IRT / RTSM
Interactive Response Technology / Randomisation and Trial Supply Management
An IRT — increasingly called an RTSM — is the system that executes a trial's randomisation and runs its investigational product supply chain. When a site confirms an eligible subject, the system assigns the treatment arm according to the randomisation design, tells the site which blinded kit to dispense, decrements inventory, and triggers resupply when a site or depot runs low. The class descends from the telephone-keypad IVRS of the 1990s through web-based IWRS to today's platforms, and the name RTSM describes the two jobs plainly: randomisation, and trial supply management.
What this page does not claim
A system class is not a product. SPEQ describes what a CTMS or a LIMS is; the vendor directory at /tools lists the products that implement one, and a GAMP category is a property of an implementation, not of a class.
What a IRT / RTSM actually is
An IRT — increasingly called an RTSM — is the system that executes a trial's randomisation and runs its investigational product supply chain. When a site confirms an eligible subject, the system assigns the treatment arm according to the randomisation design, tells the site which blinded kit to dispense, decrements inventory, and triggers resupply when a site or depot runs low. The class descends from the telephone-keypad IVRS of the 1990s through web-based IWRS to today's platforms, and the name RTSM describes the two jobs plainly: randomisation, and trial supply management.
It is arguably the highest-consequence system in clinical operations, because its failures are unrecoverable in a way most system failures are not. A mis-executed randomisation cannot be re-run; a subject assigned outside the algorithm, a stratification factor applied wrongly, or a broken blind is permanent damage to the trial's statistical integrity — the property ICH E9 exists to protect. The randomisation list itself is generated outside the system by an independent statistician under the statistical plan; the IRT's job is to execute that list faithfully, and the loading and verification of the list is one of the most tightly controlled activities in study start-up.
The blinding architecture is the system's defining engineering problem. The IRT knows everything — who is on which arm, which kit contains active and which placebo — and must reveal each fraction of that knowledge only to the roles entitled to it: blinded site staff see kit numbers, unblinded pharmacists may see more, unblinded supply managers see treatment-level inventory, and the sponsor's blinded team must be provably unable to infer assignments even from patterns in resupply. Emergency unblinding is the controlled exception: an investigator who needs the assignment to treat the subject can obtain it at any hour, and the event — who, when, why — is permanently recorded and notified.
On the supply side, the IRT is the clinical face of a GMP operation. Kits it assigns were manufactured, labelled, and released under investigational-product GMP — in the EU, Commission Delegated Regulation (EU) 2017/1569 — and the system enforces the constraints that keep dispensing lawful: expiry dates, quarantine following temperature excursions in transit, lot status, and country-specific release. Every trial demands a study-specific configuration of randomisation design, kit types, resupply algorithms, and visit schedules, so the class carries a heavier per-study verification burden than almost any other: the platform is GAMP 5 Category 4, but each study's build is tested as if the trial depends on it, because it does.
WHERE THE BOUNDARY ACTUALLY SITS
Not the clinical database. The IRT records the randomisation transaction and kit assignment; the subject's clinical data lives in the EDC, which receives the randomisation number — never the treatment arm — by integration.
EDC owns it →Not the warehouse or ERP. Physical inventory, GMP batch status, and release of investigational product are owned by manufacturing and distribution systems; the IRT holds the trial's logical view of kits and drives demand against it.
ERP & Warehouse Management owns it →Not the generator of the randomisation list. An independent statistician produces the list under the statistical analysis plan; the IRT loads, protects, and executes it, and the segregation between the two is itself a control.
Not the safety system. Emergency unblinding is executed through the IRT, but the clinical event that prompted it is assessed by the investigator and processed through pharmacovigilance workflows.
Safety / PV Database owns it →WHAT IT HOLDS, AND WHAT CROSSES ITS BOUNDARY
CORE RECORDS
- The loaded randomisation and kit lists, with the verification records from list transfer and loading
- Randomisation transactions per subject — screening, stratification values, assignment, and timestamp
- Kit assignment and dispensing records, linking each subject to each kit across the trial
- Inventory movements: shipments requested, dispatched, received, and acknowledged at depots and sites
- Expiry, quarantine, and lot-status changes, including holds from temperature excursions in transit
- Emergency unblinding events — who unblinded, when, and the recorded reason
- Role assignments distinguishing blinded from unblinded users, with the access history behind them
DATA FLOWS OUT
Randomisation numbers and kit identifiers written to the clinical database — never the treatment assignment itself while the trial is blinded
Resupply demand and shipment requests driving physical picking, distribution, and depot inventory movements
Screening, randomisation, and enrolment counts by site, feeding oversight and projection
End-of-trial randomisation documentation, unblinding logs, and IP accountability records filed as essential records
HOW THIS CLASS IS USUALLY VALIDATED
- SPEQ synthesis: the platform is configured commercial software — GAMP 5 Second Edition (2022) Category 4 — but the class behaves unlike its neighbours because every trial requires a study-specific build of the randomisation design, kit structure, and resupply logic, and any custom algorithm is treated at Category 5 rigour. The category describes the implementation, not the product, and here the implementation is remade per protocol.
- Randomisation execution is verified against the design, not the demonstration: assignments exercised across strata and blocks, edge cases at block boundaries and stratum exhaustion, and confirmation that the executed sequence reproduces the independent statistician's list exactly.
- Blinding is tested as a negative requirement — the hardest kind: scripted evidence that no blinded role, screen, report, extract, or integration can reveal or allow inference of treatment assignment, including through patterns in kit numbering and resupply.
- Supply logic earns scenario testing because its failures strand subjects: resupply triggers under enrolment surges, expiry-driven replacement, quarantine on excursion, and country-release constraints all exercised before a real depot ships a real kit.
SPEQ synthesis, not a rating. This is SPEQ’s reading of how this system class is commonly approached, offered to help you scope your own work. A GAMP category is a property of a specific implementation, not of a product class, and one deployment routinely spans several. It is not a classification service and does not replace your own documented risk assessment.
- Stage 1 · Reactive
Randomisation runs on envelopes or spreadsheets, supply is pushed to sites on a fixed schedule regardless of enrolment, and stock-outs and expired kits at site are discovered by the subject standing in the pharmacy. Unblinding events are handled ad hoc and reconstructed after the fact.
- Stage 2 · Defined
A configured IRT executes the randomisation list and tracks site inventory, with defined user acceptance testing per study. Resupply parameters are set at start-up and rarely revisited, so overage is bought as insurance and mid-study design changes are slow and nervous.
- Stage 3 · Controlled
Each study build is verified against the randomisation design with documented edge-case evidence, blinded and unblinded roles are enforced and audited, and supply settings are reviewed against actual enrolment so waste and stock-out risk are managed rather than absorbed.
- Stage 4 · Predictive
Supply strategy is simulation-driven: forecast enrolment, expiry, and depot lead times shape resupply algorithms before first subject in, and live consumption re-tunes them during the trial. Unblinding patterns, dispensing errors, and shipment excursions are trended as quality signals.
- Stage 5 · Adaptive
Randomisation and supply are a portfolio capability: validated design patterns and build libraries compress study start-up, cross-trial supply pooling cuts overage without adding risk, and the organisation can absorb an adaptive design change with the same discipline as a routine amendment.
SPEQ’s shared five-stage progression, labelled synthesis. It is not the FDA QMM rating scale and not the scored maturity-assessment domains — assess your quality system for those.
WHAT AN INSPECTION PROBES, AND WHERE IT GOES WRONG
INSPECTION SIGNALS
- Whether every subject's assignment reproduces the randomisation list exactly, and how the list's loading and verification were controlled and segregated.
- The emergency unblinding record — who obtained assignments, when, for what recorded reason, and whether the trial's blinded roles could have inferred assignments any other way.
- Investigational product accountability end to end: kits shipped, received, dispensed, returned, and destroyed, reconciled without unexplained gaps.
- Whether expired or excursion-quarantined product was blocked from dispensing by the system, or merely by procedure.
- How mid-study changes — new strata, kit types, resupply parameters — were tested and released, and what regression evidence protected the running trial.
COMMON RISKS
- A configuration defect in stratification or blocking that mis-assigns subjects silently and is discovered only at the statistical review, when nothing can be repaired.
- Blinding leaks through the side channels — kit-number patterns, unblinded reports reaching blinded staff, or an integration exporting the assignment field to the EDC.
- Resupply parameters set defensively at start-up and never revisited, producing expiry write-offs at depots while individual sites still run dry.
- User acceptance testing that exercises the happy path per study but skips block boundaries, stratum exhaustion, and re-screening edge cases.
- Emergency unblinding that is too hard at 3 a.m. — an investigator who cannot reach the code break promptly is a subject-safety failure, not a blinding success.
WHO WORKS IN IT, AND WHERE IT IS SHAPED
ROLES
- Clinical supply chain manager (unblinded)
- Biostatistician / randomisation statistician
- Site pharmacist / unblinded site staff
- Investigator and blinded site staff
- IRT study configuration specialist
- CSV analyst
DELIVERY-LIFECYCLE PHASES
[ POSITION IN THE FRAMEWORK ]
6 OF 7 DIMENSIONS · 23 LINKSRandomises subjects, assigns blinded investigational product, manages depot-to-site supply, and controls emergency unblinding — clinical operations' highest-consequence system, because a mis-executed randomisation cannot be re-run.
06 · QUALITY MATURITY — IRT / RTSM, REACTIVE TO ADAPTIVE
Randomisation runs on envelopes or spreadsheets, supply is pushed to sites on a fixed schedule regardless of enrolment, and stock-outs and expired kits at site are discovered by the subject standing in the pharmacy. Unblinding events are handled ad hoc and reconstructed after the fact.
A configured IRT executes the randomisation list and tracks site inventory, with defined user acceptance testing per study. Resupply parameters are set at start-up and rarely revisited, so overage is bought as insurance and mid-study design changes are slow and nervous.
Each study build is verified against the randomisation design with documented edge-case evidence, blinded and unblinded roles are enforced and audited, and supply settings are reviewed against actual enrolment so waste and stock-out risk are managed rather than absorbed.
Supply strategy is simulation-driven: forecast enrolment, expiry, and depot lead times shape resupply algorithms before first subject in, and live consumption re-tunes them during the trial. Unblinding patterns, dispensing errors, and shipment excursions are trended as quality signals.
Randomisation and supply are a portfolio capability: validated design patterns and build libraries compress study start-up, cross-trial supply pooling cuts overage without adding risk, and the organisation can absorb an adaptive design change with the same discipline as a routine amendment.
SPEQ’s shared five-stage progression, labelled synthesis — not the FDA QMM rating scale. Where does your organization sit? Score your quality system →
07 · REGULATORY & EVIDENCE
GOVERNING STANDARDS · 7
Derived from the 7 standards SPEQ maps to this subject, across 5 regulatory bodies: FDA, ISPE, ICH, EMA, EC.
RECORDS & OBJECTIVE EVIDENCE
- The loaded randomisation and kit lists, with the verification records from list transfer and loading
- Randomisation transactions per subject — screening, stratification values, assignment, and timestamp
- Kit assignment and dispensing records, linking each subject to each kit across the trial
- Emergency unblinding events — who unblinded, when, and the recorded reason
- Role assignments distinguishing blinded from unblinded users, with the access history behind them
COMMON INSPECTION FINDINGS
- Subject assignments that do not reproduce the independent statistician's randomisation list
- Blinding leaks through kit-number patterns, unblinded reports, or an assignment field exported to the EDC
- Expired or excursion-quarantined product blocked only by procedure, not by the system
- Investigational-product accountability with unexplained gaps between shipped, dispensed, and returned
- Emergency unblinding too cumbersome to reach promptly — a subject-safety failure
Choosing, validating, and living with IRT / RTSM
FREQUENTLY ASKED
What is the difference between IVRS, IWRS, IRT, and RTSM?
Generations of the same system class. IVRS — interactive voice response system — was the telephone-keypad original: sites dialled in to randomise a subject. IWRS moved the interaction to the web. IRT, interactive response technology, became the umbrella term as platforms merged voice and web channels. RTSM — randomisation and trial supply management — is the functional name now displacing them all, and it is the most honest: it says what the system does rather than how you talk to it. In current usage IRT and RTSM are interchangeable; a protocol may say either and mean the same system.
Why does an IRT need study-specific testing if the platform is already validated?
Because the platform being sound proves nothing about this trial's build. Every study configures its own randomisation design — ratios, strata, block structure — its own kit types and blinding scheme, and its own resupply algorithms, and a defect in any of them damages the trial in ways that cannot be repaired afterwards: a mis-stratified assignment stays mis-stratified. Platform validation covers the engine; study-level user acceptance testing must prove the configuration executes the independent statistician's list exactly, keeps every blinded role blind, and moves product correctly through the edge cases — block boundaries, stratum exhaustion, re-screened subjects — where defects actually live.
How does emergency unblinding work in an IRT?
As a controlled, always-available exception. When an investigator judges that treating a subject requires knowing the assignment, the IRT provides it — typically to the investigator directly, at any hour, without requiring sponsor permission, because ICH GCP places that decision with the person responsible for the subject's care. The system then makes the exception expensive in the right way: the event is permanently recorded with who, when, and the stated reason; defined parties are notified; and the unblinded subject is flagged for the statistical handling the analysis plan prescribes. What the system must never do is make unblinding so cumbersome that a 3 a.m. emergency goes unmet.
Is an IRT a GMP system or a GCP system?
It sits deliberately across both. Its randomisation half is GCP territory: subject assignment, blinding, and the statistical integrity of the trial under ICH E6 and ICH E9. Its supply half is the clinical face of a GMP operation: the kits it assigns were manufactured and released under investigational-product GMP, and the system enforces expiry, quarantine, and lot status at the point of dispensing. In the EU, IMP manufacture sits under Commission Delegated Regulation (EU) 2017/1569 while the trial itself runs under Regulation (EU) 536/2014 — the IRT is where those two regimes meet a single subject at a pharmacy window.