Systematic Reviews & Meta-Analysis
Judge how evidence was found, selected and combined before relying on a pooled estimate.
Quick revision ↓ Medical Dictionary
01Review question and protocol
A systematic review uses explicit methods to answer a defined question and minimise arbitrary study selection. Specify eligibility, outcomes and analysis plans before inspecting results, ideally in a registered protocol. A narrative review can be informative but may not use exhaustive or reproducible methods. A meta-analysis is statistical pooling and is not required for every systematic review; incompatible studies may be better synthesised without pooling.
02Selection and extraction
Use a documented search, deduplicate records and apply eligibility criteria consistently. Independent screening and extraction with a method for resolving disagreement reduce avoidable errors. Link multiple reports of the same study and record full-text exclusions. Extract the actual outcome definition, follow-up and analysis population, not just the abstract's most favourable number. Assess risk of bias rather than treating all eligible reports as equally credible.
03Read a forest plot
A forest plot shows study estimates and intervals alongside a pooled result when appropriate. For ratio measures, the no-effect line is one; for differences, it is zero. The weight reflects the analysis method and available information, not a moral ranking of authors. The pooled diamond's width represents an interval. Always check which direction favours which option and whether different studies measured comparable outcomes.
04Heterogeneity and pooling
Clinical heterogeneity involves differences in people, interventions or outcomes; methodological heterogeneity involves differences in design and conduct. Statistical heterogeneity concerns variation in estimated effects beyond sampling error. I-squared describes the relative contribution of heterogeneity under the model but is uncertain, especially with few studies. A random-effects model allows a distribution of effects; it does not make incompatible studies comparable or remove bias.
05Missing evidence and robustness
Published studies may differ from unpublished ones, and favourable outcomes may be preferentially reported. Funnel-plot asymmetry can raise concern but is not proof of publication bias, particularly with few studies. Examine protocols, registries and sensitivity analyses. A pooled estimate may be precise yet rest on biased evidence. Explain how conclusions change when influential studies or questionable assumptions are removed rather than presenting one pooled number as final truth.
High-Yield Revision
- A systematic review need not include a meta-analysis.
- Read the effect scale and direction before the pooled diamond.
- Pooling does not repair bias in the included studies.
Common Pitfalls
- Do not treat I-squared as a complete clinical judgement.
- Funnel-plot asymmetry has several possible explanations.
References & Further Reading
Use the full methods and relevant chapters when appraising a paper. These sources support further study; the examples and questions above are original teaching exercises.
- Oxford CEBM — Critical appraisal tools
- Cochrane Handbook — Methods for evidence synthesis
- CDC — Analysing and interpreting epidemiological data
- ASA — Statement on p-values and statistical significance
- Oxford CEBM — Likelihood ratios
- Cochrane — GRADE and summary of findings
- WMA — Declaration of Helsinki, 2024 revision
- EQUATOR — Reporting guideline library
- NIH OBSSR — Mixed-methods research resources
- Institute for Healthcare Improvement — Model for Improvement