PRISMA 2020

What PRISMA 2020 is

PRISMA stands for Preferred Reporting Items for Systematic reviews and Meta-Analyses. The 2020 statement updated the original 2009 version to reflect advances in the methods of review, including new ways of searching and screening, additional synthesis methods and the importance of reporting data and code. It was developed by an international group through a literature review, a survey, a Delphi process and a consensus meeting, and it was published with an explanation and elaboration paper that gives the rationale and an example for each item.

The aim of the guideline is transparency. A systematic review makes a claim to be a reliable summary of the evidence, and readers can only judge that claim if the report shows how the review was done. Incomplete reporting hides weaknesses and prevents others from checking or repeating the work. PRISMA lists the information that should be reported, so that a reader can see what was searched, how studies were chosen, how they were appraised and combined, what was found and how certain it is. Many journals require authors to submit a completed checklist and a flow diagram, and editors and reviewers use it to judge completeness. The method it reports is described on the page for systematic review.

The parts of the guideline

The checklist

Twenty-seven items, some with sub-items, grouped under title, abstract, introduction, methods, results, discussion and other information. Each is something a review should report.

The flow diagram

A diagram of the identification, screening and inclusion of studies, with numbers at each stage and reasons for exclusion at full-text stage. There are two templates: one for new reviews and one for updated reviews, and each can show databases and registers and other methods.

The abstract checklist

A separate checklist of items for the abstract, for reviews reported in journal abstracts or conference abstracts.

The explanation and elaboration

A companion paper that explains each item, gives the evidence for it, and provides examples of good reporting.

The 27 items

Several items have sub-items. For example, the synthesis methods item asks authors to say how studies were chosen for each synthesis, how data were prepared, how results were tabulated and visualized, which model and methods were used, how heterogeneity was explored and which sensitivity analyses were done. The full wording is in the published statement, and authors should work from the official checklist and not from a summary.

What good reporting looks like, item by item

Eligibility criteria and sources

The criteria are stated in a form that a reader could apply, with the reason for any restriction, for example language or date. Every database, register and other source is named with the date it was last searched.

Search strategy

The full strategy for each database is given, usually as a supplement, exactly as run, with limits and filters, so that it could be repeated.

Selection and data collection

The report says how many reviewers screened and extracted, whether independently, how disagreements were settled, and whether automation tools were used, with what checks.

Risk of bias and certainty

The tool is named for each design, with the number of assessors and how judgments were used. The method for rating certainty is named, and the results are tabulated.

Synthesis

The model and estimator, the software and version, the method for exploring heterogeneity, the planned sensitivity analyses and the handling of missing data are stated, so that the numbers can be reproduced.

Results of syntheses

Each synthesis is reported with its summary estimate and interval, the measures of heterogeneity and, where used, a prediction interval, with forest plots.

Discussion

The limitations of the evidence, such as risk of bias and inconsistency, are separated from the limitations of the review process, such as the restriction of the search.

Registration, funding and shared materials

The last group of items goes beyond the content of the review to its provenance. Authors report whether the review was registered, where, and under what number, and whether a protocol was prepared, with any amendments and the reasons for them. They report the sources of financial and non-financial support and the role of any funder, and the competing interests of the authors. And they report which of the data, code and other materials used in the review are publicly available and where, including the extracted data, the data used in analyses, the analytic code and the template data collection forms. This last item reflects a movement toward open evidence synthesis, in which readers can check and reuse the work. Sharing the extraction dataset and the analysis script costs little and makes errors easier to find, which is a benefit to authors as well as readers. The practice is discussed in the guide on reproducible meta-analysis.

PRISMA and journals

Most journals that publish systematic reviews endorse PRISMA, and many require the checklist at submission, either uploaded as a file or entered into the submission system. Some journals ask that the abstract follow the PRISMA abstract checklist and that the title include the words systematic review or meta-analysis, so that the paper can be found and indexed correctly. Editors and peer reviewers use the checklist as a screening tool: a review that omits the search strategies or the flow diagram is likely to be returned. Journals differ in the strictness with which they check, so authors should follow the instructions of the journal they choose. A review that follows the guideline carefully is easier to review, and a manuscript prepared to PRISMA from the start needs less revision.

What changed from 2009

Both the 2009 and the 2020 checklists have 27 items, but the 2020 version has sub-items and its content changed substantially. New or strengthened requirements ask authors to describe the methods used to assess the certainty of evidence, to report any automation tools used in selection and data collection, to give the full search strategies for every database and not just one, to cite studies that were excluded at full-text stage, and to say what data, code and other materials are available. The flow diagram was revised to separate records identified from databases and registers from those identified through other methods, and a version for updated reviews was added. The guideline also now assumes that a review may use a range of synthesis methods, with or without meta-analysis, and it asks for the methods used to explore heterogeneity, to test sensitivity and to assess reporting bias. The aim was to match the guideline to current practice.

The flow diagram

The flow diagram is the visual summary of study selection. For a new review it shows in order the number of records identified from each database and register, and from other methods such as citation searching and websites; the number removed before screening as duplicates or for other reasons; the number screened by title and abstract and the number excluded; the number of reports sought for retrieval and those not retrieved; the number assessed for eligibility at full text and the number excluded with reasons; and the number of new studies and reports included. A second column of the template handles other methods. A reader can check that the numbers add up, see how selective the criteria were, and see how many included studies came from searches other than the main databases. The diagram should be accurate, because arithmetic that does not agree is a sign of problems in the conduct or reporting of a review.

Using the checklist well

The checklist is a tool for authors and not a form to be completed at the end. It works best when it is used from the start, as the plan for the report, and again at the end, as a check. For each item, the author finds where in the manuscript the information is, and writes the page or section number in the checklist. If it is not there, it should be added. A checklist filled in with page numbers that do not lead to the information is worse than none, because it gives a false impression of completeness, and editors notice.

The guideline allows for items that do not apply. A review without a meta-analysis will not report the effect measure used for pooling, and an item about a registered protocol does not apply if none exists, in which case the authors say so. Where an item is not met, the honest course is to state this, with the reason, and not to leave it blank. The aim is to make the report complete, and not to make the review look better than it is. See the guide on reporting a meta-analysis.

Extensions

PRISMA has several extensions for particular types of review or parts of reporting. PRISMA-P covers protocols, PRISMA-S covers the reporting of searches, PRISMA-ScR covers scoping reviews, PRISMA-DTA covers diagnostic test accuracy reviews, the extension for network meta-analysis covers reviews that compare multiple interventions, and PRISMA-IPD covers reviews using individual participant data. Others cover equity, harms, abstracts and further topics. An extension adds items to the main checklist, and a review of that type should use both. The guide on PRISMA extensions explains which fits which review. Guidelines for other designs, such as MOOSE for observational meta-analyses, may also apply.

Common misunderstandings

• PRISMA is not a quality measure. A complete report can describe a poor review, and a high-quality review can be poorly reported. PRISMA addresses reporting only. Tools such as AMSTAR 2 and ROBIS assess the conduct and risk of bias of reviews.
• PRISMA is not a method. It does not say how to search, screen, appraise or synthesize, although it lists what should be reported about each.
• It is not only for meta-analyses. It applies to systematic reviews, whether or not studies are pooled.
• A checklist is not a substitute for reading the guideline. The explanation and elaboration paper gives the reasoning and examples.
• It does not apply to all reviews. Scoping reviews, qualitative syntheses and other designs have their own guidance.

Support with PRISMA reporting

Frequently asked questions