Vitalis Therapeutics Inc. — Enrolment and site performance for the VitaFlow (VTX-401) pivotal program as at 15 October 2026: 1,684 randomized of 2,480, 76 behind the planned curve, across 214 enrolling sites. Includes the diagnosis and the recovery plan.
Contents
1. Position at 15 October 2026
1,684
Randomized
68%
Of target
76
Behind curve
214
Sites enrolling
| Measure | Value |
|---|---|
| Randomization target | 2,480 across two pivotal trials |
| Randomized to date | 1,684 (68%) |
| Planned to date | 1,760 |
| Variance to curve | 76 behind (96% of plan) |
| Screened | 2,910 |
| Screen failures | 1,142 |
| Currently in screening | 84 |
| Screen-failure rate | 40% of completed screening |
| Sites planned / activated / enrolling | 260 / 231 / 214 |
| Randomizations per enrolling site per month | 2.2 |
How to read a set of numbers like this. The headline — 76 behind curve —
tells you there is a problem but nothing about where it is. Enrolment shortfall has exactly three
possible causes, and they call for different and mutually exclusive responses:
- Not enough people are being screened → a referral and site-productivity problem.
- Too many of those screened are failing → an eligibility or pre-screening problem.
- Not enough sites are open → a start-up and activation problem.
2. The Curve
| Month | Plan | Actual | Delta | Cumulative | Note |
|---|---|---|---|---|---|
| Jul 2026 | 180 | 142 | -38 | 142 / 180 | Activation ramp — 96 sites live at month end |
| Aug 2026 | 420 | 361 | -59 | 503 / 600 | Ramp continues; 168 sites live |
| Sep 2026 | 640 | 598 | -42 | 1,101 / 1,240 | 214 sites enrolling |
| Oct 2026 (to 15th) | 520 | 583 | +63 | 1,684 / 1,760 | First month ahead of plan |
The shortfall accumulated in the first two months and is now closing. October is the first month ahead of plan, driven by sites that activated in August reaching steady-state productivity.
The shape matters more than the total. A program 76 behind after four months of a
27-month enrolment window is not in trouble; a program 76 behind and widening would
be. The deficit here was built entirely during activation ramp — 96 sites live in July
against a plan built on more — and monthly run-rate has now caught up.
The forecast risk is therefore not the current gap. It is whether the remaining 46 planned sites come online, because the curve from here assumes they do.
The forecast risk is therefore not the current gap. It is whether the remaining 46 planned sites come online, because the curve from here assumes they do.
3. Where the Randomizations Come From
| Tier | Criterion | Sites | % sites | Randomized | % rand. | Note |
|---|---|---|---|---|---|---|
| High performers | ≥12 randomized | 42 | 18% | 812 | 48% | 18% of activated sites delivering 48% of randomizations. |
| On plan | 6–11 randomized | 78 | 34% | 601 | 36% | Performing to the assumption the enrolment curve was built on. |
| Below plan | 1–5 randomized | 94 | 41% | 271 | 16% | 41% of sites delivering 16% of randomizations. The remediation population. |
| Activated, none randomized | 0 | 17 | 7% | 0 | — | Activated but not yet productive — 7% of activated sites, inside the published ~11% non-activation norm, but 17 sites of paid start-up returning nothing. |
| Total activated | — | 231 | 100% | 1,684 | 100% |
Read the first and third rows together. 42 sites —
18% of those activated — are producing
48% of randomizations. Meanwhile
94 sites, 41% of the network, are
producing 16%.
This distribution is normal and is the single most useful fact in clinical operations. Enrolment is not evenly distributed across sites and never has been. A plan that assumes the average site delivers the average number will be wrong in both directions: it will under-resource the high performers who could take more, and it will keep paying for a long tail that will not improve.
This distribution is normal and is the single most useful fact in clinical operations. Enrolment is not evenly distributed across sites and never has been. A plan that assumes the average site delivers the average number will be wrong in both directions: it will under-resource the high performers who could take more, and it will keep paying for a long tail that will not improve.
17 sites are activated and have randomized nobody. Published benchmarks put site non-activation at roughly 11% globally; at 7% this program is inside the norm. That does not make it costless — each one carries completed contracting, IRB submission, site initiation and training.
4. Why Participants Fail Screening
| Reason | Count | Share | Assessment |
|---|---|---|---|
| Prior GLP-1 receptor agonist exposure within 90 days | 384 | 34% | The largest single cause, and a direct consequence of requiring a washout in a market where GLP-1 use is widespread. Shortening it risks carry-over on the primary endpoint. |
| HbA1c ≥6.5% or diagnosed diabetes | 297 | 26% | Non-negotiable. A weight-management indication cannot be confounded by glycemic effect. |
| BMI below threshold at screening visit | 168 | 15% | Largely a referral-quality problem — sites pre-screening on self-reported rather than measured BMI. Avoidable waste. |
| Body weight unstable in the prior 90 days | 121 | 11% | Design integrity: an unstable baseline makes the endpoint uninterpretable. |
| Other exclusion criteria | 94 | 8% | Pancreatitis history, MTC/MEN2 family history, recent bariatric surgery, cardiovascular event within 180 days. |
| Withdrew consent during screening | 78 | 7% | Participant choice. Monitored for clustering by site, which would indicate a consent process problem rather than ordinary attrition. |
| Total screen failures | 1,142 | 100% |
Three-quarters of screen failures are structural and cannot be touched. Prior GLP-1
exposure (34%), diabetes or elevated HbA1c
(26%) and unstable weight
(11%) all protect the interpretability of the
primary endpoint. Relaxing any of them would trade a recruitment problem for an analysis
problem, and the analysis problem is the one that ends programs.
One category is genuinely avoidable. 168 participants — 15% of all failures — failed on BMI below threshold at the screening visit. BMI is measurable before a visit is booked. Those are consented participants, site visits, and staff hours spent on people who were never eligible. That is the cheapest fix on the recovery list and it is a pre-screening discipline issue, not a protocol issue.
One category is genuinely avoidable. 168 participants — 15% of all failures — failed on BMI below threshold at the screening visit. BMI is measurable before a visit is booked. Those are consented participants, site visits, and staff hours spent on people who were never eligible. That is the cheapest fix on the recovery list and it is a pre-screening discipline issue, not a protocol issue.
The GLP-1 washout deserves particular note because it is a genuine strategic bind. The 90-day requirement exists to prevent carry-over effect on the primary endpoint. But the trial recruits from a population in which GLP-1 use is now common — so the criterion that protects the science is the same criterion that shrinks the eligible pool. The protocol holds at 90 days and the program absorbs the cost. The Protocol Summary §7 classifies it as schedule-critical rather than fixed for exactly this reason: it could move, and moving it would be a mistake.
5. Diagnosis
Against the three possible causes at §1:
| Possible cause | Evidence | Verdict |
|---|---|---|
| Not enough people being screened | 2,910 screened for 1,684 randomizations, with 84 in flight. Screening volume is adequate to the target. | not the constraint |
| Too many screen failures | 40% of completed screening, mid-range against a published 20–80% span. Three-quarters of failures are structural. | not the constraint |
| Not enough sites open and productive | 29 of 260 planned sites not activated; a further 17 activated but randomizing nobody; 94 below plan. | this is the constraint |
The shortfall is a site problem, not a screening problem, and that determines the entire
recovery strategy. 46 sites are either unactivated or
activated and unproductive — roughly 18%
of the planned network contributing nothing.
Recovery therefore runs through activation and site remediation. It does not run through eligibility criteria, and a program that reached for the criteria here would weaken its own endpoint to solve a problem the criteria did not cause.
Recovery therefore runs through activation and site remediation. It does not run through eligibility criteria, and a program that reached for the criteria here would weaken its own endpoint to solve a problem the criteria did not cause.
6. Recovery Plan
| Action | Owner | By | Rationale |
|---|---|---|---|
| Activate the 29 unstarted sites | Dr. R. Molyneux | Q4 2026 | Highest-yield lever available. Start-up cost is already partly sunk. |
| Remediate or replace the 17 zero-randomizing sites | Dr. R. Molyneux | Q1 2027 | Site visits to establish whether the constraint is referral flow, staffing, or pre-screening quality. Replacement only where it is not fixable. |
| Draw from the reserve site list | Meridian Clinical Research | Q1 2027 | 34 pre-qualified sites held in reserve at protocol finalization for exactly this. |
| Improve pre-screening quality | Meridian Clinical Research | Q4 2026 | 168 screen failures were BMI-below-threshold, measurable before a screening visit is booked. Pure waste and the cheapest thing on this list to fix. |
| Central pre-screening registry | Dr. R. Molyneux | Q1 2027 | Identifies washout candidates early so the 90-day GLP-1 clock starts before referral rather than at the screening visit. |
The reserve list at item three is worth noting as a planning practice: 34 sites were pre-qualified at protocol finalization and deliberately held back. Qualifying a site takes months; a program that begins qualification only when it discovers it is behind has already lost the time it is trying to recover.
What is deliberately not on this list. No protocol amendment. No relaxation of the
washout, the HbA1c threshold, or the BMI criterion. No extension of the enrolment window, which
would push last-patient-in and therefore database lock, the NDA date, and every dependency in
§19 of the Charter.
Each of those would show up in the next reporting period as an improvement. Two of the three would cost the filing.
Each of those would show up in the next reporting period as an improvement. Two of the three would cost the filing.
Forecast
A dashboard without a forecast is a rear-view mirror. Four scenarios, against a target of 2,480 randomizations.
| Scenario | Implied rate | Last patient in | Assessment | Assumption |
|---|---|---|---|---|
| Current run-rate continues | 480 / month | Mar 2027 | on plan | Assumes no further site activation and no attrition in current performers. |
| Recovery plan delivers in full | 545 / month | Feb 2027 | ahead | 29 unstarted sites activated plus reserve draw. |
| Recovery plan delivers half | 510 / month | Mar 2027 | on plan | The planning case. Most recovery plans deliver partially. |
| No recovery, current sites only | 395 / month | Jun 2027 | at risk | Assumes below-plan sites do not improve. Last-patient-in slips a quarter, which moves database lock and the filing date. |
The planning case is the third row, not the second. Recovery plans deliver partially —
sites activate late, some of the remediated sites do not improve, and reserve sites take their own
22 weeks to come online.
Forecasting on full delivery of a recovery plan is how a program reports green until the month it reports red. The plan of record assumes half.
Forecasting on full delivery of a recovery plan is how a program reports green until the month it reports red. The plan of record assumes half.
Only the fourth scenario threatens the filing, and it does so indirectly. A June 2027
last-patient-in slips database lock by a quarter, which slips topline, which slips the Gate 5
submission — and the submission date sets the action date, the launch, and every commercial
assumption downstream of it (DEP-05).
That is the escalation trigger named at §7. Not “enrolment is behind” — enrolment has been behind since July and nobody escalated. The trigger is any credible forecast in which last-patient-in moves the filing date, which is a different and much narrower condition.
That is the escalation trigger named at §7. Not “enrolment is behind” — enrolment has been behind since July and nobody escalated. The trigger is any credible forecast in which last-patient-in moves the filing date, which is a different and much narrower condition.
7. Governance Position
Enrolment shortfall is R-05 in the RAID Log, rated Medium, with a horizon of June 2027. It is materializing at a manageable scale and within the response recorded when it was registered — site over-recruitment against plan, clinic partnerships, and a pre-screening registry.
It is reported to the Development Committee at the monthly checkpoint rather than escalated, because nothing about it currently requires a decision the Chair cannot make. It becomes an escalation if any of three things happen:
- The monthly run-rate stops improving for two consecutive periods.
- Recovery requires contingency beyond the Chair's delegated authority.
- Any credible forecast shows last-patient-in moving the Gate 5 submission date.
The third is the one that matters. Enrolment does not have its own deadline — it has
a deadline because everything downstream of it does. Last-patient-in sets last-patient-out, which
sets database lock, which sets topline, which sets NDA submission
(DEP-05). A four-week enrolment slip is not a four-week problem if it consumes float that
does not exist.