Your note becomes a signed, cited, attributed grade in seven steps.

Below, one toxicity makes the whole trip. It’s the same walkthrough your monitor gets at a monitor visit, and the same one your IT reviewer can take apart.

  • 01

    Day 8 of cycle 3. Loose stools, more than yesterday.

    the note
  • 02

    Loose stools, more than yesterday

    extraction
  • 03

    CTCAE v6.0
    GI-DIAR-3

    the criterion
  • 04

    G3

    the grade
  • 05

    pembrolizumab
    probable

    attribution
  • 06

    signed by your
    investigator

    signature
  • 07

    IRB
    sponsor
    FDA

    the reports

Figure 1. One toxicity, seven steps, from the note as written to the reports that leave your site.

The note

The note as written. Voice, text, or EHR.

ENC-2026-0412 · Amara Osei · Cycle 3 Day 1

S: Pt reports loose stools since cycle 2 day 3, now about 6 stools/day over baseline. Endorses dizziness on standing, resolving after a few seconds. Denies fever or blood in stool. Oral intake fair, pushing fluids.

O: BP 104/68 sitting, 92/60 standing. HR 88. Afebrile. Abdomen soft, non-tender. Skin turgor slightly reduced.

A/P: Likely treatment-related diarrhea on FOLFOX. Loperamide per protocol, oral hydration. Recheck labs today.

Extraction

Negations and clinician-stated grades are kept whole.

ENC-2026-0412 · Amara Osei · Cycle 3 Day 1

S: Pt reports loose stools since cycle 2 day 3, now about 6 stools/day over baseline. Endorses dizziness on standing, resolving after a few seconds. Denies fever or blood in stool. Oral intake fair, pushing fluids.

EXTRACTED TERMS

Diarrhea

Dizziness

Criteria

For each term, Burna retrieves the governing criterion from CTCAE v5.0 or v6.0, matched to the study's version. 850 terms, both versions, one engine.

Diarrhea

CTCAE v6.0

Gastrointestinal disorders

G1Increase of fewer than 4 stools per day over baseline; mild increase in ostomy output compared to baseline.
G2Increase of 4 to 6 stools per day over baseline; moderate increase in ostomy output; limiting instrumental ADL.
G3Increase of 7 or more stools per day over baseline; hospitalization indicated; limiting self-care ADL.

Grade + citation

Where the evidence runs thin, Burna returns a 'cannot grade' state and names the gap.

G3

Grade 3 · Diarrhea

Increase of 7 or more stools per day over baseline; hospitalization indicated; limiting self-care ADL.

CTCAE v6.0 · GI-DIAR-3

cannot grade · returned when evidence is insufficient, with the gap named

Attribution

Comorbidities, concomitant medications and regimen composition are carried into the scoring, because an event that is true may still have another cause. The investigator sees the reasoning, and decides.

ATTRIBUTION · WHO-UMC + KRAMER

OxaliplatinProbableKramer 6
5-FUPossibleKramer 4

42 regimen profiles · 26 organ system classes

Review + signature

The clinician accepts, edits, or rejects. The signature is 21 CFR Part 11 aligned. The audit trail records what the AI suggested and what the human decided.

G3Diarrhea · Grade 3CTCAE v6.0 · GI-DIAR-3
AcceptEdit

Signed: Dr. N. Adeyemi · 21 CFR Part 11

AI suggested G3 · clinician accepted

The reports come out of the signed record.

From the signed record, Burna generates the IRB report, the sponsor report and the FDA submission. The narrative arrives already written, and the packet arrives assembled.

IRB report

Site-facing safety report, built from the signed grade, its criterion and its attribution.

Sponsor report

Cross-site graded events with the evidence span and the scoring behind each attribution.

FDA submission

MedWatch-ready narrative that arrives already written.

Real time is structural.

Adding real time to a batch system is a rebuild. Burna was built on real-time infrastructure, so a grade 3 event at one site reaches the programme in seconds.

Grade 3 eventseconds, not monitor visitsAll sites

Figure 2. Propagation.

We chose this scope, and we hold to it.

The engine grades adverse events and scores attribution. Dose decisions remain with the treating clinician; response assessment remains with RECIST readers. That focus is what keeps every citation exact and every output inspectable.

Bring a de-identified note to the demo. Watch the cascade run on your own words.