The Silent Stockout: No One Owns The Data That Would Have Caught It

By Neeraj Shah, Founder & CEO, Boston Insights

The systems that decide whether a patient receives their investigational medicine have five owners in five different reporting lines — IT, clinical operations, manufacturing, data management, and often an external CRO.

Not one of them owns the only question that matters: do the systems agree?

How The Failure Shows Up

A coordinator opens the IRT for a routine kit assignment. The system quietly assigns a kit that isn’t there. The site calls the depot. The depot’s warehouse system shows stock the IRT says doesn’t exist. Someone starts a spreadsheet. Ninety minutes later, a patient goes home without a dose.

The postmortem blames logistics — a courier, a customs hold, a low forecast. Logistics is easy to blame because it’s visible. A truck. A border. A shelf.

But often the physical supply chain worked perfectly. The drug was in the building. What drifted was the data describing it — where it was, who it belonged to, and which protocol version governed its release.

That is not a logistics failure. It is an ownership failure.

Five Systems, Five Clocks

Clinical supply runs on systems that were never designed as one:

• IRT/RTSM — patient enrollment and randomization and drug dispensing logic
• CTMS — site and staff status
• EDC — patient visit and eligibility data
• ERP/WMS — physical inventory, lots, and expiry
• Regulatory tracker — protocol version control

Each is well governed in isolation. Each has its own owner, change log, and definition of “current.”

None can answer the question that matters in real time: does the drug that physically exists match the patient entitled to receive it, under the protocol in effect, at this site, today?

Most organizations reconcile these systems manually, periodically, optimistically. The moment something changes faster than that cycle, the systems disagree.

No one notices until a patient is standing at the site.

Why Now: An Edge Case Became The Norm

Protocol amendments drive most cross-system drift. And amendments are rising fast.

Seventy-six percent of Phase 1–4 protocols now require at least one substantial amendment — up from 57% in 2015, averaging 3.3 per protocol (Tufts CSDD).

Every amendment re-encodes dosing and eligibility logic that must be pushed separately into the IRT, CTMS, and EDC — on different timelines, by different vendors.

Clinical sites now run on mismatched protocol versions for an average of 215 days per cycle. That’s seven months of two systems being both “correct” and mutually contradictory.

The cost is real: one Phase 3 amendment carries a median cost of about $535,000, and the industry spends $7 billion to $8 billion a year on them. Most of that buys retraining and reconfiguration, not science.

None of them buy the one thing that would prevent the next failure: a person accountable for whether the systems agree.

The Three Gaps Behind A Missed Dose

Walk the data trail back from the empty-handed patient.

1. Site-status lag. The CTMS shows a site active; the IRT still flags it “pending initiation.” Kits sit unreleasable in the depot, invisible to anyone forecasting demand.
2. Lot data staleness. Expiry, temperature holds, and quarantine live in the warehouse system — but reach the IRT only at the last batch sync. A lot quarantined this morning can still be assigned this afternoon.
3. The shadow spreadsheet. A clinical supply manager’s file tracks what the systems can’t explain. It’s often more accurate than any system — because a person patches gaps in real time.

The organization’s real source of truth is an unaudited, un-versioned file on one person’s laptop. And it stays invisible until that person is on leave the week an amendment goes live.

The Stakes Are No Longer Line-Level

A missed dose can be a protocol deviation — and, depending on the drug, a patient safety event: a missed titration, a gap in a washout, a break in a regimen the protocol was built to protect. Sponsors must track, explain, and often report these deviations. “The two systems didn’t agree” is not an answer an inspector, an IRB, or a patient’s family will accept.

It’s also a data integrity exposure. The ALCOA+ principles that govern GCP data apply to release drug data just as they do to a case report form. A dispensation record that can’t be reconciled across systems fails “consistent” and “accurate” by definition.

And the stakes are rising with the trial models the industry is betting on:

• Direct-to-patient trials ship medicine to homes and need address, eligibility, and temperature data to hold in real time.
• Adaptive designs amend protocols on a rolling basis, compounding the drift.
• Cell and gene therapy treats every patient as a single batch — a data error isn’t a kit shortage, it’s a wasted, irreplaceable dose.

Cross-system data agreement is no longer an IT discipline. It’s the difference between a trial model that works and one that can’t be trusted.

Don’t Buy A Model You Can’t Trust

Most organizations find these gaps only after a dose is missed, because each system reports its own data as clean. The disagreement lives in the space between systems — and nothing watches that space.

There’s a maturity curve here:

1. Descriptive: Each system tells you, after the fact, what it did.
2. Real-time: You can see the status across systems as it changes.
3. Predictive: You can see where data is about to disagree, before it hurts.

Most organizations sit at stage one for the data that matters — even when their inventory dashboards look real-time. Visibility into boxes on shelves is not visibility into whether the data on those boxes matches the patient waiting for them.

The instinct is to buy a predictive model. That’s backward. An algorithm on top of five disagreeing systems doesn’t predict — it guesses confidently. AI on bad data just automates confusion faster.

The fix isn’t better forecasting. It’s ownership.

Start With The Org Chart, Not The Technology

Assign an owner to the patient-drug-protocol match. Their job is to certify that IRT, CTMS, EDC, and warehouse systems agree — not to own any single one. Four owners in four reporting lines is, in practice, no ownership at all.

Then make the agreement visible and continuous:

• Reconcile continuously, not periodically. If an amendment or lot change can update one system without triggering a check in the others the same day, that gap is the next missed dose.
• Put change control on configuration, not just the protocol. An IRT build is a controlled artifact — versioned, approved, dated — tied to the amendment it implements.
• Instrument the disagreement. Flag where systems diverge — a site-status mismatch, a kit assigned against an expired configuration — before dispensation, not after.
• Retire the shadow spreadsheet by making the official systems trustworthy enough that no one needs it. Its existence is the clearest sign of where your real source of truth lives.

The Unglamorous Layer Clinical Supply Depends On

Data governance won’t headline a keynote. It isn’t glamorous. But it’s what makes every downstream investment — forecasting, optimization, simulation — trustworthy rather than merely fast.

A sponsor that solves cross-system agreement will prevent more missed doses, more deviations, and more inspection findings than one that buys a smarter model for the same broken systems.

The patient who left with no medicine wasn’t failed by the depot. They were failed by five systems that were each individually correct — and by an organization that never assigned anyone the job of noticing when they stopped agreeing.

It’s a solvable problem. Today, it’s still an unowned one.

Key Takeaways

• A missed dose is rarely a logistics problem. It’s five systems each holding a “correct” version of the truth that no longer agree — and no one owning the agreement.
• Protocol amendments are the main driver of drift, and they’re accelerating. Seventy-six percent of trials are now amended; sites run on mismatched versions for about seven months; one amendment costs roughly $535,000.
• The fix is not more forecasting or AI. It’s ownership — one accountable owner for the patient-drug-protocol match.
• Three moves do most of the work: continuous reconciliation, change control on configuration, and monitoring that flags divergence before it causes harm.
• Data governance isn’t compliance overhead. It’s the foundation that makes every downstream investment worth trusting.

Article Sources:

1. Getz K, Smith Z, Botto E, Murphy E, Dauchy A. “New Benchmarks on Protocol Amendment Practices, Trends and their Impact on Clinical Trial Performance.” Therapeutic Innovation & Regulatory Science. 2024;58(3):539–548. doi:10.1007/s43441-024-00622-9.
2. Getz KA, Stergiopoulos S, Short M, Surgeon L, Krauss R, Pretorius S, Desmond J, Dunn D. “The Impact of Protocol Amendments on Clinical Trial Performance and Cost.” Therapeutic Innovation & Regulatory Science. 2016;50(4):436–441. doi:10.1177/2168479016632271.
3. Getz K. “Doubling Down on Protocol Amendments and Deviations.” Pharmaceutical Outsourcing. 2022.

About The Author:

Neeraj Shah is founder and CEO of Boston Insights. Neeraj has been a pharmaceutical supply chain practitioner who led the strategy and execution of global supply chain digital transformation initiatives at world’s leading  biopharmaceutical companies, including Bristol Myers Squibb, Celgene Corp, Shire Pharmaceuticals, and Biogen for over 20 years.

He founded Boston Insights in 2023 to develop the industry's first SaaS-based clinical supply chain control tower that provides real time visibility of global clinical supply networks and establishes resilient and a semi-autonomous supply chain.

He also designed and built a SaaS-based data-governance platform that monitors data inconsistencies across multiple systems in the clinical supply chain value stream that can help stakeholders teams to ensure data consistency across systems and prevent the patient from missing a drug dispensation.