A CONSORT flow diagram shows how many people in a randomised trial were assessed for eligibility, randomised, and then allocated, followed up and analysed in each group. The CONSORT 2025 template has four stages: enrolment, allocation, follow-up and analysis. Draw one column per group, give a reason for every exclusion, and check that each column reconciles from randomisation to analysis.
Participant flow, including the diagram, is item 22 of the 30-item CONSORT 2025 checklist. The 2010 checklist numbered it 13. Item 22a asks for three counts in each group: how many were randomised, how many received the intended intervention and how many were analysed for the primary outcome. Item 22b asks for losses and exclusions after randomisation, with reasons.
This post is about the figure itself. Writing the same numbers as sentences is covered in how to write the Results section of a clinical paper, and choosing between CONSORT, PRISMA and STROBE in reporting guidelines explained.
Draw the CONSORT flow diagram in four stages, one column per group
Download the current template from the SPIRIT–CONSORT Group rather than copying a diagram from an older paper. The CONSORT 2025 statement kept the four stages of the previous diagram and did not rename them. Enrolment sits above the split into groups; the other three stages repeat in every column.
| Stage | Entry in the 2025 template | Who you count | What to write beside the count |
|---|---|---|---|
| Enrolment | Assessed for eligibility | Everyone checked against the eligibility criteria | Total |
| Enrolment | Excluded | Everyone assessed but not randomised | Counts for not meeting inclusion criteria, declined to participate and other reasons |
| Enrolment | Randomised | Everyone randomly assigned | Total |
| Allocation | Allocated to intervention | Everyone assigned to that group | Received and did not receive the allocated intervention, with reasons for non-receipt |
| Follow-up | Discontinued intervention | People who stopped the intervention early | Reasons |
| Follow-up | Lost to follow-up for primary outcome | People whose primary outcome was never measured | Reasons |
| Analysis | Analysed for primary outcome | People in the primary analysis, in their allocated group | Total, and any imputation of missing outcomes |
| Analysis | Excluded from analysis | Allocated people left out of the primary analysis | Reasons |
Put each exclusion in the box where it happened
Anyone removed before randomisation belongs in Enrolment. Anyone removed after it stays in the column of the group they were allocated to.
- Ineligible, declined or not randomised for another reason: Enrolment, "Excluded", under the matching category.
- Randomised but never started the allocated intervention, or given the other group's intervention: Allocation, "did not receive allocated intervention", in the original column.
- Stopped the intervention early but kept attending assessments: Follow-up, "Discontinued intervention". This person is not lost and is analysed if the primary outcome was measured.
- No primary outcome measurement, for any reason: Follow-up, "Lost to follow-up for primary outcome", with the reason.
- Left out of the primary analysis for any reason, including a missing primary outcome or ineligibility found after randomisation: Analysis, "Excluded from analysis", with the reason. Someone lost to follow-up is therefore counted again here, as in the example below.
Mixing up the two follow-up counts breaks the arithmetic. Count discontinuations as losses and the Lost box no longer reconciles with the Analysed box, because people who stopped the intervention but were assessed are still analysed. One person can appear in both counts: someone who stopped the intervention and then missed the outcome visit. Say so in a note, so the totals can still be checked.
A worked CONSORT 2025 diagram for an invented two-group trial
The study below is invented to illustrate the structure; its numbers are not real data. Adults discharged from hospital were randomised to a structured telephone follow-up programme or to usual care, with the primary outcome measured at 12 weeks. Copy the text into your drawing software, one line per box, and replace the numbers with your own.
Enrolment
Assessed for eligibility (n = 412)
Excluded (n = 142): not meeting inclusion criteria (n = 97); declined to participate (n = 38); other reasons (n = 7)
Randomised (n = 270)
Allocation
Telephone follow-up. Allocated to intervention (n = 136). Received allocated intervention (n = 131). Did not receive allocated intervention (n = 5): withdrew before the first call (n = 3); readmitted before the start date (n = 2)
Usual care. Allocated to intervention (n = 134). Received allocated intervention (n = 134). Did not receive allocated intervention (n = 0)
Follow-up
Telephone follow-up. Discontinued intervention (n = 9): asked to stop the calls (n = 7); moved to another service (n = 2). Lost to follow-up for primary outcome (n = 6): no response at 12 weeks (n = 4); withdrew consent to data collection (n = 2)
Usual care. Discontinued intervention (n = 0). Lost to follow-up for primary outcome (n = 8): no response at 12 weeks (n = 7); moved out of the area (n = 1)
Analysis
Telephone follow-up. Analysed for primary outcome (n = 130). Excluded from analysis (n = 6): no primary outcome data
Usual care. Analysed for primary outcome (n = 126). Excluded from analysis (n = 8): no primary outcome data
Note: additional analysis limited to participants who received the whole programme or usual care and had the primary outcome measured: telephone follow-up n = 120; usual care n = 126. Of the 6 lost in the telephone group, 2 had not received the programme, 2 had discontinued it and 2 had completed it.
The 130 analysed in the telephone group include 3 people who never started the programme and 7 who stopped it early, because their outcome was measured. That is why the Analysed box is compared with the Allocated box, not with the Received line.
Make every column of the diagram balance
Run each sum below on your own numbers before the figure goes anywhere.
| Check | Sum in the example | Result |
|---|---|---|
| Assessed minus excluded equals randomised | 412 − 142 = 270 | Balances |
| Exclusion categories add up to the Excluded total | 97 + 38 + 7 = 142 | Balances |
| Allocated boxes add up to Randomised | 136 + 134 = 270 | Balances |
| Received plus did not receive equals Allocated | 131 + 5 = 136; 134 + 0 = 134 | Balances |
| Reasons inside each box add up to its count | 3 + 2 = 5; 7 + 2 = 9; 4 + 2 = 6; 7 + 1 = 8 | Balances |
| Analysed plus excluded from analysis equals Allocated | 130 + 6 = 136; 126 + 8 = 134 | Balances |
| The additional-analysis count in the note can be rebuilt from the boxes | 131 − 9 − 2 = 120; 134 − 0 − 8 = 126 | Balances, using the note |
Two numbers outside the figure must agree with it as well. The group sizes heading your baseline table should equal the Allocated boxes, or a footnote should say why they differ. The number analysed in your primary outcome table should equal the Analysed boxes; CONSORT 2025 item 26 asks for that number, by group, for every outcome.
Label the analysis box by what it contains
Intention-to-treat (ITT) analysis means analysing every randomised participant in the group to which they were allocated. The CONSORT 2025 explanation and elaboration sets the limit: once any randomised participant is excluded, for example for a missing outcome, the analysis is no longer strictly intention-to-treat.
In the worked telephone follow-up example, 130 of 136 and 126 of 134 were analysed, so the box says "Analysed for primary outcome", not intention-to-treat. If you imputed missing outcomes, the Analysed box can show all 136 and 134, with the number imputed stated beside each. CONSORT 2025 item 21c asks how missing data were handled, so the diagram and that Methods sentence must tell the same story.
The CONSORT 2025 explanation and elaboration describes a per-protocol population as participants who completed the study with no major protocol deviations. It also advises stating exactly who each analysis includes, and in which group, rather than relying on labels such as "per protocol" or "modified intention-to-treat". Give each additional analysis its own line under the Analysis stage, stating who it includes, as the example note does. A reader can then rebuild 120 from the boxes above.
Adapt the layout for multi-arm, cluster and crossover trials
CONSORT 2025 covers all randomised trials but concentrates on the two-group parallel design. Other designs rearrange the same boxes.
- Three or more groups: add one column per group below Randomised. The Allocated boxes still add up to the randomised total.
- Cluster randomised trials: the unit randomised is the cluster, such as a practice, ward or school. Give the number of clusters and the number of participants in each box, because losing one cluster removes all of its participants.
- Crossover trials: participants are randomised to a sequence, such as A then B or B then A, so the columns are sequences. Add a follow-up row for each period and record each loss in the period in which it happened.
- Randomised pilot and feasibility trials: the CONSORT extension for these trials has its own flow-chart template.
Before drawing a design-specific version, check the SPIRIT–CONSORT Group extensions page. In September 2026, the only CONSORT extension it lists as built on the 2025 checklist is the one for trials in children and adolescents. The CONSORT 2025 authors recommend using the existing version of every other extension until it is aligned.
Checks that catch a diagram disagreeing with the paper
Alongside the balancing sums, tick each line before you submit and after every revision round.
- No loss that happened after randomisation sits in the Enrolment "Excluded" box.
- Participants who discontinued but were assessed are not counted as lost.
- Every box that asks for reasons gives them.
- The Analysed boxes match the number analysed in the primary outcome table and the Results text.
- The baseline table's group sizes match the Allocated boxes, or a footnote explains the difference.
- The abstract's randomised and analysed totals appear in the diagram.
- No analysis that excludes randomised participants is labelled intention-to-treat.
- The caption and the text quote the same counts as the latest version of the figure.
Where the flow diagram goes in the manuscript
The flow diagram belongs in the main text, as the decision table in preparing supplementary materials sets out. When a revision changes any count, update the diagram, its caption and every sentence that quotes it together, as described in figure version control across revision rounds. For non-randomised designs, methods for a retrospective chart review applies the same stepwise attrition to records.
The Directive Publications author guidelines ask you to match your manuscript to the relevant EQUATOR checklist; for a randomised trial that is CONSORT on the EQUATOR Network. The CONSORT 2025 explanation and elaboration gives the rationale and published examples for item 22. If a diagram in one of our published articles does not balance, report the problem to us and say which box is wrong.
For a systematic review rather than a trial, how to draw a PRISMA flow diagram fills the PRISMA 2020 template box by box, counting records, reports and studies separately.