A Reviewable Change Record for Clinical Documents
By Kamila Kiriek, MD, PhD, Chief Medical Officer, Skaldi
A reviewable change record shows, for every edit to a clinical document, what the text said before, what it says now, which source the change follows from, why it was made, and who decided. Version history and tracked changes cover only the first two. The rest is what lets a reviewer check a revision without redoing the author's work, and what lets someone reconstruct the decision long after the people involved have moved on.
This article describes what such a record contains and how to keep one without adding a second job to every revision. It does not set out which changes need regulatory submission or approval; the function that owns each document decides that.
When you need one
Whenever a document is revised because something else changed. The common cases:
- a protocol amendment that carries into the consent form, the statistical plan, case report forms and manuals
- an Investigator's Brochure update that carries into the protocol and the consent form
- a safety finding or a regulatory request that changes several documents at once
- a correction to one document that exposes the same error in others
In each case the question a reviewer, an auditor or a new team member will ask is the same: why does this document say this now? The expectation behind that question is part of good clinical practice. ICH E6(R3) expects trial records to allow the conduct of the trial to be reconstructed and evaluated, and ICH E3 section 9.8 asks the clinical study report to describe changes in the conduct of the study or planned analyses. For electronic records maintained in place of paper records under United States regulations, 21 CFR Part 11 sets requirements for audit trails. None of these documents prescribes a format for a change record. They set the outcome: the reasoning must be recoverable.
What a reviewable record contains
One entry per change, with these fields:
| Field | Content |
|---|---|
| Location | Document, version and section |
| Before | The wording that is being replaced |
| Proposed | The new wording |
| Source | The document, version and section the change follows from |
| Reason | One sentence: why this source requires this edit |
| Decision | Accepted, edited, rejected, or further assessment requested |
| Decided by | Name or role, and date |
| Open question | Anything the source does not settle, written as a question |
The last field matters as much as the others. A record that only lists what was changed implies that everything else was checked and found fine. A record that also lists what could not be decided from the source is the honest version.
An illustration. A protocol amendment adds a clinic visit and a blood sample on Day 8 to a schedule that previously had visits on Days 1 and 29. The consent form entry reads:
- Location: ICF v1, section "What will happen during the study?"
- Before: "After your Day 1 visit, your next clinic visit will be on Day 29."
- Proposed: "After your Day 1 visit, you will return to the clinic on Days 8 and 29."
- Source: Protocol v2, Schedule of Assessments
- Reason: The consent form's visit schedule still reflects Protocol v1.
- Decision: Accepted, by the reviewing physician, with the date
- Open question: Does the added visit change the stated blood volume or time commitment? The amendment does not specify.
Everything a reviewer needs is in one place, and the open question travels with the change instead of living in someone's inbox.
How to keep one
1. Start the record when the source changes, not when the edits are done
Write the first entries from the comparison of the old and new source. A record assembled afterwards from tracked changes describes what was edited; it cannot show what was considered and left alone.
2. Tie every entry to a version
"Protocol, Schedule of Assessments" is not a source. "Protocol v2, Schedule of Assessments" is. The same applies to the document being changed. If either moves to a new version while the work is in progress, the entry is re-checked against it.
3. Write the reason as a sentence a stranger could follow
"Per amendment" is not a reason. "The amendment adds a visit on Day 8; this sentence lists the visit days" is. The test is whether someone outside the team could confirm the edit from the entry and the source alone.
4. Record decisions, including the ones that say no
A proposed edit that the reviewer rejects stays in the record with the reason. So does an edit the reviewer rewrote. Rejections are evidence that review took place and carry information the next amendment will need.
5. Keep open questions open until someone answers them
An open question is closed by a named person with the missing information, and the answer is recorded with its own source. It is not closed by the deadline.
6. Check the finished documents against the record
Before the package goes out, confirm that each accepted entry appears in the final document as recorded, that nothing changed that is not in the record, and that the same fact reads the same way everywhere it appears. This is the consistency check, and the record is what makes it possible: without it, consistency means reading everything against everything.
7. File it with the documents it explains
The record belongs with the amendment documentation in the Trial Master File, alongside the approved versions. A record kept only in a working folder is lost at the first reorganisation.
Common mistakes
- Relying on tracked changes. They show the edit and the editor. They do not show the source or the reason, and they disappear when changes are accepted.
- Summarising instead of recording. "Section 4 updated for the new visit schedule" cannot be verified. The before and after wording can.
- One reason for many edits. "Updated per Amendment 2" on forty entries hides the three that had nothing to do with Amendment 2.
- Dropping what was not changed. If a section that the impact assessment identified as potentially affected was reviewed and left as it is, an entry saying so, with the basis, saves the next reviewer from assessing it again.
- Letting the record and the documents diverge. A late edit made directly in the document, outside the record, is the usual cause. The fix is procedural: no edit without an entry.
- Recording a decision without a decider. "Approved" with no name or role is a status. It is not a record of review.
What remains a specialist's decision
A change record documents decisions. It does not make them. In particular:
- Whether the proposed wording is right. The owner of the document decides: the responsible physician for medical content, the statistician for analyses, regulatory affairs for submissions. Accepting an entry is an editorial decision about wording in that document. It is not a medical or regulatory approval of the change.
- What the open questions' answers are. They need information the source does not contain.
- Whether the change must be submitted, notified or approved before it is used. That depends on the document, the region and the nature of the change, and regulatory affairs determines it.
- Whether the documented process meets the organisation's quality requirements. Quality assurance decides, against the organisation's own procedures.
What the record does is make each of those decisions quicker and safer to take, because the person deciding sees the change, its source and its reason together, and sees plainly what is still unknown.
Related reading
- Protocol amendment impact assessment: which documents to check
- Updating the ICF after a protocol amendment
- Managing protocol amendments across clinical documentation
- Audit trails in clinical trial documentation
Sources
- ICH E6(R3), Guideline for Good Clinical Practice.
- ICH E3, Structure and Content of Clinical Study Reports, section 9.8.
- 21 CFR Part 11, Electronic Records; Electronic Signatures.
🔄 Skaldi connects a protocol amendment to proposed, source-linked revisions in the linked ICF, with expert review.
See the amendment exampleProtocol changed. Documents in step.
When a source changes, every affected document follows. Pilots start with protocol amendments.