Circular Evidence Network Plot Generator

What the plot shows

A circular evidence network places nodes at equal intervals around a circle, preserving their input order. Teal node area represents patients; semi-transparent curves connect directly compared pairs, with line width proportional to studies. Horizontal labels sit outside the circle, and a two-column legend for patients and studies appears at the lower right. Unlike a conventional chord diagram, this chart uses separate dots and curves rather than outer sectors and ribbons.

Patient totals and study counts are independent inputs. Prepare each node’s patient total separately from the number of studies for each comparison. The tool never estimates patients by summing edge weights. Nodes have a minimum visible size, so zero and very small positive patient counts may use equally sized markers; consult the values and legend rather than judging those markers alone.

How to use it

1. Start with the synthetic 29-node sample. The CSV template download contains the same sample with ASCII node names in every language.
2. Replace the node names and patient totals, then edit pair endpoints and study counts. Keep both English headers and the blank separator line.
3. Drop or select a CSV file to generate the chart automatically. Alternatively, expand the CSV editor, edit the text, and select Generate chart.
4. Check the status message, node count, and nonzero pair count. After changing editor text, regenerate before exporting; exports are disabled while edits are pending.
5. Choose light or dark mode, then download a 2× PNG or vector SVG. Both include the current theme background, labels, and legend. Scroll horizontally inside the chart on narrow screens.

CSV format and validation

This shortened example is synthetic, not original clinical data from the reference image. It is not the full downloadable sample.

name,patients
TreatmentA,2000
TreatmentB,1000
TreatmentC,500
TreatmentD,0

source,target,studies
TreatmentA,TreatmentB,8
TreatmentA,TreatmentC,3
TreatmentB,TreatmentC,1
TreatmentC,TreatmentD,0

• Supply exactly two sections: name,patients, a blank line, then source,target,studies. Keep the second header even if there are no pair rows.
• Use 1–40 nodes and at most 300 pair rows, including rows with zero studies. Uploaded files must not exceed 1 MB (1,048,576 bytes).
• Both numeric fields accept integers from 0 to 9007199254740991. Blank values, negatives, decimals, and scientific notation are invalid.
• Node names must be nonempty and unique. Every endpoint must exist in the node table; self-loops are rejected.
• Enter an undirected pair only once. Reversing its endpoints still creates a duplicate and triggers an error; counts are not merged or added automatically.
• UTF-8 labels are supported. Quote names containing commas, and escape an embedded double quote by doubling it. Multiline labels and control characters are not supported.
• Zero patients retain a minimum marker. Zero studies draw no line. Input order is retained rather than sorted by patient count.

Interpretation and appropriate uses

A larger node means only a larger entered patient total; a thicker curve means only more entered studies. Neither implies better treatment effects, higher evidence quality, statistical significance, or a superior treatment ranking. Legend marks share the drawing scales, with ticks derived from the supplied data. Scales can change between datasets, so absolute marker sizes should not be compared across separate exports without checking the legends.

Use the figure to communicate the structure of direct comparisons in systematic reviews, network meta-analysis reports, or evidence-mapping presentations. It can highlight densely connected comparisons or sparsely connected nodes. It does not perform statistical analysis, pool effect sizes, rank treatments, or test heterogeneity or inconsistency. It cannot recover authentic patient and study counts from a screenshot.

Data preparation, privacy, and export

For multi-arm studies, independently aggregate patient totals and pairwise study counts according to your analysis plan. Avoid double-counting: the tool neither splits multi-arm studies nor corrects duplicated participants or comparisons. Enter aggregate statistics only, not patient names, medical records, or patient-level data.

CSV parsing, rendering, and image generation take place locally in the browser; chart data is not uploaded to a server. PNG is useful for slides, while SVG supports scalable page layouts. If very long labels cause export failure, shorten them and regenerate. Check counting rules, zero values, labels, and legends before publication.

Related tools

For general node metrics and relationship visualization, see the Bubble Chord Diagram Generator. For pairwise links arranged along a horizontal baseline, try the Arc Diagram Generator. Their CSV schemas differ, so templates are not interchangeable.