What is the CONSORT checklist?

The CONSORT checklist is a reporting standard for randomised controlled trials, listing the minimum information a trial report must contain. CONSORT 2025, published on 14 April 2025, comprises 30 items covering the title, abstract, open science apparatus, methods, results and discussion, alongside a participant flow diagram. CONSORT governs what is reported, not how a trial is designed or conducted.

That distinction carries most of the practical weight. A trial can satisfy every item and still be a poor trial; a strong trial can fail half the items because nobody wrote the numbers down. This page covers where the checklist came from, what its 30 items ask for, what changed in the 2025 revision, how to check a manuscript against it, which items fail most often, what reviewers say when they have failed, and what the evidence does and does not show about whether the checklist works.

Where the CONSORT checklist came from

CONSORT stands for Consolidated Standards of Reporting Trials, and the word “consolidated” records a merger. Two groups worked on trial-reporting standards independently in the early 1990s — the Standards of Reporting Trials group and the Asilomar Working Group on Recommendations for Reporting of Clinical Trials in the Biomedical Literature — and each published a proposal that journals could plausibly have adopted separately. A joint meeting consolidated the two, and the first CONSORT statement appeared in JAMA in 1996 under Begg and colleagues.

Three revisions followed: CONSORT 2001, CONSORT 2010 (Schulz, Altman and Moher), and CONSORT 2025 (Hopewell and colleagues, JAMA, 2025). Each revision was driven by accumulating methodological evidence rather than by editorial taste, and each was published simultaneously across several journals — CONSORT 2025 appeared in the BMJ, JAMA, The Lancet, Nature Medicine and PLOS Medicine on the same day, which is itself a dissemination tactic, since a guideline that lives behind one paywall gets ignored.

CONSORT is maintained alongside SPIRIT, the corresponding standard for trial protocols. SPIRIT 2025 and CONSORT 2025 were developed together by a single merged group — the two executive groups met in Oxford in January 2020 and formed one group — and their items are aligned where the protocol and the report ask for the same thing. SPIRIT 2025 was published separately, on 28 April 2025, two weeks after CONSORT 2025. The practical consequence for an author is that a protocol written to SPIRIT supplies most of the methods section a CONSORT-compliant paper needs.

What the 30 items of CONSORT 2025 cover

CONSORT 2025’s 30 items run in manuscript order, so the checklist can be read as a table of contents for a trial report. Item 1a concerns the title and item 1b the structured abstract: the title must identify the report as a randomised trial, so that it can be found by anyone searching for one, and the abstract must summarise the trial’s design, methods, results and conclusions.

Items 2 to 5 form a new open-science section. Item 2 asks for the trial registry, registration number and date of registration; item 3 asks where the protocol and the statistical analysis plan can be accessed; item 4 asks where and how the de-identified participant data, the statistical code and any other materials can be accessed; item 5a asks for sources of funding and item 5b for the financial and other conflicts of interest of the authors. The methods items open at item 8, which asks whether patients or members of the public were involved in the design, conduct or reporting of the trial.

The methods items carry the bulk of the checklist. They ask for the trial design and any important changes made after it commenced (item 10), eligibility criteria for participants and, separately, for sites and the people delivering the intervention (item 12b), the interventions in replicable detail, the outcomes with their definitions and timepoints, how harms were assessed (item 15), how the sample size was determined, sequence generation, allocation concealment, implementation, blinding, and the statistical methods — including who is included in each analysis and in which group (item 21b), how missing data were handled (item 21c), and which analyses were pre-specified rather than post hoc (item 21d).

The results items ask for participant flow, recruitment dates, baseline characteristics, intervention delivery and concomitant care (item 24), the numbers analysed and the numbers with available data at the outcome timepoint together with effect estimates and their precision (item 26), harms, and ancillary analyses distinguished as pre-specified or exploratory (item 28). The discussion items ask for an interpretation consistent with the results and, at item 30, the trial’s limitations.

What changed between CONSORT 2010 and CONSORT 2025

CONSORT 2025 added seven items, revised three, deleted one, absorbed material previously scattered across CONSORT extensions, and reorganised the remaining items into manuscript order — taking the checklist from 25 items to 30. The development process ran a scoping review, a three-round Delphi survey with 317 participants, and a consensus meeting of 30 international participants — a scale worth knowing, because it is the answer to a reviewer who treats the checklist as one committee’s opinion.

The seven additions name what a decade of meta-research found missing. Data sharing arrives as item 4, which asks for the plan for de-identified participant data, statistical code and materials — the reporting counterpart of a data availability statement that names a repository and an accession rather than promising availability on request. Conflicts of interest arrive at item 5b. Patient and public involvement arrives at item 8. Eligibility criteria for sites and intervention deliverers arrive at item 12b, which matters because a trial of a surgical or behavioural intervention is a trial of the people delivering it. Harms assessment arrives at item 15 as a methods item, not only a results item. Analysis population and missing data arrive at items 21b and 21c. Intervention delivery and concomitant care arrive at item 24.

Three items were completely revised: protocol reporting now includes access to the statistical analysis plan, item 10 now asks explicitly for outcomes and analyses that were not pre-specified, and item 26 now separates the number analysed from the number with available data. One item was deleted — the standalone generalisability item, whose content moved into the limitations item at 30.

The participant flow diagram

The CONSORT flow diagram is a required figure showing the number of participants at each of four stages: enrolment, allocation, follow-up and analysis, with losses and exclusions and their reasons given for each arm. CONSORT 2025 keeps the diagram alongside the 30-item checklist rather than treating it as optional decoration.

The diagram is the single most information-dense object in a trial report, and it is arithmetic, which means it can be checked. Numbers assessed for eligibility minus those excluded should equal those randomised. Randomised should equal the sum of the arms. Each arm’s randomised total minus discontinuations should reconcile with the number analysed, and where it does not, the gap is either an exclusion the text has not explained or an analysis population the methods have not defined.

Two recurring defects live here. The first is a diagram whose analysis boxes disagree with the denominators in the results tables — often because a per-protocol population was used without saying so. The second is a diagram that gives losses to follow-up without reasons, which converts a checkable statement into an unfalsifiable one; where dropout differs between arms, the reasons are the evidence that the difference is not informative. Neither defect requires expertise to find, which is why both are standard early checks in desk screening at scale.

The CONSORT extensions

CONSORT extensions adapt the core checklist to designs and interventions the parallel-group standard does not fully cover, and using the wrong one is a real reporting failure rather than a formality. More than twenty exist. The ones a clinical author meets most often are the extension for cluster randomised trials, the extension for non-inferiority and equivalence trials (Piaggio and colleagues, JAMA 2006, updated 2012), the extension for pragmatic trials, CONSORT for abstracts, CONSORT-PRO for patient-reported outcomes, CONSORT Harms, and CONSORT-AI for trials of interventions involving artificial intelligence.

Extensions add items; they do not replace the core checklist. A cluster randomised trial reports the 30 core items and, in addition, the rationale for cluster randomisation, the number of clusters, the average cluster size and the intracluster correlation coefficient used in the sample-size calculation. A non-inferiority trial reports the core items and, in addition, the non-inferiority margin with its derivation and both analysis populations.

Choosing the extension is a decision made at the design stage, not at submission. A trial that randomised clinics but analysed patients has a reporting problem that the cluster extension will expose and a statistical problem the extension cannot fix. The same holds for competing risks in a trial with mortality and a non-fatal primary endpoint: the checklist asks for the estimand, and a Kaplan–Meier curve censoring death does not supply it.

What the CONSORT checklist is not

The CONSORT checklist is not a quality-assessment instrument, and treating a completed checklist as evidence of trial quality is the most common misuse. Risk-of-bias tools such as Cochrane’s RoB 2 assess whether a trial’s results are likely to be biased; CONSORT assesses whether the report says what happened. A trial with unconcealed allocation that describes its unconcealed allocation clearly satisfies the item and remains at high risk of bias.

CONSORT is also not the right guideline for most study designs. Observational studies use STROBE, systematic reviews use PRISMA 2020, diagnostic accuracy studies use STARD 2015, preclinical animal work uses ARRIVE 2.0, and trial protocols use SPIRIT 2025. Applying CONSORT to a single-arm phase I study or a cohort analysis produces a checklist full of not-applicable rows and no useful discipline.

Finally, CONSORT is not a journal policy. Journals endorse it in their instructions to authors with widely varying force — some require a completed checklist with page numbers at submission, some mention it, some check nothing — so read the specific instructions for the specific journal rather than assuming, as part of ordinary journal submission preparation.

How to check a manuscript against the CONSORT checklist

Checking a manuscript against the CONSORT checklist is a reconciliation exercise, not a reading exercise, and it works because most items ask for something that can be located or shown absent in seconds. Work through the following, in this order, before the prose is polished.

Reconcile the flow diagram against the results tables. Take each arm’s analysed count from the diagram and find the same number as a denominator in the primary outcome table. Where they differ, the manuscript owes a sentence naming the analysis population.

Compare the primary outcome in the methods against the result the abstract leads with. These should be the same outcome, at the same timepoint, in the same population. Where they differ, the paper has an endpoint pre-specification problem that no amount of editing will hide, because the registry record is public.

Find the sample-size paragraph and check it supplies its inputs. CONSORT 2010 item 7a asked for one thing — “How sample size was determined” — and a paragraph that gives only the resulting number answers none of it. The assumed effect, its source, the assumed variability or event rate, the alpha, the target power and any attrition inflation should all be present. A paper that instead computes power from the observed effect has produced a post hoc power calculation that carries no information.

Check that sequence generation and allocation concealment are described separately. “Patients were randomised using a computer-generated list” describes generation only and leaves concealment unstated.

Check that every effect estimate carries its precision. CONSORT 2010 item 17a requires, verbatim, “For each primary and secondary outcome, results for each group, and the estimated effect size and its precision (such as 95% confidence interval)”, and CONSORT 2025 item 26 carries the requirement forward, adding both absolute and relative effect sizes for binary outcomes. A results section built on p-values alone fails the item on every line.

Check that additional analyses are labelled. CONSORT 2010 item 18 asks for “Results of any other analyses performed, including subgroup analyses and adjusted analyses, distinguishing pre-specified from exploratory”, and CONSORT 2025 asks the same at item 28. An unlabelled subgroup analysis will be read as post hoc, and a paper reporting several of them invites a multiple comparisons objection it could have pre-empted with one clause.

Check harms reporting has denominators. Per-arm counts with denominators, including for participants who received no intervention, and the same seriousness definitions applied in both arms.

The items that most often fail

Two CONSORT items fail more often than the rest, and both have measured consequences rather than merely aesthetic ones.

Allocation concealment. Schulz and colleagues showed in JAMA in 1995 that trials with inadequately concealed allocation produced treatment-effect estimates exaggerated by around 41%, and those with unclearly described concealment by around 30%. Concealment is not blinding and not sequence generation: it is the mechanism preventing the person enrolling a participant from knowing which arm comes next. Sequentially numbered opaque sealed envelopes, a central telephone or web randomisation service, and pharmacy-controlled allocation are describable mechanisms; “randomly allocated” is not.

Outcome reporting. Chan and colleagues, comparing published trial reports against their protocols in JAMA in 2004, found that at least one primary outcome was changed, introduced or omitted in 62% of trials, and that statistically significant outcomes were substantially more likely to be fully reported than non-significant ones. This is the empirical basis for CONSORT 2025’s revised item 10, which asks explicitly for outcomes and analyses that were not pre-specified.

A third failure is less studied but easy to spot: baseline characteristics presented with p-values comparing the arms. Significance testing of baseline imbalance in a randomised trial tests a null hypothesis known to be true by design, and imbalance matters through its prognostic strength rather than its p-value — which is a question about confounding, not about testing.

What a peer reviewer says when the checklist has been mishandled

Peer reviewers rarely cite CONSORT item numbers. They restate the missing item as a question about what the authors actually did, and the phrasings recur across specialties.

“The method of allocation concealment is not described.” “The flow diagram does not account for all randomised participants; please reconcile the numbers analysed with Table 2.” “The primary outcome in the manuscript differs from that registered; please explain when and why this changed.” “It is unclear whether the analysis was by intention to treat; please state which participants were included in each analysis and in which group.” “Please report effect estimates with confidence intervals rather than p-values alone.” “Subgroup analyses should be identified as pre-specified or exploratory.” “Harms are reported narratively; please provide per-group counts with denominators.” “The sample size calculation does not state the assumed effect or its source.” “The conclusion is not supported by the primary outcome result.”

The last of these is the one that decides papers, because it is not a reporting complaint at all — it is the reviewer saying the abstract claims more than the trial showed, which is the failure an overclaim check exists to catch before submission. A reviewer will forgive a missing item and ask for it. A reviewer will not forgive a conclusion that survives only because the missing item was missing.

Does the CONSORT checklist actually improve reporting?

Evidence that CONSORT improves reporting exists and is weaker than its reputation. A Cochrane methodology review led by Turner and colleagues in 2012 found reporting completeness higher in journals endorsing CONSORT than in journals that did not, for several individual items including allocation concealment. The review’s own assessment was that the evidence is observational: journals that endorse CONSORT differ from journals that do not in ways other than endorsement, and no randomised comparison of endorsement exists.

Two limitations deserve stating plainly. First, endorsement is not enforcement — studies of adherence in endorsing journals routinely find that a substantial proportion of core items remain unreported, and a required checklist is often completed with page numbers that do not contain the item. Second, and more fundamentally, a reporting guideline can only change reports. Better description of an unconcealed allocation is better reporting and the same trial.

There is also a live critique that checklists encourage a compliance posture in which authors satisfy the letter of each item while leaving the substantive judgement — whether the estimand matches the question, whether the analysis population was decided before the data were seen — untouched. That critique is fair, and the answer is not to abandon the checklist but to treat it as the floor. CONSORT establishes that a primary outcome is stated. It does not establish that the stated outcome is the one the trial was designed around, and finding that requires reading two parts of the paper against each other, which is where structured clinical trial review earns its place.

How to use the CONSORT checklist while writing

Use the CONSORT checklist as an outline before drafting rather than as an audit after submission, because roughly a third of its items ask for facts that are easy to record while a trial runs and near-impossible to reconstruct afterwards. Recruitment start and end dates, the reason each participant discontinued, who generated the allocation sequence, and whether an analysis was decided before or after unblinding all decay quickly in institutional memory.

Write the methods section from the protocol and the statistical analysis plan directly, in checklist order. Where the trial deviated, write the deviation as its own sentence at item 10 rather than silently reporting what was done — a stated, explained change draws a reviewer comment; an unstated one that a reviewer finds in the registry can sink the paper.

Complete the checklist document itself last, with real page numbers, and verify each one by opening the page. A checklist pointing at pages that do not contain the item is a worse signal than no checklist. The same applies to the data availability and ethics statements, whose specificity is checked directly under ethics and data sharing review, and to any subgroup analysis by sex, ancestry or socioeconomic group, which now carries its own reporting expectations under health equity reporting.

Related

Clinical trial review · Trial registration and endpoints · Post hoc power · Multiple comparisons · Journal submission

Review my manuscript

Frequently asked questions

What does CONSORT stand for?

CONSORT stands for Consolidated Standards of Reporting Trials. The name records the 1990s merger of two independent trial-reporting initiatives into a single standard, first published in JAMA in 1996. The current version, CONSORT 2025, is a 30-item checklist plus a participant flow diagram specifying the minimum content of a randomised trial report.

How many items are in the CONSORT checklist?

CONSORT 2025 contains 30 items, up from 25 in CONSORT 2010. The 2025 revision added seven items, revised three completely, deleted one — the standalone generalisability item, whose content moved into the limitations item — and reorganised the remainder into manuscript order. A participant flow diagram accompanies the checklist and is required rather than optional.

What is CONSORT 2025?

CONSORT 2025 is the current version of the reporting standard for randomised trials, published on 14 April 2025 by Hopewell and colleagues simultaneously in the BMJ, JAMA, The Lancet, Nature Medicine and PLOS Medicine. It was developed through a scoping review, a three-round Delphi survey of 317 participants, and a consensus meeting of 30 international participants.

What is the CONSORT checklist used for?

The CONSORT checklist is used to write and to check a randomised trial report, specifying the minimum information the report must contain so that readers can judge the trial’s validity and reproduce its methods. Many journals ask authors to submit a completed checklist with page numbers indicating where each item appears in the manuscript.

Is the CONSORT checklist a quality score?

CONSORT is a reporting checklist, not a quality score. It assesses whether a trial report states what was done, not whether what was done was sound. Risk-of-bias tools such as Cochrane’s RoB 2 assess bias. A trial with unconcealed allocation that describes that clearly satisfies the item and remains at high risk of bias.

Do I have to follow the CONSORT checklist?

Journal requirements for the CONSORT checklist vary. Many biomedical journals endorse CONSORT in their instructions to authors, some require a completed checklist at submission, and enforcement ranges from strict to nominal. Check the specific journal’s instructions rather than assuming. Following the checklist regardless pre-empts most of the revision requests its items were written to prevent.

What is the CONSORT flow diagram?

The CONSORT flow diagram is the required figure accompanying the checklist, giving participant numbers at four stages — enrolment, allocation, follow-up and analysis — with exclusions, discontinuations and their reasons for each arm. Because it is arithmetic, it can be reconciled against the denominators in the results tables, and disagreement between the two is a common reviewer finding.