Pharmacovigilance literature monitoring, every article screened with a reason
Every journal article screened with a written reason, and reportable cases at intake in about a day.
See one case, screen by screen ↓The rules ask for worldwide literature to be screened every week for every product. A reviewer runs the saved searches in each database, reads the local journals the databases do not index, removes the duplicates, then reads every title and abstract to find the few that describe an identifiable patient with a suspected reaction. Most of what comes back is efficacy data, reviews, animal studies or another company’s brand.
The few that matter cannot wait. Day 0 is the day the article is found, and a serious case is due in 15 days. Full text has to be ordered, a Spanish or Japanese case report translated, the patient checked against cases already in the safety database, and every exclusion written down — because the weekly log is what an inspector asks for.
Estimated split for one weekly run across three products, by hand and with the solution.
Six specialist agents search, screen, read, extract and check every article; a literature reviewer confirms each case.
Week 41. The saved searches ran on Monday at 06:00 for Veltrimab, Corventa and Talmirex. Here is what happens next, screen by screen, in the working solution.
Maya opens the solution to 612 articles ready to screen. The search found 741 hits — MedIndex 438, LitBase 263 and 40 from 8 local journals — and 129 duplicates or already-screened articles are gone. Below that: a full text ordered last week has arrived, and last week’s log is waiting for her signature.
“Week 41 search finished — 612 articles to screen. 741 hits from 3 sources · 129 duplicates already removed.”
She presses one button. The abstract screener reads 612 titles and abstracts and finds 37 possible cases, a reason for each. The full-text reader reads 31 articles in full, translates 4 and orders 6 more. The case extractor builds the cases; the QC sampler draws exclusions for a second look.
Every step is shown as it runs, with what it found.
The week comes back sorted: 9 articles describe a reportable case — 10 patients, 8 of them serious, first report due Oct 20. 14 more carry safety information for signal review. 583 were excluded, each with a reason, and 58 of them are drawn for the QC sample.
“9 articles describe a reportable case — 10 patients, 8 serious, first report due Oct 20.”
The article list shows each decision with its reason — “One patient · liver injury · hospitalised” — the four case criteria, the stage and the days left to report. Day 0 is the search date, Oct 5, unless the full text was needed. Serious cases are due in 15 days, non-serious in 90. Special situations are tagged — pregnancy exposure, drug interaction, medication error, overdose — and so are the Spanish and Japanese originals that were translated.
LIT-41-0007: drug-induced liver injury after the third dose of veltrimab in a 47-year-old woman with plaque psoriasis. The four minimum criteria are highlighted in the abstract — the patient, the reporter Hanne Vestergaard, MD, veltrimab 150 mg subcutaneously, and ALT 12 times the upper limit with six days in hospital. Seriousness: hospitalisation. 14 days left.
“Possible new risk. Liver injury is not in the Veltrimab product information — a copy went to signal review (Omar Haddad) in addition to the case.”
No match in 14,280 cases or 23,386 earlier articles — but the conference version of the same report, LIT-41-0080, is linked to this article so the patient is not reported twice. The journal is not on the regulator’s literature list, so the case is ours to submit.
Age, sex, weight, medical history and amlodipine as a concomitant medicine; dose and route, indication and start date; the reaction coded in MedDRA as drug-induced liver injury, hepatobiliary disorders; outcome recovered, drug withdrawn, dechallenge positive. Each field shows how sure the agent is, cited fields link to their passage, and a draft narrative is ready to edit.
Fields without a source passage are left blank.
Maya confirms and presses Create case. The fields are checked against their passages again, the PDF is attached as the literature reference, and LC-26-0412 goes to case intake with Day 0 Oct 5 kept and the report due Oct 20. Sam Patel, the case intake lead, is notified.
A case reaches intake only after a person confirms it.
58 of 583 exclusions are drawn at random — 10%. Each shows the abstract and the agent’s reason, such as “No adverse event — efficacy only”: the article reports response and remission rates only. She agrees, or sends it back with “Disagree — send back”, which re-opens the article. This week, 58 of 58 agree.
The week 41 log holds what an inspector asks for: the search strings and hits per source, the duplicates removed, every decision, the reportable articles with their seriousness and due dates, and the QC result. Maya signs with her password, and the meaning of the signature is recorded with it. Cases still with reviewers carry over on their own clock.
“I confirm the weekly literature screening for Week 41 is complete and correct.”
The same week, seen by the five people who carry it — what it looked like by hand, and what it looks like now.
Six specialist agents search, screen, read, extract and check every article; a literature reviewer confirms each case.
Runs the saved weekly searches for every product across MedIndex, LitBase and 8 local journals, then removes duplicates and articles already screened.
Reads every title and abstract and decides: reportable case, needs full text, safety information or not reportable — always with a reason and the passage it relied on.
Reads the full article when the abstract is not enough, translates non-English articles and confirms whether the four criteria are met.
Draws a random sample of the week’s exclusions for a second person to confirm, and records the result in the weekly log.
Builds one case per identifiable patient: patient, product, dose, reaction coded in MedDRA, seriousness, outcome, reporter and a draft narrative — every field cited.
Looks for the same patient in the case database and in earlier articles — conference versions, preprints and translations of the same report.
Confirms cases and the QC sample. The agents propose; a named person decides.
Ask about any article, case or week — or tell it what to change. Answers point back to the passage.
Why is the liver injury article reportable?
It meets all four minimum criteria in the abstract: the patient, a 47-year-old woman; the reporter, Hanne Vestergaard, MD, corresponding author; the product, veltrimab 150 mg subcutaneously; the event, ALT 12× the upper limit, admitted for six days. It is serious (hospitalisation), so it is due Oct 20. Liver injury is not in the product information, so a copy also went to signal review.
Which cases are due first?
Day 0 for this week is the search date, Oct 5. Serious cases are due in 15 days, on Oct 20. The one non-serious case, oesophageal candidiasis, is due Jan 3.
Show me articles that mention pregnancy
Two matches this week: LIT-41-0058, neonatal neutropenia after in-utero exposure to veltrimab — reportable, translated from Spanish; and LIT-41-0052, a national pregnancy registry with group data only — safety information.
Add CRV-208 to the Corventa search terms
Done — CRV-208, the development code, is now a Corventa search term. It applies from Monday’s run and is recorded in the audit trail. I checked last week’s results with it: 3 extra hits, none describing a patient.
13 screens from the working solution, on its sample data. Pick one to see it large.
The schedule, the rules, the QC sample and who signs are settings, changed in plain words and versioned.
| Setting | Default | Choose from |
|---|---|---|
| Run the weekly search | Monday · 06:00 CET | Sunday 22:00 · Monday 06:00 · Tuesday 06:00 CET |
| Order full text when the abstract suggests a case | On | On · Off (when on, Day 0 moves to the date the full text gives all four criteria) |
| Assume our product when no brand is named | On | On · Off (unless the article names another company’s brand or batch) |
| Skip journals the regulator already monitors | Off | On · Off (for substances on the regulator’s literature list, the regulator enters those cases) |
| Sample of exclusions each week | 10% | 5 · 10 · 20 % |
| Minimum sample | 20 articles | 20 · 30 articles |
| If a sampled exclusion is wrong | Re-screen the whole week for that product | Re-screen the week for that product · Send only that article back |
| Signs the weekly log | Dr. Maya Chen | Dr. Maya Chen · Omar Haddad |
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 products and articles · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published requirement (EU GVP Module VI — weekly literature screening, 15-day clock for serious cases). People, products and journals named on this page are characters and examples in the working solution.
The weekly screening of worldwide and local medical literature for reports of suspected adverse reactions to your products. Articles that describe an identifiable patient with a suspected reaction become individual cases with a reporting deadline; the rest is excluded with a reason or passed on as safety information.
The abstract screener checks the four minimum criteria — an identifiable patient, an identifiable reporter, a suspect product and a suspected reaction — and flags special situations such as pregnancy, overdose or medication error. Each decision carries a reason and the passage it relied on. When it is unsure, it orders the full text instead of excluding.
Yes. The full-text reader translates non-English articles and keeps the original-language text with the translation. In the working solution, a Spanish and a Japanese case report became reportable cases in the same week.
The duplicate checker looks for the same patient in your case database and in earlier articles, including conference versions, preprints and translations. Possible duplicates are linked and go to a person; they are never merged automatically.
Yes. A case goes to intake only after a named reviewer confirms it, a second person confirms a random sample of exclusions, and the weekly log is signed electronically. The agents propose; people decide.
A weekly screening log for every week: the search strings and hits per source, the duplicates removed, every decision with its reason, the reportable articles and their cases, the QC result and the signature. It exports as a PDF.
Yes. Products and search terms, the weekly schedule, full-text ordering, translation, special situations, signal copies, the QC sample size and who confirms and signs are all settings. Search term changes are versioned and need approval.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your screening procedure, product list and search strings, and a past week of results — and it goes live once every quality gate has passed. We will walk you through it on your own literature first.
Every side-effect report becomes a cited case in minutes, filed with days to spare.
1-min filmSignal ManagementEvery potential safety signal gets its evidence assembled in days, each number traced to its query.
1-min filmPeriodic Safety ReportsEvery periodic safety report is fully drafted within two weeks of data lock, each number traced to its query.
1-min filmWe’ll run Safety Literature Screening on a past week of your own search results.