CONSORT is the checklist journals expect randomized controlled trials to follow. This walks the items reviewers check first — the flow diagram, the pre-specified primary outcome, the intention-to-treat analysis — with a self-check for each.
CONSORT — the Consolidated Standards of Reporting Trials — is the reporting checklist and flow diagram ↗ for parallel-group randomized trials, maintained through the EQUATOR Network ↗. Many trial-publishing journals require a completed checklist and a flow diagram at submission, and an editor can confirm in minutes whether they are present. Because a trial makes the strongest claims in medicine — cause, not just association — the reporting bar is correspondingly high, and a missing item is a fast desk-rejection.
Identify the trial as randomized in the title (item 1) and report its registration (item 23); a prospectively registered protocol is what lets a reviewer confirm the primary outcome was not chosen after seeing the data.
Pre-specify one primary outcome (item 6). This is the single most consequential item. A trial with several endpoints and no designated primary invites the reader to suspect the most favorable one was promoted after the fact.
State how the sample size was determined (item 7) — the effect you powered to detect, and the assumptions behind it.
Self-check. Does your registered primary outcome match the primary outcome in the manuscript, worded the same way and measured at the same time? A silent switch is the difference reviewers look for hardest.
Describe how the sequence was generated (item 8), how allocation was concealed (item 9), and who was blinded (item 11). These are the machinery that makes a trial a trial. Allocation concealment in particular — that the person enrolling a patient could not foresee the next assignment — is what protects randomization from being quietly undone at the bedside.
Self-check. Can a reader tell, from your methods alone, that assignment was truly random and concealed for every participant? A subset randomized differently — by household, by clinic, by an inconsistent table — is exactly the flaw that forced the reanalysis of a landmark trial, which we walk through in catching what the record already knew.
The participant flow diagram (item 13) is CONSORT's signature requirement: numbers randomized, receiving intended treatment, lost to follow-up, and analyzed, for each arm. A reviewer who cannot reconstruct the denominator cannot trust the result. Report which analysis set was used (item 16) — intention-to-treat keeps the benefit of randomization; a per-protocol analysis presented as primary quietly gives it back.
For each outcome, give the effect size and its confidence interval (item 17), not a bare p-value.
Self-check. Do the arm totals in your flow diagram reconcile with the numbers analyzed in every table? Counts that do not add up are a deterministic catch — the kind recomputed automatically, as in how journals catch statistical errors.
Report all harms, not only benefits (item 19). Address limitations and the risk of bias (item 20), and keep the conclusion inside what the trial showed. The most reliable way to lose a reviewer's trust is to lead the abstract with a significant secondary endpoint when the primary was negative — reporting a trial for the result you wanted rather than the one you pre-specified.
Self-check. If your primary endpoint was not met, does your abstract say so plainly, in its own sentence, before any secondary result? For extensions — cluster trials, non-inferiority, pilot and feasibility studies — confirm you are reporting against the right CONSORT variant.
Complete the official checklist and draw the flow diagram before submission — both are published in advance, so a gap is preventable rather than fatal. A structured pre-submission review checks them alongside the statistics: RigorMD maps a trial to CONSORT item by item as one of six scored domains, recomputes the reported numbers deterministically, and has two independent engines assess whether the conclusion is calibrated to the pre-specified result. It flags; it does not certify, and it does not replace peer review or a statistician's input on design.
See a full sample report →, read how the engine works, or review pricing — the pre-submission review is $25. For other designs, see the STROBE, TRIPOD, and PRISMA walkthroughs.