The Trial Master File: eTMF Structure, the DIA Reference Model, and Inspection Readiness

The Trial Master File is the evidence that a clinical trial was run the way the protocol and the regulations said it would be. If a document is not in the TMF, an inspector treats it as if the activity never happened. That single rule explains most of what follows: why timeliness matters as much as completeness, why an empty placeholder is sometimes worse than a missing one, and why a clean filing structure with bad metadata still fails inspection.

This article covers how to structure an electronic TMF (eTMF), how the DIA TMF Reference Model maps the content, how to run quality control so the file is inspection-ready continuously rather than in a panic before an audit, and the finding patterns that recur across GCP inspections. The aim is that you can read this, then go build or run a TMF, answer interview questions on it, and hold your own in front of an inspector.

What the TMF is and why it is required

The TMF is the body of essential documents that, taken singly and as a whole, let an outside reviewer judge how a trial was actually run and how trustworthy its data are. That principle, that a defined set of records must be enough to reconstruct and evaluate the trial, is the operative expectation set out in ICH E6, the governing standard.

The regulatory basis is ICH E6(R2) Good Clinical Practice, Section 8 (“Essential Documents for the Conduct of a Clinical Trial”), and its successor ICH E6(R3), which restructures the essential-records expectations around a risk and quality-by-design philosophy. The R3 Principles and Annex 1 (interventional trials) reached Step 4 in January 2025; Annex 2 (additional considerations for nontraditional designs such as decentralised elements, pragmatic trials, and real-world data) reached Step 4 on 3 June 2026, completing the R3 set. Reaching ICH Step 4 is the guideline body’s own sign-off on the finished text, not automatic law in any single region, and the two annexes illustrate why that gap matters: in the EU, Annex 1 became applicable on 23 July 2025, while Annex 2 was endorsed by the CHMP on 25 June 2026 and only becomes applicable on 15 January 2027. Check your own region’s adoption date rather than assuming a freshly Step 4 text is already binding there. R3 keeps the same essential-records intent but speaks of records being “fit for purpose” and quality being built into design, which changes how you justify your TMF Index and your QC effort rather than the list of documents itself. ICH E6(R2) Section 8.1 sets the core obligation: essential documents are the records that, whether examined one at a time or as a full set, make it possible to reconstruct how the trial was conducted and to gauge the reliability of the data it produced. Their purpose is to show that the investigator, sponsor, and monitor met Good Clinical Practice and every applicable regulatory requirement.

Two layers of regulation sit underneath this. In the United States, 21 CFR Part 312 (IND regulations) and 21 CFR Part 11 (electronic records and signatures) govern, with Part 312.62 requiring the investigator to maintain records and 312.57/312.58 covering sponsor records and FDA access. In the European Union, the Clinical Trials Regulation (EU) No 536/2014 explicitly names the TMF in Article 57 and requires it to be readily available and directly accessible to inspectors, with Commission Implementing Regulation (EU) 2017/556 detailing inspection procedures and Commission Delegated Regulation (EU) 2017/1569 addressing investigational medicinal product manufacture. EU 536/2014 also sets the archiving floor: the TMF must be retained for at least 25 years after the end of the trial, longer than most other jurisdictions.

The risk rationale is simple. The TMF is the only durable proof that participants were properly consented, that the drug given matched the protocol, that safety events were reported on time, that the data feeding the marketing application is traceable to a source. When an inspector cannot reconstruct what happened from the TMF, they cannot trust the data, and untrustworthy data can sink a submission. A disorganized or incomplete TMF is, in inspection terms, a data integrity problem dressed as a filing problem.

The TMF has two halves that must reconcile:

• Sponsor TMF: held by the sponsor or CRO, covering trial-level and country-level oversight, vendor management, safety, and the documents the sponsor generates or receives.
• Investigator Site File (ISF), sometimes called the site TMF: held at each investigator site, covering site-specific conduct, the local copy of consent, delegation, training, and source-adjacent records.

These are mirror images for many document types. A signed delegation log lives in the ISF; the sponsor’s filed copy lives in the sponsor TMF. Reconciliation between the two is a recurring inspection theme.

eTMF versus paper, and what “electronic” actually demands

An eTMF is a validated computerized system used to capture, manage, store, and retrieve essential documents electronically. Moving from paper to electronic does not relax any GCP requirement; it adds the requirements of 21 CFR Part 11 and EU Annex 11 on top.

If you run an eTMF, you must be able to demonstrate:

• Validation of the system under a documented lifecycle. The category and rigor follow GAMP 5 risk-based principles. See GAMP 5 CSV framework and computer software assurance (FDA).
• Audit trails that are secure, computer-generated, time-stamped, and capture who did what and when to each record, including document status changes, metadata edits, and deletions. See audit trail design and review.
• Access controls with role-based permissions, unique user IDs, and no shared logins. See electronic signatures implementation.
• True copy integrity when paper originals are scanned in. A scanned document only replaces the paper if the copy process is verified to produce a complete and accurate reproduction with the same content and meaning. See static, dynamic records and true copies.
• System clock control so that filing dates and audit-trail timestamps are reliable. See time stamps and system clock control.

A common trap: organizations run a “hybrid” TMF where some documents are electronic and some remain paper, with no single index that tells you which is where. Hybrids are allowed, but you must control the boundary. See hybrid paper-electronic records.

The DIA TMF Reference Model: what it is and how to use it

The DIA TMF Reference Model is an industry-developed reference taxonomy. It originated within the Document and Records Management community of the Drug Information Association (DIA), an industry association, not a regulator, and since April 2022 it has been published and maintained by CDISC (Clinical Data Interchange Standards Consortium). The name “DIA TMF Reference Model” persists in common usage, but current stewardship sits with CDISC. It is voluntary. It exists because ICH E6 lists essential documents but does not prescribe a filing structure, and before the model every sponsor and CRO invented their own, which made TMF transfers and inspections painful.

The model gives you a standardized hierarchy so that anyone, including an inspector or a new CRO inheriting a study, can find a document in a predictable place.

Structure of the model

The reference model organizes documents into a three-level hierarchy:

Each artifact carries metadata attributes in the model: a unique artifact number, the artifact name, a definition, sub-artifacts where relevant, and flags for whether it is expected at trial / country / site level, whether it is “core” or “recommended”, and ICH E6 cross-references. The model also supplies an exchange mechanism (a defined file and metadata format) so a TMF can be transferred between systems without losing structure.

Versions matter. The model is periodically revised (major published versions have included 3.x releases). Always state which version you built your TMF Index against, because artifact numbering and definitions shift between versions. When you migrate or merge studies, version mismatch is a real source of mis-filing.

How to use the model in practice

Do not adopt the reference model verbatim and call it done. Use it as the starting taxonomy, then produce a study-specific TMF Index (sometimes called the TMF Plan content list) by tailoring:

1. Start from the current reference model as your master artifact list.
2. Remove artifacts that cannot apply to this trial. A trial with no central lab does not need the Central and Local Testing zone; document the rationale for exclusion rather than silently dropping it.
3. Add artifacts the model does not cover. Novel modalities and decentralized designs generate document types the model lags on, e.g. eConsent system validation evidence, wearable device data agreements, home-nursing vendor oversight. See eConsent and decentralized trials.
4. Set the expected count and level for each artifact. Is this expected once at trial level, once per country, or once per site? This is what later drives completeness metrics.
5. Map ownership. For each artifact, name the responsible role (sponsor, CRO function, site).

The output is the backbone of the TMF Plan, the controlled document that governs the whole file.

The TMF Plan: the governing document

Before documents are filed, write the TMF Plan (also called the TMF Management Plan). It is the SOP-level contract for how the TMF will be run for this trial, and inspectors ask for it early. See how to write an SOP.

A solid TMF Plan covers:

• Scope and ownership: which system(s), sponsor vs CRO vs site responsibilities, the named TMF owner.
• Structure: the reference model version used and the study-specific TMF Index, including documented exclusions.
• Expected document list with timeliness targets: when each artifact type must be filed relative to the triggering event.
• Filing and QC process: who indexes, who QCs, what the QC criteria are, and the sampling approach.
• Completeness, timeliness and quality metrics and how they are reported.
• Access and security: roles and permissions matrix.
• Hybrid handling: how paper and electronic are reconciled, and where the master copy lives.
• Transfer and reconciliation: how the file moves if a CRO changes, including the exchange format.
• Archiving and retention: format, location, retention period per region, and migration plan if the system is decommissioned.
• Roles and responsibilities (RACI).

Acceptance criterion for the plan itself: a reviewer who has never seen the study should be able to predict, for any document type, where it goes, who files it, by when, and how it is QC’d.

What goes in the TMF: the essential documents

ICH E6 Section 8 groups essential documents by trial phase: before the clinical phase begins, during the conduct of the trial, and after completion or termination. Here is a working set of the high-frequency, high-risk artifacts, with the zone they map to and the filing trigger.

Two principles cut across this list. First, every “received” document needs evidence of receipt and the date, because timeliness is judged against the trigger event. Second, blank/version-controlled documents (the consent template, the sample CRF) must always be filed with the correct, IRB-approved version, because filing a superseded ICF version is a classic and serious finding.

Filing workflow, step by step

The day-to-day job is converting a stream of incoming documents into correctly classified, QC-passed, timely TMF records. Run it as a defined process, not ad hoc.

1. Capture. The document arrives (signed PDF, scanned wet-ink page, system export). For paper, scan to a true-copy standard: full page, legible, all annotations and signatures captured, color where color carries meaning. Verify the scan against the original before destroying paper, and record the true-copy verification.
2. Classify / index. Assign the artifact type from the study TMF Index, plus metadata: site number, country, document date, version, language, and any sub-artifact. Correct classification is where most downstream errors originate.
3. Set the document date correctly. The document date is the date the document was effective or signed, not the date you scanned it. Audit-trail timestamps record the filing event separately. Confusing the two corrupts timeliness reporting.
4. QC. A second person (segregation of duties) checks legibility, correct classification, correct version, complete signatures and dates, and that the document is the right one for that placeholder. QC pass or reject is recorded; rejects route back with a reason.
5. Approve / finalize. The record moves to a “final” or “approved” status. Status itself is metadata an inspector can see.
6. Reconcile against the expected list. Mark the placeholder satisfied. If the artifact is expected per site, the system tracks one slot per site so a gap is visible.
7. Handle exceptions. Where a document genuinely does not exist or is delayed, file a documented placeholder or a “missing document” note explaining why and when it is expected. Never leave a silent empty slot.

The same workflow drawn as a decision path, including the two places a document can leave the happy path:

Acceptance criteria for a single filed record: correct artifact classification; correct and current version; legible and complete (all pages, signatures, dates present); accurate document date; QC passed by a second person; filed within the timeliness target; metadata fields complete; audit trail intact.

Completeness, timeliness, and quality: the three TMF metrics

Inspectors and sponsors judge a TMF on three dimensions. Run all three as live metrics, not a one-time pre-inspection scramble.

Completeness

Completeness = filed expected documents / total expected documents, per the study TMF Index. The denominator is the tailored expected list, which is why getting the TMF Plan right matters: an unrealistic or padded expected list distorts the percentage.

A worked example for a single site at a given point in time:

Ninety percent looks fine until you ask which two are missing. A missing monitoring visit report and a missing IRB amendment approval are not equal in risk. This is why completeness alone is not enough; weight by criticality.

Timeliness

Timeliness = documents filed within target / documents filed, where the target is days from the trigger event to the file date. Typical internal targets are tiered by document risk, for example:

Set the targets in your TMF Plan; do not invent a single number here, because organizations differ. The key inspection concept is the contemporaneous standard from ALCOA+: records made at the time the activity occurred. A monitoring report written four months after the visit is a timeliness failure and an ALCOA+ (“contemporaneous”) failure at once. See ALCOA+ in detail and good documentation practices.

Quality

Quality is whether the filed documents are actually correct: right version, legible, complete signatures, correct classification, no superseded versions masquerading as current. A TMF can be 100% complete and timely and still fail quality if half the documents are misfiled or the consent version is wrong. Quality is measured through QC sampling and error-rate tracking.

A practical combined dashboard view that holds up in front of an inspector:

TMF health benchmarks across the trial lifecycle

A single snapshot dashboard, like the one above, tells you where the file stands today. It does not tell you whether today’s numbers are normal for where the trial is. A completeness score that would be a serious problem two months after first patient in is unremarkable during start-up, when most artifacts have not been triggered yet, and a completeness score that looked fine mid-study is a real gap if the trial is heading into database lock. Judge the metrics against the phase, not against one fixed target applied all the way through.

The pattern across all four phases is the same: completeness and timeliness targets tighten as the trial matures, and an organization that only ever reports one trial-level number never sees whether it is tracking that curve or drifting off it.

QC processes: how to keep the file inspection-ready

Continuous QC is the difference between a TMF you can show with a day’s notice and one that needs a six-week remediation before anyone outside the team can look at it.

Risk-based QC sampling

You cannot 100% QC a large TMF forever, so sample by risk:

• 100% QC for the highest-risk artifacts: informed consent versions, IRB/IEC approvals, regulatory approvals, delegation logs, safety reports.
• Statistically meaningful sampling for medium-risk routine documents, with a defined sample size and acceptance limit (for example a defined number of documents per site per quarter with an error threshold that triggers escalation and a wider review).
• Trigger a 100% review of an artifact type or a site when sampling shows an error rate above your acceptance limit, which is the same logic as any quality sampling plan. See statistics in quality: Cpk and control charts.

A worked sampling decision makes this concrete. Suppose a site has 240 medium-risk filed documents this quarter and your TMF Plan sets a sample of 32 with an acceptance limit of 1 error (accept on 0 or 1 finding, escalate to 100% review of that artifact type at 2 or more). The QC reviewer pulls 32, finds 2 misclassifications in monitoring visit reports, so the artifact type fails the sample. The action is not “fix those two”: it is a 100% review of monitoring visit reports at that site, a root-cause look at why classification drifted (often a coordinator using the wrong artifact code), and a check of whether the same pattern shows at other sites. Record the sample size, the acceptance limit, the result, and the triggered action so the sampling decision is defensible rather than ad hoc.

Periodic TMF review (the formal health check)

Run scheduled TMF reviews at milestones: study start-up complete, first patient in, ongoing at a defined cadence (often quarterly), database lock, and pre-inspection. Each review checks completeness against the expected list, timeliness trends, open placeholders, and quality sampling, and produces a documented report with CAPA-style actions and owners. A TMF review with no follow-through is theater; the actions and their closure are what an inspector reads. See CAPA effectiveness verification.

Reconciliation

Reconcile the sponsor TMF against each Investigator Site File at monitoring visits and at close-out. The monitor confirms the ISF holds what it should and that sponsor-side copies match. Document the reconciliation; “I checked it” with no record is, by the TMF’s own rule, the same as not having checked it.

TMF transfer at a CRO change: the reconciliation runbook

Changing CRO or bringing a study in-house mid-trial is where TMFs quietly break. Documents are lost or duplicated, reference-model versions do not line up, metadata and audit trail drop on export, and the reconciliation is never formally closed. Run the transfer as a controlled, signed-off project, not an IT file copy.

1. Write a transfer plan and freeze the source. Name the sending and receiving owners, the cut-off date, and the exact scope (which studies, which sites, sponsor TMF and ISF copies). Put the source eTMF into a read-only or controlled state at cut-off so nothing changes mid-transfer.
2. Agree the index mapping first. Both sides confirm the reference-model version each system uses and produce an artifact-to-artifact mapping. Version mismatch (for example a 3.1 index into a 3.2 index) is the single biggest source of mis-filing, so resolve every unmapped or renumbered artifact before any document moves.
3. Export with metadata and audit trail intact. Use the model’s exchange format where both systems support it. The export must carry the document, its metadata (artifact type, document date, version, site, status), and the audit trail. A bulk PDF dump that loses metadata and history is not an acceptable transfer of a GxP record.
4. Verify counts and a content sample on import. Reconcile expected counts per artifact and per site against the source. Then QC a risk-based sample of imported records for correct classification, correct document date, intact audit trail, and readability. High-risk artifacts (consent versions, IRB approvals, safety reports) get 100% verification, not sampling.
5. Reconcile against the expected list, not just against the old system. Confirm the receiving TMF’s expected-document configuration matches the study TMF Index so completeness metrics restart correctly. A transfer that reconciles system-to-system but breaks the expected-list denominator hides gaps.
6. Resolve and close every exception. Track missing, duplicated, or mis-mapped documents to closure. Duplicates are deduplicated with a documented rule (keep the QC-passed final, retire the draft); gaps are chased from the source or rationalised.
7. Sign off the transfer. A documented, dual-signed reconciliation statement (sending and receiving owners, plus QA) records that the scope transferred completely, the sample passed, and exceptions are closed. Until that sign-off exists, the receiving organisation cannot claim the TMF is complete and an inspector will treat the transfer as open.

Acceptance criteria for a completed transfer: artifact and site counts reconcile to the source; high-risk artifacts are 100% verified; metadata and audit trails are intact and reviewable; the expected-document list is reconfigured to the study TMF Index; every exception is closed; and a dual-signed reconciliation statement exists. The receiving organisation owns the TMF the moment it accepts the transfer, including any inherited gaps, which is exactly why the sample and the sign-off matter. See quality in technology transfer for the analogous discipline on the manufacturing side.

Roles and responsibilities

The non-negotiable principle: the people who QC must be independent of the people who file. Segregation of duties is both a data integrity control and an inspection expectation.

Archiving and retention

When the trial ends, the TMF moves from active management to archive, but it must stay complete, legible, and accessible for the full retention period.

Retention floors differ by region, and you keep the longest applicable:

• EU CTR (EU) No 536/2014: at least 25 years after the end of the trial for the sponsor and investigator TMF. The archiving obligation sits in Article 58 (archiving of the clinical trial master file), distinct from Article 57, which governs the TMF’s contents and accessibility to inspectors.
• UK Clinical Trials Regulations: amendments to the Medicines for Human Use (Clinical Trials) Regulations 2004 took full effect on 28 April 2026 and raised the TMF retention floor from 5 years to at least 25 years after the trial concludes (Regulation 31A of the amended 2004 Regulations), bringing the UK in line with the EU CTR floor. The 25-year TMF requirement applies to trials whose application was submitted on or after 28 April 2026; trials approved before that date keep the prior 5-year TMF retention under the regulation as it stood immediately before the amendment, though participant medical records connected to those trials now fall under the new 25-year floor regardless of the application date. Where the trial’s data still support a pending UK marketing authorisation application at the point the 25-year clock would otherwise expire, Regulation 31A(7A) extends retention to at least 2 years beginning the day after that authorisation is granted. Check the current transitional guidance on GOV.UK for the trial in front of you, since which regime applies turns on the submission date, not the trial’s start or end date.
• US 21 CFR 312.62: records retained for 2 years after a marketing application is approved for the indication, or 2 years after the IND is discontinued and FDA notified, whichever is later. Longer retention is common in practice.
• ICH E6 expects retention for at least 2 years after the last approval or discontinuation, and per applicable regulatory requirements, which in the EU and UK means the 25-year floor.

Archiving requirements that trip teams up:

• Readability over time. A PDF/A or equivalent durable format, with a plan to migrate if formats or systems become obsolete. A 25-year obligation outlives most software. See data migration validation and backup, restore, and disaster recovery validation.
• Audit-trail preservation. The audit trail is part of the record and must be archived and remain reviewable. Archiving the documents but losing the metadata and audit trail is a data integrity gap.
• Access control through retirement. Even archived, the TMF must be retrievable on request and protected from alteration. Decommissioning an eTMF system without a validated migration is a known failure mode.

Common mistakes and recurring inspection findings

These patterns recur across GCP inspections and audits. None of them require naming any company; they are generic.

• Documents missing or filed late. The single most common TMF finding. Late filing breaks the contemporaneous principle and signals that oversight was not real-time.
• Wrong or superseded ICF version filed. Filing an old consent version, or being unable to show which version each participant signed, is a high-severity finding because it goes to participant protection.
• Empty placeholders with no explanation. A blank slot with no note explaining absence reads as a gap. An inspector cannot tell “not applicable” from “lost.”
• Misclassification. Documents filed under the wrong artifact, so completeness looks fine but the document cannot be found where it belongs. Inspectors test this by asking for a specific document and watching how long retrieval takes.
• Sponsor TMF and ISF do not reconcile. Versions differ, signatures present in one and not the other, monitoring reports filed centrally but not at site.
• Audit trail not reviewed or not available. For eTMF, failing to demonstrate audit-trail review, or an audit trail that does not capture metadata changes, is a Part 11 / Annex 11 finding. See operationalizing audit trail review.
• True copy process not verified. Scanned documents accepted as originals without a documented verification that the copy is complete and accurate.
• No documented TMF Plan or expected-document list. Without an expected list there is no defensible completeness denominator, so the sponsor cannot prove the file is complete.
• eTMF not validated, or shared logins / weak access control. Treating the eTMF as “just document storage” rather than a GxP computerized system.
• CRO transfer gaps. When a study changes CRO, documents are lost, duplicated, or mis-mapped across reference-model versions, and reconciliation is never closed.
• Metrics that are not acted on. Reporting 88% completeness for months with no remediation is worse than not measuring, because it shows the organization knew and did nothing.

The meta-finding behind most of these: the TMF was treated as a clerical archive to be assembled before an inspection, instead of a continuously maintained quality record. The fix is process and metrics, not a heroic clean-up sprint.

How GCP inspectors classify TMF findings: critical, major, and minor

Not every TMF gap carries the same weight in an inspection report, and it helps to know how inspectors actually grade what they find.

European and UK GCP inspectorates, national competent authorities coordinated through EMA, and the MHRA in the UK, score each individual finding into one of three tiers based on its effect on subject rights, safety, and wellbeing, and on the reliability of the trial data:

• Critical: a significant, unjustified departure with evidence that subjects’ rights, safety, or wellbeing have been or could be jeopardized, or that the trial data are unreliable, or a pattern of major findings across areas indicating a systemic quality failure, or inadequate or untimely correction of a previously reported major finding. Considered totally unacceptable; can lead to data rejection or referral for further action.
• Major: a significant, unjustified departure that has not yet become critical but could if left unaddressed, or several departures within a single area of responsibility indicating a systemic issue in that area. A serious, direct violation of GCP principles.
• Minor: a departure from requirements or GCP guidance that does not meet the critical or major threshold on its own. A number of minor findings in the same area can add up to the equivalent of a major finding, so a long list of small TMF gaps is not automatically low risk.

FDA’s Bioresearch Monitoring program works differently. It does not label individual findings critical, major, or minor. Each inspection instead resolves to one overall classification, No Action Indicated (NAI), Voluntary Action Indicated (VAI), or Official Action Indicated (OAI), and individual objectionable conditions are listed on Form FDA 483 without a per-item severity tag. A pattern of TMF-related 483 observations, particularly anything touching consent validity, safety reporting, or data reliability, is what moves an inspection from VAI toward OAI. There is no formally labeled “critical TMF finding” on an FDA form, but the underlying judgment about what pushes an inspection toward OAI tracks the same substance as a European critical finding.

Mapping the recurring findings from the section above onto that European three-tier scheme, and onto the FDA outcome each tends to push toward, is a useful way to prioritize remediation:

This table shows tendency, not a rule. The grade an inspector actually assigns depends on the specific evidence in front of them, how many subjects or sites it touched, and whether the organization had already been told and failed to fix it. Use it to prioritize remediation, not to argue with an inspector about which box a finding belongs in.

Inspection readiness for the TMF

A TMF that is run well day to day is most of the readiness work. The rest is being able to demonstrate it. See GCP audits and inspections and FDA inspection readiness.

Before an inspection:

1. Run a final completeness, timeliness, and quality review and document it, with any open items tracked and risk-assessed.
2. Resolve or rationalize every placeholder. Each gap should have a documented reason and, where relevant, a remediation date.
3. Confirm retrieval speed. Inspectors judge a TMF partly on how fast you can produce a named document. Practice retrieval against a list.
4. Verify ISF/sponsor reconciliation for the sites in scope.
5. Prepare the audit-trail review evidence for the eTMF, and confirm access for the inspector if they will view the system directly. EU 536/2014 expects the TMF to be directly accessible to inspectors.
6. Brief the people who will answer TMF questions so they can explain the structure, the metrics, and any gaps honestly.

During the inspection, the rules are the same as any GCP inspection: produce only what is requested, answer the question asked, do not speculate, and never alter or back-date a document to close a gap mid-inspection. Back-dating or fabricating a TMF document during an inspection converts a filing finding into a data integrity and potential fraud finding. See managing a live inspection.

TMF readiness for a BLA or MAA submission

Everything above applies to any GCP inspection. A submission-triggered inspection is narrower and better defined than a routine one, and it deserves its own readiness pass.

For an NDA or BLA, FDA requires the sponsor to submit a standardized electronic dataset identifying every clinical investigator site, any entity that received transferred clinical trial obligations, and the location of clinical study documentation, for all major, pivotal studies supporting the safety and efficacy claims. CDER’s Bioresearch Monitoring program uses that dataset to select which sites, and where relevant which sponsor or CRO functions, to inspect as part of the application’s review. In the EU, GCP inspections can similarly be requested by the CHMP as part of a Marketing Authorisation Application assessment, scoped to the trials supporting that application. Either way, the inspection scope is not “the trial program in general,” it is the exact studies and sites named in the submission.

That narrower, submission-linked scope changes what “TMF readiness” means before a BLA or MAA:

1. Scope the review to the submission, not the whole program. Identify every trial cited as supporting evidence, including trials that finished enrolling years earlier and may have been deprioritized once their own database was locked, and confirm each has a complete, reconciled TMF.
2. Prioritize the sites most likely to be inspected. The highest-enrolling sites, and any site tied to a protocol deviation or a data-quality signal in the submission, are more likely to be selected. Reconcile their Investigator Site Files first, not last.
3. Run a scoped TMF inspection-readiness review before the pre-BLA or pre-MAA meeting, not after an inspection is announced. Once FDA or a national inspectorate schedules a submission-linked inspection, there is rarely enough runway left to close a major gap; the review has to finish while the application package itself is still being assembled.
4. Check that consent and safety-reporting evidence line up with the analysis population. For every subject counted in the primary analysis, the TMF and ISF should support the correct consent version and a complete safety-reporting trail. This is the artifact set an inspector pulls first when the inspection exists because of the submission.
5. Confirm source-data traceability evidence for the pivotal endpoints, not just document presence. A submission-linked inspection tests whether the data in the application can be traced back to source, and the TMF’s monitoring and SDV records are part of that evidence chain. See source data and SDV in clinical trials.

Retention obligations tighten around a submission too. Under US 21 CFR 312.62, the sponsor’s retention clock is explicitly tied to marketing application approval, 2 years after approval or after IND discontinuation, whichever is later. Under the UK’s amended Clinical Trials Regulations, Regulation 31A(7A) extends TMF retention a further 2 years from the date a UK marketing authorisation is granted, where the trial’s data still support that pending application when the general retention period would otherwise expire. Submitting the application does not shorten what you have to keep; in both regimes it can lengthen it.

Worked example: tailoring the reference model for a decentralized trial

Suppose you are setting up a Phase 2 decentralized trial with eConsent, home nursing, and a wearable device. Working from the reference model:

1. Keep the standard zones: Trial Management, Central Trial Documents, Regulatory, IRB/IEC Approvals, Site Management, IP and Trial Supplies, Safety, Data Management, Statistics.
2. Add study-specific artifacts the base model handles thinly:
   • eConsent system validation summary and Part 11 assessment (file under Data Management / a computerized-systems section).
   • Home-nursing vendor qualification, contract, and training records (Third Parties / vendor oversight).
   • Wearable device data transfer agreement and data flow diagram (Data Management).
   • Remote monitoring plan and remote source data review records (Trial Management). See risk-based monitoring in clinical trials and source data and SDV in clinical trials.
3. Set expected level and count for each: eConsent validation is trial-level once; home-nursing qualification is per vendor; wearable agreement is trial-level.
4. Assign owners and timeliness targets, then load the tailored list as the eTMF’s expected-document configuration so completeness tracks against reality.

The lesson: the reference model is the map, not the territory. New trial designs always need tailoring, and the tailoring decisions are themselves documented in the TMF Plan.

Interview-ready questions and how to answer them

“What is the TMF and why does it matter?” The collection of essential documents that, per ICH E6 Section 8, permit evaluation of trial conduct and data quality. It matters because it is the evidence the trial was run compliantly; if it is not in the TMF, regulators treat it as not done, and an untrustworthy TMF undermines the data feeding the submission.

“What is the DIA TMF Reference Model and is it a regulation?” It is a voluntary, industry-developed taxonomy, not a regulation. It started within the DIA and has been maintained by CDISC since April 2022, though the “DIA TMF Reference Model” name has stuck. It standardizes the structure into zones, sections, and artifacts so documents are filed predictably and TMFs can be transferred and inspected consistently. You tailor it per study; you do not adopt it blindly.

“How do you measure whether a TMF is in good shape?” Three dimensions: completeness (filed vs expected), timeliness (filed within target days of the trigger event, tied to the contemporaneous principle), and quality (right version, legible, correctly classified, complete signatures). You run all three as live metrics with risk-based QC and periodic reviews, not a pre-inspection scramble.

“What is the difference between completeness and quality, and which is worse to fail?” Completeness is whether expected documents are present; quality is whether the present documents are correct. A 100% complete TMF can still fail if versions are wrong or documents are misfiled. Quality failures are often worse because they hide: the slot looks satisfied but the content is wrong, for example a superseded consent version.

“An inspector asks for a specific monitoring report and it is not where it should be. What does that tell them?” That classification or filing discipline is weak, and they will probe wider. Retrieval speed and accuracy are how inspectors test a TMF in real time. The right answer is to find it through metadata and explain the indexing, not to improvise.

“What extra requirements does an eTMF add over a paper TMF?” All the GCP requirements stay, plus 21 CFR Part 11 and EU Annex 11: system validation, secure time-stamped audit trails, role-based access with unique logins, verified true copies for scanned originals, and controlled system clocks. The eTMF is a validated GxP computerized system, not just storage.

“How long must the TMF be retained?” The longest applicable requirement. Under EU CTR 536/2014 it is at least 25 years after the trial ends. As of 28 April 2026, the UK moved to the same 25-year floor for trials submitted on or after that date, up from a prior 5-year UK requirement. US 21 CFR 312.62 ties retention to 2 years after marketing approval or IND discontinuation. You keep the longest and ensure readability, audit-trail preservation, and access for the full period.

“You inherit a study that is changing CRO mid-trial. What are your TMF risks?” Document loss or duplication during transfer, reference-model version mismatch causing mis-mapping, broken metadata and audit trail, and unreconciled sponsor/ISF copies. You manage it with a controlled transfer using the exchange format, a reconciliation against the expected list, and a documented sign-off that the transfer is complete.

“What is a placeholder and why do empty ones fail inspection?” A placeholder marks where an expected document goes. An empty placeholder with no note cannot be distinguished from a lost document. You either file the document, file a documented true reason it does not apply, or file a note stating why it is delayed and when it is expected.

“How does an inspector decide whether a TMF finding is critical, major, or minor?” European and UK inspectors score each finding against its effect on subject rights, safety, wellbeing, and data reliability: critical is totally unacceptable and often ties to consent validity or unreliable data; major is a serious departure or a systemic pattern within one area; minor is a departure that does not reach that bar on its own, though enough of them together add up to a major. FDA does not use those three labels on individual findings; it classifies the whole inspection NAI, VAI, or OAI and lists objectionable conditions on Form 483 without a severity tag, though the same substance, especially anything touching consent or data reliability, is what drives an inspection toward OAI.

“What is different about a TMF inspection tied to a BLA or MAA versus a routine GCP inspection?” Scope and timing. A submission-linked inspection targets the specific pivotal trials, and often the highest-enrolling or flagged sites, named in the data the sponsor submitted for that purpose, rather than the trial program broadly. Readiness has to happen before the pre-BLA or pre-MAA meeting, because there is little runway left to fix a major gap once the inspection is scheduled.

Practical tips

• Tie every timeliness clock to the trigger event, not the upload date, and store both dates so the metric is honest.
• Weight completeness by criticality. A dashboard that treats a missing administrative letter and a missing IRB approval as equal will mislead you and your inspector.
• Configure the eTMF’s expected-document list to the tailored study index at start-up; an out-of-the-box list inflates the denominator and hides real gaps.
• Run reconciliation at every monitoring visit, not just at close-out, so ISF/sponsor drift is caught early.
• Keep a short, current TMF narrative for inspection: structure, metrics, known gaps, and remediation status. Honesty about gaps beats a surprise.
• Never fix a gap by back-dating. Use a current-dated note explaining the historical situation.
• State the reference model version on the index and re-check it at any system migration or CRO change.

Related reading

• ICH E6 Good Clinical Practice
• Clinical QA and GCP data integrity
• GCP audits and inspections
• Source data and SDV in clinical trials
• eConsent and decentralized trials
• Document control fundamentals
• 21 CFR Part 11 and EU Annex 11
• ALCOA+ in detail
• FDA inspection readiness
• Risk-based monitoring in clinical trials