A CONSORT flow diagram is a standardized visual summary that tracks every participant through a randomized controlled trial, from initial screening to final data analysis. It exists to answer a deceptively simple question: what happened to all the people in this study? The diagram forces researchers to account for every individual who was assessed, randomized, treated, lost, or excluded, and it does so in a format that a reader can absorb in seconds. If you have ever read a clinical trial and wondered why the number of people at the start does not match the number at the end, the flow diagram is where those gaps get explained.
Why the Diagram Exists
The CONSORT statement (Consolidated Standards of Reporting Trials) is a set of reporting guidelines designed to make published trial results more transparent. Alongside a 25-item checklist, CONSORT includes the flow diagram as a required visual element. The diagram gives readers a broad view of how the trial was conducted and concisely reports the methods used to move participants through each phase.1PubMed Central. Evidence-based Orthodontics: CONSORT: when and how to use it Without it, a reader would need to piece together scattered numbers from the methods and results sections of a paper, often unsuccessfully.
The practical value is straightforward. When a trial enrolls 500 people but only analyzes 300, the reasons for that shrinkage matter enormously. Did 200 people drop out because of side effects? Were they excluded after randomization for administrative reasons? Did the researchers simply lose contact? Each scenario tells you something different about the reliability of the trial’s conclusions. The flow diagram lays all of this out in a single figure, making it harder for authors to quietly sweep participant losses under the rug.
The Four Stages
Every standard CONSORT flow diagram is organized into four sequential stages. Each stage captures a different phase of the trial and the participant numbers associated with it.
Enrollment
The top of the diagram shows how many people were initially assessed for eligibility and how many were excluded before randomization, along with the reasons for those exclusions. Under the CONSORT 2010 guidelines, the diagram should report the number assessed for eligibility and the number, with reasons, excluded before randomization.2PubMed Central. Reporting of participant flow diagrams in published reports of randomized trials Typical exclusion reasons include not meeting the study’s inclusion criteria, declining to participate, or having a medical condition that makes enrollment unsafe.
This stage matters because it tells you how selective the trial was. If researchers screened 10,000 people but only enrolled 200, the study population may be so narrow that the results do not generalize well to the broader population. The enrollment box also reveals whether a trial had recruitment difficulties, which can signal practical problems with the intervention or study design.
Allocation
After enrollment, the diagram shows how many participants were randomized and how they were distributed across treatment groups. A two-arm trial will show one box per group (say, drug versus placebo), each with the number of participants assigned to it. The diagram also notes whether any participants did not actually receive their allocated intervention. Someone might be randomized to the drug group but never take the first dose because of a scheduling conflict, a change of mind, or a newly discovered contraindication.
This distinction between being assigned to a group and actually receiving the treatment is critical. It feeds directly into whether the final analysis was conducted on an “intention-to-treat” basis (analyzing everyone as randomized, regardless of what they actually received) or a “per-protocol” basis (analyzing only those who completed the treatment as planned). Both approaches have strengths and limitations, and the allocation section of the flow diagram is where you first see the numbers that determine which analysis is possible.
Follow-Up
The follow-up stage accounts for participants who were lost after allocation. Lost to follow-up means the research team could not contact or collect data from a participant at one or more time points. The diagram should show the number followed up and the number lost to follow-up, per arm, along with reasons.2PubMed Central. Reporting of participant flow diagrams in published reports of randomized trials Discontinued interventions are also captured here: a participant who stops taking the study drug because of nausea is different from one who simply moves away and becomes unreachable.
High loss to follow-up is one of the clearest red flags when evaluating a trial. If one group loses substantially more participants than the other, it suggests something about the intervention is driving people away, and the remaining participants may not represent the original randomized group. The flow diagram makes this asymmetry visible at a glance.
Analysis
The bottom of the diagram shows how many participants in each group were included in the primary analysis. It should also note any exclusions from analysis and the reasons for them. This is where a reader can verify that the numbers add up. If the enrollment box says 500 were randomized, the allocation boxes say 250 per group, and the analysis boxes say 220 and 210, a reader can immediately see that 70 participants were lost somewhere in between and check the follow-up section to understand why.
When the analysis numbers match the analysis denominators reported in the trial’s tables, readers can have greater confidence that the results were reported consistently. A mismatch between the flow diagram and the statistical tables is a warning sign that something was not tracked carefully.
How Often the Diagram Falls Short
Despite being a core requirement of CONSORT, flow diagrams are frequently incomplete. A survey of trials published in periodontology and implantology journals found that only 56% of flow diagrams were fully complete, while 44% were missing at least one required element.3PubMed Central. Reporting of flow diagrams in randomised controlled trials published in periodontology and implantology: a survey The most commonly missing pieces were reasons for loss to follow-up (absent in about 70% of diagrams in that sample) and exclusions from the analysis (absent in roughly 86%).
A separate review of early-phase dose-finding trials found even starker gaps. Flow diagrams in those studies were most deficient in reporting reasons for loss to follow-up (reported in just 2% of diagrams) and reasons for not receiving the allocated intervention (14%).4PubMed. Reporting quality of CONSORT flow diagrams in published early phase dose-finding clinical trial reports: Improvement is needed In other words, authors often include the boxes and the numbers but leave out the explanatory text that gives those numbers meaning.
Even the enrollment section, which might seem like the easiest part to get right, has reporting issues. A study examining flow diagrams across multiple fields found that while 98% reported the total number randomized and 81% reported the number assessed for eligibility, far fewer reported the specific reasons people were excluded before randomization: only 57% noted exclusions for not meeting inclusion criteria, and 66% noted exclusions for declining to participate.2PubMed Central. Reporting of participant flow diagrams in published reports of randomized trials The big-picture numbers tend to be present, but the detail that makes the diagram truly useful for critical appraisal is often stripped away.
Reading a Flow Diagram as a Non-Researcher
You do not need to be a statistician to get value from a CONSORT flow diagram. A few quick checks can tell you a lot about whether to trust a trial’s conclusions.
Start at the top. How many people were screened versus how many were randomized? A steep drop-off suggests the trial’s population was highly selected. That does not invalidate the study, but it should temper how broadly you apply the results. If a blood pressure drug was tested only in people who had already failed two other medications and had no kidney problems, the findings may not reflect what happens in a more typical patient population.
Then look at the allocation boxes. Are the groups roughly equal in size? They should be, unless the trial deliberately used an unequal randomization ratio (like 2:1, where twice as many people get the active treatment). If the groups are unexpectedly unequal, something went wrong with randomization or reporting.
Next, check the follow-up stage. Is the dropout rate similar between groups? And are the reasons for dropout similar? If 15% of the drug group dropped out due to adverse events but only 3% of the placebo group did, that imbalance tells you something about tolerability that the trial’s headline result might not emphasize.
Finally, compare the analysis numbers to the randomized numbers. A trial that randomized 400 people but analyzed only 250 lost nearly 40% of its sample, which weakens the power of randomization and makes the results less reliable. The smaller the gap between randomized and analyzed, the more confidence you can have.
Extensions for Non-Standard Trial Designs
The classic CONSORT flow diagram was designed for simple parallel-group trials, where each participant is randomized to one intervention and stays there. Many real-world trials use more complex designs, and the flow diagram needs to be adapted accordingly.
In cluster randomized trials, the unit of randomization is not an individual but a group, like a hospital, a school, or a village. The CONSORT extension for cluster trials requires the flow diagram to track both clusters and individual participants through each stage.5PubMed Central. CONSORT statement: extension to cluster randomised trials This matters because losing one cluster (say, a clinic that drops out of the study) can mean losing dozens or hundreds of participants at once. A standard flow diagram that only counted individuals would obscure this pattern.
Crossover trials, where each participant receives both interventions in sequence, present a different challenge. The flow diagram needs to capture not just who dropped out, but at which stage of the crossover they dropped out. A participant who completes the first treatment period but drops out before the second contributes partial data, and the diagram should make that visible. The CONSORT extension for crossover trials addresses more complicated variations as well, including designs that compare three or more interventions and cluster crossover trials.6BMJ. CONSORT 2010 statement: extension to randomised crossover trials
Other extensions exist for trials involving non-pharmacological interventions (like surgery or physiotherapy), trials reporting patient-reported outcomes, and trials of herbal medicines. Each extension tweaks the flow diagram’s structure to capture information specific to that trial type while preserving the core four-stage framework.
Creating a Flow Diagram
Researchers building a CONSORT flow diagram generally have three options: draw it manually, use a template-based tool, or generate it programmatically from study data.
Manual creation using software like PowerPoint or a dedicated diagramming tool is still common, especially for smaller trials. The CONSORT group provides a blank template that researchers can fill in. The main risk with this approach is transcription error: someone reads a number from a database, types it into a text box, and gets it wrong. With a complex trial involving multiple time points and subgroups, these errors accumulate.
Template-based online generators offer a middle ground. You enter the numbers for each stage, and the tool produces a formatted diagram. This reduces formatting work but still requires someone to pull the correct numbers from the trial database.
Programmatic generation ties the diagram directly to the trial’s data. One approach demonstrated in a published study used commercially available relational database software to code patient-level data so that CONSORT diagrams were automatically generated as participants moved through enrollment and follow-up waves.7PubMed Central. Automated Generation of CONSORT Diagrams Using Relational Database Software In that trial, which enrolled over a thousand participants, the diagrams were generated weekly for study team meetings, tracking follow-up at one, six, and twelve months. Contact rates ranged from 88% to 92% across those time points.7PubMed Central. Automated Generation of CONSORT Diagrams Using Relational Database Software The advantage of this approach is that the diagram always reflects the actual data. There is no moment where a human transcribes a number incorrectly.
More recently, statistical programming languages like R have been used to build CONSORT diagrams directly from raw enrollment data. This approach eliminates transcription errors and ensures the figure matches the analysis denominators in the trial’s statistical tables exactly.8MetricGate. CONSORT Flow Diagram Generator: Build a Trial Diagram For large multicenter trials where data are constantly being updated, automated pipelines that regenerate the diagram whenever the underlying dataset changes are becoming standard practice.
The Diagram as an Ongoing Monitoring Tool
Most people encounter CONSORT flow diagrams as static figures in published papers, but they can serve a more dynamic role during trial conduct. The weekly-generation approach described above illustrates this well: by producing updated flow diagrams throughout the study, the research team could spot problems in real time. If follow-up rates at six months suddenly dropped in one study site, the team could investigate immediately rather than discovering the gap years later during manuscript preparation.
This monitoring function is recommended by CONSORT itself. The flowchart is designed to guide authors to the details required in the results section, and researchers are strongly encouraged to use it throughout the trial to track and illustrate outcomes as they develop.9PubMed Central. The CONSORT statement When the diagram is only assembled at the end, after all data are locked, researchers sometimes discover discrepancies between their databases and their narrative results that are difficult to resolve retroactively.
Adapting the Framework Beyond Clinical Trials
Although CONSORT was built for randomized trials in human medicine, its flow-diagram structure has been borrowed and modified for other research contexts. In preclinical animal research, for instance, investigators have used modified CONSORT guidelines alongside the ARRIVE framework (designed for reporting animal experiments) to critically analyze the quality of published studies.10PubMed Central. Osteogenic Potential of Dental Mesenchymal Stem Cells in Preclinical Studies: A Systematic Review Using Modified ARRIVE and CONSORT Guidelines The logic is the same: you want to know how many subjects started, how many were in each group, how many were lost, and how many contributed to the final analysis.
Systematic reviews and meta-analyses use a similar flow-diagram concept through the PRISMA framework, which tracks the number of studies identified, screened, assessed for eligibility, and included. If you have seen a PRISMA flow diagram and wondered where the idea came from, CONSORT was the template. The underlying principle, that transparent accounting of what went in and what came out strengthens the credibility of the results, applies whether you are tracking patients, animals, or published papers.
Common Misconceptions About What the Diagram Shows
One frequent misunderstanding is that a flow diagram with high numbers at every stage automatically means the trial was well conducted. The numbers alone are not the whole story. A trial might retain 95% of its participants through follow-up, but if the flow diagram does not explain why the other 5% left, a reader cannot evaluate whether those losses were random or systematic. A diagram that reports 10% dropout with clear, balanced reasons across groups can be more reassuring than one reporting 3% dropout with no reasons given at all.
Another misconception is that the flow diagram captures everything about participant flow. It does not. The diagram does not typically show protocol deviations that occurred during the trial (like a participant in the placebo group accidentally receiving the active drug), nor does it show the timing of dropouts within the follow-up period. A participant who drops out on day two of a twelve-month trial is very different from one who drops out at month eleven, but they may appear as the same line item in the flow diagram. Some trial reports address this with additional tables or figures, but the standard CONSORT diagram was not designed to convey that level of temporal detail.
There is also a tendency to treat the presence of a flow diagram as a stamp of quality. Journals that endorse CONSORT often require a flow diagram for publication, but the requirement does not guarantee completeness. As the reporting surveys described earlier show, a large fraction of published diagrams are missing key information. A reader who sees a flow diagram and assumes every required element is present is making an assumption that fails nearly half the time in some fields. The diagram is a tool for transparency, but only if the person filling it out actually uses it that way.
The 2025 Update
CONSORT is not frozen in time. The guidelines have been revised periodically since their first publication in 1996, with major updates in 2001 and 2010. An updated CONSORT 2025 statement has been published, reflecting evolving expectations around trial reporting.11PubMed. CONSORT 2025 Statement: updated guideline for reporting randomised trials While the core four-stage structure of the flow diagram has remained remarkably stable across revisions, each update has refined what information should be captured within those stages and how much detail is expected in the accompanying text. As trials grow more complex, involving adaptive designs, platform protocols, and decentralized recruitment, the flow diagram will need to continue evolving to capture participant journeys that no longer follow a simple linear path.