Regulations and compliance · Practical guide

Audit trail review: justifying scope and record selection

Start with events that can change the meaning of a record, including amendments, deletions, repeats and critical configuration changes. Verify that the review view reliably includes them. Its filter forms part of the control and needs its own assessment.

Table of contents

At a glance

Can an audit trail review be limited to selected events?

Review scope should reflect data criticality, possible manipulation or error paths and the decision to be made. A justified selection of relevant events can be suitable. An arbitrary sample or untested filter must not hide critical changes. Define and document the scope, timing and response to findings.

Scope and Applicability

The article elaborates on the methodology for selection. It does not specify a general allowable percentage or provide a blanket exemption from reviewing relevant data prior to its use. Applicable guidelines and process context remain decisive.

Sources: [1] [2]

Deriving the review from the required quality decision

Start with data and potential consequences of errors. Which change could influence a quality decision? Who is authorized to perform it and how might it otherwise be detected? From this, it follows which events during review carry particular significance. A lengthy list of all log entries is not inherently more thorough than a justified, effective selection.

Annex 11 No. 9 provides audit trails for GMP-relevant changes and deletions based on risk. They should be available, can be converted into a generally understandable form and regularly checked. A universal daily frequency or full check of each technical message does not mention the entity. Determine the appropriate frequency and selection based on data criticism, change possibilities and existing controls.

Distinguish ongoing evaluation of a process from a cross-process review. A change in outcome may need to be considered prior to the associated decision. Recurrent administrative anomalies may require a different review pathway. Specify the purpose, timing, and responsible role for each relevant level.

Sources: [1] [2] [3]

Create a selection matrix based on data, event, and risk criteria

Capture critical data objects and potential actions: creation, modification, deletion, reprocessing, or status change. Include source of the audit record and available information. A requested review is practically ineffective when the required event is not adequately visible within the used system.

Justify exclusions as well. A technical notification with no impact on the considered record may be handled differently than a change in limit value. However, the selection must not be based solely on which events are most convenient to evaluate. Inconvenient, critical events must remain visible in the risk profile.

Data domains, critical events, and review questions
Data areaCritical event in the exampleInquiry
Result Value updated upon entry Why, by whom, and what effect?
Specification Limit value or method adjusted Which processes use the new status?
Identity Probe or batch reordered Is the correct assignment maintained?
Rights Critical authorization expanded Is the award and actual use appropriate?
Audit function Data capture or filter modified Is the required coverage retained?

Sources: [1] [2] [3]

Example: a filter overlooks result changes due to a second role

A review report shows modifications made by laboratory users. A technical role may alter the same result fields, but is excluded by the filter. The report is empty despite a critical value having been changed. The mere existence of a review procedure therefore does not confirm complete coverage.

Generate controlled, known events through the relevant roles and pathways. Verify that they appear in the intended report and are clearly attributable. Document the actual filter settings. Include a case outside the time window and an unknown event type if these risks are present.

After filter correction, the detection capability is re-evaluated. Additionally, assess whether the previous incomplete report may have omitted earlier relevant changes. The scope of such a comprehensive review is determined based on the specific risk. A new functional report does not automatically address historical questions.

Test cases for a filter with multiple operator roles
Test caseActionExpected evidence
Business user Change critical value Event in the intended review visible
Technical Path Perform identical change controlled execution Relevant pathway not unreasonably excluded
Time limit Create an event at the boundary of the time window Defined temporal coverage
Filter change Intentionally modify critical selection rules Control inspection identifies the gap
Failure Interrupt report or data supply No misleading empty success report

Sources: [1] [2] [3]

Justify frequency and timing on a professional basis

Query when a erroneous change must be detected at the latest for its impact to be appropriately addressed. A review conducted long after a quality decision may be too late for certain risks. Other events may be appropriately considered in a broader context. The appropriate frequency arises from the process itself, not from an arbitrary standard value.

Consider event volume and available reviewers. A process that daily requests thousands of unstructured alerts without defined criteria may practically fail. Improve selection and presentation, without losing critical coverage. A risk-based reduction requires a transparent justification and appropriate effectiveness validation.

Define how missing or delayed reviews are detected. A signature on a plan does not constitute evidence of actual execution. Appropriate task management and escalation procedures are required. Scope and format can be integrated into existing QMS procedures.

Sources: [1] [2] [3]

Equip reviewer with criteria and investigation pathway

A reviewer requires subject-matter expertise, access to the relevant process, and criteria for identifying anomalies. A change may be legitimate and yet require explanation. The review should assess whether the rationale and authorization are traceable and whether additional data may be impacted. Relying solely on the presence or absence of justification fields is often insufficient for this purpose.

Define how an anomaly is investigated. Required are data linkage, backup of relevant information, and responsibility assignment. Technical justification and impact on the process are linked but not identical. Confirmation of a system function by an administrator does not automatically conclude the evaluation of a result change.

Practice with known cases: legitimate correction, incorrect assignment, and unexpected privileged modification. Document whether the reviewer identifies the relevant differences. This is a standalone methodological exercise, not a general prescribed audit procedure.

Sources: [1] [2] [3]

Selection and justification for changes obtained

New data fields, roles, or interfaces may alter the review scope. Therefore, evaluate filters and reports during relevant system changes. A previously reviewed report does not remain automatically suitable for every subsequent configuration. The linkage between risk mapping, selection, and actual coverage should be maintained.

Obtain the review status, review period, findings, and their assessment. A note stating "reviewed, no abnormalities" is insufficiently informative when the scope of review is not clearly defined. At the same time, a detailed individual justification is not required for every routine entry. The documentation should enable clear traceability of the intended review.

Sources: [1] [2] [3]

Fewer entries read does not mean less evidence is required

The trigger alters the scope. A routine data review, a periodic system assessment, and an investigation into a specific concern address different questions. A substantiated issue may necessitate expansion to additional time periods, users, or data sets.

Document also the selections that are not made and the rationale behind them. Technical noise may reduce the detectability of important events; however, it should not be removed indiscriminately if relevant direct interventions or erroneous processing events are visible. The selection process represents a professional judgment, not merely a convenience setting.

Sources: [1] [2] [3]

Draft: Scope and filtering of known events for review coverage

Maintain the actually tested filter status together with data, roles, and selection limits. The table also includes a deliberate failure in data input: an empty report must not be perceived as a normal or inconspicuous data set. Identify who is responsible for handling unknown events and failed evaluations during operations.

Sample filling example – internal work aid
Review fieldExample entryAcceptance question
Data Result and associated sample. Is the critical interrelationship addressed?
Action Value modification and reassignment. Do any known test events occur?
Role Technical and relevant scientific approach. No unjustify ed exclusion?
Time Specified period with edge cases. Is the selection boundary correct?
Failure Data supply is controlled and interrupted. No misleading zero result?
Assessment Reviewer explains the event and next step. Is control practically feasible?

Sources: [1] [2] [3]

Next steps

  • Define decisions and critical events.
  • Select and filter against known test events.
  • Define review timing, responsibility, and documentation.
  • Define when a finding should trigger a wider review.
Further reading:

Project experience and collaboration

Test the review filter against known events

With process owners, we prepare a small set of known changes and trace them into the review output. Is a changed result still detected when a different authorised role makes the change? Such cases test whether selection and filters answer the relevant business question. A limited application may need a configuration fix. Multiple event sources first need their relationships clarified. Review timing and assessment must suit the use of the data, even where that requires a longer report.

Frequently asked questions about Selecting events for audit trail review

Are ten percent always an adequate sample?

No. A percentage figure without reference to criticality and failure modes is not a credible justification. Certain relevant events may require a targeted, comprehensive evaluation.

Sources: [1] [2]

Can software replace the manual review?

Software can identify and flag relevant events when its selection and detection logic has been verified against known events. This does not replace professional assessment of anomalous entries and the determination of their implications. Define which automated controls provide reliable coverage and which cases require evaluation by a designated individual.

Sources: [1] [2]

Is an annual review sufficient for batch-related changes?

An annual system review does not replace the evaluation of events that could have influenced a prior batch release decision. For US Drug cGMP, the FDA initially references applicable recordkeeping requirements when specifying review frequency; where such requirements do not define a specific frequency, it considers process knowledge and risk assessment. In Question 8, the FDA recommends an audit trail review accompanied by the examination of related records, such as prior to batch release. Verify that the timing, data usage, and applicable requirements are consistent.

Sources: [1] [2]

Do we need to manually review each technical login entry?

Not at all. Determine relevant data and events from risks and quality decisions. Reason for choice, timing and limitations. Technical logs can contain important additions, but are not automatically suitable as an audit trail. The detection capability of the selected report should be checked with known critical events.

Is an empty audit trail report a good result?

Only if data input, time range, and filtering have been demonstrably functional. An empty report may also indicate a wrong selection or system failure. Use appropriate known events and completeness controls. The review should be able to justify its coverage, rather than automatically equating emptiness with the absence of relevant changes.

What should a review remark at least explain?

It should be possible to trace the considered scope, time period, appropriate regulatory or reporting standard, findings, and decisions. The specific format may vary. A general statement of "reviewed without a clear subject" is insufficient. Therefore, not every uncritical entry necessarily requires a lengthy individual justification.

How do we assess the reviewer's capability?

Use known legitimate and conspicuous changes together with the intended comparison information. Observe whether the reviewer identifies relevant differences and responds appropriately. The exercise should align with the task and workload. A participation confirmation alone does not adequately demonstrate this practical capability.

Sources and review date

Source date of original classification: 14 October 2026. Regulatory references: EU-GMP Annex 11, PIC/S PI 041-1 and FDA-Q&A on data integrity. Selection matrices and examples are standalone work aids. Targeted original reconciliation conducted on 18 October 2026: regular review in accordance with Annex 11, Section 9.

  1. European Commission: EU GMP Annex 11, Computerised Systems (2011)§ 9: Risk-based Audit Trail Framework; specific selection matrix and test events are standalone methods. Current version as of EudraLex on 14.10.2026; no universal project or deadline mandate.
  2. PIC/S: PI 041-1, Good Practices for Data Management and Integrity (2021)Inspection Guidance on Data Integrity; in particular electronic data, transmission, access, and review.
  3. FDA: Data Integrity and Compliance With Drug CGMP – Questions and AnswersFinal Guidance, December 2018; no blanket extension to all GxP regulatory areas.

Sources checked: .

Daniel Herrmann, Daniel Herrmann Consulting

Daniel Herrmann Consulting

Establish a verifiable review scope

We support you in assessing the current situation, clarifying responsibilities, and implementing appropriate evidence within operations. The type of support that is most effective depends on your system, your team, and the specific objective.

  • Audit and inspection preparation →

    Go into an audit or inspection with accessible evidence and a prepared team. We begin with a walkthrough, rank open issues by risk and rehearse the inspection with your team.

Send a project enquiry Customer testimonials →