Vitalis Therapeutics Inc. — Randomization, blinding and investigational product supply for the VitaFlow (VTX-401) pivotal program: the stratified block scheme, who is blinded to what, the IRT system as the single point of failure, the physical chain from manufacture to destruction, drug accountability, and the supply forecasting problem.
1:1
Allocation
41k
Kits manufactured
99.8%
Accountability
44%
Manufacturing overage
Contents
1. Randomization
| Element | This program |
|---|---|
| Allocation ratio | 1:1, VitaFlow to placebo |
| Method | Stratified permuted block |
| Block size | Variable (4 or 6), blinded to sites and to the sponsor |
| List generated by | Independent statistician, Meridian Clinical Research |
| List held by | Interactive response technology system; sealed copy with the independent statistician |
| List seen by | Nobody on the program team, at any point before database lock |
“Randomized” does not mean a coin flip per participant, and the two elements of the
method are doing different jobs.
Blocking ensures the arms stay balanced as enrolment proceeds. Without it, a trial that stops at an awkward moment — or a site that enrolls only four participants — can end badly imbalanced by chance.
Stratification ensures balance within the factors that predict outcome, so the comparison is not confounded by one arm happening to enroll heavier participants.
Blocking ensures the arms stay balanced as enrolment proceeds. Without it, a trial that stops at an awkward moment — or a site that enrolls only four participants — can end badly imbalanced by chance.
Stratification ensures balance within the factors that predict outcome, so the comparison is not confounded by one arm happening to enroll heavier participants.
The block size is variable and blinded, which is not fussiness. With a fixed block of four, a site that has observed three assignments can deduce the fourth. Predictable allocation is not allocation concealment, and a site that can predict the next arm can — consciously or not — time which participant it screens next.
| Factor | Levels | Why it is stratified |
|---|---|---|
| Baseline BMI category | <35 vs ≥35 kg/m² | Response magnitude differs by starting BMI. Imbalance here would confound the primary endpoint directly. |
| Prediabetes status | Present vs absent | A prognostic factor for weight response and a subgroup the label will describe. |
| Trial | 301 vs 302 | Randomization is within trial, not across the program. |
2. Blinding
| Party | Status | Note |
|---|---|---|
| Participants | Blinded | Placebo is visually and physically identical. |
| Investigators and site staff | Blinded | Including anyone assessing an endpoint. |
| Sponsor program team | Blinded | Including the Program Director and the Chief Medical Officer, until database lock. |
| CRO operational staff | Blinded | Except the independent statistician. |
| Independent statistician | Unblinded | Prepares DMC reports. Reports to the DMC, not to the sponsor. |
| Data Monitoring Committee | Unblinded | In closed session only. See the DMC Charter. |
Read the third row carefully. The sponsor's own Program Director and Chief Medical Officer
are blinded, for the whole trial, to the arm assignment of every participant.
This is the operational consequence of the argument the DMC Charter makes: the people running the trial are structurally incapable of noticing a safety signal, by design, because the alternative — letting them see the data — would let expectation contaminate the measurement. The blind data review that assigns analysis populations happens under exactly this constraint.
This is the operational consequence of the argument the DMC Charter makes: the people running the trial are structurally incapable of noticing a safety signal, by design, because the alternative — letting them see the data — would let expectation contaminate the measurement. The blind data review that assigns analysis populations happens under exactly this constraint.
Placebo is manufactured to be indistinguishable from active: same presentation, same pen, same labeling apart from the kit number. That is a real CMC obligation and not a trivial one — a placebo that looks, feels or injects differently unblinds the trial one participant at a time.
3. The IRT System
Interactive response technology sits between the randomization list and everyone who must not see it.
| Function | What it does | Note |
|---|---|---|
| Randomization | Assigns treatment on eligibility confirmation | The site enters eligibility; the system returns a kit number, not an arm. |
| Kit assignment and dispensing | Which physical kit goes to which participant | |
| Automatic resupply | Triggers shipment when site stock falls below threshold | Removes a forecasting judgment from 231 sites and centralizes it. |
| Expiry management | Blocks dispensing of expired kits; triggers replacement | Kits expire; a 68-week trial outlives several supply lots. |
| Emergency unblinding | Investigator-initiated, logged, irreversible | The single point at which blinding can be broken. Every use is recorded, reported and reviewed for clustering. |
The IRT is the single point at which blinding can break, and it is a piece of software.
It holds the list. It knows every assignment. Access control, role definition, audit trail and validation under 21 CFR Part 11 are therefore not IT hygiene — they are the trial's integrity, implemented in configuration.
A misconfigured report that returns arm rather than kit number, a role granted to the wrong person during a mid-study update, an export that includes a column it should not: any of those unblinds a trial silently, and it may not be discovered until someone notices a report that could only have been produced with knowledge nobody was supposed to have.
It holds the list. It knows every assignment. Access control, role definition, audit trail and validation under 21 CFR Part 11 are therefore not IT hygiene — they are the trial's integrity, implemented in configuration.
A misconfigured report that returns arm rather than kit number, a role granted to the wrong person during a mid-study update, an export that includes a column it should not: any of those unblinds a trial silently, and it may not be discovered until someone notices a report that could only have been produced with knowledge nobody was supposed to have.
Emergency unblinding is available to the investigator when knowing the assignment would change clinical management. Each use is logged and irreversible. Clustering is itself a signal — several unblinding events at one site suggests either a safety pattern or a site misunderstanding when the provision applies.
4. The Physical Chain
| Stage | Who / where | Note |
|---|---|---|
| Manufacture and packaging | Aldergate Biologics | Active and placebo produced and packaged to be indistinguishable. |
| Labeling | Aldergate, to clinical label specification | Investigational labeling differs from commercial: it carries the protocol number, kit number, storage conditions, expiry and the caution statement — and no indication of arm. |
| Depot storage | Regional depots, validated cold chain | 2–8°C throughout. |
| Shipment to site | Triggered by IRT, temperature-monitored | Each shipment carries a data logger; an excursion quarantines the consignment pending assessment. |
| Site storage | Site pharmacy, monitored and logged | Temperature excursions at site are the most common supply deviation. |
| Dispensing | To participant, against IRT assignment | |
| Return and reconciliation | Used and unused kits returned | Every unit accounted for. |
| Destruction | Certified, documented | Certificates filed to the TMF. |
Investigational labeling is not commercial labeling, and the difference is the point. A
clinical kit carries the protocol number, the kit number, storage conditions, expiry, the caution
statement that it is limited to investigational use — and nothing that indicates which
arm it is.
Everything a participant or a nurse could use to infer treatment has been deliberately removed. That is why the label is a controlled document reviewed by regulatory and quality, not a production detail.
Everything a participant or a nurse could use to infer treatment has been deliberately removed. That is why the label is a controlled document reviewed by regulatory and quality, not a production detail.
| Measure | Value |
|---|---|
| Kits manufactured | 41,600 |
| Manufacturing overage | 44% above dispensed forecast |
| Kits shipped to sites | 33,180 |
| Kits dispensed | 28,940 |
| Kits returned | 27,610 |
| Kits destroyed | 26,800 |
| Accountability | 99.8% |
| Unreconciled units | 54 |
| Temperature excursions | 23 — 2 resulted in product rejection |
Excursion data is what turns 23 excursions into 2 rejections. Without the thermal-cycling
work described in the Stability Program, every one of those 23
consignments would have been rejected automatically, because there would be no evidence on which
to assess whether the product was actually affected.
That is a study run years earlier paying for itself in inventory nobody had to destroy.
That is a study run years earlier paying for itself in inventory nobody had to destroy.
5. Drug Accountability
Every unit of investigational product must be accounted for from manufacture to destruction. Not most units. Every unit.
99.8% is a good figure and 54 unreconciled units is
still 54 open items.
Accountability is not primarily an inventory control. It is evidence about two things a regulator cares about: that participants received what the record says they received, and that investigational drug did not leave the trial. A kit that cannot be located is, formally, a kit whose fate is unknown — and unknown fate for an unapproved drug is a GCP finding regardless of how obviously it was simply discarded.
Accountability is not primarily an inventory control. It is evidence about two things a regulator cares about: that participants received what the record says they received, and that investigational drug did not leave the trial. A kit that cannot be located is, formally, a kit whose fate is unknown — and unknown fate for an unapproved drug is a GCP finding regardless of how obviously it was simply discarded.
Reconciliation runs at three levels — the site reconciles what it received against what it dispensed and returned; the depot reconciles shipments against returns; and the sponsor reconciles the whole chain against IRT records. All three must agree before a site can be closed out.
6. The Forecasting Problem
This is the part a program manager actually owns, and it is harder than it looks.
| Constraint | Why it bites |
|---|---|
| Supply must precede enrolment | A site cannot activate without stock. Product is manufactured against enrolment that has not happened yet. |
| Enrolment is running behind | 76 behind curve, with 46 planned sites not yet enrolling. Supply was positioned on the original curve. |
| Kits expire | Clinical supply carries an expiry. A 68-week treatment period outlives supply lots, so re-supply and re-labeling are continuous activities, not a one-time build. |
| Duration varies by participant | Someone who discontinues at week 12 consumes a fraction of what a completer does — and discontinuation rates are exactly what the trial is measuring. |
| Under-supply stops the trial; over-supply is destroyed | Hence 44% overage — a deliberate, costed decision, not an estimating error. |
The overage figure is the honest expression of the trade. 44% of
manufactured kits were never going to be dispensed. That looks like waste on a cost report and it
is the cheapest insurance in the program: a site that cannot dispense stops enrolling, and an
enrolment pause on a trial already 76 behind curve costs far
more than the product it would have saved.
It is the same shape of decision as the launch supply build at CD-05 — spending against an outcome that may not arrive, because the cost of being short is asymmetric.
It is the same shape of decision as the launch supply build at CD-05 — spending against an outcome that may not arrive, because the cost of being short is asymmetric.
The second row carries a live tension. Supply was positioned against the original enrolment curve; enrolment is behind, so some sites hold stock they are consuming slowly while the reserve sites in the recovery plan will need stock they do not yet have. IRT-triggered resupply handles the routine case. It does not handle a network whose shape is changing.
7. What Can Go Wrong
| Failure | Consequence |
|---|---|
| Blinding compromised via IRT configuration | The most serious. May invalidate affected data and, if widespread, the trial. |
| Distinguishable placebo | Unblinds participants individually and progressively. Often discovered late. |
| Predictable block size | Allows a site to anticipate allocation and, in principle, to select who gets screened next. |
| Expired product dispensed | A protocol deviation for every affected participant, and a question about every other control at that site. |
| Unreconciled accountability | A GCP finding at inspection, and it cannot be fixed retrospectively. |
| Stock-out at an enrolling site | Enrolment stops at a productive site — the thing the whole recovery plan is trying to avoid. |
Note that the first three are not operational failures. They are design failures that only
show up operationally.
A predictable block size is decided when the randomization scheme is written. A distinguishable placebo is decided in formulation. An IRT that can expose arm is decided in configuration and user-acceptance testing.
By the time any of them surfaces as an event, the decision that caused it is years old and the participants affected are already enrolled. This is the same pattern as the endpoint hierarchy and the market access assumption — the decisions that determine whether a trial can be believed are taken long before anyone could know they mattered.
A predictable block size is decided when the randomization scheme is written. A distinguishable placebo is decided in formulation. An IRT that can expose arm is decided in configuration and user-acceptance testing.
By the time any of them surfaces as an event, the decision that caused it is years old and the participants affected are already enrolled. This is the same pattern as the endpoint hierarchy and the market access assumption — the decisions that determine whether a trial can be believed are taken long before anyone could know they mattered.