Regulations and compliance · Practical guide

GMP review by exception: validating exception reports

Review by exception directs attention to predefined data or actions that need closer examination. PIC/S describes a validated exception report and identifies this approach as a possible focus for audit trail review in PI 041-1, sections 5.6.2 and 9.6. Establish whether the report reliably detects relevant exceptions, including process deviations that generate no technical error.

Table of contents

At a glance

When is review by exception suitable for a GMP process?

Review by exception uses a validated exception report to identify predefined unusual data or actions for further review. PIC/S discusses it for audit trail review in sections 5.6.2 and 9.6. Demonstrate rule coverage, complete data input and successful execution before relying on the report. Staff still need to evaluate the exceptions and perform controls outside those rules.

  • Independently evaluate the technically appropriate rules, their technical implementation, and the reviewer's actions.
  • Known exceptions and incomplete or missing runs with independent scheduled occurrences are distinguished.
  • Maintain traceability of operating procedures, data scope, findings, and decisions within the workplace.

What "Review by Exception" means

PIC/S describes an exception report as a validated search tool for predefined unusual data or actions requiring further attention or investigation (5.6.2, footnote 4). Section 9.6 identifies unusual or unauthorised activity as a focus for audit trail review.

The procedure supports a defined portion of the review process. It does not automatically replace the evaluation of all required data or the professional release decision. The following matrices and review steps are standalone aids for specific implementation.

Sources: [1] [2]

What a blank exclusion list indicates

An empty list may indicate that none of the defined exceptions were identified. However, the same display can also occur when a source fails to provide data or when the incorrect time period was evaluated. Therefore, validation evidence must distinguish between a fully executed, non-critical run and the absence of a review basis.

First establish which decision the report supports and which errors it must reveal. Then test three things. Does the rule address the process risk? Is it implemented correctly? Can the reviewer interpret the result and its limitations?

Sources: [1] [2] [3]

Define rules for normal, error, and failure cases

Record for each rule the rationale, the required data, and the expected response. The exception catalogue should also indicate which controls remain outside the report. New or unknown states must not be treated as normal conditions without evaluation.

The following matrix serves as a fillable work template. Populate it with specific data sources, thresholds, and responsible roles within your process. Your selection must justify how both professional identification and completeness of the assessed inventory are verified.

Own rule matrix: Detection and limit of statement per test area
Area to assessPrepared CaseExpected reaction
Normal case Known occurrence within the defined limits No unjustifyecd report
Professional Exception Critical post-hoc value change Correct procedure together with context appears
Data gap Expected process or area is missing Incomplete basis will be visible
Unknown state Not yet assigned relevant event type Controlled clarification instead of silent exclusion
Failure of the analysis Reporting job or data supply fails Failed run remains distinguishable from a valid result
Reviewer's action Known exception requiring additional information Evaluation, investigation, or escalation is possible

Sources: [1] [2] [3]

Test case: A laboratory instrument fails unexpectedly

The example involves two laboratory areas. For Area A, four test events are prepared, including one known critical change; for Area B, three events are prepared, including an additional critical change. During the full run, seven inputs and two exceptions must be traceable and assignable.

Subsequently, the supply of B in the test environment is interrupted. The internal requirement states: The missing area must be visibly apparent as an incomplete basis. A report containing only the four events from A and including one exception must not create the impression that both areas have been fully reviewed.

Upon restoration, the reference is reconciled again. Specifically verify that delivered events are complete, correctly attributed, and free from unresolved duplicate processing. This example describes intended conditions of an exercise and does not represent an actual product release or inspection.

Example: Target values for the Two-Range Test
RunExpected receipts / ExceptionsAdditional inspection statement
A and B completely 7 Events / 2 known exceptions Both sources and both exceptions assigned
B interrupted Only 4 events / 1 known exception available from A No complete overall status
B delivered Again 7 distinct events / 2 known exceptions No missing or unexplained duplicate events

Sources: [1] [2] [3]

Six technical root causes deliberately distinguished

In addition to source loss, the time period, filter, job status, permissions, or data format may alter the outcome. Modify only the specified condition in an appropriate test environment. An independently prepared reference establishes which process must remain detectable and which status is expected during a disruption.

The matrix supplements the technical rules with different technical failure pathways. It is not a requirement to manipulate production records. The required control response is derived prior to testing from the intended use.

Own failure matrix: six different technical failure paths
Manipulated test inputRiskExpected inspection conclusion
Source missing No data instead of no exception Incomplete run detected
Incorrect time period Significant event excluded Time period verified
Filter too tight Known exception disappears Coverage demonstrated
The job did not run An old report appears current Execution status visible
Permission is missing Subpopulation is evaluated Access gap identified
Format changed Event no longer interpreted Schema change detected

Sources: [1] [2]

Secure implementation with an independent reference

A limited comparison period can help evaluate both the previous review and the new exception logic against the same appropriate cases. Select not only a period with no incidents, but also include controlled rare error cases. The duration and scope of this approach shall be determined according to your needs.

The expected exceptions must be derived independently. A copied filter in a subsequent analysis could reproduce the same error. A subject-matter reference list or an independently validated comparison method provides a more robust basis.

A recognized control event initially confirms only the relevant test pathway. Therefore, link rule tests, population comparison, and the actual action of the reviewer. Document the statement each piece of evidence supports and the remaining limitations.

Sources: [1] [2] [3]

Integrate reviewers and rule maintenance into operations

The reviewer requires the process, relevant data, regulatory basis, and access to the necessary context. Define when a finding is accepted, corrected, or further investigated. Document the scope of consideration, current regulatory status, findings, and decisions; a red code alone is insufficient.

Retain false alarms and missed cases in routine maintenance. A shorter reporting list is only beneficial if critical detection remains intact. Therefore, changes to sources, rights, formats, or filters are evaluated for their impact on the review process.

A well-maintained reference dataset containing known normal, error, and failure cases can support future assessments. Identify who is responsible for the technical implementation and professional guidelines, and specify when the dataset must be updated due to new processes or events.

Sources: [1] [2] [3]

Next steps

  • Define relevant exceptions and decision purposes.
  • Detection, completeness, and failure cases testing.
  • Reviewer, escalation and release criteria to be defined.
  • Monitor regulatory changes and continue to assess detection performance.
Further reading:

Project experience and collaboration

Test exception rules against known findings and failures

With process owners, we prepare known normal cases and relevant exceptions and establish expected responses independently of the report. With the system team, we test how missing sources, changed filters and incomplete access become visible. A defined process may need a compact set of rules. Complex data paths need further completeness controls. We also test whether a reviewer can identify the appropriate investigation from a finding. Fewer notifications alone do not demonstrate success.

Frequently asked questions about Review by Exception

Does a blank report automatically indicate a defect-free batch?

No. First, the data scope, successful processing, and suitability of the rules must be established. The overall decision includes the necessary additional substantiation.

Sources: [1] [2]

Is a Review by Exception equivalent to a random sample?

No. An exception report identifies defined relevant events. A sample selection selects data based on a predefined methodology; both approaches require a justified rationale.

Sources: [1] [2]

Must report rules be verified upon changes?

Relevant changes require an impact assessment and, if necessary, renewed validation. An originally suitable filter does not remain automatically permanently suitable.

Sources: [1] [2]

Can a Review by Exception replace a complete data verification?

For a validated exception, the report may focus on predefined anomalies. PIC/S specifies the validated Exception Report and Review by Exception (5.6.2, 9.6). However, the rules must fully cover the relevant question and enable detection of deviations. Controls outside the rules and professional release decisions remain necessary. Clearly define this boundary.

Sources: [2] [1]

Is a parallel phase always required?

No. A parallel phase represents a possible introduction method. Crucially, appropriate evidence of rule and data coverage must be established. Controlled error scenarios, independent comparisons, and relevant real-world data can contribute to this. The duration and scope should be defined based on actual evidence requirements, rather than adopting an arbitrary standard period.

How do we avoid excessive irrelevant exceptions?

Analyze the root cause and improve rules or displays based on technical criteria. After modifications, continue to review known critical cases. A simple reduction in reporting volume may lead to the loss of important detection capabilities. The objective is to achieve a practical selection with proven coverage, not a minimally sized list.

What information does a reviewer need in an exception list?

It requires the relevant process, pertinent data, regulatory basis, and appropriate access to context. A technical error code alone is insufficient. The information should enable a justified decision or escalation. The scope and level of protection shall be determined by the task and the actual process.

How do we verify that an empty report is indeed meaningful?

Use controlled reference data with known exceptions and verify source, time period, filters, job status, and authorizations. A valid blank report must be distinguishable from an incomplete or failed run. Conduct manipulative test cases in an appropriate test environment, not uncontrolled at production records.

Does a detected control event justify complete report coverage?

No. It first demonstrates that a specific test pathway functions. Additional rules, data sources, and populations may contain other errors. Combine positive controls with appropriate completeness and rule tests. The review decision must reflect the actually demonstrated scope and remaining limitations.

Sources and review date

The sources provide the technical foundation. Tables, checklists, and examples are derived methodological tools, not verbatim regulatory requirements. Manufacturer information must be referenced to the specific version and configuration in use.

  1. European Commission: EU GMP Annex 11, Computerised Systems (2011)Current version as per EudraLex inventory as of 14.10.2026. Application, risks, operation, and data lifecycle; specific testing proposals are derived independently.
  2. PIC/S: PI 041-1, Good Practices for Data Management and Integrity (2021)PI 041-1, 1 July 2021: Section 5.6.2 with footnote 4 relating to the validated Exception Report, Section 9.6 regarding risk-based Audit Trail review, and Section 9.8 concerning data review. The specific rule matrix in the article is a separate work aid.
  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

From an open point to a traceable approach.

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.

  • Computer system validation →

    Bring your GxP systems into validated operation, from laboratory software, MES and eQMS to ERP. We define the scope, plan risk-based tests using GAMP 5 and hand the evidence to QA for release.

Send a project enquiry Customer testimonials →