MLR review automation that checks every claim at its cited page before committee
Every claim checked against its reference before committee.
See one case, screen by screen ↓Software that takes the checking work out of medical, legal and regulatory review of promotional material. MLR Review Automation pins every claim on the piece, opens each cited reference at its page and compares the value, population, endpoint and timepoint, checks the claim against the current label, fair balance and brand rules, and routes the piece by risk. Reviewers decide each finding and sign.
How to choose AI solutions for life sciences →
Updated
Before a detail aid, an HCP email or a TV spot can be used, medical, legal and regulatory reviewers must confirm that every claim says exactly what its source says, that risk is as prominent as benefit, and that the safety information matches the current label. Today a reviewer opens each cited reference by hand, finds the page and compares the number, the population, the endpoint and the timepoint — fourteen times for one detail aid. Each round of comments with the agency adds days, and pieces queue for the weekly committee.
The errors that get through are the costly kind: a descriptive subgroup shown as the study result, an exploratory endpoint shown as efficacy, a “safety profile similar to placebo”, safety text from a superseded label. Overstated efficacy, minimised risk and omitted risk information are what OPDP untitled and warning letters cite most often, and FDA judges the net impression of the whole piece.
Estimated split of a piece's days in review, from the 29-day baseline in the working solution's sample data to the 13.5-day target — our estimate, not a measurement.
Six agents pin every claim and check it at its cited page; medical, regulatory and legal reviewers decide and sign.
P-2417, a four-page Veltrimab HCP detail aid for gastroenterologists — “Remission at 12 weeks and at 1 year” — comes in from the agency, Wrenfield Health, with its references. Here is what happens next, screen by screen, in the working solution.
Maya has pieces to sign before Thursday's committee. The review queue shows every piece with its tier, stage, open findings and who has signed. P-2417 is in reviewer check on the full-committee lane, due Oct 8, with 4 red and 6 amber findings. Two new pieces — a Talmirex HCP email from Priya Raman and a Veltrimab rep-triggered email — have arrived and wait for “Pin and check claims”.
“I spend my week checking footnotes, not judging claims.”
The four tablet pages are on screen with 14 claims underlined and pinned: 4 not supported, 6 to fix, 4 supported. Pin 6 is open. “2 out of 3 patients stay in remission at 1 year” is checked against VLT-302, opened at page 9 with the passage highlighted: the 67% comes from the 141 patients already in remission at week 12, a descriptive subgroup. The piece-against-source table marks the population and the analysis with a cross.
“The label and the primary analysis report 46% vs 19% at week 52 in week-12 responders.”
The extractor finds 14 claims across 4 pages and pins each one. The reference verifier opens every cited page: VLT-301 p.6 says the same — supported; p.7 week-2 relief rests on an exploratory endpoint; one anchor pointed at page 5 when the result is on page 6. The label checker flags safety information from the superseded label, and the fair balance checker flags risk text far less prominent than the benefit. The risk router has the piece at score 72 — full committee, Thursday 10:00.
Each pin turns green, amber or red as its page is checked. The last one is the subgroup.
Maya clicks “Use approved claim”. CL-036 replaces the subgroup line: “At week 52, 46% of patients were in clinical remission vs 19% with placebo†”, with its footnote on the re-randomised population. Pin 6 turns green and is marked Fixed, and the change is applied to the draft v3 and recorded.
Every red finding offers an approved claim or the label wording — or “Send to agency”, or “Dismiss…” with a reason.
“The #1 prescribed biologic for new UC patients” is checked against the data on file, DOF-VLT-0412, at page 1: new-to-brand prescriptions among advanced therapies, GI-office prescribers only, Jun–Aug 2026 — 24.6% against 22.1%. The population, the comparator and the period do not match. The suggested change qualifies it: “Most new prescriptions among advanced therapies in GI offices, Jun–Aug 2026.*”
“Show me the population behind ‘#1’.” — Lena Ortiz decides comparative wording.
The Important Safety Information block matches label rev 03/2025 word for word. The current label, rev 09/2026, adds Warning 5.4 Hepatotoxicity, added Sep 18, 2026 — missing here. The fix swaps in the standard safety information module for rev 09/2026, approved by Omar Haddad on Sep 22.
The label checker uses only the label version marked current.
On the marked version the reading is 41 against the team's threshold of 70. Risk carries 14% of the words, set in 8 pt against a 22 pt benefit headline, at 45% contrast, on the last page after the references — with no direction to the full Prescribing Information. The finding asks for the safety information under the claims, 11 pt, at full contrast.
“A reviewer makes the call — the reading is evidence, not a verdict.” The checker never passes or fails a piece on its own.
Maya applies the remaining suggestions she has read — any she unticks stays open — and every pin is green: 10 accepted, 0 with agency, 0 dismissed, 0 open. The signature records its meaning and asks for her password before it is written.
“I reviewed this piece for medical accuracy and confirm each claim is supported by the cited evidence and consistent with the current label.”
Omar Haddad signs the regulatory review, label rev 09/2026 confirmed, and Lena Ortiz the legal review. Only then does the packager build the Form FDA 2253 package: the completed form, the final piece clean and annotated, the current Prescribing Information and the 7 annotated references. P-2417 leaves Thursday's agenda; Omar files at first use, Oct 12.
A package is built only after every required signature.
The committee screen shows where the router placed each piece in review and Thursday's agenda — open judgement calls only. Beside it, the label change watch: Veltrimab rev 09/2026 added Warning 5.4 Hepatotoxicity and retired claim CL-019, and 7 approved, live pieces carry the old safety information. One click re-checks them: five revisions to swap the safety information, two withdrawals for a person to confirm.
“Committee time is for judgement calls only.”
The same detail aid, seen by the five people who carry it — what their week looked like, and what it looks like now.
Six agents pin every claim and check it at its cited page; medical, regulatory and legal reviewers decide and sign.
Finds every claim, statistic and reference anchor in a piece — email, PDF, deck or TV script — records each statistic as value, comparator, endpoint and timepoint, and pins it on the piece.
Opens each cited reference at the anchored page, finds the passage and checks the value, the comparator, the population and whether the endpoint was primary, key secondary or exploratory; re-anchors a claim that points at the wrong page.
Compares each claim with the current Prescribing Information, and the safety information word for word with the current label version, naming any missing warning.
Measures how prominent risk is next to benefit — share of words, type size, contrast, placement and, for audio, words a second — and checks the required statements and links for the channel.
Checks approved phrasing, trademarks, absolute and superiority words, comparative wording and channel rules against the brand guide.
Scores each piece and routes it to fast track, standard review or full committee, then builds the Form FDA 2253 package on approval.
Medical, regulatory and legal sign. The agents propose; a named person decides.
Ask about any piece, claim, reference or label — or say what to change. Every answer points to its page.
Why is the week-2 claim not allowed?
Because the week-2 result is exploratory: 22% vs 14% with no rectal bleeding, nominal p=0.03, not controlled for multiplicity — the authors say no conclusion on early onset should be drawn. The label describes no timepoint before week 12, so presenting it as an efficacy claim fails the consistent-with-labeling test. The closest approved claim is CL-031: clinical response at week 12, 58% vs 33%.
Which live pieces still carry the old Veltrimab safety information?
Seven approved, live pieces still carry the rev 03/2025 safety information, which has no hepatotoxicity warning — among them the HCP website home, the dosing card and rep-triggered email #3. Five are proposed for revision; two for withdrawal — one also says “similar to placebo”, the other uses retired claim CL-019.
Can we say “#1”?
Only with the population the data covers. DOF-VLT-0412 measures new-to-brand prescriptions among advanced therapies, GI-office prescribers only, Jun–Aug 2026 — 24.6% vs 22.1%. Suggested wording: “Most new prescriptions among advanced therapies in GI offices, Jun–Aug 2026*” with the footnote. Lena Ortiz decides comparative wording.
Flag “well tolerated” unless the adverse-event rates are beside it
Added. The rule “well tolerated” needs the rates beside it is now a check in Settings, and the fair balance agent applied it to the pieces in review: P-2402 slide 22 says “well tolerated” with no rates — discontinuations were 4.2% vs 3.9%. It is flagged for Thursday.
13 screens from the working solution, on its sample data. Pick one to see it large.
Routing, signers and the checks the agents run are settings. Every change is versioned and on the audit trail.
| Setting | Default | Choose from |
|---|---|---|
| Fast track below | score 30 | score 20 · score 30 · score 40 — one reviewer when nothing is high |
| Full committee from | score 60 | score 50 · score 60 · score 70 |
| Always full committee | On | Broadcast (TV, radio) and first pieces of a launch |
| Fair balance threshold | 70 | 60 · 70 · 80 — for email and digital channels, shown as the line on the reading |
| Who signs, per brand | Veltrimab: Dr. Maya Chen, Omar Haddad, Lena Ortiz | Medical: Dr. Maya Chen or Dr. Sana Iqbal · regulatory: Omar Haddad or Dana Okafor · legal: Lena Ortiz or Ben Adler — three signatures on standard and committee, one on fast track |
| Checks the agents run | 7 of 8 on | Reference page match · Consistent with the label · Current safety information · Fair balance reading · Exploratory and subgroup data · Absolute and superiority words · Trademarks and approved phrasing · Accessibility (off) |
| Add a rule | Described in a sentence | The assistant turns it into a check for your approval |
| Label and reference sources | Prescribing Information checked nightly | Current versions from the regulatory label store · reference library indexed by page |
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 pieces · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (EY 2024 benchmark, 24 days average to approval; Veeva Pulse Content Metrics 2021, 21 days). Regulatory context: 21 CFR 202.1(e) on fair balance and consistency with labeling; Form FDA 2253 at first use under 21 CFR 314.81(b)(3)(i) and 601.12(f)(4). The 29-day baseline is the fictional company's own in the working solution. People, companies, brands, studies and the agency named on this page are fictional — characters and sample data in the working solution.
Software that takes the checking work out of medical, legal and regulatory review of promotional material. MLR Review Automation pins every claim on the piece, opens each cited reference at its page and compares the value, population, endpoint and timepoint, checks the claim against the current label, fair balance and brand rules, and routes the piece by risk. Reviewers decide each finding and sign.
The reference verifier opens the cited reference at the anchored page, finds the passage and compares the piece with the source row by row: value, population, analysis and timepoint. It names the page for every verdict, moves an anchor that points at the wrong page, and never calls a subgroup or exploratory result supported as an efficacy claim.
Yes. The label consistency checker compares the safety information word for word with the label version marked current and names any missing warning. When a new label version is approved, the label change watch re-checks every approved, live piece and proposes revise or withdraw for a person to confirm.
The fair balance and safety checker measures risk against benefit: share of words, type size, contrast, placement and, for audio, words a second, plus the required statements and links for the channel. It shows a reading against your threshold as evidence for the reviewer and never passes or fails a piece on its own.
Yes, on approval and only after every required signature: the completed form, the final piece clean and annotated, the current Prescribing Information and the annotated references. The regulatory reviewer files it at first use.
Yes. The agents never approve or sign. Each finding is accepted, sent to the agency or dismissed by a named reviewer — a high finding needs a written reason — and medical, regulatory and legal reviewers sign with the meaning of their signature: one on fast track, three on standard and committee.
Yes. The fast-track and committee score thresholds, the broadcast rule, the fair balance threshold, the signers per brand and each check the agents run are settings, versioned and on the audit trail. A new rule can be described in a sentence and becomes a check once you approve it.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your review SOPs, brand guides, label versions, claims library and a sample of past pieces with their references — and it goes live once every quality gate has passed. We will walk you through it on your own pieces first.
Every sentence of the reply linked to its approved source, off-label parts evidence only.
1-min demoInvestigator-Initiated Study ReviewProposals decided in half the time with a cited review packet.
1-min demoMedical Insights ManagementEvery field insight becomes a cited theme within days, with an owner for the next step.
1-min demoWe'll run MLR Review Automation on a sample of your own past promotional pieces and their references.