Community medicine chapters
Reporting guidelines, systematic reviews and meta-analysis
Which reporting guideline goes with which study design, how a systematic review differs from a narrative review, the steps of a systematic review, and how a meta-analysis pools results, with forest plots, heterogeneity and publication bias.
- Match the guideline to the design: CONSORT (randomised trials), STROBE (observational studies: cohort, case-control, cross-sectional), PRISMA (systematic reviews and meta-analyses), MOOSE (meta-analyses of observational studies); also STARD (diagnostic accuracy) and CARE (case reports).
- A systematic review answers a focused (PICO) question with a comprehensive, reproducible search and quality assessment; a meta-analysis statistically pools their results.
- Results are shown in a forest plot (the diamond is the pooled effect); a funnel plot checks for publication bias; I² measures heterogeneity.
On this page
Overview
Good research must be reported clearly and completely, so readers can judge it and repeat it. Reporting guidelines are checklists (often with flow diagrams) for each study design.
Systematic reviews and meta-analyses bring together all the good evidence on a question. They sit at the top of the hierarchy of evidence and underpin evidence-based medicine.
Reporting guidelines
| Guideline | Study type |
|---|---|
| CONSORT | Randomised controlled trials (checklist and participant flow diagram) |
| STROBE | Observational studies: cohort, case-control and cross-sectional |
| PRISMA | Systematic reviews and meta-analyses (flow diagram of study selection) |
| MOOSE | Meta-analyses of observational studies |
| STARD | Diagnostic accuracy studies |
| CARE | Case reports |
| COREQ / SRQR | Qualitative research (interviews and focus groups / all qualitative) |
| SPIRIT | Trial protocols |
| CHEERS | Economic evaluations |
The EQUATOR Network keeps a library of reporting guidelines.
CONSORT = RCTs. STROBE = observational. PRISMA = systematic reviews. MOOSE = meta-analysis of observational studies.
These are reporting checklists. They improve the reporting of a study, not the study's design itself, and they are not quality scores.
Narrative vs systematic reviews
| Feature | Narrative (traditional) review | Systematic review |
|---|---|---|
| Question | Broad | Focused (often PICO) |
| Search | Not stated; selective | Comprehensive, explicit, reproducible (several databases, grey literature) |
| Selection | Unclear; author's choice | Pre-defined inclusion and exclusion criteria |
| Quality assessment | Usually none | Formal risk-of-bias assessment |
| Synthesis | Qualitative summary | Structured; may include meta-analysis |
| Risk of bias | High | Low |
Steps in a systematic review
- Frame a clear question using PICO: Population, Intervention (or exposure), Comparison, Outcome.
- Write and register a protocol (for example on PROSPERO).
- Search widely: PubMed/MEDLINE, Embase, Cochrane Library and others, plus grey literature (theses, reports, unpublished trials) to reduce publication bias.
- Select studies against pre-set criteria, usually by two reviewers independently.
- Extract data using a standard form.
- Assess quality (risk of bias) of each study.
- Synthesise results: narrative and, if studies are similar enough, meta-analysis.
- Interpret and report (PRISMA), including the certainty of the evidence (for example GRADE).
The Cochrane Collaboration produces and updates systematic reviews on health care.
Meta-analysis
A meta-analysis is the statistical pooling of results from two or more similar studies to produce a single, more precise summary estimate.
Advantages: larger combined sample, more power and precision, can settle conflicting results, can explore differences between studies.
Limitations: "garbage in, garbage out" (poor studies give poor pooled results), publication bias, and heterogeneity (combining studies that are too different).
The forest plot
- Each study is a square (its result; size shows the study's weight) with a horizontal line (its confidence interval).
- The vertical line is the line of no effect: 1 for ratios (RR, OR) or 0 for differences.
- The diamond at the bottom is the pooled (overall) result; its width is the pooled confidence interval.
- If the diamond does not touch the line of no effect, the pooled result is statistically significant.
In a forest plot, the diamond shows the pooled effect; a study whose line crosses the line of no effect is not significant on its own.
Line of no effect drawn at RR = 1
Study A: RR 0.80 (95% CI 0.60–1.05) → crosses 1 → not significant alone
Study B: RR 0.75 (95% CI 0.62–0.90) → does not cross 1 → significant
Pooled diamond: RR 0.78 (95% CI 0.68–0.89) → entirely below 1
→ overall the treatment reduces risk by about 22% (1 − 0.78), and the pooled result is significant
Heterogeneity
Heterogeneity means the study results differ more than would be expected by chance.
- Tested with Cochran's Q and measured by I²: roughly 25% low, 50% moderate, 75% high.
- With low heterogeneity a fixed-effect model may be used; with more heterogeneity a random-effects model is used, and the reasons for differences are explored.
Publication bias and the funnel plot
Publication bias: studies with positive or significant results are more likely to be published, so reviews may overestimate effects.
A funnel plot plots each study's effect against its size (or precision). Without bias, it looks like a symmetrical inverted funnel. Asymmetry (for example missing small negative studies) suggests publication bias.
Funnel plot → publication bias. I² → heterogeneity.
Evidence-based medicine (EBM)
EBM is "the conscientious, explicit and judicious use of current best evidence in making decisions about the care of individual patients" (Sackett), combining best research evidence, clinical expertise and patient values.
Steps: Ask (a PICO question), Acquire (search), Appraise (critically assess), Apply, Assess (evaluate the outcome).
Expect: matching each design to its reporting guideline (CONSORT, STROBE, PRISMA, MOOSE, STARD, CARE), PICO, systematic vs narrative reviews, the steps of a systematic review, what a meta-analysis is, reading a forest plot (squares, lines, diamond, line of no effect), heterogeneity and I², and the funnel plot for publication bias.
Quick revision
- CONSORT: RCT. STROBE: observational. PRISMA: systematic review. MOOSE: meta-analysis of observational studies.
- STARD: diagnostic accuracy. CARE: case reports. COREQ: qualitative.
- Systematic review: PICO question, comprehensive search, criteria, quality appraisal.
- Meta-analysis: statistical pooling; forest plot with a pooled diamond.
- I²: heterogeneity. Funnel plot asymmetry: publication bias.
- EBM: evidence + expertise + patient values; ask, acquire, appraise, apply, assess.
Written by The House Job from Excel Community Medicine and Park's Textbook of Preventive and Social Medicine (named in the UHS PGET syllabus), with guidelines from the EQUATOR Network.
FAQ
Which reporting guideline is used for randomised controlled trials?
CONSORT.
Which reporting guideline is used for case-control and cohort studies?
STROBE.
What does the diamond in a forest plot show?
The pooled overall result of the meta-analysis and its confidence interval.
How is publication bias detected in a meta-analysis?
By looking for asymmetry in a funnel plot.