OPBSResearch
Must Know
- Evidence-based medicine combines individual clinical expertise with the best external evidence.
- A narrative review summarises studies chosen by expert opinion. It is inherently biased.
- Narrative reviews suit broad overviews and history but not answering a specific clinical question.
- A systematic review is a protocol-driven investigation in which published studies are the subjects.
- The prospective, published protocol is what separates a systematic from a narrative review.
- Systematic reviews have a narrow scope and are sensitive to publication bias.
- Meta-analysis is the quantitative synthesis of data from a systematic review.
- Pooled reanalysis applies the same methods to unpublished raw data from the original investigators.
- Meta-analysis raises power and precision of the effect estimate.
- A statistically significant pooled result may not reach the MCID.
- Neither systematic review nor meta-analysis removes bias in the included studies.
- A review is only as good as the studies it contains.
Review Types
| Review | Selection | Use | Main weakness |
|---|---|---|---|
| Narrative | Author choice | Broad overview, state of the art, technique | Selection and interpretation bias. May lag new evidence |
| Systematic | All relevant studies by protocol | Answers one focused PICO question. Basis of AAOS guidelines | Narrow. Publication bias |
| Meta-analysis | Systematic review plus pooled statistics | More power and precision | Heterogeneity, publication bias |
| Network meta-analysis | Direct and indirect comparisons | Compares A with C via B | Uncertainty compounds with network complexity |
Doing a Systematic Review
- Start with a PICO question, then a protocol and search strategy, ideally with a librarian.
- Check PROSPERO and search engines for overlapping reviews first.
- Register the protocol prospectively in PROSPERO to prevent overlap and reporting bias.
- PRISMA-P sets the elements of a review protocol. PRISMA developed from the QUOROM statement.
- Search several databases plus conference abstracts and grey literature.
- At least two independent reviewers screen titles and abstracts, with a third to arbitrate.
- Reviewers should be blinded to authors, journal and year where possible.
- Study selection is shown in a PRISMA flow diagram.
- Quality equals risk of bias.
- Jadad scores trials, Newcastle-Ottawa scores nonrandomised studies, GRADE rates certainty of evidence.
- Cochrane rejects summary quality scores and uses tabulated risk of bias domains.
Reporting Guidelines
| Study type | Guideline |
|---|---|
| RCT | CONSORT |
| Observational | STROBE |
| Systematic review | PRISMA |
| Protocols | SPIRIT, PRISMA-P |
| Diagnostic or prognostic | STARD, TRIPOD |
| Case report | CARE |
| Clinical guideline | AGREE, RIGHT |
| Animal preclinical | ARRIVE |
| Quality improvement | SQUIRE |
| Economic evaluation | CHEERS |
Meta-Analysis Tools
- Pooled effects (RR, OR, risk difference) are usually weighted by sample size or precision, such as inverse variance weighting.
- More precise studies get more weight.
- Heterogeneity is the statistical diversity of treatment effect among included studies.
- Small, nonrandomised and negative studies are less likely to be published.
- Time-lag bias arises when studies with one result are published faster than others.
- Language limits in searching can cause detection bias.
- An asymmetric inverted funnel plot suggests publication bias.
- Meta-regression and subgroup analysis explore causes of heterogeneity.
| Tool | What it shows | Limitation |
|---|---|---|
| Cochran Q | Weighted squared differences of study effects from pooled effect | Underpowered with few studies, overpowered with many |
| Higgins I squared | Percentage (0 to 100) of variation due to between-study heterogeneity | No confidence level unless given with a CI |
| L’Abbé plot | Intervention event rate against control event rate | Visual only |
Reading a Forest Plot
- The vertical line is no effect (the null hypothesis).
- Each square is a study estimate. Square size reflects its weight.
- Horizontal lines are each study’s CI.
- The diamond is the pooled estimate. Its centre is the effect and its width the 95 percent CI.
Network Meta-Analysis
- Transitivity means studies are similar enough for direct comparisons to estimate indirect ones.
- Incoherence is disagreement between direct and indirect estimates.
- More complex networks give less certain indirect estimates. More studies per comparison give more certainty.
- Rankograms and SUCRA scores display treatment rankings but may not reveal bias or incoherence.
More on examples
- A meta-analysis of three RCTs found double-bundle ACL reconstruction gave less KT-1000 laxity, but without a clinically meaningful benefit.
- Published orthopaedic literature grows each year, faster than one surgeon can read.
- Covidence, EndNote, RefWorks and Zotero help manage a review.
Exam Traps
- Meta-analysis does not reduce bias. It only adds precision.
- A methodologically sound systematic review can still overestimate treatment effect through publication bias.
- I squared does not depend on the number of studies. Cochran Q does.
Related pages. Level of Evidence, Grades of Recommendation
Source
Patterson JT, Morshed S. Systematic Reviews and Meta-Analysis. In. Aaron RK, editor. Orthopaedic Basic Science. Foundations of Clinical Practice. 5th ed. Philadelphia, Wolters Kluwer, 2021.
Practise Type X questions on this topic
Reviewed by Professor Phong Tran, Head of Orthopaedic Surgery, Western Health. Last updated 17 October 2026.