Protocol deviation management that classifies every deviation the same way, at every site and every CRO
Every protocol deviation classified in minutes, the same way at every site, with its reason cited.
See one case, screen by screen ↓A deviation is logged by a site or a CRO, and its importance is decided weeks later at the monthly deviation review. Until then nobody knows whether a participant may leave the per-protocol set, whether a serious breach clock is running, or whether a site has done the same thing three times.
Each team reads the classification spec its own way. In the sample study, Ashgrove’s CRAs call a Week 16 PASI by an untrained assessor not important because the visit was inside the window, and call a single late self-injection important. The medical monitor re-reads the protocol, the spec and old decisions for every case — and at database lock someone still has to build the Appendix 16.2.2 listing and the §10.2 summary by hand.
Estimated split for one deviation, by hand and with the solution. “By hand” means waiting for the next monthly review meeting.
Six specialist agents read, classify, compare, assess, watch and list; the medical monitor decides where opinions differ.
At Site 114, the Week 16 PASI for participant 114-0087 was done by a study coordinator who is not on the assessor training log. The CRO says not important. Here is what happens next, screen by screen, in the working solution.
Maya opens the solution on VLT-302, a Phase 3 plaque psoriasis study across 64 sites — 40 run by Ashgrove Research, 24 in-house. A cause map shows 418 deviations, 57 important, in 196 of 612 participants. Seven calls need her today; five are cases where a CRO and the agent disagree.
“Every deviation classified the same way, whichever team logged it.”
1,192 deviations across three studies, one classification spec per study, CRO and in-house in one place. Each row shows the team’s proposal and the agent’s call side by side — “Not important → Important 0.94” — with per-protocol and breach-check flags. At the top: DV-302-0412, Week 16 PASI done by an untrained assessor.
The deviation record sits on the left, the passages that matter highlighted: the PASI and sPGA were done by the coordinator, J. Moreno, while the trained sub-investigator was on leave; the visit was on Day 113, inside the ±3-day window. In the middle a balance weighs Ashgrove’s “not important” against the classifier’s “important”, and the two ICH E3 questions are answered — reliability of key data: yes; participant safety and rights: no.
Evidence lands on the side it supports — each point numbered back to its source.
Show the run, and every step is there with its time. The deviation reader read the visit report. The classifier matched spec row EA-03 and cited protocol §8.1.2 — important, confidence 0.94. The precedent finder found 3 signed decisions on similar deviations. Because Ashgrove proposed not important, the agreement check sent it to Maya. The impact assessor flagged that 114-0087 may leave the per-protocol set for the Week 16 PASI 75 and sPGA analyses, and notified Sam Patel.
“Read, classified and checked against 3 precedents in 41 s · 4 agents”
Beside the record: the protocol passage — PASI and sPGA must be done by an assessor who has completed the sponsor’s training and is listed on the site’s log — and spec row EA-03 with its action, a per-protocol review by the statistician for a primary assessment. The site’s training log lists Dr. A. Brooks and Dr. R. Ellison, and no entry for J. Moreno. The CRA’s visit report adds that this is the third time since August.
Training log: “No entry for J. Moreno.”
Maya keeps “Important” and signs. The reason is drafted from the evidence: the Week 16 PASI and sPGA were done by a coordinator not on the assessor training log; visit timing was within the window but does not change the assessment. Her name, the meaning — “I confirm the classification of this protocol deviation” — and her password are recorded together.
Had she overruled the classifier, the screen would say: “You are deciding differently from the classifier. Your reason becomes a precedent for future classifications.”
The activity trail records each step after the signature: the agreement check sent the decision and reason to Ashgrove Research and notified Dana Okafor; the impact assessor confirmed the per-protocol flag for 114-0087 and notified Sam Patel; the precedent finder added the decision to the precedent library. A signed decision can be reopened — with a reason, kept in the audit trail.
Site trends show important deviations per 100 participant-visits. Site 114, Hillcrest Dermatology, has risen to 3.1 — above the 2.0 threshold — with three assessments by staff not on the training log in 58 days, and two more participants reach Week 16 before Oct 30. The alert proposes one action: Create retraining action. Consistency by team shows the classifier at 96 % against the medical monitor’s final call, Ashgrove’s CRAs at 79 %.
Alerts propose actions; people decide.
The Appendix 16.2.2 listing is grouped by site, as ICH E3 asks, and built from confirmed decisions only. With Maya’s signature, 53 of 57 important VLT-302 deviations are confirmed, and 114-0087 shows as confirmed. Next to it: the Section 10.2 draft, with a table of important deviations by cause and citations back to the log, and the DV domain tabulation.
Sync from CTMS brings in three new records — from CTMS, EDC and IRT. They are classified in 21 seconds and checked against 11 signed decisions. Two, agreed not important at 0.90 or higher, are confirmed automatically. The third — site 188 dispensed kit 4417 instead of the assigned 4471 — is important and waits for Maya, and the impact assessor flags 188-0014 for per-protocol review.
“3 deviations classified in 21 s. Two confirmed automatically; DV-302-0421 — a wrong kit — needs your decision.”
The same deviation, seen by the five people who carry it — from the CRA who logs it to the medical writer who lists it.
Six specialist agents read, classify, compare, assess, watch and list; the medical monitor decides where opinions differ.
Reads each deviation from CTMS, EDC, IRT and visit reports and writes one clean record: participant, visit, occurrence and awareness dates, what happened, and the system it came from — keeping the site’s own words as the description.
Assigns category, subcategory and importance against the study’s classification spec, answering the two importance questions and citing the spec row and the protocol section breached, with a confidence.
Finds up to five of the closest signed decisions on similar deviations in the same study and programme, from any CRO, and says how many agree with the proposed classification.
For each important deviation, flags whether the participant may leave the per-protocol set, whether a serious breach assessment is needed for EU and UK sites with its 7-day clock, and whether a site rule fires.
Recalculates important deviations per 100 participant-visits by site, cause and team each night, and raises an alert with one proposed action when a rule in Settings fires.
Builds the Appendix 16.2.2 listing by site, drafts the §10.2 summary with citations in the house CSR style, and prepares the DV domain tabulation.
Signs where opinions differ. The agents propose; a named person decides.
Ask in plain words — why a deviation was classified the way it was, what is happening at a site, which breaches are open — or tell it what to change.
Why does the agent say DV-302-0412 is important?
Because the Week 16 PASI — the primary efficacy assessment — was done by a study coordinator who is not on the assessor training log. Protocol §8.1.2 requires a trained assessor, and spec row EA-03 classifies this as important. All 3 earlier signed decisions on the same rule were important. Ashgrove called it not important because the visit was inside the window — timing is fine, but it does not fix who assessed. Confidence 0.94. The decision is yours.
What is happening at Site 114?
Site 114 (Hillcrest Dermatology, Austin) has had 3 assessments by staff not on the training log in 58 days — Week 8, Week 12 and Week 16. Its important-deviation rate rose from 0.8 to 3.1 per 100 visits since May, the highest in VLT-302. Two more participants reach Week 16 before Oct 30, so retraining now protects the primary endpoint.
Any possible serious breaches?
One open: DV-302-0410 at Site 163, Valencia — an SAE reported to the sponsor 4 days after site awareness. As an EU site it went to Omar Haddad for a serious breach assessment on Oct 2; the 7-day clock ends Oct 9. Importance is agreed: important under SR-01.
Alert me when a site has 3 assessor deviations in 60 days
Done — I added a trend rule: “Assessor not on the training log: 3 at one site within 60 days”. It alerts you and Lena Ortiz and proposes retraining. It would already fire for Site 114 (3 in 58 days). You can switch it off in Settings.
13 screens from the working solution, on its sample data. Pick one to see it large.
The classification spec, who decides and when a person must look are settings — per study, changed in plain words.
| Setting | Default | Choose from |
|---|---|---|
| Classification specification | VLT-302 v3.0 · 46 rows | Upload a new version — open and confirmed classifications are re-checked against it |
| CRO and agent disagree | Dr. Maya Chen | Dr. Maya Chen · Lena Ortiz |
| Possible serious breach (EU and UK sites, 7-day clock) | Omar Haddad | Omar Haddad · Lena Ortiz |
| Per-protocol impact | Sam Patel | Sam Patel · Priya Raman |
| Confirm when CRO and agent agree on “not important” | On | On · Off — important deviations always go to a person |
| Minimum agent confidence | 0.90 | 0.80 to 0.99 |
| Quality sample of automatic confirmations | 10 % | 5 · 10 · 20 % each week |
| Trend alerts | Same cause at a site 3 times in 60 days · above 2.0 per 100 participant-visits · late SAE reports above twice the study rate | Add a rule |
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 · by hand = the next monthly review meeting. Sources: ICH E3 Q&A (R1) · FDA draft guidance on protocol deviations (Dec 2024). People, companies and products named on this page are fictional — characters and sample data in the working solution.
Software that takes each protocol deviation from the site or CRO log, classifies its category and importance, routes the ones that need judgement to the right person, and keeps the record ready for the clinical study report. The Protocol Deviation Classifier & Trend Analyst adds agents that read, classify, compare with precedents, assess impact, watch trends and write the CSR listing.
The classifier matches the deviation to a row of your study’s classification specification, cites the protocol section breached and answers two questions: could it significantly affect the reliability of key data, and could it significantly affect the participant’s rights, safety or well-being. It gives a confidence and shows the closest signed decisions on similar deviations.
Yes. Every important deviation, every disagreement between a team and the agent, and every call below the minimum confidence goes to a named person for a signed decision. Only agreed “not important” deviations confirm automatically, and a weekly quality sample re-checks them.
It flags them. The impact assessor marks participants who may leave the per-protocol set and notifies the statistician, who decides. For EU and UK sites it flags possible serious breaches to clinical quality with the 7-day notification clock.
Yes. Trends show important deviations per 100 participant-visits by site, the top causes, and how consistently each team classifies against the medical monitor’s final call. Alerts fire on the rules you set and propose one action, such as retraining.
Yes. It builds the Appendix 16.2.2 listing grouped by site, drafts the §10.2 summary with citations back to the log, and prepares the DV domain tabulation — from confirmed decisions only.
Your CTMS deviation log, EDC, IRT and monitoring visit reports, plus the protocols, amendments and classification specification for each study. Source systems are read, never changed.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your protocols, classification specifications and a sample of past deviations — and it goes live once every quality gate has passed. We will walk you through it on your own deviations 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 run it on a sample of your own deviations and classification spec.