Measure eTMF completeness against the study's expected evidence

A large document count can conceal a small, important gap

An electronic trial master file (eTMF) contains thousands of documents and reports steady upload activity. The study team still struggles to answer whether the expected evidence for one site milestone is present, correct and usable.

eTMF completeness should describe the state of expected records, not the volume of files stored. A duplicate upload increases document count without closing a missing-record requirement. An incorrect version can make a folder look populated while leaving the underlying expectation unresolved.

ICH E6(R3), Appendix C, describes management of essential records, including identification, version history and retrieval. The study's applicable requirements and filing plan determine the expected evidence in practice. ICH E6(R3), Appendix C.

Define the expected record before measuring it

Create a filing plan that reflects the study, countries, sites, responsibilities and milestones. Identify which records are expected at each point and who is accountable for providing and reviewing them. A generic taxonomy helps organisation but does not establish applicability by itself.

Each expectation needs a stable identity and enough context to evaluate fulfilment. Useful fields include study, site where relevant, record type, required version or period, responsible party, due basis and acceptance state. Some records may be held in another authorised repository; record their location and retrieval arrangements rather than demanding unnecessary duplicate copies.

Document the basis for an expectation being not applicable. A reviewer should be able to distinguish a justified exclusion from a missing file that someone removed from the denominator to improve the score.

Use several states instead of one completion flag

Separate receipt from review. A file may be uploaded but unreadable, misclassified, unsigned where a signature is required, or linked to the wrong site. These states need different actions and should not all appear as complete.

Define these states in the filing procedure and system configuration. The labels should match how the study operates, and reports should disclose which states count towards each metric.

Worked example: 98 uploads do not establish 98% completeness

Consider an illustrative site milestone with 100 applicable record expectations. The repository contains 98 uploaded files, but ten are duplicates of records already represented. Only 88 distinct expectations have a file associated with them.

Of those 88, six remain under quality review and four have unresolved issues. On the stated assumptions, 78 expectations have accepted evidence. Receipt coverage is therefore 88%, while accepted-record completeness is 78%. Neither number should be presented without its definition.

Now consider a simpler display that shows 98 files and highlights only two visibly empty folders. It can hide the duplicate and review-state problems. A useful dashboard lets the reviewer inspect the missing expectations, their milestone relevance and the accountable owner.

Suppose two of the unresolved expectations are prerequisites for a forthcoming site decision. The operations team should see them prominently even if the portfolio-wide percentage is high. The authorised study roles assess the consequence; the eTMF percentage does not grant permission to proceed.

These numbers are illustrative reporting logic. They are not a regulatory threshold or an acceptable level of missing evidence.

Add timeliness without rewriting history

An accepted record can still have arrived late relative to the study's approved expectation. Preserve the due basis, receipt time and review completion time separately. Uploading a file today should not make an earlier milestone appear fully evidenced at the time it occurred.

Define how changes to due dates are authorised. Retain the original date and reason for revision where relevant. Otherwise, repeatedly moving targets can create an apparently healthy dashboard while concealing unresolved work.

Review delays by cause: late generation, delayed transfer, incorrect classification, review backlog or unresolved content. Those categories point to different improvements and are more useful than treating every overdue item as a filing-team failure.

Connect discrepancies to accountable resolution

Give each issue an affected expectation, evidence reference, description, owner and resolution state. The reviewer should be able to request a correction without deleting the original submission or losing the reason it was inadequate.

Where a replacement file is supplied, preserve the version relationship. Avoid an unrestricted “replace attachment” action that makes the earlier evidence disappear from the review history. Corrections should remain understandable to a person who was not involved in the exchange.

The CTMS evaluation guide explains how these states affect operational milestones. An interface should transmit the agreed meaning, not convert every upload event into readiness.

Protect access across filing and review

Define permissions by study responsibility and information sensitivity. A person who needs to resolve a missing administrative record may not need access to unrelated participant information or restricted study documents. Review search, previews, exports and notifications as part of the access model.

The clinical data privacy guide helps map the necessary fields and purposes. Apply those decisions to connected reporting tools as well as the repository itself; an unrestricted completeness export can reveal more context than the dashboard screen.

Plan access changes when staff or service providers leave. Retain attribution and required records while removing current permissions. Historical evidence should not depend on an active account belonging to its original uploader.

Demonstrate retrieval with an independent reviewer

Ask someone unfamiliar with the filing activity to retrieve the evidence for one milestone. They should be able to identify the expected records, locate accepted versions, understand open issues and explain where any externally held evidence resides.

The audit-evidence readiness guide offers a broader retrieval approach. For the eTMF, test whether context survives export and archive: document identity, version, relationships and review history may matter as much as readable file content.

Include a closed study in the test. A system that works during active operations may become difficult to use after permissions change, integrations stop or a supplier contract ends. Verify the actual retained access method rather than relying on a promise that files can be exported later.

Improve the denominator through controlled change

A study amendment, new site or changed responsibility can alter expected records. Assess and approve the corresponding filing-plan changes, then identify affected expectations. Keep an explanation of what changed and why, especially when a trend report spans several plan versions.

Start a pilot with one study and a defined milestone. Compare the expected-record list with the actual operating process, resolve ambiguous ownership and test duplicate detection. Expand once the team can explain the denominator and act on the gaps.

Measure accepted completeness, timeliness and unresolved issues separately. Use the measures to direct work and investigate recurring causes, not to reward teams for uploading more files or excluding difficult expectations.

Summary and next actions

eTMF completeness becomes meaningful when the expected evidence, review state and time basis are explicit. A useful workflow helps study teams resolve the important gaps and retrieve the record of what happened.

• Define the applicable record expectations for one study milestone with accountable owners.
• Separate received, accepted, overdue and unresolved states in the reporting model.
• Test duplicate files, wrong versions and changed expectations without erasing history.
• Run an independent retrieval exercise for an active and a closed study, then address the observed gaps.