ICH E6(R3) is direct on this point: sponsors may delegate trial-related duties and functions to a CRO, but accountability for those duties stays with the sponsor regardless. That principle isn't new in theory. What's changed is where it lands in practice squarely on how sponsors actually oversee Trial Master File records once most of the day-to-day filing happens inside a CRO's systems, not the sponsor's own.
Most sponsors currently oversee that CRO-held TMF through a familiar set of tools: completeness dashboards from the vendor's system, periodic spot-checks, spreadsheet exports of artifact-level listings, sample-based QC exercises timed around key milestones. Under E6(R3)'s Appendix C and Section 4.2.3, that toolkit leaves four specific gaps open and each one is exactly the kind of gap an EMA inspector operating under the 23 July 2025 implementation date is positioned to test for.
A sponsor relying on vendor reports often can't answer, for a given trial, which essential records sit in the CRO's eTMF versus the sponsor's own, versus CTIS, versus a quality or safety system and what access exists to each. Without that consolidated view, "we oversee our CRO" is a claim without a map behind it.
Even when a CRO's completeness percentage looks healthy, connecting a specific filed document back to the CTMS milestone, deviation, or risk signal it's supposed to support usually takes manual cross-referencing. That gap makes it hard to answer a very specific and very reasonable inspector question: does what was filed actually match what happened in the trial?
Section 4.2.3 asks for review of trial-specific data and metadata, including audit trails, to be planned, risk-based, and documented not reconstructed from an export pulled together after the fact.
A sponsor that can only point to a vendor's own periodic QC exercise, without its own documented review process layered on top, doesn't have direct evidence that this requirement was met from the sponsor's side.
When a CRO's reported TMF completeness doesn't match what a sponsor's own CTMS or risk data suggests should be true, spreadsheet-based oversight has no built-in way to triage that discrepancy someone has to notice it, investigate manually, and resolve it outside of any documented process. That's a gap that tends to surface at the worst possible time: right before, or during, an inspection.
None of these four gaps mean a CRO is doing its job poorly. A CRO can run an excellent TMF operation and a sponsor can still fail to demonstrate oversight of it, because oversight isn't the same thing as trusting that the work is being done well it's being able to show, from the sponsor's own systems, that essential records are understood, current, and connected to trial conduct. Appendix C and Section 4.2.3, read together, are effectively asking sponsors to prove that delegation hasn't quietly become abdication.
| Vendor Dashboard Alone | Periodic Manual Spot-Checks | Sponsor-Side CTMS↔eTMF With an Embedded Agent | |
|---|---|---|---|
| Consolidated view across CRO, sponsor, CTIS, quality, safety systems | No single-vendor view only | Partial, rebuilt manually each time | Yes governed pointers link across systems |
| Links filed documents back to CTMS milestones and risk signals | Rarely, without manual work | Possible, but labor-intensive | Automatic, based on structured metadata |
| Evidence of planned, documented audit-trail review | Only the vendor's own process | Depends on how rigorously it's logged | Built into an ongoing, documented review stream |
| Resolving vendor-vs-sponsor data mismatches | No defined process | Ad hoc, when someone notices | Flagged and routed to an owner as it's detected |
| What an inspector actually sees | A completeness percentage | A reconstructed explanation | A navigable, evidenced oversight record |
Cloudbyz runs CTMS and eTMF on a shared Salesforce data, security, and audit foundation, giving sponsors a natural place to centralize oversight of CRO-held records without pulling every file on-premise.
CTMS holds the authoritative view of studies, countries, sites, milestones, monitoring plans, deviations, and risk signals. eTMF manages the sponsor-held portion of the TMF, structured against the TMF Reference Model. Integration with CRO eTMFs and other provider repositories runs through governed connections exchanging metadata, status, and pointers rather than raw file transfers.
Inside that structure, the AI eTMF Agent auto-classifies incoming sponsor-side documents, ingests and normalizes metadata feeds from CRO repositories, and reconciles what CTMS records as having happened against what the CRO's system reports as filed. When a provider feed clearly maps a document to a CTMS milestone, country, or site, the agent can associate and tag the corresponding sponsor-side reference directly.
When it detects the kind of pattern Section 4.2.3 is concerned with frequent late version changes, inconsistent country assignments, missing audit-trail evidence on high-risk documents it escalates that to the right Regulatory, Clinical Operations, or Quality owner, with the decision and action logged as auditable.
How much of the four-gap exposure this actually closes depends on how many CRO and provider systems are genuinely connected into the sponsor-side model but the shift from a vendor's completeness percentage to a sponsor-owned, evidenced record of oversight is what Section 4.2.3 is asking for either way.