Table of contents
At a glance
How do we assess assurance needs after a small configuration change with potentially serious consequences?
Define the changed behaviour and identify which earlier evidence it affects. Assess failure consequences, dependencies and existing controls. Retain evidence that remains applicable and test the gaps. A minor configuration change can have a substantial effect. FDA’s CSA guidance recommends a risk-based approach. It calls for neither an automatic full retest nor an exemption for small changes.
- Evaluate the impact of changes based on line count or ticket size.
- Consider affected existing functions and data states.
- Support decisions with appropriate evidence and a clear rationale.
1. Describe previous and new behavior
A technical change note such as "Operator adjusted" does not clarify which business decision has been modified. Therefore, describe the previous process, the desired new process, and the rationale for the change. Identify the affected users, data, and subsequent decisions.
For a small and medium-sized enterprise (SME), a brief comparison may suffice if it is precise. A detailed ticket description with numerous technical details provides little value if the impact on the process remains unclear. Specifically inquire whether a previous inspection is waived, a edge case is handled differently, or a manual decision is replaced by automated processing.
Sources: [1]
2. Separate Guidance, Requirements, and Reporting Questions
The FDA-CSA-Guidance of February 3, 2026 deals with production and QMS software in the medical device sector. Their recommendations support a risk-based approach within the relevant QMSR framework. They do not replace binding requirements and are not transferable to software as a medical product or pharmaceutical GMP applications.
Section V.A.(3) outlines reporting requirements for products with an approved PMA- or HDE-approval application. The FDA specifies a 30-day notification requirement for certain production and QMS software changes with predictable safety implications; for other changes, an annual report may be appropriate. However, the suitability of this approach depends on the nature of the change: 21 CFR 814.39 distinguishes between manufacturing process changes and modifications to design, performance, or manufacturing location. Therefore, reporting obligations should be evaluated separately from software assurance needs, and the 30-day deadline should not be considered a general requirement for all software changes.
3. Systematically evaluate dependencies
Our matrix focuses on statements that may become uncertain due to changes. A function that remains unchanged can be affected if it uses the same calculation or identical master data. Similarly, existing records may be presented or processed differently without new data being entered.
First, assess the plausible impact. If technical dependencies are unclear, identify this gap and obtain the necessary information. Unknown does not imply unimpacted. Conversely, every theoretically possible connection does not necessitate a full test of all functions.
| View direction | Question to ask | Possible evidence |
|---|---|---|
| Business rule | Which decision changes, and was it permitted or prohibited? | Targeted Boundary and Error Scenarios |
| Common Logic | Which additional functions utilize the same building block? | Rationale selection of affected regression tests |
| Existing data | Are previous states re-evaluated or overridden? | Comparison of representative existing records |
| Interface | Do format, meaning, or order change? | Review transition and error handling |
| Operation | Do rights, work processes, or return pathway change? | Assess availability and handling |
Sources: [1]
4. Example: a modified tolerance rule
A manufacturer uses a software function that compares a measured distance against a predefined tolerance. For this example, the technical requirement is that values from 9.5 to 10.5 mm, inclusive of both limits, are acceptable. An unintended configuration change inadvertently replaces "less than or equal to" with "less than," resulting in the upper limit being incorrectly interpreted.
The following target matrix consciously uses values below, on and above the limits. The figures are sample assumptions and not a specification of a real medical device. In the actual project, method, measurement resolution and technical acceptance criteria must be established before testing expectations are derived from this.
| Test result | Expected decision in the process | Significance for changes | Additional note |
|---|---|---|---|
| 9.4 mm | Not permitted | Lower outer boundary remains detected | No requirement for silent rounding to 9.5 |
| 9.5 mm | Permissible | Lower limit inclusive | Check configured comparison rule |
| 10.0 mm | Permissible | Value within the range | Not enough as a limit test |
| 10.5 mm | Permissible | The upper limit must remain inclusive. | The modified operator would violate this expectation |
| 10.6 mm | Not permitted | Upper outer limit remains detected | Also detect erroneous expansion too |
Sources: [1]
5. Select additional tests with justified rationale
The threshold matrix evaluates the specific rule but does not yet address every possible use of its results. Query whether a subsequent process accesses the status, whether an export displays the same value, and whether previously stored results are recalculated. Select the relevant additional cases from this list.
Existing relevant evidence may continue to be valid. If, for example, an independent user management remains unchanged and no verifiable impact is demonstrated, a tolerance change does not need to automatically trigger all authorization tests. Document the rationale. What is essential is coverage of the actual impact, not the execution of a maximally large or minimally large number of additional tests.
Sources: [1]
6. Treat unexpected effects as findings
If the review indicates that 10.5 mm is deemed unacceptable, the deviation from the specification must first be documented. The root cause could lie in configuration, implementation, test data, or a misinterpretation of a technical requirement. Do not retrospectively alter the expectation merely to achieve a successful run.
Additionally, assess whether the same rule has been previously applied and whether any prior results may be affected. Correcting the operator does not automatically address this question. The closure should link the root cause, the change, appropriate retesting, and an evaluation of potential prior impacts. Such an assessment should be proportionate to the specific significance.
Sources: [1]
7. Link audit evidence to actual product availability
An appropriate test in a validation environment does not by itself demonstrate that the exact same configuration was delivered. Assign the tested state to the actual change implemented. Consider which parameters, data, or authorizations are relevant during transfer and how a misconfigured delivery is detected.
In addition, define the approach to handling issues following a change. While a system failure may technically occur, previously processed data or decisions cannot be automatically reversed. Therefore, the usage decision should take into account the actual transition scenario. The individuals responsible shall be determined based on established authorisation and competence; a generic role hierarchy is not prescribed.
8. Checklist for a justified change approval
The closure should demonstrate which function was modified, which previous statements remain valid, and which additional evidence has been generated. A ticket status of "completed" is only sufficient as part of the documentation when the key information remains properly verified and linked.
The following internal checklist can complement an existing change management procedure. It avoids both undue leniency toward minor changes and an automatic full retest. Both extremes may obscure the central question: Is the modified version adequately protected for its intended use?
- Previous and new behavior described technically.
- Product and software scope, as well as separate reporting questions, have been categorized.
- Evaluated potential failure consequences and critical controls.
- Taking dependencies and existing data into account.
- Accepted evidence and additional assessments justified.
- Investigations including possible effects from prior occurrences addressed.
- Associated with the approved and provided system, together with the user decision.
Sources: [1]
Project experience and collaboration
Base the scope of testing on the actual change
For a planned change, we compare old and new behaviour with the system team. Process owners identify affected decisions, IT traces dependencies and Quality assesses failure consequences. A single well-defined case may warrant focused evidence. Changes to further functions or assumptions require broader testing. Any regulatory reporting obligations are assessed separately. A successful software test does not resolve that question.
Frequently asked questions about Assessing software changes under FDA CSA
Is a one-line change automatically considered trivial?
No. A single operator can alter a significant decision. Assess the involved behavior and its potential consequences. The technical magnitude of the change may assist in planning, but it does not replace a professional evaluation of impact.
Sources: [1]
Do we need to perform a complete revalidation after every update?
Not at all. Check which previous statements are affected and which proofs continue to bear. Additional activities should close the relevant gaps. Unclear dependencies must be clarified instead of automatically treating them as non-following or as an occasion for all tests.
Sources: [1]
Are configuration changes less significant than program code?
Not fundamentally. Configuration can control calculations, authorizations, or automated decisions. Its impact can therefore be significant. Assess the specific function and intended use; the nature of the technical change alone does not determine the assurance requirement.
Sources: [1]
Why are values midway within the acceptable range insufficient?
They do not demonstrate whether the limits are correctly handled. In the example from this article, a value of 10.0 mm would also be accepted under a faulty modification. Only the value of 10.5 mm clearly distinguishes between an inclusive and an exclusive upper limit.
Sources: [1]
Can we change the expectation following a failed test?
Yes, if the original test expectation was demonstrably wrong or if there is a technically justified change of requirement. Keep original target, finding, correction reason and necessary re-examination comprehensible. Adjusting the expectation only to the observed result does not produce suitable proof.
Sources: [1]
What happens to existing data following a regulatory amendment?
Check whether they remain unchanged, are presented anew, or are reprocessed. This effect may influence the evaluation of prior results. Successful verification of new entries does not automatically address questions regarding existing records.
Sources: [1]
Must the supplier provide all necessary documentation?
Supplier information can significantly contribute. Verify their actual coverage for your configuration and usage. Remaining internal process or transition questions must be resolved. A general release notice does not automatically replace a specific impact assessment.
Sources: [1]
Is there a general 30-day deadline for software modifications?
No. The CSA Guidance addresses a specific reporting procedure for medical devices with an approved PMA or HDE application. For certain changes to manufacturing processes, a 30-day notification under 21 CFR 814.39(f) may apply; the Guidance includes within this category changes to production and QMS software with predictable safety implications. For other changes, an annual report may suffice. This is not a general deadline for software tickets. Whether this procedure applies or whether an alternative PMA supplement is required must be evaluated based on the nature of the change and the approval pathway.
Is a technically successful rollback sufficient?
It may form part of a remedial action plan, but does not automatically address the consequences of already processed data or decisions. Assess the status following data retrieval and, if necessary, implement corrective actions. Impact on the process is equally important as technical recovery.
Sources: [1]
How do small teams justify their decisions effectively?
Use a compact mapping of change, effect, existing evidence, additional tests, and conclusion. Reference retained details. This ensures decision-making is traceable, without repeating the same information across multiple disjointed reports.
Sources: [1]
Sources and review date
Source date: 7 October 2026. Regulatory requirements, guidelines, and industry practices must be categorized separately. Tables, figures, and case examples are to be considered standalone reference materials; they do not constitute completed product testing.
- FDA: Computer Software Assurance for Production and Quality Management System SoftwareFinal guidance of 3 February 2026. Non-binding recommendations for production and quality management system software in the medical device sector.
- 21 CFR Part 820 – Quality Management System RegulationQMSR, effective since 2 February 2026: §§ 820.1, 820.7 and 820.10. The CSA-Guidance refers for software validation to the included ISO-13485:2016 subsections 4.1.6, 7.5.6 and 7.6 as well as for records on 4.2.5; § 820.35 contains additional recording requirements. The ISO-Verweiseen are documented here via the FDA-Guidance, no independent full text comparison of the ISO standard.
- eCFR: 21 CFR 814.39 – PMA supplementsAs of October 18, 2026: Review sections (a), (b), and (f); delineation of PMA post-approval changes, annual reports, and 30-day notifications for specific process change scenarios. No general software change period.
Sources checked: .
Jointly develop the specific evidence process
We support you in identifying a specific software modification, selecting appropriate validation evidence, and documenting the user decision in a traceable and justifiable manner.
- 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.

