What is the PRISMA checklist?

The PRISMA checklist is a 27-item reporting guideline that specifies what a systematic review or meta-analysis must state in print. PRISMA 2020, published in 2021, covers seven sections from title to data availability, and its authors are explicit that the checklist reports a review rather than judging it: it is not a quality assessment tool.

That distinction is where most misuse begins. This page sets out what PRISMA stands for, the structure of the 27 items and their sub-items, the flow diagram, the extension family, what the checklist deliberately does not do, how incomplete PRISMA reporting is detected in a submitted manuscript, and what reviewers write when it has been handled badly.

What PRISMA stands for and what it replaced

PRISMA stands for Preferred Reporting Items for Systematic reviews and Meta-Analyses. The acronym has been stable since 2009, when the PRISMA statement replaced QUOROM (Quality of Reporting of Meta-analyses), the 1999 Lancet guideline for reporting meta-analyses of randomised controlled trials.

PRISMA 2020 is the current version, published simultaneously in BMJ (2021;372:n71), Systematic Reviews, PLOS Medicine, the Journal of Clinical Epidemiology and the International Journal of Surgery, with a companion explanation and elaboration paper (BMJ 2021;372:n160) that gives a worked exemplar for every item. Citing “PRISMA guidelines” without a year is now ambiguous, because the 2009 and 2020 checklists both contain 27 items but do not contain the same 27 items — certainty assessment, competing interests and data availability are new in 2020, and the search-strategy item changed from one database to all of them.

The seven sections and the 27 items

The PRISMA 2020 checklist is organised into seven sections, and the item numbering runs continuously across them.

Title is item 1: identify the report as a systematic review. Abstract is item 2, which defers to a separate 12-item PRISMA 2020 for Abstracts checklist rather than restating requirements inline. Introduction is items 3 and 4 — rationale and objectives, the latter usually expressed as a structured PICO question.

Methods carries items 5 to 15 and is the largest section: eligibility criteria (5), information sources (6), search strategy (7), selection process (8), data collection process (9), data items (10a and 10b), study risk of bias assessment (11), effect measures (12), synthesis methods (13a to 13f), reporting bias assessment (14) and certainty assessment (15).

Results carries items 16 to 22, most of which report the outcome of a corresponding Methods item: study selection (16a and 16b), study characteristics (17), risk of bias in studies (18), results of individual studies (19), results of syntheses (20a to 20d), reporting biases (21) and certainty of evidence (22).

Discussion is item 23, split into interpretation, limitations of the evidence, limitations of the review process, and implications (23a to 23d). Other information closes with registration and protocol (24a to 24c), support (25), competing interests (26) and availability of data, code and other materials (27).

Why the sub-items matter more than the headline number

The PRISMA 2020 checklist expands well past 27 lines once sub-items are counted, and the sub-items are where reviews fail. The statement itself is explicit that its seven sections hold 27 items, some of which contain sub-items — so the count that matters when completing it is not 27.

Item 13 is the clearest case. Synthesis methods split into six sub-items covering which studies were eligible for each synthesis, how data were prepared, how results were tabulated or visualised, which methods were used to synthesise and why, how heterogeneity was investigated, and what sensitivity analyses were run. A single sentence saying “a random-effects meta-analysis was performed” satisfies none of the six.

Item 24 works the same way: 24a asks for the registration name and number or an explicit statement that the review was not registered, 24b for where the protocol can be accessed, and 24c for any amendments made after registration. Silence does not satisfy 24a — the checklist requires the negative statement, which is why “not registered” is a compliant answer and an empty field is not. The same discipline applies to primary studies, where registration and endpoint concordance is checked against the registry entry rather than against the manuscript’s own account of itself.

The abstract has its own checklist

Item 2 is one line in the main checklist and a full second document in practice. PRISMA 2020 for Abstracts is a 12-item checklist covering what a structured abstract must state: the report should be identified as a systematic review, with the objectives, the eligibility criteria, the information sources with the date each was last searched, the risk of bias and synthesis methods, the number of included studies and participants, the summary results with effect estimates and precision, the limitations of the evidence, the interpretation, the funding and the registration number.

This matters more than its length suggests, because the abstract is what most readers read and what indexing services carry. Three failures are routine. The abstract reports a pooled estimate with no interval, so the reader cannot see the precision. The abstract states a conclusion with no certainty qualifier, while the full text grades the evidence as low — the two documents then say different things. And the abstract omits the registration number that item 24a supplies in the body, which is the easiest of the twelve to fix and among the most often missed.

Registration, the protocol and item 24

Item 24 exists because a review written after its results are known is a different object from a review written before. Prospective registration — PROSPERO for most health-related reviews, with Open Science Framework and similar registries used elsewhere — fixes the question, the eligibility criteria, the outcomes and the planned synthesis in a dated public record.

The reporting obligation is narrower than the methodological one and is often confused with it. PRISMA does not require registration. It requires that the manuscript say what happened: the register and number if registered (24a), where the protocol can be read (24b), and what changed after registration and why (24c). A review that was never registered is fully compliant with 24a by saying so in one sentence, and a review that was registered but quietly changed its primary outcome is non-compliant with 24c no matter how good the rest of the reporting is.

Outcome switching is the case reviewers escalate. When the registered primary outcome and the reported primary outcome differ and no amendment is declared, the reviewer cannot distinguish a defensible protocol change from a result-driven one, and will usually say so in those terms.

The flow diagram is part of the guideline

The PRISMA flow diagram ships as part of the PRISMA 2020 statement rather than as a decorative extra — item 16a asks authors to describe the search and selection process “ideally using a flow diagram” — and its 2020 revision changed more than the boxes. The statement supplies templates in several variants, including versions for an original review and for an update of an existing review, and expects authors to pick the one that matches the work rather than to redraw the diagram freehand.

Two structural changes matter when reading a diagram. First, PRISMA 2020 separates identification via databases and registers from identification via other methods — citation searching, websites, organisations and direct contact with authors — instead of the single “additional records identified through other sources” box of the 2009 version. Second, PRISMA 2020 distinguishes records (database entries), reports (documents) and studies (the unit of research), because one study can generate several reports and one report can describe several studies. A diagram that reports a single number at each stage has usually collapsed that distinction, and the collapse is what makes the arithmetic irreconcilable later.

What the PRISMA checklist is not

PRISMA 2020 is a reporting guideline, and its authors bound it twice in the statement itself. On quality: “PRISMA 2020 should not be used to assess the conduct or methodological quality of systematic reviews; other tools exist for this purpose.” On method: “PRISMA 2020 is not intended to guide systematic review conduct, for which comprehensive resources are available.”

Two named tools do the work PRISMA declines. AMSTAR 2 (Shea et al., BMJ 2017) rates the methodological quality of a systematic review across 16 items, seven of which its authors identify as critical domains. ROBIS (Whiting et al., Journal of Clinical Epidemiology 2016) assesses risk of bias in a review across four domains in a three-phase process. Writing “quality was assessed using PRISMA” is a category error, and an experienced methodological reviewer will name it as one.

PRISMA 2020 also states its own scope: it “has been designed primarily for systematic reviews of studies that evaluate the effects of health interventions, irrespective of the design of the included studies.” Reviews of prevalence, prognosis, diagnostic accuracy, mechanism or qualitative evidence are outside that primary design target, which is precisely why the extension family exists.

The extension family

The PRISMA family maintains a published list of officially developed extensions, and choosing the right one is a reporting decision made before the first search, not after peer review.

The four with the widest use are PRISMA-P (Moher et al., Systematic Reviews 2015), a 17-item checklist for review protocols; PRISMA-ScR (Tricco et al., Annals of Internal Medicine 2018), which carries 20 essential reporting items plus 2 optional ones for scoping reviews; PRISMA-S (Rethlefsen et al., Systematic Reviews 2021), a 16-item checklist covering literature searches alone; and PRISMA-DTA (2018) for reviews of diagnostic test accuracy.

The most common extension error is choosing PRISMA when the work is a scoping review. A scoping review maps the extent and nature of a literature and does not synthesise effects, so several core PRISMA items — effect measures, synthesis methods, certainty of evidence — have no honest answer, and filling them in anyway produces a checklist full of “n/a” that a reviewer reads as either the wrong guideline or the wrong study design. PRISMA-ScR exists so that the reporting matches the question actually asked.

Beyond those sit PRISMA-NMA for network meta-analyses (2015), PRISMA-IPD for individual participant data (2015), PRISMA-Harms (2016), PRISMA-Equity (2012) for reviews with a health equity focus, PRISMA-Children and Adolescents, PRISMA-Complex Interventions, PRISMA-Living Systematic Reviews, PRISMA-EcoEvo, PRISMA-COSMIN, and extensions for acupuncture, moxibustion and Chinese herbal medicines. Extensions supplement the core checklist; PRISMA-S in particular replaces nothing, because item 7 of PRISMA 2020 still applies and PRISMA-S expands what a complete answer to item 7 looks like.

Which items go missing, and why those items

Citing PRISMA and adhering to PRISMA are different acts, and the gap between them is not random. The items that go missing cluster in the Methods and Results sections, and they share a property: each one demands work rather than words.

Writing “we searched PubMed, Embase and Scopus” takes a line. Reproducing the full Boolean strategy for each source, with the date run and every filter applied, takes an appendix and forces the authors to admit which limits they used. Saying “studies were excluded at full text” takes a clause. Listing those studies with a reason for each takes a table and exposes every borderline judgement to challenge. Saying “the review followed PRISMA” demands nothing at all. Reporting the registration number, or stating plainly that no registration exists, surrenders the protection of ambiguity.

Certainty assessment behaves the same way. Items 15 and 22 ask for a stated method — usually GRADE — and a judgement per outcome, which means grading down for risk of bias, inconsistency, indirectness, imprecision and publication bias, and defending each downgrade. A review can be internally competent and still skip this entirely, and many do.

The pattern is worth stating directly, because it is the practical argument for the checklist and it is stronger than any argument from journal policy: the items most often absent are precisely the items that would let a reader check the review without trusting its authors. A review missing all four is not necessarily wrong. It is unauditable, which for a synthesis is close to the same thing. The same asymmetry shows up in primary research, where data and accession reporting is the item most often reduced to a formula.

How incomplete PRISMA reporting is detected in a manuscript

Detecting PRISMA gaps in a submitted review is arithmetic and cross-referencing, not judgement, and six checks find most of it.

Reconcile the flow diagram against itself. Records identified, minus duplicates removed, minus records screened out, minus full texts excluded, must equal the number of studies included. It frequently does not, and a discrepancy of even one study means the numbers were typed rather than derived.

Reconcile the flow diagram against the tables. The count of included studies in the diagram, the count in the abstract, the number of rows in the characteristics table, and the number of studies in the forest plot are four independent statements of the same quantity. In manuscripts with a reporting problem, at least two disagree.

Look for a reproducible search, not a described one. Item 7 and PRISMA-S require the full search strategy for every source, with the date each was run and any limits or filters applied. “PubMed, Embase and Scopus were searched using terms related to X” is a description, not a strategy, and nobody can rerun it.

Check item 24a for an affirmative statement. Either a register name and registration number appear, or the manuscript states that the review was not registered. Absence is the most common finding and is a failure of the item regardless of whether registration occurred.

Search for the certainty items. Items 15 and 22 require a stated method for assessing certainty of evidence, usually GRADE, and the resulting judgements per outcome. A Methods sentence naming GRADE with no corresponding judgement in Results satisfies item 15 and fails item 22, and that split is the common shape of the failure.

Check item 16b: the list of studies excluded at full text, with a reason for each. This is the item that lets a reader test whether a study they consider relevant was seen and rejected, or never found. A review that omits it cannot be audited by anyone outside the author team.

A seventh check applies to the submitted checklist document itself. Journals typically ask for a completed checklist with page numbers, and the fastest tell of retrospective completion is a page number that points to a page not containing the item, or a run of “n/a” entries against Methods items that are not optional. Comparable submission-stage requirements across journal policies are increasingly checked by editorial staff before a manuscript reaches reviewers at all.

What reviewers say when the checklist is mishandled

Reviewer comments on PRISMA compliance are formulaic, which makes them predictable and therefore preventable. The recurring phrasings are worth reading as a list.

“The search strategy is not reproducible; please provide the full strategy for each database, including the date run and any limits applied.” “The numbers in the PRISMA flow diagram do not reconcile with the text or with Table 1.” “Studies excluded at full-text screening are not listed, and no reasons are given.” “The review does not appear to have been registered, and no protocol is provided; please state this explicitly.” “No assessment of certainty of the evidence is reported.” “The authors state that quality was assessed using PRISMA; PRISMA is a reporting guideline, and a risk of bias tool such as AMSTAR 2 or ROBIS is required.” “The completed checklist cites page numbers that do not correspond to the relevant content.” “Outcomes reported here differ from those registered in the protocol, and the change is not declared.”

The last two are the ones that damage credibility beyond the item concerned, because both suggest the checklist was completed to satisfy a submission form rather than to describe the work. Responding to them well means correcting the record rather than defending it, which is the same principle that governs any response to reviewers. Where a gap is genuine and unfixable — a database that could not be searched, a protocol amendment made without documentation — stating it as a limitation under item 23c is a compliant answer and a stronger one than an evasion. Reviewers raising these points are making methods objections, not stylistic ones, and they usually decide the paper.

Completing the checklist without gaming it

Complete the PRISMA checklist while writing the review, not at submission. The items are ordered to match the structure of a manuscript, so an item that cannot be filled in during drafting is usually signalling that the underlying work is missing rather than that the sentence is missing.

Three practices distinguish a checklist that reflects the manuscript from one that decorates it. Fill in page and line numbers rather than section names, because a section name cannot be checked and a line number can. Record a protocol amendment when it happens, since item 24c asks for amendments and no author reconstructs them accurately from memory a year later. And write the negative statements explicitly — no registration, no funding, no competing interests, no meta-analysis performed and why — because PRISMA treats an absent answer and a negative answer as different.

Checklists also do not extend to interpretation. PRISMA item 23a asks for a general interpretation of the results in the context of other evidence, and no reporting guideline prevents a review from stating a conclusion its data do not support. That failure mode is a separate one, checked by overclaim discipline rather than by any item number.

Limitations and what is changing

PRISMA 2020 measures reporting only, and a completely reported review can still be a badly conducted one: a search of two databases, reported in full with dates and limits, satisfies item 7 while remaining an inadequate search. Complete reporting makes that inadequacy visible, which is the honest claim for the checklist, and it is a smaller claim than the one usually made for it.

Measurement of adherence is itself slippery, and published adherence figures should be read with suspicion rather than arithmetic. Studies differ on whether they score the 27 headline items or the full expanded set of sub-items, whether a partially addressed item counts, whether sub-items are scored independently or collapsed into their parent, and whether the assessment covers the whole checklist or a selected subset of domains. Two studies of the same literature can therefore report very different percentages without either being wrong. Any adherence figure worth citing names the tool and the scoring rule that produced it; a bare percentage does not mean much on its own.

The family is also consolidating. Work is under way, led by members of the PRISMA 2020 authorship group, on a unified statement that harmonises content across the core guideline and several of its extensions, together with tooling intended to make completion and checking less manual. Authors choosing an extension should expect the boundary between the core checklist and its extensions to move, and should cite the version they actually used rather than “PRISMA” unqualified.

Related

Journal policies on AI in peer review · Research methods concepts · Reviewer objections on methods · Epidemiology review

Checked before submission by data availability and accession reporting, which reports whether stated sources, registrations and identifiers are present and resolvable rather than assumed.

Review my manuscript

Frequently asked questions

What is PRISMA in a systematic review?

PRISMA in a systematic review is the reporting guideline the manuscript follows: a 27-item checklist plus a flow diagram specifying what the review must state about its search, selection, synthesis and limitations. PRISMA governs what appears on the page, not how the review was conducted or how good it is.

What does PRISMA stand for?

PRISMA stands for Preferred Reporting Items for Systematic reviews and Meta-Analyses. The current version, PRISMA 2020, was published in 2021 across five journals including BMJ (2021;372:n71). PRISMA replaced QUOROM, the 1999 meta-analysis reporting guideline, when the first PRISMA statement appeared in 2009.

How many items are in the PRISMA checklist?

The PRISMA 2020 checklist contains 27 items across seven sections, and many items divide into sub-items — synthesis methods alone splits into six (13a to 13f). A separate PRISMA 2020 for Abstracts checklist contains 12 items. The 2009 version also had 27 items, but they were not the same 27.

What is the PRISMA 2020 statement?

The PRISMA 2020 statement is the 2021 update to the PRISMA reporting guideline for systematic reviews and meta-analyses. PRISMA 2020 added items on certainty of evidence, competing interests and data availability, expanded the search-strategy item to cover every source, and revised the flow diagram to separate databases and registers from other search methods.

What is the PRISMA checklist used for?

The PRISMA checklist is used to report a systematic review completely enough that a reader can reproduce and audit it. Authors complete it while drafting, journals request it at submission with page numbers, and reviewers use it to locate omissions — most often the full search strategy, excluded studies, registration and certainty assessment.

What does a completed PRISMA checklist look like?

A completed PRISMA checklist is a table listing each of the 27 items against the page and line number where the manuscript addresses it, including explicit negative statements such as “review not registered” or “no meta-analysis performed”. Page numbers that point to content not containing the item are the standard sign of retrospective completion.

What kind of guideline is the PRISMA checklist?

The PRISMA checklist is a reporting guideline, not a quality assessment tool and not a method for conducting reviews. The PRISMA 2020 statement says it “should not be used to assess the conduct or methodological quality of systematic reviews; other tools exist for this purpose”. AMSTAR 2 rates methodological quality across 16 items with seven critical domains, and ROBIS assesses risk of bias in a review across four domains.