Audit trail review is the control that turns an audit trail from a feature into evidence. Recording who changed what and when is only half of it; you also need to show that someone competent read those entries before the data were relied on. This post sets out a risk-based procedure you can adapt: what to review, who reviews, when, and what the rules say, with the in-force texts kept apart from the July 2025 Annex 11 draft.
The draft is a consultation text and is not adopted. Where it is cited below, the wording is “the draft expects”. The January 2011 Annex 11 and 21 CFR Part 11 are the texts in force.
What the rules say about audit trail review
Part 11, section 11.10(e), asks for four things from an audit trail, and a review procedure should test all four:
- It is secure, computer-generated and time-stamped, and logs, independently of the operator, the timing of each entry or action that creates, changes or removes an electronic record.
- It covers what operators did, not just the final results.
- Changes do not obscure previously recorded information, so the earlier value stays visible.
- It is kept for at least as long as the underlying record and is available to FDA for review and copying.
Part 11 does not itself say who must read the audit trail or how often. That is where the 2011 Annex 11 comes in. Its audit trail section expects a risk-based audit trail of GMP-relevant changes and deletions, a documented reason for any change or deletion, and an audit trail that can be turned into an intelligible form and is regularly reviewed.
The draft is far more specific. It expects a documented review procedure for each system or system type that says who reviews, what is reviewed and when (12.5), reviewers who were not involved in the activity (12.6), targeted rather than exhaustive review (12.7), and review before batch release unless a later check can be justified (12.8).
The 2003 discretion and its limits
FDA’s 2003 guidance on Part 11 scope and application says the agency intends to exercise enforcement discretion for the Part 11 audit trail provisions: 11.10(e), 11.10(k)(2) and the matching parts of 11.30. Teams sometimes hear that as “audit trails are optional”. They are not, for four reasons.
- The guidance keeps every applicable predicate rule requirement in place, including documentation of dates, times and sequence of events, and the duty that changes must not obscure earlier entries.
- It recommends deciding whether to use audit trails, or other physical, logical or procedural measures, on the basis of predicate rule needs, a documented risk assessment and the effect on product quality, safety and record integrity. It adds that audit trails are particularly appropriate where users create, modify or delete regulated records in normal operation.
- Discretion covers enforcement of those specific Part 11 provisions. It is not an exemption, and the guidance is non-binding.
- For an EU-regulated site, Annex 11 section 9 stands on its own, and the draft would tighten it.
The practical reading: if your system generates audit trails for GMP data, you have little to gain from not reviewing them, and a good deal to explain if an inspector finds a problem you could have found yourself. Our 21 CFR Part 11 requirements guide covers the full list of what the discretion does and does not relax.
Data review vs system review
These are two different reviews. Treat them as separate lines in your procedure, because they have different owners and different timing. This split is a working method, not a term from the regulations.
| Aspect | Audit trail review of data | Audit trail review of the system |
|---|---|---|
| Question | Was this batch, run or result changed, and was the change justified? | Were controls, settings, users or the clock changed, and was that authorized? |
| Unit of review | A batch, sample set or record package | A period, such as a month or a release cycle |
| Typical entries | Edited results, reprocessed data, deleted or repeated runs, overridden steps | Privilege changes, audit trail setting changes, time changes, method or specification edits, failed logins |
| Reviewer | Someone independent of the work, usually QA or a second analyst role | System owner and QA, with an administrator not involved in GMP work supplying the data |
| Timing | Before the record is relied on, and before batch release (draft 12.8) | Scheduled, with frequency set by risk |
| Output | Review record for the batch or run | Periodic review record feeding the periodic system review |
If you only do data review, a changed configuration can alter results quietly. If you only do system review, you can miss a justified-looking edit on one critical batch. You need both, sized by risk.
A risk-based audit trail review procedure in five steps
1. Define the review scope per system
Start from the data criticality assessment: which records and actions matter to product quality and patient safety, and which entries would signal a problem. The draft says reviewing every entry may not be effective and asks for targeted reviews built around local processes (12.7). Write down the filters, for example edits after approval, deletions, reprocessing and changes to critical fields.
2. Set who reviews and what they need
The reviewer must be independent of the activity (draft 12.6) and trained on both the process and the audit trail screen. If your only qualified person is also the operator, use a peer from another group, and record why. Give reviewers sort and search tools or an export to a tool that supports them, as the draft expects (12.4).
3. Set when the review happens
Tie the review to a point in the process, not to a calendar guess. For batch-related records, the draft expects review before batch release unless a later check is justified (12.8), and expects that reviews with a direct effect on release are available to the Qualified Person at release time (12.10). Be careful with the clock: the trigger is the release decision for that batch, not the date of the last system backup or the end of the quarter.
4. Verify the reason, not just the entry
An entry that says an edit was made is not a finding. The draft says a key element is verifying why a change was made (12.7), and that systems should prompt for and store the reason at the time of change (12.2). Reject generic reasons that could cover anything, and compare the stated reason against the batch or run record.
5. Investigate, document and close
Any significant variation from the expected outcome is investigated and recorded (draft 12.5). Route confirmed problems into your deviation and CAPA procedure, record clean reviews too, and feed the results to periodic review so the review scope adapts.
Audit trail review red flags by system type
These patterns are illustrative, not a complete list. Add the ones that match your processes and the failure modes you have already seen.
| System | Red flags worth a filter |
|---|---|
| Chromatography data system (CDS) | Repeated injections of the same sample, aborted or deleted sequences, manual changes to peak integration, processing method edits after acquisition, results reprocessed after reporting, system time changes |
| Laboratory information management system (LIMS) | Result edits after approval, specification or limit changes, retests with no matching investigation, deleted or re-created samples, status changes out of sequence |
| Electronic batch record (eBR) | Edits to critical process parameters, step overrides, late entries, repeated entries for the same step, changes after a signature, generic reasons for change |
| Instrument or equipment controller | Method or recipe changes, alarm acknowledgments without a comment, calibration data edits, audit trail or time setting changes |
| Spreadsheet or in-house tool | No audit trail at all, which is itself the finding, and a signal to replace the tool or compensate with controlled procedural measures |
Across all of them, check three entry types: deletions, changes made close to a release or reporting deadline, and anything done by a user acting outside their normal role. The draft’s list of what review should detect covers repeated activities, errors, omissions, unauthorized deviations and loss of data integrity (12.7).
Reviewer independence and the administrator problem
Independence has two sides. The reviewer should not be the person who generated the data (draft 12.6). The administrator who can change audit trail settings, or the system clock, should not be someone involved in GMP activities, and those changes should themselves be audit-trailed (12.3). If one person can do the work, change the settings and review the results, the review proves very little.
Small sites often cannot split every role. Where that is true, document the compensating measure: a second reviewer from QA, review of the administrator’s own entries by management, or a vendor-held log. State the gap and the reasoning in the procedure.
Making audit trail review workable
- Configure before you review. Verify at qualification that audit trails are on, locked, time-stamped, and show old and new values. Test the audit trail with the same rigor as a calculation. Our post on ALCOA+ as system controls explains why completeness and attribution depend on it.
- Use tools. The draft encourages them (12.5). Saved filters or exception reports keep a review focused on the entries that matter.
- Keep the electronic copy usable. The draft expects a complete, searchable and sortable electronic copy that includes the audit trail (12.9). Flat, locked files such as a static PDF do not meet that expectation.
- Record the review. State the system, period or batch, reviewer, filters used, findings and actions. An unrecorded review cannot be shown to an inspector.
- Revisit the scope. If deviations show a new failure mode, add it to the filters.
Electronic signatures sit in the same chain: a signature applied after an unexplained edit is a problem the review should catch. See the 21 CFR Part 11 electronic signature rules for the controls around signing.
To put this into a procedure you can validate, the 21 CFR Part 11 Toolkit includes an Audit Trail Management and Review Procedure, an Audit Trail Review Record and an Audit Trail Configuration Verification Checklist, among 105 templates (68 Word and 37 Excel) for $99. The download is named “Computer System Validation Toolkit”.
Audit trail review FAQ
Do I have to review every audit trail entry?
No. The draft Annex 11 says reviewing all entries may not be effective and expects targeted, risk-based review of critical processes and data. Your procedure should say what is in scope and why.
How often should audit trails be reviewed?
The 2011 Annex 11 says regularly, with no interval. The draft expects timeliness according to process risk, and before batch release unless later detection is justified. Set the frequency by risk for system reviews and tie data reviews to the release decision.
Can the system administrator review the audit trail?
Only if independent of the activity reviewed, and an administrator who can change audit trail settings should not be involved in GMP activities at all, as the draft expects. For system reviews, an administrator can supply the data while QA or the system owner reviews it.
Does the 2003 FDA discretion mean I can skip audit trail review?
No. It concerns enforcement of specific Part 11 provisions. Predicate rule duties still apply, FDA recommends risk-based decisions on audit trails, and the EU Annex 11 audit trail section is unaffected.
What should be in an audit trail review record?
The system, batch or period, reviewer and independence, filters applied, entries examined, findings with their justification, and the actions taken or the deviation reference.


