CSR patient safety narratives, drafted from the study data with every sentence traced to its record
Every serious event gets a narrative drafted in under a minute, each sentence traced to the data.
See one case, screen by screen ↓Every participant who died, had a serious adverse event or an event of special interest, or stopped study drug for an adverse event needs a narrative in section 14.3.3 of the clinical study report. In the sample study that is 247 participants. Each narrative pulls from demographics, exposure, adverse events, concomitant medications, medical history, labs, vital signs and disposition — and each one has to read in the same template and the same house style.
The data keep moving until lock. A weekly SDTM transfer corrects an onset date or a lab unit, and every narrative it touches has to be found, rewritten and re-checked — then reconciled with the safety database, where onset, outcome and causality can disagree with the clinical data. By hand that is about four hours of writing per narrative, then QC and medical review, then rework each time the data change.
Estimated split for one narrative, by hand and with the solution, in hours. The totals follow the working solution’s typical figures: about 4 hours of writing and 65 minutes of QC and medical review by hand; with agents, under a minute to draft, 18 minutes of writer check and 12 minutes of medical review.
Five specialist agents find, build, write, check and polish every narrative; a safety physician reviews and signs.
CRV-301, Corventa against docetaxel in stage IV non-small cell lung cancer. Transfer 14 has just arrived, nine days before database lock. Here is what happens next, screen by screen, in the working solution.
Sam opens the solution to the state of the study: 12 participants to draft, 27 drafted, 9 held on a data flag, 42 in medical review and 157 signed — 64 % of 247. Database lock is on 16 Oct, the CSR is due on 13 Nov, and 90 narratives are still to sign. The narrative wall shows one tile per participant, by trigger: deaths, special interest, serious adverse events, stopped treatment.
“Transfer 14 arrived at 08:12 and every narrative it touched is updated.”
The trigger finder confirms 247 participants meet a trigger and 12 have no draft. The profile builder builds 12 timelines from DM, EX, AE, CM, MH, LB, VS and DS — 1,896 records. The writer drafts all 12 in the plan’s template, the data checker checks 204 facts and flags 2 differences, and the style editor applies the house style and MedDRA 29.1 terms.
“12 narratives drafted · 10 routed for review · 2 data flags.”
A 64-year-old woman at site 1043 in Lyon, on Corventa, with Grade 3 interstitial lung disease — special interest, serious and the reason she stopped treatment. The timeline puts the four doses, the hold on Day 84 and the stop on Day 99 against the adverse events, the SpO₂ fall to 89 %, the steroids and the hospital stay. Below it, the narrative: 289 words, 13 sentences, every sentence cited.
Sam clicks the sentence about the chest CT. The panel shows its two sources: the Day 84 imaging record — bilateral ground-glass opacities, no pulmonary embolism — and the vital signs with SpO₂ 89 % on room air, each value highlighted and one click from its listing.
“Click any sentence — or any bar on the timeline — to see the exact study records behind it, with the values used highlighted.”
Dates, study days, doses, grades, outcomes and causality are checked one by one against transfer 14. The onset date now matches the safety database after the site’s correction. All 14 ICH E3 12.3.2 elements are present, each mapped to its sentence; post-mortem findings are marked not applicable.
Transfer 14 moved the onset of the lung inflammation, so sentence 4 shows the change: “Day 82 (1 Jun 2026), 18 days” becomes “Day 80 (30 May 2026), 16 days”. Lena Ortiz, the lead medical writer, has added the lung adjudication outcome to the causality paragraph and asks Sam to confirm the wording. Sam accepts the change and resolves the comment.
“The site corrected the onset date on Oct 5 after query Q-1043-118. Transfer 14 carries the correction; the safety database already said 30 May.”
The signing screen shows what Sam is signing: 41/41 facts, 14/14 ICH E3 elements, 0 open changes. Sam signs as Dr. Sam Patel, safety physician, with a password, and the meaning of the signature is recorded with it.
“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 is locked. If a later transfer changes a record it cites, it reopens as a tracked change in the reviewer’s view — the way transfer 14 reopened a signed narrative when a dose date moved by a day. Death narratives take a second signature from Dr. Maya Chen, the medical lead.
“Signed and locked · 301-1043-0217 is ready for section 14.3.3.”
301-0614-0402, Grade 3 keratitis: the medicine record for the lubricating eye drops has no dose or frequency. The draft says “dose not recorded” instead of guessing, 26 of 27 facts match, and signing waits until the difference is settled — by a query to the site or a note on why it does not change the narrative.
“Send query to the site” · “Not needed — note why”
The dashboard plots signed narratives against plan, with a forecast of all signed by Oct 21 — after the final refresh at lock and ahead of the CSR on Nov 13. In the last 30 days, 25 % of reviewed narratives needed no change, 38 % changed wording, 21 % clinical interpretation, 9 % a data correction and 7 % causality wording.
“Data corrections by reviewers are rare because the checker catches differences first — most edits are wording.”
The same study, seen by the people who write, check, sign and feed its narratives — what their week looked like, and what it looks like now.
Five specialist agents find, build, write, check and polish every narrative; a safety physician reviews and signs.
Applies the narrative plan to each transfer and lists every participant who needs a narrative, with the reasons — merging overlaps so each participant gets one narrative covering every triggering event, and reporting new, changed and removed participants against the previous transfer.
Assembles one timeline per participant across exposure, events, tests, medicines and disposition, computes study days from the first dose and keeps the record of 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 and MedDRA preferred terms with the reported term in brackets — citing every sentence to the records it came from.
Checks every date, study day, dose, grade, value, outcome and causality in a draft against the cited record, compares serious events with the safety database case, and flags each difference with a suggested query.
Applies the house style and keeps terms consistent across all 247 narratives — the same drug names, units and phrasing.
Reviews and signs. The agents propose; a named person decides.
The safety physician, the writers and the data team can ask in plain words — or tell it what to change in the plan.
Which narratives are blocked, and why?
9 narratives wait on a data decision — none is blocked by writing: outcome differs from the safety database (4), lab value without a unit (2), eye-drop record with no dose, AE end date differs from the safety database, medicine record without a start date. The first of yours is the eye-drop record with no dose — query the site or settle it with a reason.
Summarise the lung inflammation cases on Corventa
7 participants have lung inflammation (interstitial lung disease or pneumonitis): 6 on Corventa, 1 on docetaxel. All were confirmed by the lung adjudication committee. In 301-1043-0217 the CT on Day 84 showed bilateral ground-glass opacities, saturation fell to 89 % and Corventa was stopped on Day 99.
Add a trigger for Grade 3 or higher liver events
Added Liver events, Grade 3 or higher to the narrative plan (ALT or AST above 5 × upper limit, or bilirubin above 3 × upper limit). Against transfer 14 it matches 9 participants, none with a narrative yet. Lena Ortiz is notified to approve plan version 2.2.
Will we be ready before the CSR date?
Yes, with margin. 157 of 247 are signed. At 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. Risks: 9 narratives held by a data difference, and 19 death narratives that still need both signatures.
13 screens from the working solution, on its sample data. Pick one to see it large.
Who signs, how the text reads and what happens when the data change are settings for each study.
| Setting | Default | Choose from |
|---|---|---|
| Deaths | Two signatures: safety 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 | Dr. Sam Patel and Dr. Priya Raman, shared 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 |
| Study day with the calendar date | On | “Day 80 (30 May 2026)” at first mention in each paragraph |
| Brief narratives for deaths clearly unrelated to study drug | On | ICH E3 12.3.2 allows them · 29 narratives use this |
| Include normal ranges with every lab value | On | On · Off |
| Keep treatment arm hidden in the text | Off | Off after unblinding for the primary analysis |
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 (a published industry case study: about 4 hours per narrative by hand). Narrative content follows ICH E3 §12.3.2 and §14.3.3. People, companies and products named on this page are fictional — characters and sample data in the working solution.
A short clinical account of one participant’s death, serious adverse event, event of special interest or discontinuation for an adverse event, written for section 14.3.3 of the clinical study report as ICH E3 12.3.2 describes. It covers the participant, treatment, the event timed to dosing, tests, action with study drug, treatment, outcome and causality.
The trigger finder applies your narrative plan to each transfer to find who needs a narrative. The profile builder assembles one timeline per participant from DM, EX, AE, CM, MH, LB, VS, DS and other domains, and the narrative writer drafts in your template with a citation on every sentence. The data checker then checks each fact against its record.
Yes. Onset dates, outcomes and causality of serious events are compared field by field with the safety database case, which it reads but never changes. Each difference is flagged with a suggested query and must be settled by a person before the narrative can be signed.
The solution shows every changed record and the narratives it touches. Only the affected sentences are re-drafted, and the change arrives in the reviewer’s view as a tracked edit with its reason. A signed narrative reopens the same way.
No. The narrative writer has no causality opinion of its own; it writes causality only as the investigator, the sponsor and the adjudication committee recorded it.
Yes. The narrative plan holds your triggers, your template sections and your house style rules, such as study day with the date at first mention, MedDRA preferred terms and brief narratives for deaths clearly unrelated to study drug. Changes to the plan need the plan owner’s approval.
Yes. A named safety physician reviews and signs every narrative, with the meaning of the signature recorded, and death narratives need a second signature from the medical lead. Signing locks the narrative, and every accepted change is on the audit trail.
The Agentic Solution Engine builds and deploys it from your requirements — your narrative plan, template, house style and a sample of your study data — and it goes live once every quality gate has passed. We will walk you through it on your own data first.
A trial synopsis becomes a checked first-draft protocol in days, in your own approved language.
1-min filmTMF CheckEvery trial document is read, filed in the right place and checked, so the file is inspection-ready every day.
1-min filmCSR WriterResults sections and patient narratives are drafted within days of database lock, every number linked to its table.
1-min filmWe’ll draft narratives from a sample of your own study data, in your template.