CONSORT flow diagram generator for randomized trials

CONSORT (Consolidated Standards of Reporting Trials) is a reporting guideline for randomized trials, produced by an international group of trialists, statisticians and journal editors. Its checklist lists what a trial report should contain, and its flow diagram shows what happened to every participant, from the first assessment to the final analysis. Many journals expect a flow diagram with a randomized trial manuscript, and it is a good habit even when you are writing an internal report.

The diagram is useful because it exposes how representative and how complete a trial was. A reader can see how many people were screened to find the randomized group, why people were excluded, whether the arms lost participants at different rates, and how many were in the final analysis. Large or uneven losses are a warning about bias, and a reader can only judge that if the numbers are shown.

The standard layout has four stages. Enrolment shows the number assessed for eligibility and the number excluded, with reasons such as not meeting inclusion criteria, declining to take part or other reasons. Allocation shows how many were randomized and how many in each arm, and, where relevant, how many received the allocated intervention. Follow-up shows how many in each arm were lost to follow-up or discontinued, with reasons. Analysis shows how many in each arm were analysed and how many were excluded from analysis, with reasons. This generator covers the numbers most reports lead with: assessed, excluded, randomized per arm, lost and analysed.

Plan for the diagram at the start of the study, not at the manuscript stage. Record the number screened, and the reason for each screen failure, in a screening and enrolment log or an eligibility screening eCRF. Record the date and arm of each randomization; the randomization list generator and the guide to randomization software explain how allocation works. Record the reason when a participant stops, in a structured field and not only free text, so the reasons can be counted.

Then compare the diagram with your plan. The sample size calculator says how many evaluable participants you needed; the dropout-adjusted calculator shows how many you planned to enrol to allow for losses. If the real losses in the diagram are higher than the plan assumed, say so in the discussion and show what it means for power.

Capture keeps the study data in one place, with field-level audit trail and exports by site, so the counts behind a diagram come from the same system as the rest of the data rather than from several spreadsheets. See clinical trial randomization software for how randomization is handled in a study, and real-time dashboards for monitoring progress while the trial runs. Data export for statisticians is covered in EDC for biostatisticians.

In the demo numbers, 320 people were assessed and 120 excluded (80 did not meet criteria, 30 declined and 10 for other reasons), leaving 200 randomized, 100 in each arm. In follow-up, 12 were lost in one arm and 15 in the other, and 88 and 85 were analysed. Two things stand out to a reader: the arms lost similar numbers, which is reassuring, and the analysed total is lower than the randomized total, so the authors should explain who was excluded from the analysis and why, and state whether the analysis was by intention to treat.

If the losses had been 5 in one arm and 30 in the other, a reader would immediately ask whether the intervention caused people to leave. That is the diagram doing its job. When you draft your own, read it as a sceptical reviewer would, and write the explanation of any imbalance into the text before they have to ask.

The generator draws a two-arm, parallel-group diagram with the counts you give it. It does not draw the full CONSORT 2010 layout for every optional box (for example, "received allocated intervention" and detailed per-reason breakdowns), nor extensions for cluster trials, crossover trials, non-inferiority and equivalence trials, pilot and feasibility studies, or multi-arm designs. Those have their own recommended diagrams. It does not check that your numbers are true, only that they are consistent with each other.

The output is intended for drafting and is not a substitute for the CONSORT checklist itself, which covers far more than the diagram: how the trial was designed, how randomization was concealed, how outcomes were assessed, and how missing data were handled. If you are preparing a manuscript for a journal or a regulatory report, have a statistician confirm the analysis populations and definitions, and follow the journal's current reporting guidance. This tool is a demonstration and planning aid, not validated software.