PRISMA stands for Preferred Reporting Items for Systematic Reviews and Meta-Analyses. It is not a research method — it is PRISMA 2020, the 27-item reporting checklist that specifies what a finished systematic review or meta-analysis must report, not how to search, screen, or synthesize evidence. The rules for conducting a review live in the Cochrane Handbook; a PRISMA-compliant review can still rest on flawed methodology.
What PRISMA is — and isn’t
The current version, PRISMA 2020, replaced the original 2009 statement. It was developed by an international working group led by Matthew Page and David Moher and published simultaneously on 29 March 2021 in the BMJ (the most commonly cited version, DOI 10.1136/bmj.n71), with concurrent publication in PLOS Medicine, Journal of Clinical Epidemiology, Systematic Reviews, International Journal of Surgery, and several other journals — a deliberate dissemination strategy so the update would reach different disciplinary audiences at once.
PRISMA 2020 consists of four parts, maintained at prisma-statement.org:
- A 27-item checklist covering what to report in the title, abstract, introduction, methods, results, discussion, and “other information” sections of a systematic review.
- An expanded checklist giving worked reporting recommendations for each item.
- A flow diagram template documenting the number of records identified, screened, excluded (with reasons), and ultimately included.
- A companion explanation and elaboration paper with rationale and exemplars for every checklist item.
Because it is a reporting standard, satisfying PRISMA is a necessary but not sufficient condition for a good systematic review. A review can be PRISMA-compliant on paper — every checklist item ticked — while still resting on a biased search or an inappropriate meta-analysis. The methodological rules for actually conducting a review live elsewhere, principally the Cochrane Handbook for Systematic Reviews of Interventions. PRISMA and the Cochrane Handbook are complementary, not interchangeable: PRISMA governs the write-up, the Handbook governs the conduct.
The PRISMA 2020 checklist: all 27 items
The checklist groups its 27 items under seven sections. A handful of items (10, 13, 16, 20, 23, 24) have several lettered sub-parts covering closely related reporting elements, but each is still counted as a single numbered item — which is why the standard is described as “27 items,” not more. The table below summarises what each item requires; for the exact wording and worked exemplars, use the official checklist and explanation & elaboration paper at prisma-statement.org.
| # | Section | What to report |
|---|---|---|
| 1 | Title | Identify the report as a systematic review. |
| 2 | Abstract | Report per the separate PRISMA 2020 for Abstracts checklist (objectives, methods, results, limitations, funding). |
| 3 | Introduction — Rationale | Explain the review’s context within existing knowledge and why it’s needed. |
| 4 | Introduction — Objectives | State the review’s questions or objectives explicitly. |
| 5 | Methods — Eligibility criteria | Specify inclusion/exclusion criteria and how studies were grouped for synthesis. |
| 6 | Methods — Information sources | List every database, register, website and other source searched, with the date each was last searched. |
| 7 | Methods — Search strategy | Present the full search strategy for every source, including filters and limits. |
| 8 | Methods — Selection process | Describe how studies were selected, how many reviewers were involved, whether they worked independently, and any automation tools used. |
| 9 | Methods — Data collection process | Describe how data were extracted, how many reviewers were involved, and how extracted data were confirmed. |
| 10 | Methods — Data items | List and define every outcome and other variable (participant characteristics, interventions, funding, assumptions about missing data) for which data were sought. |
| 11 | Methods — Risk of bias assessment | Specify the method and tool used to assess risk of bias, and how many reviewers were involved. |
| 12 | Methods — Effect measures | Specify the effect measure used for each outcome (e.g. risk ratio, mean difference). |
| 13 | Methods — Synthesis methods | Describe how studies were grouped for synthesis, how missing data and effect estimates were handled, how results were presented, the meta-analysis method (if used), and any heterogeneity or sensitivity analyses. |
| 14 | Methods — Reporting bias assessment | Describe the method used to assess bias from missing results (e.g. selective non-reporting) within each synthesis. |
| 15 | Methods — Certainty assessment | Describe the method used to assess certainty (confidence) in the body of evidence for each outcome. |
| 16 | Results — Study selection | Report the number of records/reports identified, screened, and excluded, ideally using a flow diagram, plus reasons for excluding studies that appeared to meet the criteria. |
| 17 | Results — Study characteristics | Cite each included study and present its characteristics. |
| 18 | Results — Risk of bias in studies | Present risk-of-bias assessments for every included study. |
| 19 | Results — Results of individual studies | For every outcome, present summary statistics and effect estimates with precision (e.g. confidence intervals) for each study and each synthesis. |
| 20 | Results — Results of syntheses | Summarise the characteristics and risk of bias of studies contributing to each synthesis; present all statistical results, including heterogeneity and sensitivity findings, for each meta-analysis or other synthesis. |
| 21 | Results — Reporting biases | Present an assessment of reporting bias for each synthesis. |
| 22 | Results — Certainty of evidence | Present the certainty/confidence assessment for each outcome. |
| 23 | Discussion | Interpret the results in the context of other evidence, discuss limitations of the included studies and of the review process, and discuss implications for practice, policy and future research. |
| 24 | Other information — Registration and protocol | State the registration name/number (or that the review wasn’t registered), where the protocol can be accessed, and any amendments made after registration. |
| 25 | Other information — Support | Describe funding sources and the role of funders/sponsors. |
| 26 | Other information — Competing interests | Declare any competing interests of the review authors. |
| 27 | Other information — Availability of data, code and materials | State which of the protocol, search strategies, data, analytic code and other materials used in the review are publicly available, and where. |
Source: Page MJ, McKenzie JE, Bossuyt PM, et al. “The PRISMA 2020 statement: an updated guideline for reporting systematic reviews.” BMJ. 2021;372:n71. doi.org/10.1136/bmj.n71.
Database vs. register: what item 6 actually asks for
Checklist item 6 (Information sources) is a common source of confusion because PRISMA 2020 explicitly split ‘databases’ and ‘registers’ into separate categories that PRISMA 2009 did not distinguish as clearly. They are not interchangeable, and a review that searches only one when it should have searched both has a genuine, reportable gap:
- A database is a bibliographic index of published literature — MEDLINE/PubMed, Embase, CINAHL, PsycINFO, Scopus, Web of Science. Searching these finds published journal articles, conference papers and other indexed literature.
- A register is a record of studies themselves, not of publications about them — clinical trial registers such as ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform (ICTRP), and review registers such as PROSPERO. Registers surface ongoing or completed-but-unpublished studies that a bibliographic database search will miss entirely, which is exactly why searching them matters for detecting publication bias (checklist items 14 and 21).
PRISMA 2020 requires authors to report every database and every register searched, with the date each was last searched (item 6), plus the full search strategy used for each one (item 7). A methods section that says only ‘we searched PubMed, Embase and PROSPERO’ without separating what each source contributes, or that omits a trial register altogether, is the kind of item-6 gap reviewers and editors are specifically trained to flag.
The PRISMA flow diagram
Checklist item 16 asks authors to report the results of the search and selection process, ideally using a flow diagram. The PRISMA 2020 flow diagram has three stages — Identification, Screening, and Included — and comes in two structural variants: one for new systematic reviews and one for updated reviews that carry forward studies from an earlier version, each offered in a “databases/registers only” and a “databases/registers plus other methods” sub-form. Official downloadable templates (Word, CC BY 4.0) are maintained at prisma-statement.org/prisma-2020-flow-diagram. The free PRISMA2020 R package and its companion Shiny web app (estech.shinyapps.io/prisma_flowdiagram) generate a compliant diagram — including an interactive HTML export — from a short set of counts you supply.
The stages and the boxes each one requires are fixed by the standard; only the counts are study-specific:
- Identification. Records identified from databases and registers, plus records identified through other methods (websites, organisations, citation searching). Records removed before screening — duplicates, records marked ineligible by automation tools, records removed for other reasons.
- Screening. Records screened → records excluded. Reports sought for retrieval → reports not retrieved. Reports assessed for eligibility → reports excluded, with a reason stated for each one.
- Included. Studies included in the review; reports of included studies.
Every count must reflect the review’s actual numbers, and every full-text exclusion needs its own stated reason — an unexplained gap between “reports assessed for eligibility” and “studies included” is one of the more common PRISMA compliance failures (see below).
Why PRISMA exists
PRISMA’s predecessor, QUOROM (1999), addressed reporting for meta-analyses of randomised trials specifically. PRISMA 2009 broadened the scope to systematic reviews generally, and PRISMA 2020 updated the checklist to reflect advances in how reviews identify, select, appraise, and synthesise studies — search reporting, risk-of-bias assessment, and certainty-of-evidence rating chief among them.
The underlying problem PRISMA addresses is straightforward: incomplete reporting makes a systematic review impossible to appraise. A reader cannot judge whether a review’s conclusions are trustworthy without knowing how comprehensive the search was, how many studies were excluded and why, how risk of bias was assessed, and whether the synthesis method matched the data. Before structured reporting standards existed, these details were routinely missing, and reviews that called themselves “systematic” often weren’t distinguishable in practice from narrative literature reviews with a more authoritative-sounding label. PRISMA gives editors, peer reviewers, and readers a fixed checklist against which to hold a submission accountable, and it gives review teams a structure that reduces the chance of silently omitting something material.
How to Use PRISMA for a Systematic Review
PRISMA is formally a reporting standard, applied when you write up a review — but working from it from the start avoids reconstructing information later. A practical sequence:
- Register your protocol before screening begins, commonly with PROSPERO, and draft it against the PRISMA-P (2015) 17-item checklist so it captures eligibility criteria and the planned search and analysis methods up front.
- Design and document your search strategy per database as you run it, not after the fact, using PRISMA-S for guidance; save the literal, re-runnable search string for every source (checklist item 7).
- Track every record through screening in a form that maps directly onto the flow diagram — Identification, Screening, Included — so counts don’t need to be reconstructed retrospectively; record a reason for every full-text exclusion (item 16).
- Choose and apply a risk-of-bias tool matched to your studies’ design — RoB 2 for randomised trials, ROBINS-I for non-randomised studies of interventions — and record who assessed what (item 11).
- Decide the synthesis method before seeing the results: meta-analysis if studies are clinically and statistically similar enough, SWiM-guided narrative or tabular synthesis if not, specified in the protocol rather than chosen afterward (item 13).
- Rate the certainty of evidence per outcome, typically with GRADE, if the review reports one (items 15 and 22).
- Work through all 27 checklist items against the manuscript draft before submission, attach the completed checklist (most journals require it), and complete the flow diagram with the review’s actual counts using the official template or the PRISMA2020 R/Shiny tool.
- State registration and protocol details, funding, competing interests, and data/code availability in the “other information” section (items 24–27) — these are the items most often left incomplete because they feel administrative rather than scientific.
The systematic review process PRISMA documents
PRISMA does not define these steps — it specifies how to report them. The steps themselves are standard systematic review methodology, most authoritatively described in the Cochrane Handbook.
1. Protocol and registration
A systematic review protocol sets out the research question (commonly framed with PICO or PICOS — Population, Intervention, Comparator, Outcome, Study design), the eligibility criteria, and the planned search and analysis methods, before the review begins. Registering the protocol publicly — most commonly with PROSPERO, the international prospective register of systematic reviews maintained by the Centre for Reviews and Dissemination at the University of York (launched February 2011) — creates a time-stamped record of the planned methods. This is the same underlying purpose as preregistration in primary research: it lets a reader later check whether the review’s methods or outcomes changed after the fact, and it reduces duplicate effort by making planned-but-unpublished reviews visible to other teams. Protocols themselves have a dedicated reporting standard, PRISMA-P (2015), a 17-item checklist for protocol completeness.
2. Search strategy
A defensible systematic search covers multiple bibliographic databases (rarely just one) plus, where relevant, grey literature, trial registries, and reference-list checking, using a documented, reproducible search strategy — not an ad hoc set of queries. The PRISMA-S extension (2021, 16 items) specifically addresses search reporting: it requires the exact search strategies for every database, copied and pasted as run, rather than a paraphrased summary, so the search can be re-run or audited.
3. Screening
Records are screened against the protocol’s eligibility criteria in two stages — title/abstract, then full text — with reasons recorded for every full-text exclusion. The Cochrane Handbook recommends at least two reviewers screen independently, to reduce single-reviewer error and bias. The result of this stage is exactly what the PRISMA flow diagram exists to summarise: records identified, duplicates removed, records screened, records excluded, full-text reports assessed for eligibility, reports excluded (with reasons), and studies included in the final review.
4. Data extraction and risk-of-bias assessment
Included studies are read in full and their data pulled into a standardised extraction form, ideally by more than one reviewer independently, to limit transcription error. Each included study is then assessed for risk of bias using a tool matched to its design — commonly the Cochrane RoB 2 tool for randomised trials or ROBINS-I for non-randomised studies of interventions — rather than a single generic quality score applied indiscriminately across designs.
5. Synthesis
Where included studies are similar enough in population, intervention, and outcome measurement, results are pooled statistically in a meta-analysis. Where they aren’t — a common outcome, not a failure — reviewers use a structured narrative or tabular synthesis instead; the SWiM (Synthesis without Meta-analysis) guideline (Campbell et al., BMJ 2020;368:l6890) exists specifically because non-meta-analytic synthesis was, historically, reported far less transparently than meta-analysis was. Many reviews then rate the overall certainty of their evidence per outcome using the GRADE framework (high/moderate/low/very low), which combines risk of bias, inconsistency, indirectness, imprecision, and publication-bias considerations into a single certainty rating.
PRISMA extensions
The core 2020 checklist is written for systematic reviews of intervention effects. Real reviews often need more specific guidance, which is why a family of extensions exists — each is a distinct reporting checklist, not an optional add-on to the main one:
- PRISMA-P (2015) — protocols, 17 items.
- PRISMA-ScR (2018) — scoping reviews, 20 essential items plus 2 optional items. Use this, not the core checklist, if the review is a scoping review mapping the breadth of a literature rather than answering a focused effectiveness question.
- PRISMA-S (2021) — search reporting, 16 items.
- PRISMA-DTA (2018) — diagnostic test accuracy studies.
- PRISMA-NMA — network meta-analyses comparing multiple treatments via direct and indirect evidence.
Several further extensions exist (individual participant data, harms reporting, equity-focused reviews, abstract-only reporting) — the current full list is maintained at prisma-statement.org/extensions, worth checking directly rather than relying on a static list, since new extensions are added periodically.
How this differs from choosing a review type
PRISMA governs reporting once you already know what kind of review you’re conducting. Deciding between a systematic review, a scoping review, and a narrative literature review in the first place is a separate, upstream question — driven by whether you have a focused, answerable question (systematic review), you’re mapping the extent and nature of evidence on a broader topic (scoping review), or you’re providing expert synthesis without a pre-registered, exhaustive methodology (narrative review). CASRAI’s dictionary defines each of these — systematic review, scoping review, narrative review — and a dedicated comparison of the three is covered separately; this guide assumes you’ve already made that choice and are conducting (or reporting) a systematic review specifically.
Common compliance gaps
A few patterns show up repeatedly in PRISMA-noncompliant submissions:
- Mislabeling. A narrative review of selected papers, described as “systematic” without a registered protocol, documented search, or dual screening, is not a systematic review regardless of the label — this is explicitly called out as a counter-example on CASRAI’s own systematic review term.
- Incomplete flow diagrams. Omitting exclusion counts or reasons at the full-text stage, or not distinguishing “records” from “reports” from “studies” (a distinction PRISMA 2020 tightened relative to 2009).
- Undocumented search strategies. Reporting “we searched PubMed and Google Scholar” instead of the literal, re-runnable search string per database, which PRISMA-S exists to prevent.
- Meta-analysis where synthesis without meta-analysis was more honest. Pooling clinically or methodologically heterogeneous studies into a single effect estimate because a table of results is less satisfying to present, rather than because pooling was statistically and clinically appropriate.
- Protocol deviations left unreported. Changing eligibility criteria, outcomes, or analysis methods after registration without disclosing the change and the reason for it.
Frequently asked questions
What is the full form of PRISMA?
PRISMA stands for Preferred Reporting Items for Systematic Reviews and Meta-Analyses. It is the name of both the reporting standard and the checklist it comprises — not an acronym for a research method, a piece of software, or a statistical technique.
Is PRISMA required by journals?
Many biomedical and health-sciences journals require or strongly recommend PRISMA-compliant reporting (typically a completed checklist submitted alongside the manuscript) as a condition of considering a systematic review or meta-analysis for publication, and PRISMA is endorsed by a large number of journals and organisations, listed at prisma-statement.org. Requirements vary by field and publisher — check the specific journal’s author guidelines rather than assuming compliance is either universal or optional.
Does following PRISMA mean the review used good methods?
No. PRISMA confirms that a review reported its methods and findings completely enough to be appraised — it does not certify that the search was comprehensive, the risk-of-bias assessment was correct, or the synthesis method was appropriate. Methodological quality is a separate question, generally assessed with a tool like AMSTAR 2 or judged directly against Cochrane Handbook guidance.
Do I need to register a systematic review protocol before I can publish?
Registration itself is not a universal legal requirement, but it is increasingly expected: many journals and funders treat prospective registration (commonly via PROSPERO) as a marker of a credible, non-post-hoc review, and some medical journals require it or a stated reason for its absence. See CASRAI’s guide to preregistration of a study protocol for the mechanics of registering in general.
What’s the difference between PRISMA 2020 and PRISMA-ScR?
PRISMA 2020 is written for systematic reviews of intervention effects; PRISMA-ScR (2018) is a separate 20-item checklist purpose-built for scoping reviews, which map the breadth of evidence on a topic rather than answering a single, focused effectiveness question. Using the core checklist for a scoping review — or vice versa — produces a mismatch between what was reported and what the review actually is.
Can I use PRISMA outside health and medical research?
Yes. PRISMA originated in, and remains most deeply embedded in, biomedical and health-sciences publishing, but the checklist itself is discipline-agnostic and is now used in fields including environmental science, education, criminology, and social policy — sometimes with field-specific adaptations. PROSPERO’s own registration scope has broadened over time beyond pure medicine to accept protocols in these adjacent areas where there is a health-related outcome.
