PRISMA 2020: A Practical Guide to Reporting Systematic Reviews and Meta-Analyses in 2026

The PRISMA 2020 statement was published simultaneously in the BMJ, PLOS Medicine, the Journal of Clinical Epidemiology, and several other journals in March and April 2021. The update was significant: a separate 12-item abstract checklist, new requirements for disclosing automation tools used in screening, mandatory full search strategies for every database searched, and added items covering certainty of evidence and data availability that simply did not exist in the 2009 version. Most major medical journals now list PRISMA 2020 as a submission requirement for systematic reviews and meta-analyses, and many ask authors to upload the completed checklist as a supplementary file.

Despite that, a 2025 meta-epidemiological study examining adherence to the expanded PRISMA 2020 checklist found that no systematic review in their sample met all criteria, and average adherence across all items was approximately 42.64 percent. Earlier work looking at reviews published before 2020 had already identified 11 items with adherence below 67 percent. The gap between what journals say they require and what they actually receive in submitted manuscripts remains wide, and peer reviewers are catching it more consistently as familiarity with the 2020 version has grown.

Why Medical Journals Are Enforcing This More Strictly Now

Systematic reviews occupy a specific and influential place in evidence hierarchies. Clinicians use them to justify treatment decisions. Guideline bodies use them to set recommendations. Funding agencies use them to identify research gaps. A systematic review that is poorly reported cannot be critically appraised, cannot be replicated, and cannot be trusted. The clinical stakes are different from those attached to a primary observational study, and journals have become more aware of that distinction.

The broader context matters here. The reproducibility conversations that have reshaped clinical trial reporting over the past decade have now reached evidence synthesis. Calls for preregistration of systematic reviews in PROSPERO, transparency about inclusion and exclusion decisions, and availability of underlying screening data all reflect the same pressure: the scientific record needs to be auditable. PRISMA 2020 responds to that pressure by building many of those transparency requirements directly into the reporting standard.

In practice, many journals that previously accepted a PRISMA 2009 checklist with a few ticks and page numbers have moved to requiring the expanded PRISMA 2020 checklist, and some editorial offices now verify at the desk-review stage that the checklist matches the manuscript. Journals in the BMJ Group, Wiley, Elsevier, and Springer Nature have all updated their author instructions to specify PRISMA 2020 for systematic reviews. The Cochrane Handbook, updated in 2024, aligns with PRISMA 2020 and adds Cochrane-specific guidance on top.

What Actually Changed from PRISMA 2009 to PRISMA 2020

Authors who wrote systematic reviews before 2021 and have not refreshed their understanding of PRISMA often assume the 2020 version is a light revision. It is not. The checklist still contains 27 items organized across seven sections: title, abstract, introduction, methods, results, discussion, and other information. But a substantial number of those items were modified or extended, and several important additions were made.

The most visible change is the addition of a standalone 12-item abstract checklist. The 2009 version referenced the need for a structured abstract but provided no checklist for it. PRISMA 2020 corrects that by specifying exactly what the abstract must include: title identification as a systematic review, background and objectives, data sources, eligibility criteria, appraisal methods, synthesis methods, results (including number of studies and participants), limitations, conclusions, and registration information. Research tracking adherence to PRISMA 2020 abstract items has found near-zero compliance for three of them specifically: risk of bias summary, funding and competing interests, and protocol registration. These are not obscure sub-items. They are items explicitly listed on a separate checklist that most journals are now attaching to their author instructions.

Within the main checklist, the search strategy item changed significantly. PRISMA 2009 required authors to present a full search strategy for at least one database. PRISMA 2020 requires full search strategies for all databases, registers, and websites searched, not just the primary one. Authors who publish a full MEDLINE search and summarize the Embase and CINAHL strategies in a sentence are not compliant with the 2020 standard. The rationale is defensible: databases have different controlled vocabulary, and a reader who wants to update the review or audit its search cannot do so without the actual queries.

Automation tools received their own place in the 2020 checklist. If your team used any software to assist with title-and-abstract screening, full-text retrieval, deduplication, or data extraction, that must now be disclosed in the methods alongside the human screening process. Tools such as Covidence, Rayyan, DistillerSR, and similar platforms need to be named. This change reflects the significant expansion of semi-automated evidence synthesis since 2009.

The protocol and registration item was moved from the Methods section to a new “Other Information” section at the end of the manuscript, with an added sub-item asking authors to describe any amendments to the registered protocol. This matters because protocol amendments are a common mechanism for introducing bias in systematic reviews, and transparency about them is necessary for readers to judge whether the analysis was truly prospective.

The Flow Diagram: More Structure Than Most Authors Use

The PRISMA flow diagram is the most visually familiar element of the standard, and it is also one of the most frequently completed incorrectly. PRISMA 2020 introduced four diagram templates rather than one. Which template applies depends on two factors: whether this is a new review or an update of a previous one, and whether the search included databases only or also incorporated other sources such as citation searching, trial registries, grey literature, or expert consultation.

Reviews that searched sources beyond electronic databases are now expected to report records from database searches and records from other sources separately in the identification phase, then combine them for the deduplication and screening stages. The practical effect is that authors who ran citation searches in Web of Science, contacted researchers for unpublished data, or searched grey literature databases such as OpenGrey or SIGLE cannot simply pool all their records from the start. They must show where the non-database records entered the flow and in what numbers.

The most common flow diagram error identified in audits is mismatched numbers: the total records identified does not equal the sum of screened records plus those removed before screening, or the number of included studies in the diagram does not match the number analyzed in the results. These arithmetic discrepancies are often the result of teams building the diagram from memory at the end of the project rather than populating it from the actual screening database. Generating the diagram directly from the screening tool, if the software permits, eliminates most of these inconsistencies.

A second common problem is missing exclusion reasons in the full-text assessment phase. PRISMA 2020 requires authors to list the specific reasons why full-text articles were excluded, with counts for each reason. Vague categories such as “not eligible” or “does not meet criteria” are not adequate. Reviewers and editors expect to see specific exclusion reasons tied to the eligibility criteria the team defined in the protocol.

Where Methods Sections Most Often Fall Short

The methods section carries the largest share of PRISMA 2020 items and, accordingly, accounts for most of the compliance failures identified in audits. Several patterns recur across specialty areas and journal tiers.

The eligibility criteria item requires authors to specify inclusion and exclusion criteria clearly, including the types of studies, participants, interventions, comparators, and outcomes. Many reviews describe the PICO framework in general terms in the introduction and then fail to operationalize it precisely in the methods. Authors sometimes note that “randomized controlled trials in adults with type 2 diabetes were included” without specifying how they handled cluster-randomized trials, crossover designs, minimum follow-up duration, or the age threshold for “adults.” These ambiguities matter when readers want to understand why certain studies were included and others were not.

Risk of bias assessment is reported in most reviews, but the reporting is frequently incomplete in two ways. First, authors often name the tool they used (RoB 2 for randomized trials, ROBINS-I for observational studies, QUADAS-2 for diagnostic accuracy) without describing how many reviewers rated each study, whether they worked independently, and how disagreements were resolved. Second, results are often summarized as overall risk of bias ratings without the domain-level breakdown that the tools are designed to produce. Both pieces are required by PRISMA 2020.

The reporting bias assessment item covers methods used to detect risk of bias across studies rather than within individual studies. For meta-analyses, this typically means describing whether funnel plots were used and whether any statistical tests for asymmetry (such as Egger's test) were planned. Many review teams conflate risk of bias within studies and reporting bias across studies, or skip the latter entirely when they have fewer than ten studies in the pool. Acknowledging that the sample size prevents formal testing is acceptable. Omitting the item is not.

Synthesis and Certainty: The Two Sections That Separate Reviews

The synthesis methods and certainty of evidence items are where PRISMA 2020 most clearly reflects methodological advances in evidence synthesis since 2009. Both items tend to separate well-conducted reviews from those that treat meta-analysis as a mechanical aggregation step rather than a judgment-heavy analytical process.

For the synthesis methods, PRISMA 2020 requires authors to describe all methods for synthesizing results, including the rationale for any quantitative synthesis. That means explaining why a meta-analysis was appropriate or why results were synthesized narratively. If meta-analysis was used, the statistical model must be specified (fixed-effect or random-effects, with the specific estimator for random-effects models, such as DerSimonian-Laird or REML). Heterogeneity measures should include at minimum the I-squared statistic and some estimate of between-study variance. Subgroup analyses, sensitivity analyses, and meta-regression approaches should be pre-specified in the protocol and described here.

The certainty of evidence item asks authors to describe their approach to rating confidence in the body of evidence at the outcome level, not just the study level. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework is by far the most commonly used approach in medical reviews, and many journals in clinical medicine and public health expect it. GRADE ratings assess certainty across five domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias. A meta-analysis that reports a pooled effect size without a GRADE table or narrative GRADE assessment is incomplete under PRISMA 2020, even if the statistical analysis is impeccable.

The confusion among many authors is that certainty of evidence is not the same as quality of evidence at the study level. It is a judgment about how much confidence readers should have in the synthesized estimate, after considering the entire body of evidence together. Many reviews still skip it, either because the team is unfamiliar with GRADE or because the peer reviewers in their specialty have not historically required it. As journals update their instructions, that tolerance is narrowing.

The Open Science Items Most Reviews Still Ignore

PRISMA 2020 includes a new item in the Other Information section asking authors to indicate whether data, analytic code, and other materials used in the review are publicly available and, if so, where. For many authors who learned systematic review methods before 2020, this feels like an unfamiliar addition. It reflects a broader shift across biomedical publishing toward treating evidence synthesis products as data assets with their own reproducibility requirements.

In practice, this item covers several distinct things. The screening database, which records all citations identified and the inclusion or exclusion decision for each, can be deposited in repositories such as the Open Science Framework or Figshare. The data extraction spreadsheet or form can be deposited similarly. If the meta-analysis was run in R or Stata, the analytic code should be available. None of this is currently mandatory at most journals, but the expectation is moving in that direction. Reviews published in journals like the Cochrane Database of Systematic Reviews, BMJ Evidence-Based Medicine, and Campbell Systematic Reviews have been ahead of this trend for years.

The competing interests item for the review authors themselves is separate from study-level conflict of interest reporting, which PRISMA 2020 also requires in the results section. Many authors declare no competing interests for themselves while faithfully documenting funding sources in the included primary studies. PRISMA 2020 treats both as required. A pharmaceutical company-funded systematic review of a drug in which the review authors have no personal financial interest is different in risk from a review conducted by a team with consultancy relationships with the manufacturer. Readers need both pieces to interpret the work.

Extensions in Development: NMA, Scoping Reviews, and Rapid Reviews

The PRISMA family of guidelines has always included extensions for specific review types that require additional reporting guidance beyond the core statement. The extension for network meta-analysis (PRISMA-NMA), published in 2015, and the extension for scoping reviews (PRISMA-ScR), published in 2018, were both developed before PRISMA 2020 and have not yet been updated to reflect the 2020 changes or subsequent methodological advances in their respective areas.

A scoping review published in JBI Evidence Synthesis in March 2025 identified the specific gaps between the existing PRISMA-NMA guidance and current network meta-analysis practice, particularly around automated search and synthesis tools, node-splitting methods for assessing inconsistency, and the use of treatment rankings. The working group leading the PRISMA-NMA update is using this scoping review as the basis for a revised checklist. A parallel process is underway for the scoping review extension, and for the first time a formal reporting guideline for rapid reviews is being developed as part of the same initiative. Authors conducting scoping reviews or rapid reviews should note that interim guidance recommends applying the main PRISMA 2020 checklist as closely as their methods allow, with any deviations justified.

Separately, there is growing interest in using large language models to assist with PRISMA compliance checking. An arXiv preprint published in late 2024 described a system for automated PRISMA 2020 adherence checking using LLMs. The accuracy of such systems varies substantially across checklist items, and none of them should be treated as a substitute for human review of the completed checklist. Their potential value is in flagging items that the manuscript does not address at all, reducing the chance that authors miss something before submission. No journal has yet announced that AI-based PRISMA screening is part of its editorial workflow, though that may change.

What to Do If Your Review Was Written Before 2020

Some authors are preparing to submit reviews that were conducted under PRISMA 2009 guidance, or that were updated from a protocol registered before the 2020 statement was published. The question of which checklist applies is worth clarifying with the target journal before submission, because policies differ.

As a general rule, journals that now specify PRISMA 2020 in their instructions expect authors to use the 2020 checklist regardless of when the review was conducted. If your search was run in 2019 and you cannot now provide full search strategies for all databases because the records are no longer reproducible exactly, describe that limitation explicitly. If your protocol was registered before PRISMA 2020 and did not include a GRADE assessment, explain that in the methods and conduct one if the data allow it. The checklist should reflect what you did, with honest acknowledgment of what the pre-2020 protocol constrained.

Updated systematic reviews present a specific challenge because PRISMA 2020 provides a separate flow diagram template for updates. The diagram must show how many records were identified in the original review, how many new records were identified in the updated search, and how both streams progressed through screening and inclusion. Teams that run a combined search across the original and updated periods without distinguishing the streams will need to use the combined diagram template rather than the update-specific one.

A Pre-Submission PRISMA Audit

The most reliable way to use PRISMA 2020 is to download the official checklists from the PRISMA statement website before you begin writing, treat them as a writing scaffold rather than an afterthought, and complete the checklist column with actual page numbers as each section is drafted. Authors who try to fill in the checklist after the manuscript is finished often discover that they addressed an item implicitly but not in the right section, or that they made an assumption without stating it explicitly anywhere.

The abstract checklist deserves particular attention because it is completed separately and is frequently overlooked entirely. When you write the abstract, open the 12-item abstract checklist alongside it and verify that each item is addressed. The registration number for the review protocol should appear in the abstract, not just in the main text or the Other Information section. This is the item with nearly zero compliance in recent audits, and it takes about thirty seconds to fix.

One final preparation step that often gets skipped: confirm the appropriate PRISMA extension for your review type before you complete the checklist. A scoping review completed with the standard PRISMA 2020 checklist and without a PRISMA-ScR checklist will draw a reviewer comment at virtually any journal that regularly handles evidence synthesis. The extensions are short and publicly available. There is no practical justification for submitting without the right one.

Systematic reviews take months or years to conduct. The PRISMA checklist takes hours to complete properly. The disproportion between the investment and the step that is most often skipped is striking. Submitting a well-conducted review with an incomplete checklist is the equivalent of preparing a careful clinical trial and omitting the CONSORT flow diagram. The work may be sound. The submission is not.

Further Reading