Protocol design optimization that fixes avoidable amendments before the committee meets
Avoidable amendments fixed before the committee meets, each flag tied to a past amendment.
See one case, screen by screen ↓A draft protocol goes out for comment to six functions — clinical science, biostatistics, safety, clinical operations, medical writing, regulatory — and the comments come back by email and in tracked changes. The protocol lead reconciles them, checks the synopsis against the body and the Schedule of Activities by eye, and builds a pack for the Protocol Review Committee. Nobody has time to reread the archive of past summaries of changes before the meeting.
So the same design choices return: an inclusion criterion stricter than the target label, a washout longer than the standard, a screening window too short for a centrally read colonoscopy, an estimand that leaves intercurrent events to the analysis plan, a product with a liver signal and no stopping rules. Each one surfaces after first participant in, as a substantial amendment — ethics and agency approvals again, re-consent, and months of study time.
Estimated split, in working days, for one Phase 3 draft from first circulation to the committee, by hand and with the solution.
Six reviewer agents read every section and cite the past amendment behind each flag; owners decide, the lead signs.
VLT-305 is a Phase 3 study of Veltrimab in moderately to severely active Crohn's disease: 640 participants, about 180 sites in 22 countries, 52 weeks. Draft v0.9 runs to 148 pages and meets the Protocol Review Committee on Oct 14, in seven days. Here is what happens next, screen by screen, in the working solution.
The board opens on eight drafts in committee order, from NWB-118 on Oct 9 to VLT-307 on Nov 13. Each column is one protocol, its ten sections coloured by the strongest open finding. VLT-305 stands out at 73% amendment risk, with three findings in §5 Population and three in §8 Assessments & schedule. Across the board, 18 of 26 open findings cite a past amendment.
“Risk = chance of at least one avoidable substantial amendment, estimated from 148 past amendments.”
Inclusion 3 requires failure of at least two biologic classes: 22% likely to force an amendment, beside VLT-204 Amendment 2 — 4.1 months, $162K, 64 re-consented. The 12-month washout in exclusion 7 sits beside CRV-301 Amendment 1, which followed 112 screen failures. “History of any malignancy” sits beside VLT-302 Amendment 1, where 37 screen failures were treated basal cell carcinoma.
“Inclusion criterion 3 was broadened … Enrolment was 41% of plan at month 6, and 58% of screen failures were due to criterion 3.”
One click opens VLT-204 Amendment 2 itself. The rationale reads that the original criterion was not required by the target label, and investigators at 31 of 52 sites reported eligible patients being referred elsewhere. The amendment cost $162K directly, 4.1 months from decision to last approval, 52 ethics committee approvals and 64 participants re-consented.
In the review, the finding shows the protocol line, the precedent, design standard NWB-STD-014 §3.2 and ICH E8(R1) §3.3. It also notes that the criterion clashes with the randomisation stratum “prior biologic exposure (yes or no)” — every participant would be “yes”. Maya accepts: the new wording, “conventional therapy (corticosteroids or immunomodulators) or at least one biologic class”, goes in as a tracked change. Risk falls from 73% to 67%, expected cost from $656K to $556K.
“17% once every fix is accepted.”
The lead reviewer compares the synopsis, the body text and the Schedule of Activities for every procedure and visit. The co-primary ileocolonoscopy is at Week 12 in the synopsis and §8.2, but Table 2 places it at Week 10 — and sites follow the table; NXL-115 recorded 18 deviations before its amendment. Six mismatches in all, each fixable in Table 2. Hours on site per visit show Week 4 at 6.5 hours and a burden index of 1.38× the Crohn's study median.
Estimands go to Priya Raman, safety to Dr. Omar Haddad, feasibility to Lena Ortiz, consistency to Dana Okafor. Omar has already accepted the IGRA retest wording; Priya has raised the placebo endoscopic-response assumption for the committee. Still open: the 28-day screening window (18%), the primary estimand that leaves intercurrent events to the analysis plan (14%) and the missing liver stopping rules (11%).
Every decision is logged with who, when and why — “the record the committee and inspectors see.”
Maya clicks “Generate pack”. The lead reviewer collects the decisions and reasons, draws the risk profile by section, writes the summary and the questions for the committee, and checks every citation resolves. The pack for Dr. Sam Patel's committee reads 74% at draft, 67% now; two accepted, one raised for the committee, nine still open, ten past amendments cited.
“Every statement links to a finding and its sources.”
When she tries to sign, the solution stops her: the 28-day screening window, the 12-month washout and the primary estimand are still open, each with its owner and likelihood. Once every high finding has a decision, she signs the disposition log with its meaning — “I confirm these dispositions as protocol lead” — and the draft goes to the committee.
“The committee receives this draft once every high finding has a decision.”
CRV-215, a Phase 2 dose-ranging study in HCM, arrives as draft v0.6 for the Nov 4 committee. “Run the crew” sends its 104 pages to the five reviewers in parallel. The lead reviewer then compares the synopsis, body and schedule, merges duplicate findings, estimates each likelihood from the 148 past amendments, ranks the findings and assigns owners.
The lead reviewer ranks; it never decides a finding or signs.
The dashboard tracks amendment risk at draft against risk at committee sign-off, findings by area of the protocol, decisions by reviewer and where findings cluster by section and therapy area. On the sample portfolio, §5 Population in immunology carries the most findings per protocol, and the Consistency check has the highest acceptance.
The same draft, seen by the people who own its sections — what review looked like before, and what it looks like now.
Six reviewer agents read every section and cite the past amendment behind each flag; owners decide, the lead signs.
Reads inclusion and exclusion criteria, washouts and population definitions, and compares each criterion with the target label, design standard NWB-STD-014 and every past amendment that changed a similar criterion. Estimates the likelihood that a criterion forces a substantial amendment.
Reads the Schedule of Activities and section 8, totals participant hours and invasive procedures per visit, and compares them with the therapy-area median and NWB-STD-021. Flags visits over 4 hours and procedures with no endpoint or safety reason.
Checks the five estimand attributes in ICH E9(R1) for each primary and key secondary objective, and flags any intercurrent event without a stated strategy and any assumption without a benchmark source.
Reads the investigator brochure and earlier study results for the product and confirms that every known or suspected risk has monitoring, a stopping or interruption rule and a reporting path, citing the earlier signal and NWB-STD-042.
Compares screening windows, sample shipping, vendor turnaround and country plans with site feasibility history — screening times, colonoscopy waits and courier times from 1,240 site records.
Compares every endpoint and procedure across the synopsis, the body and Table 2, merges duplicate findings, estimates each finding's likelihood from the amendment archive, ranks them, assigns the owner by area and writes the committee pack.
Decide each finding; the lead signs. The agents propose; a named person decides.
Ask in plain words — which findings matter most, what a precedent says, whether the schedule agrees — or tell it what to change. Every answer points to its source.
Has this biologic criterion forced an amendment before?
Yes. VLT-204 Amendment 2 (Mar 2024) broadened the same criterion — failure of at least two biologic classes — to conventional therapy or one biologic. Enrolment was 41% of plan at month 6, and 58% of screen failures came from that criterion. It cost $162K directly and 4.1 months. VLT-305 inclusion 3 repeats the wording; the eligibility reviewer estimates a 22% chance it forces an amendment.
What is wrong with the VLT-305 estimand?
Section 3.2 says intercurrent events “will be handled as described in the statistical analysis plan”. ICH E9(R1) expects the strategy for each intercurrent event to be stated in the protocol. CRV-301 did the same and the agency asked for the strategies after first participant in — Amendment 2, $610K and 4 months. Suggested: a composite strategy for rescue medication or stopping for lack of efficacy, and treatment policy for other reasons. Owner: Priya Raman.
Are the liver safety rules adequate?
No — VLT-305 has no hepatic stopping rules. In VLT-204, 2.1% of participants on Veltrimab had ALT above 3× ULN versus 0.6% on placebo, and Amendment 4 had to add stopping rules mid-study and re-consent 211 participants. The safety standard requires them when an earlier study shows a liver signal. The same gap is in the extension study VLT-306. Owner: Dr. Omar Haddad.
Add a rule: flag screening windows under 35 days when an endoscopy is needed
Done. The feasibility reviewer will flag any screening window under 35 days when screening needs an endoscopy with central reading. It is based on VLT-204, where the median wait for a colonoscopy slot was 19 days plus 6 days of central reading. It applies from the next review and already matches VLT-305. You can switch it off in Settings.
13 screens from the working solution, on its sample data. Pick one to see it large.
What the reviewers raise, who decides it and what holds a draft back are settings — changed in one place, on the record.
| Setting | Default | Choose from |
|---|---|---|
| Minimum likelihood of an amendment | 3% | 3% · 5% · 10% — findings below it go to a watch list instead of the owner |
| Block “Send to committee” while high findings are open | On | On · Off |
| Re-run the crew on every new draft version | On — decisions carry over when the line is unchanged | On · Off |
| Eligibility & population findings go to | Dr. Maya Chen | Dr. Maya Chen · Priya Raman · Dr. Omar Haddad · Lena Ortiz · Dana Okafor |
| Endpoints & estimands findings go to | Priya Raman | Dr. Maya Chen · Priya Raman · Dr. Omar Haddad · Lena Ortiz · Dana Okafor |
| Safety monitoring findings go to | Dr. Omar Haddad | Dr. Maya Chen · Priya Raman · Dr. Omar Haddad · Lena Ortiz · Dana Okafor |
| Review rules | Four rules, e.g. “Flag visits longer than 4 hours” (NWB-STD-021 §4.1) and “Flag primary estimands without an intercurrent-event strategy” (ICH E9(R1)) | Add a rule |
| What the reviewers read | Amendment archive, design standards, ICH E6(R3), E8(R1), E9(R1), ICH M11 template, site feasibility history | Re-index any source |
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 protocols · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (Tufts CSDD, Getz et al., Ther Innov Regul Sci 2016 — 57% of protocols with at least one substantial amendment, 45% of those avoidable, $535K median direct cost per Phase 3 amendment and about 3 months added; Tufts CSDD 2024 study of 950 protocols — 3.3 amendments per amended protocol). ICH E6(R3), E8(R1) and E9(R1) are named as context. People, companies, products and studies named on this page are fictional — characters and sample data in the working solution.
Checking a draft protocol for design choices that will cost enrolment, participant burden or an amendment, before it is final. Protocol Design Optimization does it with six reviewer agents that read each draft section by section against your own past amendments, your design standards and ICH E6(R3), E8(R1) and E9(R1), and show every draft's amendment risk on one board.
Each finding is matched to the past amendment it resembles — what changed, how many months it took, what it cost — and the lead reviewer estimates its likelihood from the archive. Amendment risk is the chance of at least one avoidable substantial amendment; on the sample archive it is estimated from 148 past amendments.
Yes. It reads your summaries of changes from past protocols, classified by area, avoidability, direct cost and delay, and cites them by study and amendment number. When a study is amended, the new summary of changes is read, classified, confirmed by a person and added to the archive for the next review.
ICH E6(R3) for quality by design and operational feasibility, ICH E8(R1) for critical-to-quality factors and ICH E9(R1) for estimands, plus your own design standards. Drafts are split into ICH M11 sections, and the synopsis, body and Schedule of Activities are compared for every procedure and visit.
No. The agents propose changes; they never edit the controlled document. When an owner accepts a suggestion it goes in as a tracked change, and statistical and safety suggestions always go to the biostatistician and the safety physician.
Yes. A named owner decides every finding — accept, reject with a written reason, or raise to the committee. High findings must be decided before the draft can be sent, the protocol lead signs the disposition log, and the committee chair approves the protocol.
Yes. The minimum likelihood to raise a finding, whether open high findings block the committee, re-review on every new version, the owner for each area, your review rules and the sources the reviewers read are all settings.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your design standards, protocol template and past summaries of changes — and it goes live once every quality gate has passed. We will walk you through it on one of your own draft protocols 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 Protocol Design Optimization on one of your draft protocols, against your own past amendments.