- The credibility of a meta-analysis depends on alignment between the evidence, methods, and inference, with studies sufficiently similar in population, endpoints, and design to justify pooling.
- Meta-analysis combines study-level effect estimates, whereas pooled analysis combines participant-level datasets into one cohort; these approaches are complementary but not interchangeable.
- Both statistical heterogeneity and clinical heterogeneity should be assessed, as formally correct analyses can still create unwarranted confidence when the underlying evidence is weak or poorly aligned.
- Transparency, critical appraisal, and integrity checks are essential because meta-analyses inherit limitations such as selective reporting, inconsistent outcomes, inadequate follow-up, and potential data integrity problems.
- Meta-analysis is not definitive truth; its value depends on whether it addresses a meaningful question using coherent, combinable evidence that produces robust and generalisable findings.
Source:
[1] Lee MMY, Kaul S, Krumholz HM, Pocock SJ, Lam CSP, Zannad F. What makes a meta-analysis believable? The Lancet. 2026;408(10560):1143-1146. doi: 10.1016/s0140-6736(26)01087-1.