← Drug Development Suite Execute · Vitalis Therapeutics Inc.

Central Laboratory & Testing Plan

Download Excel
148k
Samples expected
95%
Received
4d
Median turnaround
1.5%
Shipments with an excursion
Contents
  1. Comparability, Not Cost
  2. What Runs Where
  3. Chain of Custody
  4. Method Validation, and Why It Cannot Move
  5. Sample Reconciliation
  6. Retained Samples, and the Limits of Consent

1. Comparability, Not Cost

Every laboratory value that enters the analysis of this program is produced by one laboratory, on one instrument platform, against one set of reference ranges. Kestrel Laboratories runs it, and the reason is not price.

Position at 15 October 2026
Samples expected to date148,800
Samples received141,260 — 94.9%
Unusable on arrival1.1%
Shipments3,940, of which 61 had a temperature excursion (1.5%)
Median turnaround, receipt to result4 days
260 local laboratories would produce 260 sets of reference ranges, and no pooled analysis could survive it.

A laboratory value is only interpretable against the range of the laboratory that produced it. Two sites reporting the same number, measured on different platforms against different ranges, are not reporting the same result — one may be normal and the other flagged. Pool them and the count of abnormal findings becomes an artifact of which sites enrolled fastest.

That is the whole argument, and it is why the central laboratory is a scientific decision that happens to also be a procurement one. It is not cheaper than local testing. It is the only arrangement in which the safety tables mean anything.

2. What Runs Where

Test groupWhere it runsWhy thereThe part that matters
Efficacy endpoints — body weight, waist circumferenceSite, standardized equipmentPhysical measurement, not an assay⚠ Equipment is supplied and calibrated by the program, not the site. A site's own scale is not acceptable when the primary endpoint is a weight change.
Metabolic panel, HbA1c, lipidsCentralPooled across sites in the analysisOne method, one instrument platform, one reference range. This is the core of the central-lab argument.
Pharmacokinetics and anti-drug antibodiesCentral, specialist assayBioanalytical method under GLP-equivalent controlMethod validated before first sample; a mid-study method change requires bridging.
Safety chemistry and hematology for clinical decisionsLocal, with central confirmation⚠ An investigator needs a result in hours to treat a participant, not daysThe local result drives care; the central result drives analysis. Both are recorded, and they will not always agree.
Pregnancy testingLocal, point of careRequired before dosing; no time for shipmentResult documented in source before each dose in the relevant population.
Cardiovascular sub-study biomarkersCentral, batchedAdded under CR-02; analyzed together at study endSamples accumulate frozen and are assayed in one run, which removes inter-run variability.
One entry above surprises people: some tests run in both places, deliberately.

An investigator seeing a potassium result that might require intervention cannot wait four days for a central result. So safety chemistry runs locally for clinical decisions and centrally for analysis. Both values are real, both are recorded, and they will sometimes disagree.

The discipline is that the local result governs the participant's care and the central result governs the analysis, and neither is quietly substituted for the other when one is more convenient. A study that resolves those disagreements by preferring whichever value reads better has compromised its own safety dataset.

The first row is a different kind of care. Body weight is the primary endpoint of this program, and it is measured on equipment the program supplies and calibrates rather than on whatever scale a site happens to own. A primary endpoint measured on 260 uncalibrated instruments would carry a measurement error comparable to the treatment effect being sought.

3. Chain of Custody

Seven handoffs between a participant's arm and a number in an analysis dataset. Every one is a place a sample can be lost, and a lost sample cannot be replaced because the visit cannot be repeated.

StepOwnerWhat must be true
CollectionSiteLabeled with participant ID and timepoint from the kit label, never handwritten from memory.
ProcessingSiteCentrifuge, aliquot, freeze within a protocol-defined window. ⚠ The window is the most commonly missed step in the whole chain.
StorageSiteMonitored freezer with continuous temperature logging.
ShipmentCourierDry ice, temperature logger in every shipment, tracked.
ReceiptCentral labCondition on arrival recorded, including logger download. A shipment arriving warm is a documented event, not a discard.
AnalysisCentral labWithin assay-defined stability limits from collection.
RetentionCentral labResidual samples stored against the consent that was given.
Processing is where most losses actually happen, and it is the step with the least glamour.

A sample must be centrifuged, aliquoted and frozen within a protocol-defined window — typically measured in minutes to an hour. Miss it and the analyte degrades, quietly, in a tube that looks exactly like every other tube. The failure is invisible until the assay runs, weeks later, and by then the participant has moved on to their next visit.

1.1% of samples arrive unusable, and that number is reported rather than absorbed. Zero would not mean perfection; it would mean the definition had been relaxed.

Temperature excursions are handled the same way. 61 of 3,940 shipments arrived with a logger showing an out-of-range period, and each one is assessed rather than discarded. Most excursions are brief and fall within the analyte's demonstrated stability; the ones that do not are documented as losses. Treating every excursion as a rejection would discard usable data, and treating none as one would accept degraded data — the assessment is the control.

4. Method Validation, and Why It Cannot Move

PrincipleWhy
One method for the life of the study⚠ Changing a bioanalytical method mid-study requires a bridging study to demonstrate the two produce comparable results. It is doable, expensive, and it is why the method is validated before the first sample rather than after.
Reference ranges are the laboratory's, not the site'sA value is abnormal relative to the range of the laboratory that produced it. Mixing local and central ranges in one analysis produces abnormality counts that mean nothing.
The local result treats the participant; the central result is analyzedBoth are real and they will sometimes disagree. ⚠ Recording both, and never quietly preferring the more convenient one, is the discipline.
A sample without a matching record is not dataAn unlabeled or mismatched sample cannot be attributed to a participant or a timepoint, and is discarded. The loss is permanent because the visit cannot be repeated.
Changing a bioanalytical method mid-study requires a bridging study, and that is the reason the method is locked before the first sample rather than improved as you go.

If the assay changes, every result produced before the change and every result after it are measured on different instruments. Demonstrating they are comparable is its own piece of work — a cross-validation on shared samples — and until it is done the dataset cannot be pooled.

This is the same shape as the analytical method transfer that failed in March at Aldergate, and for the same underlying reason: a validated method carries undocumented practice, and moving it exposes what was never written down. The difference is that a manufacturing method transfer delays a batch, and a bioanalytical method change threatens an analysis.

The cardiovascular sub-study biomarkers illustrate the opposite discipline. Those samples accumulate frozen and are assayed in a single batched run at study end, which removes inter-run variability entirely. It costs storage and it delays the result — and for a sub-study whose purpose is to pre-empt a post-marketing requirement rather than to inform an interim decision, that is exactly the right trade.

5. Sample Reconciliation

Expected, received, analyzed, retained — four counts that must reconcile, and the gaps between them are all meaningful.

MeasureCountWhat a gap means
Expected148,800Derived from the protocol schedule multiplied by participants enrolled. ⚠ A denominator computed in advance, not counted from what arrived.
Received141,260 (94.9%)The shortfall is missed visits, missed collections and participants who discontinued. Every one should have a matching entry in the clinical database.
Unusable1.1% of receivedProcessing window missed, hemolysis, insufficient volume, label mismatch. Permanent loss.
Queries raised2,180Discrepancies between the laboratory record and the clinical database. ⚠ A zero here would mean nobody is reconciling.
The expected count is computed from the protocol, not from what turned up — and that is the whole control.

A laboratory reconciliation measured against samples received is complete by definition, always, and detects nothing. Measured against what the protocol required, the 5.1% shortfall becomes visible and every part of it has to be explained by something in the clinical database — a missed visit, a discontinuation, a collection the site did not perform.

This is precisely the same principle as the Trial Master File's expected-document list, and the same reason it works: a denominator fixed in advance is the only kind that can produce a finding.

Where the laboratory record and the clinical database disagree, both are investigated rather than one being corrected to match the other. A sample recorded as drawn with no corresponding visit in the database is as much a signal as a visit with no sample — and in a program of 2,480 participants, the reconciliation is how a systematic site-level problem becomes visible before it becomes a finding.

6. Retained Samples, and the Limits of Consent

QuestionAnswerWhy it constrains the program
What may be retained?Residual volume after protocol-specified testingStorage is cheap and the samples are irreplaceable, so retention is the default where consent permits it.
For how long?As stated in the consent the participant signed⚠ Not as long as the sponsor would find useful. The consent is the limit.
For what purpose?Only purposes described in that consentA future question the sponsor did not anticipate cannot be answered from these samples unless the consent contemplated it.
Can it be broadened later?Only by re-consenting the participantWhich, for a participant who completed the study two years ago, is usually not practical. The consent language written at study start determines what is possible for the life of the samples.
This is the clearest example in the program of a decision whose cost arrives years after it is made, and it is made by whoever drafts a paragraph of the consent form.

A narrowly written retention clause forecloses future research on samples that already exist and cannot be recreated. A broadly written one that participants did not properly understand is worse — it is a consent failure, and consent failures are not remediable after the fact.

The program's position is deliberately conservative: retention for purposes stated plainly, for a stated period, with a clear right to withdraw. It forecloses some future work, and it is the only position that survives being examined.

The retained samples outlive the program. They sit at Kestrel under the terms of a consent signed years earlier, governed by an agreement that has to survive the program's closure — which is why sample disposition is a named closure deliverable alongside the post-marketing obligations and the trial master file transfer.