Table of contents
At a glance
How do you justify the scope of regression testing?
Derive regression tests from the change, intended use, potential failure consequences and dependencies. Include unchanged functions that the change could affect. Existing evidence may remain applicable where its relevance to the new version is justified. A fixed test percentage, universal full retest or GAMP category alone does not justify the scope.
- The modified code and the affected technical process are not equivalent.
- Functions that have not been revalidated require a clear and traceable demarcation.
- A single test performed demonstrates its testing scope; it does not automatically validate the entire change.
1. Regression addresses a different question than the change study.
The change test verifies whether the intended new function operates as expected. A regression test assesses whether existing functionality has been adversely affected by the change. Both tests may be combined within the same test case, but must be clearly identifiable in the test coverage decision.
For GMP in human medicines, Annex 11 addresses validation and controlled changes. ICH Q9(R1) provides a risk management framework and GAMP supports a risk-based life cycle approach. The publicly available ISPE material cited here neither reproduces the full paid guide nor prescribes a particular testing technique.
2. From the technical scope of changes to the technical data pathway
A reusable import module can be utilized across multiple workflows. A modified library can impact functionalities whose underlying code remains unchanged. Conversely, a cleanly isolated function may exist outside the actual scope of influence. Therefore, the statement "This form has not been modified" still does not constitute a sufficient boundary definition.
Capture the relevant dependencies from input data, configuration, and processing down to use in the process. This includes, as applicable, roles, interfaces, and error handling. The following dedicated matrix enables verifiable justification for the selection; it does not prescribe a universal document format.
| Area | Potential impact | Test or release decision |
|---|---|---|
| Altered number import | Decimal separator misinterpreted | Test targeted new and previously authorized formats |
| Unchanged calculation | Incorrect input value generates incorrect result | Verify the technical calculation pathway with imported value |
| Unaltered report | Other formatting or incorrect data version | Verify relevant result presentation |
| Error handling | Invalid value is set to zero | Investigate negative case with defined reaction |
| Access rights | Change affects the common workflow | Clarify dependencies; where appropriate, perform testing of relevant controls |
| Separate shipping process | No common component or data pathway has been demonstrated. | Justify non-repetition with specific boundaries defined |
3. Uncertainty is a selection rationale, not evidence of low risk
Assess potential consequences of failures and the reliability of your information. The absence of a reported failure does not automatically imply that an unknown common component is non-critical. Lack of visibility may necessitate additional clarification or appropriate testing.
ICH Q9(R1) links the appropriate level of formality to uncertainty, importance and complexity, among other factors. Limited resources alone cannot justify a low risk rating. Identify the actual gap in the evidence and address it in proportion to the uncertainty.
4. Example: The value 1.25 must not be converted to 125
An import function changes. Its specification accepts the literal strings “1,25” and “1.25” in their respective input formats. The following process step multiplies the parsed value by 80. The expected result is 100. If the import removes the comma and reads 125, it produces 10,000.
The error response defined below is part of this specified expected behavior: A blank or non-numeric mandatory field entry is rejected and does not generate a technically valid result. This is not a generic product function or universal default applicable to all import processes. The test must demonstrate actual system behavior and the required data linkage.
| Test case | Input and Prerequisites | Expected result | Purpose of the evidence |
|---|---|---|---|
| New approved format | Literal “1,25” in the decimal-comma format, factor 80. | 1.25 × 80 = 100 | Intended Change |
| Current allowable format | Literal “1.25” in the decimal-point format, factor 80. | 1.25 × 80 = 100 | Existing behavior remains appropriate |
| Unaltered sequential function | Imported value 1.25; calculation unchanged | Report shows 100 with correct record referencing | Technical dependence verified |
| Missing mandatory entry | No numerical value | Rejection; no usable result | No silent conversion to 0 |
| Non-numeric mandatory statement | abc | Rejection; no usable result | Error handling |
| Identifiable error | Comma removed, resulting value 125. | 10,000 instead of 100 | The test must detect deviations. |
5. Purposefully reuse existing documentation
Not every existing validation evidence automatically loses its validity upon a change. However, it is essential to verify that the version, configuration, environment, and specific application context remain appropriate. A manufacturer's report may cover relevant standard functions, but no assessment of customer-specific combinations should be assumed without thorough evaluation.
Document the supporting evidence for each record and identify where new evidence is required. A previously passed inspection does not prove that a newly affected function continues to operate as intended. Similarly, the age of a record alone is not sufficient justification for its complete replacement without a technical evaluation.
6. Tests to be performed such that their conclusions remain clear and interpretable
The current system status and the test data used must be clearly identifiable. A predicted result lacking independent technical justification may contain the same error as the application itself. Therefore, in the example, the calculation is specified independently and is not derived from the observed system value.
Public ISPE documentation outlines different testing approaches and critical thinking. This does not imply a lack of verifiable results. Automated and investigative methods may be appropriate when the purpose, actual execution, and relevant findings remain credible and traceable. The approach serves as a means of generating evidence, not as a substitute for it.
7. Joint decision on coverage and remaining uncertainty
Summarize the areas assessed, covered by appropriate existing evidence, or logically excluded based on rationale. Evaluate open findings in terms of their potential impact. A green import test shall not obscure an unclear reporting outcome.
The further use of the modified system is governed by the controlled change procedure and the relevant responsibilities. There is no universal signature template applicable to every change. The scope of review may also expand during execution if new dependencies or issues become apparent.
8. Checklist for a Justified Regression Test Selection
A concise and traceable justification for selection is more helpful than an extensive test folder lacking a clear link to the change. The scope of expertise must reflect the actual system, not an arbitrary percentage specification.
The following internal assistance document complements your change management procedure. It is not a fixed GAMP template and does not guarantee detection of every possible deviation.
- Changes and intended behavior clearly described.
- Common components, data pathways, and configuration considered.
- Impacted on unaltered functions has been assessed.
- Identified critical failure consequences and uncertainties.
- Further appropriate existing evidence assigned.
- New positive, boundary, and error cases appropriately selected.
- Specific areas not being repeated are clearly justified.
- Findings, remaining uncertainties, and further use determined.
Project experience and collaboration
Recheck critical dependencies after a change
With your team, we map a change to its affected functions and business dependencies. We examine which existing tests exercise those links and where a new case is required. A limited configuration change may need different priorities from a platform migration. Automation helps where data, expected results and the tested version are controlled. A large test count does not establish coverage. We document the selection so it can inform subsequent changes.
Frequently asked questions about GAMP 5 regression testing
Must the entire system be retested after each update?
Why do we test functions whose code remained unchanged?
Since input data, shared components, or configurations may have changed. In the example of this article, the calculation remains unchanged but still produces an incorrect result due to a wrongly imported value. Unchanged source code therefore does not constitute sufficient evidence of non-impact.
Is there a prescribed regression testing frequency?
No. Annex 11 does not specify a fixed percentage for regression testing. It requires controlled changes (No. 10) and a justified, documented risk assessment for the validation scope (No. 1). Ten percent of existing tests may miss important dependencies; many tests may merely repeat a narrow set of scenarios. Select based on technical coverage and risk.
Does the GAMP category alone determine the test scope?
No. It neither replaces the intended use nor the specific impact assessment. A generic category number does not indicate which data pathways and decisions your change affects. The public GAMP presentation emphasizes a risk-based lifecycle approach, not a universal test package per label.
Can previous tests continue to serve as valid evidence?
Yes, provided their suitability for the new environment remains justified. Verify the relevant version, configuration, and application scenario. Not every change invalidates all prior evidence; however, a newly impacted function cannot be adequately covered solely by an older, previously passed test.
Are the manufacturer's regression tests sufficient?
You can provide important evidence. Assess the actual scope and alignment with your intended use. Customer-specific configurations, interfaces, or processes may require additional validation evidence. A general reference to the vendor's package does not substitute for this mapping.
Sources: [1]
Can regression tests be automated?
Must all expected results be independently calculated?
You need a reliable technical basis for the purpose. For calculation functions, an independent calculation can be particularly helpful. However, the sources do not require to represent any expected behavior as a numerical calculation. It is important not to deduce the objective uncritically from the observed system behavior.
What do we do regarding unknown technical dependencies?
Treat them as uncertainties. Assess the architecture or select appropriate tests that cover the potential scope of impact. Missing information does not constitute evidence of low risk. At the same time, a blanket end-to-end test is not automatically the only appropriate response.
Sources: [2]
When is a non-reproducibility traceable?
When the area is clearly defined based on the actual scope of changes and appropriate evidence, identify the basis, for example, the absence of shared components and data pathways. "Not changed" or "previously error-free" alone is often insufficient for a credible impact assessment.
Sources and review date
Source date: 7 October 2026. The technical examples relate to human medicinal product GMP. The referenced EU-GMP/ICH sources and public ISPE materials are used. A comprehensive explanation of the full GAMP guideline is not provided. Tables and case studies are intended as working aids.
- European Commission: EU GMP Annex 11, Computerised SystemsRevision January 2011; from 30.06.2011; Principle; §§1,4,5,7,9–13,17; §12.4 Identity, Date and Time; §10 controlled changes
- EMA/ICH: Q9(R1), Quality Risk ManagementEMA/CHMP/ICH/24235/2006 Corr.2, published 31 January 2025. Q9(R1) has applied since 26 July 2023. Sections 3–4.6 and 5.1–5.3 address uncertainty, formality, the knowledge base and reassessment.
- ISPE: GAMP – Purpose and risk-based lifecycle approachOfficial Topic Overview, retrieved 07.10.2026; Introduction to GAMP guidance; Section on non-prescriptive method and GAMP 5 Second Edition
- ISPE: What You Need to Know About GAMP 5 Guide, 2nd EditionPharmaceutical Engineering, January/February 2023, Article pp. 33–36; Print page 35 / PDF page 37, Development Appendices: Testing guidance; Operation Appendices
- European Commission: EU GMP Chapter 1, Pharmaceutical Quality SystemRevision 2013; effective from 31.01.2013; §1.4(xii)–(xiv): proactive change evaluation, post-evaluation, and root cause analysis/CAPA; §1.5 Management
- European Commission: EU GMP Chapter 4, DocumentationJanuary 2011; in effect since 30.06.2011; Principle; §§4.1–4.9, in particular timely recording 4.8 and visible correction 4.9; §§4.10–4.12 Storage
Sources checked: .
Protect changes with appropriate test scope
We support you in translating technical changes into a traceable selection of regression tests and appropriate existing validation evidence.
- GAMP 5 consulting →
Understand how GAMP 5, EU GMP Annex 11 and 21 CFR Part 11 apply to your systems. We translate the relevant requirements into proportionate validation work and evidence your team can defend in an audit.

