estimate than any single careful study. Standard corrections (funnel plots, trim‑and‑fill) help but cannot fully fix literatures corrupted by selection, heterogeneous designs, or p‑hacking. — If true, this changes how journalists, courts, regulators and policymakers should treat meta‑analyses: they must assess literature quality and bias patterns rather than assume aggregation equals truth.