CONSORT Statement

CONSORT (Consolidated Standards of Reporting Trials, 2010) is an evidence-based guideline for reporting parallel-group randomised trials. It pairs a 25-item checklist with a flow diagram that follows participants through four stages — enrolment → allocation → follow-up → analysis — recording the numbers, and reasons, for every exclusion and dropout. Extensions exist for cluster, non-inferiority, pragmatic and other designs.

The complete journey of participants through a trial: how many were screened, excluded, randomised, allocated to each arm, lost during follow-up and finally analysed. Because each box states a number and a reason, a reader can reconstruct the trial’s denominators and judge attrition for themselves.

Start at the top and follow the numbers down. Compare how many were randomised with how many were analysed in each arm — the gap is your attrition. Check the reasons given for dropouts (do they differ between arms?), and confirm the analysis numbers match an intention-to-treat denominator rather than only the per-protocol completers.

Poorly reported trials cannot be appraised — you cannot judge a bias you cannot see. CONSORT forces authors to expose the moving parts (randomisation method, blinding, attrition, analysis set) so journals, reviewers and exam candidates can scrutinise them. It standardises transparency, which is the precondition for any critical appraisal.

• CONSORT = 25-item checklist + flow diagram
• Flow stages: enrolment → allocation → follow-up → analysis
• Reporting guideline — not a risk-of-bias / quality tool

Reading a CONSORT flow diagram — open almost any modern ED RCT (e.g. a trauma or sepsis trial in the Lancet or NEJM) and the first figure is its CONSORT diagram. Say 1000 are randomised 500:500, but the analysis boxes read 470 and 440: that is roughly 6–12% attrition, and the reasons column tells you whether dropout was balanced or arm-driven. You have appraised the trial’s attrition in seconds — purely from the way it was reported. A tidy CONSORT diagram does not lower the risk of bias — it only lets you measure the attrition. Judging whether that attrition biases the result is a separate (RoB 2) step.

• Confusing reporting quality (CONSORT) with conduct quality / internal validity — they are different axes.
• Naming CONSORT when asked for a risk-of-bias tool — that is RoB 2 (or ROBINS-I), not CONSORT.
• Skimming past the flow diagram and missing the dropout numbers it exists to expose.