Strengths and weaknesses of different epidemiological study designs - systematic review and meta-analyses

Definition

Systematic reviews comprehensively and reproducibly identify, appraise, and synthesise all available evidence addressing a specific clinical question, with meta-analysis statistically pooling results from multiple studies (where sufficiently similar) to produce a single combined effect estimate with greater statistical power than any individual included study
Positioned at the top of most evidence hierarchies given their comprehensive synthesis of existing evidence, though the quality and reliability of a systematic review/meta-analysis depends entirely on the quality of the included studies and the rigour of the review methodology itself

Pooling data across multiple studies increases the effective sample size and statistical power, allowing detection of effects (or more precise effect estimates) that individual smaller studies may have been underpowered to detect reliably - but this benefit is only valid if the included studies are sufficiently similar (in population, intervention, outcome definition) to be meaningfully combined, explaining why heterogeneity assessment is a critical, non-optional component of any meta-analysis

Heterogeneity (statistical variation in results between included studies beyond what would be expected by chance alone) should be formally assessed (e.g. I² statistic) - substantial heterogeneity may indicate the included studies are too clinically or methodologically different to appropriately pool, and forcing a pooled estimate despite significant heterogeneity can produce a misleading combined result
Publication bias (positive/significant results being more likely to be published than negative/null results) can skew a systematic review's conclusions if the search and inclusion process does not adequately account for unpublished or difficult-to-find negative studies

A systematic review/meta-analysis's conclusions should be weighted by both the quality of the underlying included studies (a review of poor-quality studies produces an unreliable combined estimate, regardless of how many studies are included - "garbage in, garbage out") and by the degree of heterogeneity found
Funnel plot assessment and comprehensive, systematic search strategies (including grey literature/unpublished data where feasible) help identify and mitigate publication bias risk

Heterogeneity assessment (I² statistic) and its implications for pooling appropriateness, publication bias and funnel plot assessment, "garbage in, garbage out" - a meta-analysis cannot overcome poor-quality included studies, forest plot interpretation

A meta-analysis cannot produce a reliable combined result from poor-quality included studies - if the underlying individual studies are methodologically flawed or biased, pooling them together simply combines these flaws into a single, seemingly more authoritative but still fundamentally unreliable estimate, meaning critical appraisal of included study quality is just as essential when reading a meta-analysis as when reading any individual study

5 more sections, plus exam facts

Premium unlocks every note across every specialty, and the full exam fact library behind it.