Essential Documents in a Clinical Trial: The ICH E6(R3) 'Essential Records' Function Test, Not a Frozen Checklist

At a glance

• ICH E6(R3) renamed “essential documents” to “essential records” and redefined them by a function: a record is essential if it, individually or collectively, permits evaluation of trial conduct and the reliability of results. The skill is applying that test, not memorizing a 50-row table.
• The old ICH E6(R2) Section 8 before/during/after table is gone. R3 replaces it with Appendix C, an essentiality test (C.3.1) plus an Essential Records Table that is explicitly not exhaustive.
• Scope is risk-proportionate: which records exist depends on trial design and conduct, so two trials can have legitimately different “complete” files.
• Records live in two repositories: the sponsor TMF and the investigator/institution repository, often called the investigator site file (ISF). Knowing who holds what, and when it must exist, is half the job.
• Retention is set by applicable regulatory requirements, not by ICH alone. Under 21 CFR Part 312, that floor is 2 years past a defined endpoint, for both sponsor and investigator.
• Treat any published checklist as a minimum, not a ceiling. Audits fail on the trial-specific record nobody listed, not on the standard fifty.

For years, “essential documents” meant one thing to most clinical-ops and QA teams: the ICH E6(R2) Section 8 table, organized before / during / after, copied into an SOP and treated as the definitive list. ICH E6(R3) (effective with Annex 1 in 2025) breaks that habit on purpose. The term is now “essential records,” and the deliverable your team owns is no longer a memorized table. It is a decision test plus a map of who files what, where, and when. This article gives you both.

Essential records, defined: the E6(R3) function test

Start with what the standard actually says. ICH E6(R3) defines essential records as the documents and data (and relevant metadata), in any format, associated with a clinical trial that facilitate the ongoing management of the trial and collectively allow the evaluation of the methods used, the factors affecting the trial, and the actions taken during conduct, to determine the reliability of the results and to verify that the trial was conducted in accordance with GCP and applicable regulatory requirements.

Read that again, because the operative words are “collectively allow the evaluation.” Under ICH E6(R3) Appendix C, the assessment of whether a record is essential turns on whether it permits and contributes to the evaluation of trial conduct, in relation to investigator and sponsor compliance with GCP and the reliability of results. That is the function test. A record earns its place in the file because of what it lets an auditor or inspector reconstruct and verify, not because it appears on a list.

Appendix C, section C.3.1 lays out the criteria a record is measured against: whether it is submitted to or issued by a regulatory authority or IRB/IEC, whether it is a trial-specific procedure or plan, whether it documents informed consent, whether it contains data and metadata needed to evaluate conduct, whether it evidences staff qualification and delegation, and so on through a long but illustrative set. The criteria are the test. The list that follows is the worked example.

What R3 changed: from “essential documents” to “essential records”

Three shifts matter for your filing practice.

First, the rename is not cosmetic. “Records” is broader than “documents” and explicitly includes data and relevant metadata in any format. ICH E6(R3) Appendix C states that an essential record can be the data, plus the metadata, needed to allow appropriate evaluation of trial conduct. If your mental model of the TMF is paper-shaped, audit trails and structured data now sit inside scope.

Second, the frozen table is gone. ICH E6(R3) Appendix C, section C.3.2 presents an Essential Records Table, but C.3.3 is blunt about its status: the table is not an exhaustive list, and other trial records may also be considered essential by the sponsor or investigator. The table also notes that for some listed records, their presence and nature depend on trial design, conduct, and risk-proportionate management, and they may simply not be produced. The list is a floor and a prompt, never a ceiling.

Third, scope is risk-proportionate. ICH E6(R3) Appendix C, section C.1.1 says the nature and extent of records generated and maintained depend on the trial design, its conduct, the application of risk-proportionate approaches, and the importance and relevance of each record to the trial. Two well-run trials can therefore have different complete files. This is why “we used the standard checklist” is not, by itself, a defense.

The three-phase model, with the test applied

The before / during / after rhythm still helps you plan, even though R3 does not freeze it into a table. ICH E6(R3) Appendix C, section C.2.5 notes that some essential records should generally be in place prior to the start of the trial and may be updated during the trial, and the Essential Records Table marks those start-of-trial records with an asterisk.

So the phased view becomes: before the clinical phase, you expect the records that authorize and define the trial (protocol and approvals, IRB/IEC composition and opinion, agreements, qualification and delegation records, system validation evidence). During conduct, you expect the records the trial generates (consent forms, source records, data and metadata in the data acquisition tools, monitoring reports, safety notifications, deviations and CAPAs). After the clinical phase, you expect the records that close and account for the trial (final reports, accountability and disposition of investigational product, audit certificates, data finalization documentation).

The discipline is to run each candidate through the function test rather than asking only “is it on the list.” Ask: does this record, alone or with others, let someone evaluate how the trial was conducted and whether the data are reliable? If yes, it is essential and should be retained when produced.

Who holds what: sponsor TMF versus investigator ISF

This is where most stale checklists go quiet, and where inspectors do not. ICH E6(R3) Appendix C, section C.2.3 says essential records should be maintained in, or referred to from, repositories held by the sponsor and by the investigator/institution for their respective records, that these repositories may be called a trial master file (TMF), and that the investigator/institution repository may be called the investigator site file (ISF).

The EMA TMF guideline puts structure on that split. It describes the TMF as usually composed of a sponsor TMF, held by the sponsor organization, and an investigator TMF held by the investigator/institution, often called the ISF. Some records live in one only: the EMA guideline gives the subject identification code list as filed in the investigator TMF only, and the master randomization list as filed in the sponsor TMF only.

Two duties drive the duplication map.

Reconstruction without dependence. The EMA TMF guideline states that superseded versions of relevant documents such as the protocol, Investigator’s Brochure, and eCRF should be present in the investigator/institution TMF in a manner that enables reconstruction without needing to access the sponsor TMF, with evidence of date of receipt, review or approval where necessary, and date of implementation. The investigator’s file has to stand on its own.

Independent control. The EMA TMF guideline also states that a complete investigator TMF should be available before, during, and after the trial, and accessible under the control of the investigator/institution, independent from the sponsor. The site’s file is not the sponsor’s to hold.

ICH E6(R3) aligns on responsibility boundaries: under section 2.12.11 the investigator/institution should have control of all essential records they generate before and during conduct, while under section 3.16.3(a) the sponsor retains the sponsor-specific essential records pertaining to the trial. Each party owns its own.

A phase-by-holder decision table, built from these provisions, is the concrete deliverable for your team:

Record (illustrative)	Phase	Sponsor TMF	Investigator ISF
Signed protocol and amendments	Before / during	Yes	Yes (copy, with receipt/approval dates)
IRB/IEC approval and composition	Before / during	Yes	Yes
Master randomization list	Before / during	Yes (only)	No
Subject identification code list	During	No	Yes (only)
Signed informed consent forms	During	No (confidential to site)	Yes
Source records / case histories	During	No	Yes
Delegation of authority log	Before / during	Yes (copy)	Yes (original)
Monitoring reports	During	Yes	Site copies as applicable
IP accountability / disposition	During / after	Yes	Yes
Final / clinical study report	After	Yes	As applicable

Treat the table as a starting map keyed to your protocol, not a universal truth. Per ICH E6(R3) Appendix C, section C.2.10, some records are typically retained only by the sponsor (for example, those tied solely to sponsor activities such as data analysis) and some only by the investigator/institution (for example, those containing confidential participant information), while others may sit in both. The asterisk-marked, start-of-trial records in the Essential Records Table tell you which rows must exist before the first participant.

The working checklist: a minimum, not a ceiling

You can still hand your team a phase-organized checklist, and you should. But mark it loudly: minimum, not exhaustive. ICH E6(R3) Appendix C, section C.3.3 is the authority for that label, and the EMA TMF guideline reinforces it from the other direction, stating that the documentation listed in ICH GCP defines documents considered essential but should not be used as a definitive checklist for TMF content, and that many trials require additional documents not specifically mentioned, so the sponsor and investigator should include any documentation that facilitates reconstructing and evaluating the trial conduct.

In practice, build the checklist from the Essential Records Table, organize it by phase and by holder, and add a column that forces the function test on anything atypical to your trial. The checklist tells the team what usually shows up. The test tells them what to do when something does not.

Retention: how long, by whom

Here ICH defers, and the local regulation governs. ICH E6(R3) section 2.12.12 requires the investigator/institution to retain essential records for the required retention period in accordance with applicable regulatory requirements, or until the sponsor says they are no longer needed, whichever is longest. ICH E6(R3) section 3.16.3(a) places the parallel duty on the sponsor for sponsor-specific essential records. ICH sets the duty; it does not set the clock.

For US INDs, 21 CFR Part 312 sets the clock, and it is worth stating the tension plainly. ICH E6(R3) section 2.12.12 frames investigator retention as the longest of the regulatory period or the sponsor’s notice, an open-ended floor. 21 CFR 312.62(c) fixes a specific number: an investigator shall retain the required records for 2 years following approval of a marketing application for the indication, or, if no application is filed or it is not approved, until 2 years after the investigation is discontinued and FDA is notified. 21 CFR 312.57(c) sets the same 2-year structure for the sponsor. These do not conflict so much as stack: ICH tells you to honor the applicable regulatory requirement and the sponsor’s instruction, and Part 312 supplies the concrete minimum for trials under a US IND. Where a sponsor’s notice or another jurisdiction’s period runs longer, the longest period wins under ICH E6(R3) section 2.12.12. Do not quietly substitute the 2-year FDA figure for that “whichever is longest” rule.

Note also what 21 CFR 312.62(b) treats as records the investigator must keep: adequate and accurate case histories recording all observations and data on each individual, including case report forms and supporting data such as signed and dated consent forms and medical records. That is the FDA-side anchor for the source records and consent forms that R3’s function test would independently flag as essential.

How to decide if a new or atypical document is essential

This is the payoff of the reframe. When a record appears that is not on your checklist, do not ask whether it is on the list. Apply the test from ICH E6(R3) Appendix C, section C.3.1: does this record, individually or collectively with others, permit and contribute to the evaluation of trial conduct and the reliability of the results? Walk it against the C.3.1 criteria, for example whether it documents a trial-specific procedure, evidences qualification or delegation, captures data or metadata needed to evaluate conduct, or documents a regulatory or IRB/IEC interaction. If it meets the function, it is essential and should be retained when produced, and you decide which repository holds it using the C.2.3 and C.2.10 logic above.

ICH E6(R3) Appendix C, section C.3.1 even notes that this assessment, while important, is not required to be documented, and that a structured content list for your repositories may be used to prospectively identify essential records. In other words: bake the test into your TMF index up front, so the judgment is made by design rather than scrambled during an inspection.

Where teams get it wrong

The recurring audit finding is not a missing standard document. It is a present-but-insufficient file: a “complete” R2-style checklist, fully ticked, that still failed because a trial-specific record (an atypical computerized-system validation, a sample-handling chain-of-custody, a decoding procedure for an unusual blind) was never on the list and so was never filed. The R2-table habit trained teams to confirm presence against a fixed inventory and to stop there. The function test is precisely the correction: the inventory is the floor, and the skill is judging the records the floor never named.

Three more traps worth naming. Treating the EMA guideline’s references to “ICH GCP section 8” as still current: that wording predates R3 and now maps to Appendix C, so use the section-8 phrasing only as legacy context, not as the live structure. Letting the sponsor hold the site’s file: the EMA guideline requires the investigator TMF to be under the investigator/institution’s independent control before, during, and after the trial. And collapsing retention to “2 years” everywhere: that is the 21 CFR 312.62(c) and 312.57(c) figure for US IND trials, not a universal period, and it is overridden whenever ICH E6(R3) section 2.12.12’s “whichever is longest” points further out.

None of this is about software making you compliant or a checklist certifying your file. The standards set what must be evaluable; your team, and the sponsor, stay responsible for the judgment that gets each record into the right repository at the right time. Master the test, keep the checklist as a floor, and the stale-checklist failure stops being your finding.

Sources

• ICH E6(R3) Good Clinical Practice, version r3, ICH — https://www.ich.org/page/efficacy-guidelines
• EMA Guideline on the content, management and archiving of the clinical trial master file (paper and/or electronic), version 2018, EMA (EMA/INS/GCP/856758/2018)
• 21 CFR Part 312 Investigational New Drug Application, version 2026-04, FDA