Clinical trial lay summaries in plain words, every sentence checked against the study report
Every trial’s lay summary drafted within days, each fact checked against the report.
See one case, screen by screen ↓A summary of a trial’s results written for the public and the people who took part. Under the EU Clinical Trials Regulation (Art. 37 and Annex V) every trial run in the EU posts one in CTIS within 12 months of the end of the trial — 6 months for trials in children — covering ten required elements, in the language of each country where it ran.
How to choose AI solutions for life sciences →
Updated
Every trial run in the EU must post a summary of its results for the public within 12 months of the end of the trial — 6 months for trials in children — in the language of every country where it ran. Today a medical writer or an agency rewrites the clinical study report by hand into the ten Annex V sections, a statistician checks the numbers, a patient panel reads it, and translators work for weeks more.
The errors are small and they matter. A result of 64.7% rounded up to “about 7 out of 10”. “Treatment-emergent adverse events” left in a summary for participants. “Worked very well and is a safe treatment” in the comments section. A hazard ratio with its confidence interval that is correct and that no reader can use. A readability score alone does not catch what people misunderstand, and a number that drifts in the Polish version is hard to see at all.
Estimated split for one lay summary of a Phase 3 trial, by hand and with the solution.
Six specialist agents read, draft, grade, fact-check and translate every sentence; the writer decides each flag and reviewers sign.
October 7. Lena Ortiz, a medical writer in trial transparency, has seven lay summaries on her desk. The nearest is VLT-302 — Veltrimab in plaque psoriasis, Phase 3, 642 adults — due in CTIS on January 14, 2027. Here is her day, screen by screen, in the working solution.
VLT-302 opens one row per sentence: the study-report passage, the lay English with its reading grade and fact tick, then the German and Polish drafts. The ribbon reads 28 of 29 facts match the report, reading grade 8.9 against a target of 8, 5 of 8 reader-test questions answered, and 32 German and 32 Polish sentences read back. Three sentences are left to decide, 99 days before the deadline.
“This summary is about a study called VLT-302.” — Grade 5 · matches Title page · German and Polish read back · 302.
The side-effects sentence still says “treatment-emergent adverse events”: the checker proposes “at least one medical problem”, adds that not all of them were caused by the study medicine, and shows grade 11 dropping to 7. Below it, “about 7 out of 10 people” had almost clear skin differs from Table 11.4, which gives 277 of 428 — 64.7%. The third flag is the comments section: “Veltrimab worked very well and is a safe treatment for psoriasis.”
The fact checker cannot change text — it proposes a fix for Lena to accept, edit or dismiss.
The leaflet shows the ten required sections as questions, each sentence numbered to its citation, with “Thank you to everyone who took part in this study” at the top. The checklist on the left marks adverse reactions, overall results and comments on the outcome as flagged. The Facts panel shows the selected sentence beside the study-report passage — Table 12.2, page 63: any TEAE 249 (58.2%) on Veltrimab, 110 (51.4%) on placebo.
Readability scores miss what people misunderstand, so the reader test answers eight key questions using only the summary, then compares its answers with the study report. VLT-302 scores 5 of 8. From the draft, a reader takes away “about 7 out of 10” for the main result, cannot explain “treatment-emergent adverse events”, and concludes that one study shows the medicine is safe for everyone — because the summary says “a safe treatment”.
The reader test never sees the study report while it answers.
In NXL-125, Talmirex in adults whose depression has eased, one sentence is flagged: “The risk of relapse was reduced with Talmirex versus placebo during the randomized withdrawal phase (hazard ratio 0.64; 95% confidence interval 0.44 to 0.92).” It matches Table 11.3 and reads at grade 14. The same table gives the counts — relapse in 47 of 246 people on Talmirex (19.1%) and 71 of 244 on placebo (29.1%) — so the checker proposes: “About one in three fewer people on Talmirex had their depression come back.”
“A hazard ratio and a confidence interval mean little to most readers. Say what happened to people.”
Lena accepts. The sentence now reads at grade 7 with its tick — “19.1 vs 29.1 out of 100 — 34% fewer · Table 11.3” — and is marked accepted by Lena. The German and Polish rows re-translate and read back with the same meaning. The ribbon now reads 24 of 24 facts matching, reading grade 7.9 (from 9.2) and 8 of 8 reader-test questions answered (from 7), and the button changes from “1 to decide before signing” to “Approve and sign”.
The signing window repeats what she is signing: 24 of 24 statements match the study report, reading grade 7.9, reader test 8 of 8. The meaning is “I am the author and confirm every fact matches the final study report”, signed with her password. Next, Priya Raman’s patient panel reads it, then Dr. Maya Chen and Dana Okafor sign, and translations into the 7 other languages start from this signed English.
Posting is blocked while any fact is unchecked or any flag is open.
CRV-210, Corventa in atrial fibrillation, is in translation with 44 days to go. Against the signed English master, German, French, Spanish, Italian and Dutch are back-checked, Czech and Hungarian are being translated, and Polish waits on a question for Sam: “injection site reaction” has no locked Polish term yet. He can lock one from the glossary, then approve the translations.
The NXL-118 study report, Talmirex in major depression, arrived on October 6 with a March 30, 2027 deadline. One click on “Draft with agents” runs it: the study report reader reads the synopsis and 14 results tables into 56 facts with page and table, the Annex V drafter writes 10 of 10 sections as reader questions, the fact checker matches 21 statements with every number matching, the plain-language checker scores grade 7.8 with 1 wording flag, and the reader test answers 7 of 8. Then it waits for the writer.
“Writer decides · Waiting for Lena Ortiz”
The runway places each summary at its EU posting deadline — 12 months for adults, 6 for children, 30 for an early study. Eleven are open, three are due within 60 days and none is late this year. NXL-205 is ready to post, 21 days before its deadline, with its package approved by Omar Haddad. VLT-305, a trial in children, is with Priya Raman’s patient panel, 56 days out.
The same summary, seen by the five people who carry it from study report to CTIS — what their weeks looked like, and what they look like now.
Six specialist agents read, draft, grade, fact-check and translate every sentence; the writer decides each flag and reviewers sign.
Reads the study report synopsis and results tables — disposition, demographics, endpoints, adverse events — and pulls out every fact a lay summary needs under Annex V, each with its page or table.
Writes the ten required sections as questions a reader would ask, from the extracted facts and the plain-word glossary only, with short sentences and numbers as “about 65 out of 100 people”.
Matches every factual sentence back to its passage in the study report and compares every number, unit, group and time point; flags any mismatch, any rounding that moves a number by more than 2 points, and any fact without a source.
Scores the reading grade of every sentence and section, finds study-report terms that have a glossary entry, and flags promotional or overreaching wording with a neutral alternative.
Answers the eight key questions a participant would ask using only the summary, then marks an answer wrong if it differs from the study report or if the summary does not let a reader answer it.
Translates the English into each country’s language with the glossary terms locked for that language, then reads it back into English and compares every number, group and time point; raises a query for any term without a locked translation.
Decide each flag, sign to post. The agents propose; a named person decides.
Ask about a flag, a deadline or a reading level — or say what to change. Answers point to the study report.
Why is the relapse sentence in NXL-125 flagged?
The draft says “hazard ratio 0.64; 95% confidence interval 0.44 to 0.92” — correct, and it matches the study report, but it reads at grade 14 and most readers cannot use it. The report also gives the counts: 47 of 246 people on Talmirex and 71 of 244 on placebo had their depression come back — 19.1 and 29.1 out of 100. That is about one in three fewer. The proposed sentence reads at grade 7, and German and Polish re-translate when you accept.
Why was the results sentence in VLT-302 flagged?
The draft said “about 7 out of 10” people on Veltrimab had almost clear skin. The study report says 277 of 428 — 64.7%. Rounding to 7 out of 10 overstates the result by 5 points, so the fact checker proposes “about 65 out of 100”. The reader test also got this question wrong from the draft.
Which summaries are closest to their deadline?
NXL-205 is ready to post — Omar Haddad approved the package — with 21 days left. CRV-210 is in translation with 44 days, VLT-305 is with the patient panel with 56, and VLT-310 is in medical and legal review with 63. VLT-305 is a trial in children, so its deadline is 6 months after the end of the trial, not 12.
Require a patient read-through for every summary of a trial in children
Done. New rule: summaries of trials in children always go to Priya Raman’s panel, with the young readers version for ages 8–12, before medical sign-off. Two open summaries match today — VLT-305 and NXL-310. The rule is in Settings and recorded in the audit trail.
13 screens from the working solution, on its sample data. Pick one to see it large.
How plain is plain enough, how numbers are written and who signs what are settings, with an e-signature on every step.
| Setting | Default | Choose from |
|---|---|---|
| Highest reading grade | Grade 8 | Grade 6 to grade 10, one step at a time; summaries above it are flagged before review |
| How numbers are written | About 65 out of 100 people | About 65 out of 100 people · 65% of people · Both — 65 out of 100 (65%) |
| Statistics in plain words | On | Hazard ratios, odds ratios and confidence intervals become “x out of 100” and “about one in three fewer” |
| Flag promotional words | On | “safe”, “effective”, “breakthrough”, “worked very well” and 40 more |
| Reader test before patient read-through | On | At least 7 of 8 key questions answered from the summary alone |
| Version for young readers | On | Trials in children get a second version written for ages 8–12 |
| Early draft translations for Phase 3 trials | On | German and Polish drafted beside the English; re-run on the signed version |
| Approvers | Priya Raman · Dr. Maya Chen · Dana Okafor · Sam Patel · Omar Haddad | Patient read-through · medical accuracy · non-promotional check · translations · posting package — each chosen from the named team |
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 studies · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (TranspariMED — about 20 hours of writer time per plain language summary, agency drafts typically 2–4 weeks; Clinical Leader and Citeline — 116 of the first 234 trials required to report results in CTIS did so on time). Context: EU Clinical Trials Regulation 536/2014, Art. 37 and Annex V; European Commission, Good Lay Summary Practice (2021). People, companies, medicines and studies named on this page are fictional — characters and sample data in the working solution.
A summary of a trial’s results written for the public and the people who took part. Under the EU Clinical Trials Regulation (Art. 37 and Annex V) every trial run in the EU posts one in CTIS within 12 months of the end of the trial — 6 months for trials in children — covering ten required elements, in the language of each country where it ran.
The study report reader extracts every fact the ten Annex V sections need, each with its page or table. The Annex V drafter writes each section as a question a reader would ask, using only those facts and the plain-word glossary. The fact checker, the plain-language checker and the reader test then check every sentence before the writer sees it.
The fact checker matches each factual sentence to its source passage and compares every number, unit, group and time point. It flags any mismatch, any rounding that moves a number by more than 2 points — such as 64.7% written as “about 7 out of 10” — and any fact without a source. Posting is blocked while any fact is unchecked.
Every sentence carries its reading grade, recomputed as the writer edits, against the highest grade set in Settings — grade 8 by default. Study-report jargon, statistics such as hazard ratios and confidence intervals, and promotional words are flagged with a plain alternative. A reader test then answers eight key questions from the summary alone, because a readability score alone misses what people misunderstand.
The translator works from the signed English with glossary terms locked for each language, then reads its translation back into English and compares every number. Terms without a locked translation become a query for the translation coordinator, who approves each language.
Yes. The agents propose and cannot change a sentence themselves. The medical writer accepts, edits or dismisses every flag and signs as author; the patient read-through, medical accuracy, the non-promotional check, the translations and the posting package are each signed by a named person, and every action is on the audit trail.
Yes. The highest reading grade, how numbers are written, the approver for each step, the languages per country and rules such as statistics in plain words, the young readers version for trials in children and early draft translations for Phase 3 trials are all settings.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your house style, plain-word glossary and a sample of past study reports and approved summaries — and it goes live once every quality gate has passed. We will walk you through it on one of your own study reports 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 Clinical Trial Lay Summaries on one of your own final study reports.