CSR patient narratives drafted from the study data, every sentence traced to the record it came from
A narrative for every serious event in under a minute, every sentence traced to the data.
See one case, screen by screen ↓ICH E3 12.3.2 asks a clinical study report to carry a short narrative for each death, each other serious adverse event and each event of special interest, usually compiled in section 14.3.3. Each describes the event, its course timed against dosing, relevant tests, action taken with the study drug, treatment, outcome and the investigator's and sponsor's causality.
How to choose AI solutions for life sciences →
Updated
Every clinical study report needs a narrative for each death, each other serious adverse event and each event of special interest (ICH E3 12.3.2), usually placed in section 14.3.3. A medical writer builds each one from listings and patient profiles — demographics, exposure, adverse events, concomitant medicines, labs and vital signs, disposition — and turns them into a chronological paragraph in the sponsor's template. A study physician then checks it line by line, flipping between the narrative, the AE, EX and LB listings and the sponsor's medical review forms.
The data do not stand still. Until database lock every weekly SDTM transfer can move an onset date, add a culture result or reopen a signed narrative, and by hand the writer has to re-read every draft to find the ones it touched. A date that no longer matches the sponsor's medical review, or a causality statement nobody recorded, is how a narrative becomes a question from the regulator on the CSR.
Estimated split for the 247 narratives of a phase 3 study, in working days of 8 hours, by hand and with the solution — built around the published figure of up to 4 hours of writing per narrative and the review times shown in the working solution (18 min writer check, 12 min medical review).
Five agents find, build, write and check every narrative on the patient's timeline; a study physician signs.
Study CRV-301 compares Corventa 5.4 mg/kg every 3 weeks with docetaxel 75 mg/m² in stage IV non-small cell lung cancer: 612 participants, database lock on 16 Oct, the CSR due on 13 Nov. 247 participants need a narrative. Here is what happens to one of them, screen by screen, in the working solution.
Sam has 31 narratives waiting for him before lock. The first is 301-1043-0217, a 64-year-old woman at site 1043 in Lyon with Grade 3 interstitial lung disease — special interest, serious, stopped treatment. On the left, her four Corventa doses on Days 1, 22, 43 and 64, the dose held on Day 84 and stopped on Day 99, her adverse events, oxygen saturation against its normal band, medicines and hospital stay. On the right, the 289-word narrative as it will appear in section 14.3.3.
“CSR section 14.3.3 · 289 words · 13 sentences · every sentence cited”
He picks the sentence “On Day 84 a chest CT scan showed bilateral ground-glass opacities with no pulmonary embolism, and oxygen saturation was 89% on room air.” The timeline dims, the SpO₂ point at 89% lights up beside the dose marked Held, and a trace line drops at Day 84. Below, the records open with the values used highlighted: the Day 84 chest CT and the vital signs, SpO₂ 89% room air, flagged low.
“Sentence 5 → 2 sources · Day 84 (3 Jun 2026)”
The site corrected the onset date after query Q-1043-118, and transfer 14 carries the correction. The data checker proposes the edit in sentence 4: “Day 82 (1 Jun 2026), 18 days” becomes “Day 80 (30 May 2026), 16 days”. The sponsor's medical review already said 30 May. Sam accepts it. Lena Ortiz, the lead medical writer, has left a comment asking him to confirm the wording of the adjudication outcome in the causality paragraph; he resolves it.
“The site corrected the onset date on Oct 5 after query Q-1043-118. Transfer 14 carries the correction; the sponsor's medical review already said 30 May.”
The checks are already done: 41 of 41 facts match transfer 14 — dates, study days, doses, grades, outcomes and causality, one by one — and the onset date now matches the sponsor's medical review. All 14 ICH E3 12.3.2 elements are present, each tied to a sentence. He signs with his password, and the meaning of the signature is stated.
“I have reviewed the narrative for 301-1043-0217 against the study data and approve it for the clinical study report, section 14.3.3.”
The narrative locks and the participant's status turns to Signed. If a later transfer changes a record behind it, it reopens only as a tracked change for him to accept or reject. Next to review takes him to the following participant in his queue.
“Signed and locked · 301-1043-0217 is ready for section 14.3.3”
One tile per participant, by trigger: 41 deaths, 20 events of special interest, 142 other serious adverse events and 44 participants who stopped treatment for an adverse event. Each tile shows its stage — to draft, drafted, data flag, in medical review, signed — and the tracker below lists the main event, arm, onset, words, facts cited and data check for each. Twelve are ready to draft.
The trigger finder confirms 247 participants meet a trigger and 12 have no draft yet. The patient profile builder assembles 12 timelines from DM, EX, AE, CM, MH, LB, VS and DS — 1,896 records. The narrative writer writes 12 drafts in the plan's template, 1,543 words, every sentence cited. The data checker checks 204 facts against the data: 202 match and 2 differences are flagged. The style editor applies house style and MedDRA 29.1 terms. Ten go to reviewers; two wait on a data decision.
The data checker never edits study data — a difference always reaches a person.
301-0614-0402, a 58-year-old woman on Corventa, has Grade 3 keratitis. The narrative is drafted, but 26 of 27 facts match: the record for her lubricating eye drops has no dose or frequency, and the text marks it “dose not recorded”. The reviewer can send a query to the site, copied to Dana Okafor, the clinical data manager — or settle it with a reason that is kept on the audit trail.
“The medicine record for the lubricating eye drops has no dose or frequency.”
301-2210-0088, a 71-year-old man on docetaxel, developed Grade 4 febrile neutropenia and died of septic shock on Day 31; blood cultures grew Escherichia coli, a result that arrived in transfer 14. All 34 facts match. Sam signs first — “Sign (1 of 2)” — and Dr. Maya Chen, the medical lead, is notified to sign second.
Lena Ortiz: “Blood culture organism arrived in transfer 14 and is now in sentence 5. Deaths need your signature and Dr. Maya Chen's.”
Transfer 14 arrived on Oct 6 at 08:12: 53,216 records for 612 participants, 14 of them changed, touching 9 narratives. Five are already tracked changes ready for review, one was re-checked with the text unchanged, and three wait to update — one click re-drafts only the affected sentences. One signed narrative reopened because a dose date moved by a day. Serious events are cross-checked with the sponsor's medical review field by field: 312 compared, 14 differences found, 9 settled.
The same narratives, seen by the five people who carry them to the CSR — what their weeks before lock looked like, and what they look like now.
Five agents find, build, write and check every narrative on the patient's timeline; a study physician signs.
Applies the narrative plan to each transfer — deaths, serious adverse events, events of special interest, discontinuations — and lists every participant who needs a narrative, with the reasons. Overlaps are merged, so a participant gets one narrative covering every triggering event.
Assembles one timeline per participant from DM, EX, AE, CM, MH, LB, VS, DS, MB, PR and SU, computes study days from the first dose and keeps the record behind every value so the writer can cite it.
Writes each narrative in the plan's 8 template sections, in past tense, with the study day and date at first mention, MedDRA preferred terms and the reported term in brackets. Uses the brief format for deaths clearly unrelated to study drug, and writes transfer changes as tracked changes.
Checks every date, study day, dose, grade, value, outcome and causality in a draft against the cited record, cross-checks the onset, outcome and causality of serious events with the sponsor's medical review, and drafts a site query for each difference.
Applies the house style and keeps terms consistent across all 247 narratives — the same drug names, units and phrasing, MedDRA 29.1 terms, study day with the date.
Reviews and signs. The agents propose; a named person decides.
Ask in plain words about progress, data differences, a transfer or a participant — or say what to change in the plan.
What changed in transfer 14?
Transfer 14 arrived at 08:12 with 14 changed records. Nine touch a narrative: five show a tracked change, one was re-checked with no change to the text and three wait to update. The biggest: the site moved the onset of 301-1043-0217's lung inflammation from 1 Jun to 30 May, matching the sponsor's medical review. One signed narrative (301-2604-0231) reopened because a dose date moved by a day.
Where do causality opinions come from?
Only from the data: the investigator's assessment in the AE record, the sponsor's assessment in the medical review form, and the lung adjudication committee where it applies. The writer never adds its own opinion. In 301-1043-0217 all three agree.
Will we be ready before the CSR date?
Yes, with margin. 157 of 247 are signed. At the current pace — about 8 signatures a working day — the rest are signed by about Oct 21, after the final refresh that follows database lock on Oct 16, and more than three weeks before the CSR date of Nov 13.
Add a trigger for Grade 3 or higher liver events
Done. Liver events, Grade 3 or higher (ALT or AST above 5 × upper limit, or bilirubin above 3 × upper limit) is added to the narrative plan. Against transfer 14 it matches 9 participants — none had a narrative yet: 7 on Corventa, 2 on docetaxel. Lena Ortiz is notified to approve plan version 2.2.
13 screens from the working solution, on its sample data. Pick one to see it large.
Who signs, how narratives read and where the data come from are settings for each study.
| Setting | Default | Choose from |
|---|---|---|
| Deaths | Dr. Sam Patel + Dr. Maya Chen — study physician, then medical lead | Dr. Sam Patel + Dr. Maya Chen · Dr. Sam Patel only |
| Events of special interest | Dr. Sam Patel | Dr. Sam Patel · Dr. Maya Chen |
| Serious events and discontinuations | Shared between two reviewers by site | Dr. Sam Patel and Dr. Priya Raman · Dr. Priya Raman only |
| Signing locks the narrative | On — a later data change reopens it as a tracked change | On · Off |
| Study day with the calendar date | On — “Day 80 (30 May 2026)” at first mention in each paragraph | On · Off |
| Brief narratives for deaths clearly unrelated to study drug | On — 29 narratives use this | On · Off |
| Include normal ranges with every lab value | On | On · Off |
| Keep treatment arm hidden in the text | Off after unblinding for the primary analysis | On · Off |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample study data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (Certara, “Using SDTM data to produce patient narratives” — up to 4 hours per narrative by hand, 200–400 narratives per large trial; ICH E3 §12.3.2 and §14.3.3 on what a narrative contains and where it goes). People, the study, the study drug and participants named on this page are fictional — characters and sample data in the working solution.
ICH E3 12.3.2 asks a clinical study report to carry a short narrative for each death, each other serious adverse event and each event of special interest, usually compiled in section 14.3.3. Each describes the event, its course timed against dosing, relevant tests, action taken with the study drug, treatment, outcome and the investigator's and sponsor's causality.
The trigger finder applies your narrative plan to each transfer, the patient profile builder assembles one timeline per participant from the SDTM domains, and the narrative writer drafts in your template with every sentence cited to its records. The data checker then compares every fact with its record and the sponsor's medical review, and the style editor applies your house style.
Point at a sentence and the data points it came from light up on the participant's timeline, with the records below and the values used highlighted. The Checks tab shows each fact checked and each ICH E3 element tied to a sentence.
Changed records are found and only the sentences they feed are re-drafted, as tracked changes in the reviewer's view. A signed narrative reopens only as a tracked change, and the reviewer accepts or rejects it.
No. Causality comes only from the data: the investigator's assessment in the AE record, the sponsor's in the medical review form, and an adjudication committee where it applies. The writer never adds an opinion of its own.
Yes. The agents draft and check; a study physician reviews and signs every narrative, and a death narrative needs a second signature from the medical lead. Every data difference reaches a person, medical review forms are read only, and every step is on the audit trail.
Yes. The narrative plan holds your triggers, your template sections and your house style, and the sign-off rules, writing rules and data sources are settings. Adding a trigger shows how many participants it matches before it is approved by the plan owner.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your narrative plan, template, house style and a de-identified SDTM transfer — and it goes live once every quality gate has passed. We will walk you through it on your own study data first.
A checked first-draft protocol in days, in your approved language.
1-min demoeTMF Quality CheckAn inspection-ready trial master file every day, each site scored on what is due today.
1-min demoClinical Study Report AuthoringResults drafted within days of database lock, linked to tables.
1-min demoWe'll run CSR Patient Narratives on a de-identified SDTM transfer and narrative plan from one of your own studies.