6.1 PRISMA 2020 for Systematic Reviews & Meta-Analyses
Key Takeaways
PRISMA 2020 is a 27-item reporting checklist with flow-diagram templates for systematic reviews; it improves reporting but does not prescribe one required review method.
The review protocol and any registration record should be prepared prospectively where applicable, and the report must identify the registration or explain that the review was not registered.
Searches, eligibility, selection, data collection, risk-of-bias methods, synthesis decisions, reporting-bias assessment, and certainty assessment should be reported as applicable and actually performed.
PRISMA does not mandate PROSPERO for every review, Egger's test, a particular risk-of-bias tool, or GRADE in every project.
Use the appropriate extension, such as PRISMA-P, PRISMA-ScR, PRISMA-NMA, or PRISMA-IPD, when the review type calls for it.
PRISMA 2020 for Systematic Reviews & Meta-Analyses
Note
PRISMA 2020 is a reporting guideline, not a conduct manual. Its 27-item checklist and flow-diagram templates help authors explain why a review was done, how evidence was identified and synthesized, what was found, and how certain the findings are.
Scope
PRISMA 2020 is intended mainly for systematic reviews of studies evaluating health interventions, but many items apply to other systematic reviews. It applies whether or not a statistical meta-analysis is performed.
Use an extension when relevant:
| Review type | Relevant guidance |
|---|---|
| Protocol | PRISMA-P |
| Scoping review | PRISMA-ScR |
| Network meta-analysis | PRISMA-NMA |
| Individual participant data | PRISMA-IPD |
| Abstract | PRISMA for Abstracts |
| Search reporting | PRISMA-S |
An extension supplements the core guideline; it does not turn every suggested method into a universal requirement.
Title, Abstract, and Rationale
Identify the report as a systematic review. The abstract should clearly describe objectives, eligibility, information sources, methods, principal results, limitations, interpretation, and registration where applicable.
The introduction should state the rationale and an explicit objective or question, often structured by participants, interventions or exposures, comparators, outcomes, and study design.
Protocol and Registration
A prospective protocol separates decisions made before results were known from later changes. Register the protocol in a suitable registry when the review is eligible and registration is available. PROSPERO is not mandatory for every systematic review, and its eligibility and timing rules are set by PROSPERO.
PRISMA 2020 asks the report to provide registration information or state that the review was not registered, indicate where the protocol is accessible, and describe amendments. If methods changed after screening began, explain what changed and why.
Do not claim retrospective registration was prospective. A dated protocol in a repository can still improve transparency when registry eligibility is unavailable, but label timing honestly.
Eligibility and Information Sources
Define inclusion and exclusion criteria and explain how studies were grouped for synthesis. Report all databases, registers, websites, organizations, reference lists, and other sources searched, with the date each was last searched.
A reproducible search report includes full strategies for each database or at least all necessary strings and limits in supplementary material. Peer review of the search strategy may improve quality but is not a PRISMA-mandated step for every review.
Selection and Data Collection
Describe:
- how records were screened;
- how many reviewers worked independently;
- how conflicts were resolved;
- automation tools used;
- how data were collected;
- whether investigators were contacted;
- how multiple reports of one study were linked;
- assumptions about missing or unclear information.
The unit of analysis is usually the study, while the search retrieves reports or records. Several articles may describe one trial. Failing to link them can double count participants.
Outcomes and Other Variables
List and define outcomes sought and state whether all compatible results were collected. Describe other variables such as participant characteristics, intervention details, funding, and conflicts.
If only selected time points, scales, or analyses were extracted, explain the selection rule. Selecting the most favorable available result after seeing the findings creates bias.
Risk of Bias
Specify the tool and process used to assess risk of bias, including reviewer independence and conflict resolution. Choose a tool appropriate to the study design and question. Cochrane RoB 2 and ROBINS-I are examples, not mandatory PRISMA tools.
Report the assessments and explain how they affected synthesis or interpretation. Do not substitute a reporting-quality checklist for a risk-of-bias assessment unless it is appropriate for the purpose.
Synthesis
For each synthesis, explain:
- which studies were eligible;
- how data were prepared;
- how results were tabulated or displayed;
- why a meta-analysis was appropriate or not;
- the effect measure and statistical model;
- how heterogeneity was explored;
- sensitivity or subgroup analyses;
- software used.
Heterogeneity is not solved merely by choosing a random-effects model. Consider clinical and methodological differences, effect direction, precision, and the assumptions of the synthesis.
For network meta-analysis, explain network structure, transitivity, and consistency between direct and indirect evidence. Use the PRISMA-NMA extension.
Reporting Bias and Certainty
PRISMA asks authors to report methods used to assess risk of bias from missing results and methods used to assess certainty or confidence in the evidence. It does not require Egger's test or GRADE in every review.
A funnel plot or statistical asymmetry test may be uninformative with few studies and can reflect heterogeneity rather than publication bias. State the rationale and limitations of the chosen approach. GRADE is a common certainty framework, but use the method appropriate to the project and report it transparently.
Results and Flow Diagram
The flow diagram records:
- records identified;
- duplicates removed;
- records screened;
- full reports sought and assessed;
- exclusions at full-text stage with reasons;
- studies included;
- reports contributing to the review.
Counts must reconcile. Report included-study characteristics, risk-of-bias results, individual-study results, each synthesis, heterogeneity, sensitivity analyses, reporting-bias assessment, and certainty assessment as applicable.
Discussion
Interpret results in the context of other evidence. Discuss limitations of both the included evidence and the review process, including search restrictions, missing data, heterogeneity, selective reporting, and deviations. Do not treat statistical significance as certainty or causality.
Other Information
Report registration and protocol access, support and funder roles, conflicts of interest, and availability of data, code, forms, or other materials.
Applied Examples
No meta-analysis: A systematic review finds studies too heterogeneous for a pooled estimate. It remains a systematic review and can comply with PRISMA by explaining study selection, characteristics, results, and the rationale and method for narrative synthesis.
Ten-study funnel plot: The team may use an asymmetry method if justified, but PRISMA does not command Egger's test. It should report the method, assumptions, results, and limitations.
Unregistered review: The report should not invent a registration. State that it was not registered, provide any accessible protocol, and explain amendments or timing.
Important
PRISMA tells readers what to report. It does not by itself require one registry, one bias tool, one meta-analysis model, one small-study-effects test, or one certainty framework.
What does PRISMA 2020 require a review report to say about registration?
Every review must have been registered in PROSPERO before any screening
Omit registration if no meta-analysis was conducted
Register retrospectively but call it prospective
Provide registration information or state that the review was not registered, and report protocol access and amendments as applicable
A review team is deciding how to assess small-study effects and certainty. What does PRISMA 2020 require?
Egger's test and GRADE in every review
No assessment or reporting
Appropriate, justified methods reported transparently if performed; PRISMA does not mandate one test or framework
A funnel plot even when only two studies exist
A synthesis compares several treatments using direct and indirect evidence. Which extension and assumption deserve attention?
PRISMA-NMA, including transitivity and consistency
PRISMA-ScR, with no comparison assumptions
CARE, with patient consent
STARD, with reference-standard masking
Sections you finish are checked off in the contents.