How to Make a CONSORT 2025 Flow Diagram: Every Box, Every Count, and What Changed from 2010

What a CONSORT flow diagram is, and which trials need one

CONSORT is the reporting guideline for the results of a randomised trial. The CONSORT 2025 statement is a 30-item checklist plus one diagram, and the diagram has its own checklist items: 22a asks, for each group, for the numbers of participants who were randomly assigned, received the intended intervention, and were analysed for the primary outcome; 22b asks, for each group, for losses and exclusions after randomisation, together with reasons.

Those two items are the whole job of the diagram. Everything else in the template is scaffolding for them.

CONSORT 2025 was published on 14 April 2025, simultaneously in The BMJ, JAMA, The Lancet, Nature Medicine and PLOS Medicine. It is explicit about what happens to the previous version: it "supersedes the CONSORT 2010 statement, which should no longer be used", and asks journal editors and publishers to update their instructions to authors accordingly. Fifteen months on, plenty of those instructions still say 2010, and plenty of templates circulating in trial units still are.

Whether the diagram is formally required depends on where you submit. CONSORT is endorsed by the World Association of Medical Editors, the International Committee of Medical Journal Editors and the Council of Science Editors; the ICMJE Recommendations name it as an example of a reporting guideline authors should follow. The CONSORT group itself goes further than the endorsements and recommends submitting a completed checklist with the manuscript, marked up with where each item is reported. Journals that co-published the 2025 statement are the obvious place this bites first, and if you are aiming at one of them, our JAMA impact factor and figure requirements guide covers what else that submission wants.

There is also evidence the diagram changes how completely a trial gets reported. The explanation and elaboration document cites a review of trials in general medical journals finding that reporting of what happened to participants and their data was considerably more thorough in articles that included a flow diagram than in those that did not. A separate Cochrane review of 50 evaluations covering 16,604 trials found 25 of 27 CONSORT checklist items were more completely reported in CONSORT-endorsing journals than in non-endorsing ones, though the statement is careful to note that a causal effect cannot be proven from that.

What changed in the flow diagram between CONSORT 2010 and 2025

Less than the version jump suggests, and more than a quick glance catches. The top half of the diagram is unchanged in wording. The bottom half gained a qualifier that changes what the numbers mean.

Three differences, in order of how much they matter:

"Analysed" became "Analysed for primary outcome." In 2010 the box asked for a number that many trials could not honestly give. A trial reporting a primary outcome at 12 months and four secondary outcomes at three time points has a different analysis population in almost every cell of its results table. CONSORT 2025 resolves this by splitting the question: the diagram carries the primary outcome, and item 26 asks for the rest in the results table, where each outcome gets both the number included in the analysis and the number with available data at that time point.

"Lost to follow up" became "Lost to follow-up for primary outcome." Same logic, same place in the reporting. Attrition is outcome-specific, and a single figure for it was always a compression.

Discontinuation now comes first in the follow-up box. Discontinuing the intervention and being lost to follow-up are different events, and the order now matches the sequence in which they usually happen. A participant who stops taking the drug but still attends every visit is a discontinuation and not a loss, and their outcome data still exists.

The phase label also changed spelling, from "Enrollment" to "Enrolment", and the four phase labels moved from rotated text down the left margin into rounded banners between the two arms. That is a template redraw rather than a reporting change, but it does mean a diagram in the old layout is visibly the old diagram.

Every box in the CONSORT 2025 flow diagram, in order

Labels below are verbatim from the CONSORT 2025 Word template, including its spacing inside (n= ). Both arms carry identical labels.

Enrolment

• Assessed for eligibility (n= )
• Excluded (n= ): Not meeting inclusion criteria (n= ), Declined to participate (n= ), Other reasons (n= )
• Randomised (n= )

Allocation, in each arm

• Allocated to intervention (n= )
• Received allocated intervention (n= )
• Did not receive allocated intervention (give reasons) (n= )

Follow-Up, in each arm

• Discontinued intervention (give reasons) (n= )
• Lost to follow-up for primary outcome (give reasons) (n= ):

Analysis, in each arm

• Analysed for primary outcome (n= )
• Excluded from analysis (give reasons) (n= )

Four boxes ask for reasons, and those four are where the diagram earns its place. A count with no reason beside it tells a reader that people left; a count with a reason tells them whether the departures were random.

Enrolment: the numbers before randomisation

The top box is the one most often guessed at, and it is also the one CONSORT treats most leniently.

Assessed for eligibility is the number of people evaluated for potential enrolment. The explanation and elaboration document is direct about its status: this count is relevant to external validity only, is "arguably less important than the other counts", and may simply not be known. A trial recruiting through a clinic that keeps no screening log genuinely cannot produce it. What the document asks is that you report it if available — not that you reconstruct it.

That matters because a fabricated screening number is worse than a missing one. If you did not count, leave the box out and say in the text that screening numbers were not recorded. A reviewer will accept that. A number that does not match your recruitment records will not survive a data query.

Excluded takes a total and three named reasons: not meeting inclusion criteria, declined to participate, other reasons. Those three are the template’s suggestion, and "other reasons" is a bucket you should open up when you can. A trial where 200 of 340 exclusions land in "other reasons" has not reported its exclusions.

Randomised is the count the whole rest of the diagram is measured against. The explanation and elaboration document calls it the crucial count for defining trial size and for assessing whether a trial was analysed by intention-to-treat.

Allocation, follow-up and analysis: what item 22 actually asks for

The explanation and elaboration document sets out the required counts stage by stage, with the rationale for each. This is the table to build your diagram from, because it names both what goes in and what comes out at every stage.

Two additions the expanded checklist makes for particular trial types are worth knowing before you lay the diagram out, because they change its shape. For non-pharmacological interventions, where appropriate, report the number of care providers or centres performing the intervention in each group, and the number of participants treated by each provider or centre. That is a row the two-arm template has no space for, and it is easier to plan for than to retrofit.

What about trials with three or more arms?

Add a column. The template is drawn for two.

CONSORT 2025 says it provides guidance for reporting all randomised trials but focuses on the most common type, the two-group parallel design. The explanation and elaboration document is explicit that the diagram is not fixed: "Although CONSORT strongly recommends using a flow diagram to communicate participant flow throughout the study, there is no specific, prescribed format", and it will sometimes be useful or necessary to adapt the structure to a particular trial. A three-arm trial with three parallel columns below one "Randomised" box is an adaptation, not a deviation.

The checklist has more to say than the diagram does. A separate CONSORT extension covers multi-arm parallel-group trials — Juszczak et al., JAMA 2019;321(16):1610–20 — expanding on ten items of the checklist. Its recommendations are about text rather than boxes: identify the trial as multi-arm, state objectives and hypotheses referring to all treatment groups, identify the primary treatment comparisons, and report planned and unplanned comparisons completely, with the rationale and method for any multiplicity adjustment.

One wrinkle to know about: that extension, like the other CONSORT extensions, was written against CONSORT 2010. The 2025 statement says the group will work with extension leaders on aligning them, and that in the meantime readers should use the existing version of the relevant extension. So for a multi-arm trial today you are working from the 2025 checklist and the 2019 extension together, and where the extension refers to an old item number you follow the substance rather than the numbering.

CONSORT is for randomised trials. What observational studies use instead

consort diagram for observational studies is a query people type every month, and the honest answer is that there is no such thing. CONSORT covers randomised trials. Labelling a cohort or cross-sectional study’s flow diagram "CONSORT" is a category error a methods reviewer will spot on the first page.

The guideline you want is STROBE, whose item 13 covers exactly the same ground:

(a) Report the numbers of individuals at each stage of the study — e.g. numbers potentially eligible, examined for eligibility, confirmed eligible, included in the study, completing follow-up, and analysed. (b) Give reasons for non-participation at each stage. (c) Consider use of a flow diagram.

So the figure is expected, the counts are specified, and the stage names are different. Build the diagram STROBE describes, call it a participant flow diagram or a study flow diagram, and cite STROBE rather than CONSORT. For a retrospective study the same applies, with the stages renamed to match how the cohort was actually assembled.

Systematic reviews are a third case again, with their own template and their own vocabulary of records, reports and studies — our PRISMA 2020 flow diagram guide works through that one box by box.

Where to get the CONSORT 2025 template, and four routes to building one

The address changed, and this is the single most common way people end up with an outdated diagram. consort-statement.org now redirects to consort-spirit.org, the joint site the SPIRIT and CONSORT groups moved to after merging into one group. Links to the old paths are still everywhere, including in the ICMJE Recommendations, and copies of the 2010 template are still the first result on plenty of institutional pages.

The current site offers the flow diagram as an editable Word document (CONSORT_2025_flow_diagram.docx), an editable CONSORT 2025 checklist, and an expanded checklist PDF that breaks each of the 30 items into the specific elements to report.

The R route deserves a mention beyond the table if you already work in R. The consort package on CRAN, version 1.2.3 under an MIT licence, builds the diagram from a standardised disposition dataset rather than from typed labels. That means the diagram is regenerated when the data changes, which removes the entire class of error this article’s next section is about.

If you would rather not choose a tool blind, our comparison of figure tools that actually export publication-grade vector covers which of them produce EPS and PDF that survive a production department, and which only claim to.

Why your numbers do not reconcile

A methods reviewer reads a flow diagram by adding it up. Four checks catch almost everything.

Randomised = allocated in arm one + allocated in arm two. No exceptions, ever. A participant who was randomised and then found ineligible is still randomised, and still appears in an allocation box. If these two numbers differ, somebody has quietly removed people from the top of the diagram.

Allocated = received + did not receive, in each arm. This is the check that exposes a diagram built from the analysis dataset rather than from the randomisation log.

Analysed may differ from randomised, but every difference needs a reason. The explanation and elaboration document is blunt about the standard here. Participants excluded after randomisation because they were found not to meet eligibility criteria should still appear in the flow diagram. And the phrase that will not do the work: "Use of the term 'protocol deviation' in published articles is not sufficient to justify exclusion of participants after randomisation." Name the deviation and give the exact reason.

Attrition and exclusion are different things and must not be merged. Loss to follow-up is often unavoidable and happens to the participant. Exclusion is a decision the investigators made. The document notes that erroneous conclusions can be reached when participants are excluded from analysis, and that imbalances in such omissions between groups may be especially indicative of bias. Merging the two into one box hides exactly the signal a reader is looking for.

One more consistency check that sits outside the diagram: item 21b asks you to define who is included in each analysis and in which group. That sentence in your statistical methods and the "Analysed for primary outcome" box have to describe the same population. When they disagree, it is usually the diagram that was drawn first and never revisited.

Can you reproduce the CONSORT flow diagram in your paper?

Two documents govern this differently, and the difference is worth understanding before you paste a blank template into a slide deck.

The journal versions of the statement are open access. The PLOS Medicine version, including its Figure 1, is published under a Creative Commons Attribution licence, which permits unrestricted use, distribution and reproduction in any medium provided the original authors and source are credited. Copyright sits with Hopewell et al., 2025.

The SPIRIT–CONSORT website’s terms of use are narrower. They state that all materials on the site are copyrighted and may not be used except as provided in the terms, that materials may be downloaded or copied provided all copies retain the copyright and proprietary notices, and that no material may be modified, edited or taken out of context such that its use creates a false or misleading statement or impression. That is a permission to copy, not an open licence, and it says nothing explicit about adaptation.

In practice this rarely constrains an author, because of what you are actually publishing. Your flow diagram is your own figure: your counts, your arms, your reasons, built to a published structure. Cite the CONSORT 2025 statement in your reference list and the question does not arise. The licensing question only becomes live when you reproduce the blank template as-is — in teaching material, a protocol appendix, a methods handbook — and there the answer is to take it from one of the open access journal versions and carry the attribution with it.

Formatting it for submission

CONSORT publishes no artwork specification. No resolution, no width, no minimum point size, no file format appears in the CONSORT 2025 statement, the explanation and elaboration document, the expanded checklist or the flow diagram template. All four were checked. The guideline governs content; the journal governs the artwork.

Which journal, then, decides the numbers. PLOS Medicine, one of the five that co-published the statement, is specific:

Two of those rows deserve attention for this figure in particular.

The type floor is the binding constraint. A CONSORT diagram is almost entirely text, and it is usually the widest figure in the manuscript. At an 8 pt floor and a 19.05 cm ceiling, a two-arm diagram fits comfortably; a four-arm diagram with long reason lists does not, and the fix is to shorten the labels rather than to shrink the type below what the journal accepts.

It has to be vector. Boxes, rules and small text are exactly the content that turns to mush when rasterised at the wrong size. If you built the diagram in Word, export the page as PDF rather than screenshotting it, and check the result by zooming to 400% — text that stays sharp is vector, text that softens is not. Our guide to making scientific figures that survive reduction works through the type sizes and line weights this runs into first.

Frequently asked questions

What are the boxes in a CONSORT 2025 flow diagram? Assessed for eligibility; Excluded, with the three reasons; Randomised; then in each arm Allocated to intervention with received and did not receive; Discontinued intervention and Lost to follow-up for primary outcome; and Analysed for primary outcome with Excluded from analysis. The phases are Enrolment, Allocation, Follow-Up and Analysis.

What changed in the CONSORT flow diagram in 2025? Two labels gained "for primary outcome", the follow-up box lists discontinuation before loss to follow-up, and "Enrollment" is spelled "Enrolment". The top half of the diagram is unchanged in wording.

Is CONSORT 2010 still valid? No. CONSORT 2025 states that it supersedes CONSORT 2010, which should no longer be used.

Where do I download the CONSORT 2025 flow diagram template? From consort-spirit.org. The old consort-statement.org address redirects there.

Do I need a CONSORT diagram for an observational study? No — that is STROBE’s territory. STROBE item 13 asks for the numbers at each stage, reasons for non-participation, and says to consider using a flow diagram.

How do I draw a CONSORT diagram for three groups? Add a column. The template covers the two-group design, and CONSORT states that no specific format is prescribed and the structure may be adapted to the trial.

Can I publish the CONSORT flow diagram in my paper? Your completed diagram is your own figure; cite the CONSORT 2025 statement. To reproduce the blank template, use one of the open access journal versions, which carry a CC BY licence.

What resolution does a CONSORT flow diagram need? CONSORT does not say. The journal does — PLOS Medicine, for instance, asks for 300–600 dpi TIFF or EPS with 8–12 pt type.

Where to go next

If you are building the diagram now, download the 2025 template from consort-spirit.org, fill in the counts from your randomisation log rather than your analysis dataset, and run the four arithmetic checks above before you spend a minute on styling.

If your review is a systematic review rather than a trial, the PRISMA 2020 flow diagram is the equivalent document, with four templates and a different counting vocabulary.

If the diagram is written but the file is not, what Scientific Reports and the Nature Portfolio require of a figure is the closest thing to a default specification when your target journal is vague about artwork.