The CONSORT statement is an evidence-based reporting guideline that sets out the minimum information a published randomised controlled trial should contain, so that readers can judge how the trial was designed, run, and analysed. It consists of a 25-item checklist covering the title, abstract, methods, results, and discussion, together with a participant flow diagram that tracks people from enrolment through allocation, follow-up, and analysis. CONSORT does not tell authors how to run a trial; it governs how completely they describe what they did.
Why incomplete trial reporting created the need for CONSORT
Through the 1990s, reviewers repeatedly found that published trials omitted the very details needed to appraise them: how the allocation sequence was generated, whether assignment was concealed, who was blinded, and how many participants dropped out of each arm. Without those details a reader cannot tell a sound trial from an unsound one. The original Consolidated Standards of Reporting Trials guideline was published in 1996 to fix this, and the widely used revision, CONSORT 2010, refined the wording and added the explanatory paper that defines each item. The aim throughout has been transparent reporting, not a verdict on trial quality. A clearly reported trial can still be biased, and CONSORT is what lets a reader see that bias rather than guess at it. This is the same transparency logic that underpins the PRISMA 2020 reporting guideline for reviews.
What the 25-item checklist actually asks for
The checklist is organised by manuscript section, and each item maps to a concrete thing a critical reader needs. In the methods, CONSORT asks for the eligibility criteria, the settings, the intervention detail sufficient to replicate it, the pre-specified primary and secondary outcomes, how the sample size was determined, the method of random sequence generation, the allocation concealment mechanism, who was blinded, and the planned statistical methods. In the results it asks for the numbers analysed in each group, the estimated effect size with its confidence interval, and any harms. The point of demanding the effect estimate and its interval, rather than a bare p-value, is that those are the figures a later synthesis must extract; the menu of measures is set out in our guide to choosing an effect size for synthesis.
Several items are decisive for appraisal. Reporting allocation concealment separately from sequence generation matters because a properly randomised sequence can still be subverted at the point of assignment if recruiters can foresee the next allocation. Reporting who was blinded, rather than the vague word “double-blind”, matters because participants, carers, outcome assessors, and analysts can each be blinded or not, with different consequences. Reporting the numbers analysed versus randomised exposes attrition and whether an intention-to-treat principle was followed. Each of these maps directly onto a domain that a bias tool such as the RoB 2 tool for randomised trials will later interrogate.
The flow diagram
The CONSORT flow diagram is a four-stage figure: enrolment, allocation, follow-up, and analysis. It records how many were assessed for eligibility, how many were excluded and why, how many were randomised to each arm, how many received the allocated intervention, how many were lost to follow-up, and how many were finally analysed. Discrepancies between the number randomised and the number analysed are often where bias hides, which is why the diagram is reproduced so widely. It is a close relative of the PRISMA flow diagram used in reviews, though one tracks participants through a single trial and the other tracks records through a search.
How CONSORT differs from PRISMA and from risk-of-bias tools
These three instruments are constantly confused, yet they answer different questions. CONSORT governs the reporting of a single randomised trial. PRISMA governs the reporting of a systematic review or meta-analysis: a different unit of analysis, a different checklist, a different flow diagram. A risk-of-bias tool is different again, because it judges whether a trial’s result is likely to be systematically wrong, not whether the paper described its methods completely. The distinction between reporting completeness and systematic error is the same one drawn in our piece on the difference between bias and quality. A trial can satisfy every CONSORT item and still be at high risk of bias; conversely a low-risk trial can be reported so tersely that a reviewer cannot extract its data at all.
This is why reviewers care so much about CONSORT-compliant reporting in practice. The data extraction stage of a review depends on the trial having reported group sizes, event counts or means with measures of spread, and the effect estimate with its interval. When those are missing, the reviewer must either write to the trial authors or exclude the study from the quantitative synthesis, both of which weaken the review. Complete reporting is therefore a precondition for the quantitative pooling stage of a review, and for assessing the non-randomised evidence the parallel guideline is the ROBINS-I tool for non-randomised studies.
CONSORT 2025 and the family of extensions
The guideline has been updated. CONSORT 2025 revises and modernises the checklist, strengthening items on harms, open data and code, the role of any trial registry, and patient involvement, while keeping the structure familiar to anyone who has used CONSORT 2010. Alongside the main statement sits a large family of official extensions for specific designs and contexts: cluster trials, non-inferiority and equivalence trials, pragmatic trials, pilot and feasibility trials, trials of herbal or non-pharmacological interventions, the abstract, and reporting of harms. Authors choose the base statement plus any extension that matches their design, in the same way that an observational study would follow the STROBE reporting guideline for observational studies.
Using CONSORT well, and the common mistakes
For authors, the guideline is most useful before submission and ideally before the trial is written up at all, used as a structure for the manuscript rather than a box-ticking exercise at the end. Completing the checklist with the page or line number for each item, as journals increasingly require, forces honesty about what is actually present. Three mistakes recur. The first is treating CONSORT as a quality scale and summing the items into a score, which it was never designed to be. The second is reporting only that a trial was “randomised” without describing sequence generation and concealment, which leaves the most important methods item effectively blank. The third is an incomplete flow diagram that does not reconcile the numbers randomised with the numbers analysed, hiding attrition from the reader. Each is avoidable, and each materially affects whether a later review can use the trial at all, a dependency worth keeping in view from the earliest stages of planning a review.