What the Trial Master File Is For

Part of Clinical Trial Assistant (CTA) Training: The Trial Master File.

About 4 minutes. Free, in your language, no account needed.

Why this matters

The trial master file is the most misunderstood object in clinical research. People treat it as storage. It is evidence.

ICH E6(R3) is explicit that the sponsor and the investigator each maintain a file, and that together they must permit evaluation of the trial's conduct and the quality of its data. The sponsor's is the TMF; the investigator's is the investigator site file. They are not copies of each other, and the difference matters when someone asks who was supposed to hold what.

Sponsor TMF and investigator site file

Some documents belong in both. Many belong in only one.

The ISF holds what the site did: signed consent forms, source records, the site delegation log, local ethics correspondence, the site's copy of the protocol and amendments, local laboratory certification, and the site's investigational product accountability records.

The TMF holds what the sponsor did: the protocol's development and approvals, regulatory submissions, vendor contracts and oversight, monitoring reports across sites, safety reporting to authorities, and trial-level oversight records.

A document sitting in the wrong file is not a filing error. It can mean the responsible party never held the evidence that they discharged their duty.

Contemporaneous, attributable, complete

The three properties that make a TMF defensible are the same ones that make data defensible.

Contemporaneous means filed close to when it was created. A file assembled at the end of a trial is a reconstruction, and an inspector can usually tell.

Attributable means it is clear who produced the document and who filed it. Electronic systems record this automatically, which is a reason to use the system rather than a shared drive.

Complete means the set answers the questions, not that every folder has something in it. An empty folder that should be empty is fine. A folder with three versions of a document and no indication of which was in force is worse than an empty one.

The index is part of the file

A TMF that nobody can navigate is not doing its job. Most sponsors use a reference model that assigns each artefact a zone, section and artefact number, so that two people asked to file the same document put it in the same place.

Learn your study's model early. The single most useful thing a new CTA can do in their first fortnight is to read the model and then walk the file against it.

In practice

A monitor asks you for evidence that the sponsor exercised oversight of the central laboratory during

  1. You look, and the file contains the laboratory contract, the certification, and nothing else.

The laboratory performed well all year. Turnaround was good, queries were few, and nobody has a complaint. That does not help. The question is not whether the vendor performed; it is whether the sponsor discharged its duty to oversee, and the file is the only evidence of that.

What should have been there: the documented transfer of responsibilities, the agreed performance expectations, records of periodic performance review, an issue log with closures, and any audit or qualification record.

This is the most common shape of a TMF finding that surprises people. Nothing went wrong and the file cannot prove the sponsor was watching.

Sponsor file or site file

A quick test when you are unsure where something belongs: whose duty does this document evidence?

Ethics correspondence about a specific site evidences that site's compliance, so it belongs in the investigator site file, with the sponsor holding its own record of the submission it made. A monitoring visit report evidences sponsor oversight, so it belongs in the TMF, and the site typically holds a copy of the follow-up letter rather than the report.

Consent forms signed by participants are site documents and should never travel to the sponsor file. A signed consent form arriving in a sponsor's TMF is both a filing error and a personal data problem, and it happens most often when a site scans a whole folder and sends it.

Key takeaways