Practical guides · Practical guide

How to close gaps in the validation evidence for an existing system

Your system has run for years, but the current configuration has no coherent set of validation records. Old tests, supplier evidence and operating experience may still be relevant. Establish what they demonstrate, what remains unknown and where the gap affects a GMP process. QA, IT and the system owner can then prioritise the next checks.

Table of contents

At a glance

How can you validate an existing system with gaps in its evidence?

Document the current intended use, system version and data flow. Assess the available evidence against required functions and controls. Prioritise gaps by process risk and uncertainty, addressing critical workflows first. Add traceable tests dated when performed and explicitly decide which uses are acceptable. Never invent historical tests or approvals.

  • Evaluate missing documents and actually inappropriate controls separately.
  • Utilize existing evidence with appropriate qualification and clearly document its limitations.
  • New tests demonstrate the validated state today, not automatically the past.
  • Operational decisions, actions, and effectiveness controls require clearly defined responsibilities.

What "retroactively validate" means for an existing system

The EMA explicitly addresses existing systems in its publicly available Annex-11 question 6: Defining user requirements, analyzing gaps and verifying the requirements. This is a professional inventory assessment. It is not a permission to issue today's documents as supposed original documents of an earlier introduction.

Distinguish three questions: Is the current process appropriate? What evidence supports this assertion? What uncertainty exists regarding prior data and decisions? A successfully validated calculation method today can address the first question. It does not prove that the same algorithm with identical settings was used five years ago.

The guide concerns computerised processes in GMP manufacturing of human medicines. The EU index lists the January 2011 edition of Annex 11. Drafts are treated as proposals. Medical devices, clinical trials and other GxP areas need their own scope assessment. Calling a system “legacy” creates no general exemption.

Sources: [3] [2] [1]

Capture the actual state prior to documents being amended

Begin with a jointly observed process run. Demonstrate where data is generated, who makes modifications, what results are forwarded, and on what basis a quality decision is made. Capture exports, manual transfers, and local files. Even a seemingly minor auxiliary program can provide critical value to a larger process.

The technical state includes application version, relevant configuration, database, interfaces, and operating environment. The functional state includes use cases, user roles, work instructions, and key documentation. Integrate both: Which configured limit applies in which process, and who is authorized to modify it?

Preserve the existing state prior to making configuration changes. A dated configuration export or a documented as-is assessment can explain the state that was evaluated. A statement from a long-term employee may assist in identifying evidence, but is not automatically substitute for a missing test. Distinguish between reminder, directly observed state, and documented history.

Sources: [2] [3] [4]

Utilize existing evidence without extending their scope or interpretation

Review supplier assessments, installation protocols, acceptance certificates, maintenance records, change logs, failure reports, and training documents. For each, verify origin, date, subject, result, and alignment with the current system. A document may be fully acceptable, partially acceptable, or inappropriate for the specific inquiry.

A comprehensive supplier package may also include specific configurations and application-specific workflows. Do not reduce it casually to general development validation evidence. Conversely, the heading "PQ" alone provides little indication whether your interface and roles have been assessed. What matters most are the actual steps performed and the documented prerequisites.

The following workflow guide outlines typical evaluation decisions. It is not a prescribed document list. For each acceptance, briefly note which requirement is addressed and which limitations remain.

Support document: Validity of existing evidence
Existing evidenceWhat it can supportWhat else needs to be clarified
Age test with test data and result Behaviour of the previously tested function. Version, configuration and relevant changes since the review.
Supplier package Proven functions and, if applicable, agreed customer configuration. Extent of testing, requirements and customer requirements not covered.
Long-term operational history Notes on reliability, failures, and actual usage. Were any errors identified and fully documented?
Today's configuration export Current technical starting point. No automated proof of prior settings or releases.
Backup protocol Execution of a backup. Readability, completeness, and demonstrable recoverability.

Sources: [2] [3]

Identify and prioritize documentation gaps and process risks separately

A missing title page shall be evaluated differently from an unprotected calculation formula. Both may trigger a review requirement, but their immediate consequences differ. Therefore, describe each gap specifically: affected use case, missing information or ineffective control, potential impact, and existing safeguards.

Please ask the following: What decision depends on this? How could an error arise? Would it have been detected before relevant use? What uncertainty remains? "Nothing has happened so far" is insufficient if a misinterpretation had gone undetected for years.

Use traceable priorities instead of an apparently precise total score. A critical gap in evidence should not be eliminated by fulfilling numerous, trivial criteria. ICH Q9(R1) clearly states that risk management does not justify otherwise inappropriate practices. The prioritization used in this article serves to establish the order and depth of processing, not as a general risk release.

Sources: [5] [3]

Secure critical processes and openly define boundaries

For an existing risk, the protective measure is implemented prior to the cosmetic completion of the folder. Potential approaches include restricted change rights, a secured data pathway, appropriate independent monitoring, or temporary suspension of the affected function. Which measure is appropriate must be evaluated for the specific error pathway.

A second party may, for example, verify a transfer against a reliable source. However, it cannot restore deleted original data or reliably detect it, which is missing in an incomplete report. Therefore, for each transfer control, identify the demonstrable benefit and the remaining blind spot.

PIC/S specifies additional controls for existing systems but requires their suitability; the guideline itself is not a standalone legal requirement. Maintain records of responsible parties, validity period, monitoring procedures, and exit criteria for the measure. If a required condition cannot be reliably achieved with available controls, the affected usage shall be restricted or suspended. A signed risk assessment renders an inappropriate control ineffective.

Sources: [6] [5]

New tests to be planned and executed under control

Where possible, conduct testing in an appropriate test environment. Document relevant differences from the production system, such as interfaces, data volumes, or authorizations. A test performed in an empty standard installation does not automatically address questions regarding your evolved configuration.

If a test during production is required, predefine scope, test data labeling, release criteria, backup procedures, rollback strategy, and termination conditions. Ensure that test data does not enter real reports, interfaces, or release decisions. Performance or deletion tests using actual production data are not a casual data review.

Record the execution date, environment, starting condition, expected result, actual result, and deviations. The following matrix is a standalone planning aid. It does not replace a system-specific inspection procedure.

Example: Validation planning for an already deployed system
Area to assessControlled TestResult and Limit
Calculation Check known inputs and boundary cases against independent expectations. The validated calculation is documented and maintained in the current approved version.
Permission A test user without change authorization attempts to modify a configuration. Unauthorized action prevented; authorized pathway separately verified.
Interface Transfer of a unique test dataset including an invalid mandatory field. Assignment and error handling comprehensible; no real orders affected.
Recovery Restore and read a verified state in an isolated environment. Relevant data and context available; production environment remains protected.
Cancellation Stop testing when an unexpected impact on real processes is detected. Capture the event, assess the impact, and only proceed after a decision has been made.

Sources: [2] [4]

Example: Outdated laboratory software with unclear validation status

A laboratory software takes over measured values and calculates a result report. A supplier protocol is available for an earlier version. The current report structure was changed later, and several employees have far-reaching rights. Whether original data can be modified afterwards is not yet clear.

The team initiates three separate tasks: first, to identify the data flow and actual user rights; second, to review and validate the current reporting calculation; and third, to determine which prior results may be affected by the reporting change. An up-to-date calculation test does not address the third task.

The test date uses 48.0 units with a specified factor of 2.5. The expected value is 120.0 units. Additionally, an empty factor is verified: the predefined error handling is expected, implicitly resulting in a value of zero. These numbers illustrate only the test logic and are not an analytical method.

The compliance review may require actions prior to the completion of parallel document assessment. A reliable previous test is accepted if its scope of validity remains applicable. The new test receives its actual execution date. Missing historical information remains designated as uncertainty; it is not replaced with a newly written success history.

Sources: [3] [2] [4]

Who determines the use and prior impacts?

The decision falls within the existing quality process. Process owners assess the technical and scientific justification, system owners and IT evaluate the technical feasibility, and QA is responsible for the evidence of quality impact and relevant consequences. The specific decision authorities required must be clearly defined within your organization; they do not automatically follow from the project lead role.

Document the decision in such a way that it remains understandable in the future: approved scope, limitations, rationale, open issues, deadlines, and reasons for a subsequent reassessment. Limited usage must not be confused with full approval of the entire system. If new information becomes available or if a control measure is not effective, the decision must be re-evaluated.

When erroneous or incomplete quality-related data are identified, the extent of the issue and its potential impact on prior decisions must be investigated. A CSV project does not substitute this investigation. For FDA-regulated pharmaceutical processes, the Data Integrity Guidance clearly emphasizes the need to assess the scope, root causes, and implications. Whether additional product actions or communications are required is determined by the relevant responsible functions based on the specific circumstances.

Sources: [7] [6] [8]

Develop a comprehensive action plan from the gap list

Assign a responsible party and a verifiable closure criterion to each gap. "Improve documentation" is too vague. A more specific example would be: current interface status described, mandatory fields assignment confirmed, error cases reviewed and outcomes evaluated. This enables determination of whether the specific issue has been adequately resolved.

Separate technical qualification from application validation. The presence of a proven server or a functional backup does not validate the intended business process. Conversely, a successful end-user run does not substitute for necessary evidence regarding the supporting infrastructure. Both areas may leverage existing evidence and must be logically aligned.

Integrate permanent changes into the Change Control process. After completion, verify that the measures have a tangible impact in daily operations and that new changes are captured. Periodic evaluations should not repeatedly uncover unknown states. If replacement or decommissioning represents a viable solution, plan for data access and retention; the implementation of a new system does not eliminate the need for historical traceability.

Sources: [2] [7]

Checklist for the first joint inventory review

For the first meeting, you do not need a newly formatted validation folder. A clear process flow, an accurate current system status, and organized access to existing documentation are more helpful. The following checklist serves as preparation, not as a general release form.

  • GMP usage, data pathways, and relevant quality decisions are identified.
  • Applications, configuration, interfaces, and infrastructure have been identified.
  • Existing records are documented with origin, date, and scope of validity.
  • Documentation gaps, technical deficiencies, and historical uncertainties are addressed separately.
  • Immediate risks and appropriate control measures have been assessed.
  • New tests have specified acceptance criteria, appropriate environment, and a controlled procedure.
  • Impacts on prior data will be investigated separately upon receipt of relevant indications.
  • Decision authority, usage limits, and escalation procedures have been defined.
  • Measures have designated responsible parties, defined timelines, and verifiable completion criteria.
  • Efficacy, modifications, and long-term data access are embedded in operations.

Sources: [2] [3]

Project experience and collaboration

Connect the current system version to reliable evidence

Start with what the existing system is used for today. Users show us a relevant workflow, while IT and QA bring the available documents and change history. Together, we distinguish a gap in documentation from a control that does not work. Suitable existing evidence may be reused, with targeted testing for specific gaps. If a critical process lacks a sufficient basis, we assess its current state more fully and record the decision about permitted use. We do not present a test as having taken place in the past without evidence.

Frequently asked questions about Validation gaps in existing systems

Can an existing GMP system be validated retrospectively?

For existing systems, the EMA describes the definition of user requirements, a gap analysis and the verification of these requirements. Practically evaluate the actual use and supplement missing evidence. This is to be distinguished from an invented original validation. Document what has been tested today, which historical evidence can be used and which statements about the past can still not be substantiated.

Sources: [3] [4]

Must we shut down a system without complete folder structure immediately?

The document record alone does not address the issue. First, assess the intended process, existing controls, and potential consequences. A mere gap in evidence must be handled differently than an ongoing, uncontrolled risk to data or product quality. If a necessary control cannot be effectively implemented, the affected usage must be restricted or suspended. Responsible functions shall make decisions based on documented facts, not on the age of the software.

Sources: [3] [5] [6]

Which existing documents may we continue to use?

Available documentation must have verifiable origin, scope of review, and clear relevance to the system under consideration. Specifically assess version, configuration, and subsequent modifications. An older test may adequately demonstrate unchanged functionality, while a new interface requires additional validation. Document the decision to adopt, including the defined boundary; a blanket rejection of all prior documentation is no more helpful than unreviewed acceptance of such documentation.

Sources: [2] [3]

Does years of uninterrupted operation constitute valid evidence?

Operational experience is relevant information but not universal proof. What matters is whether errors would have been detected and how comprehensively incidents were documented. An undetected calculation error can persist for a long time. Leverage operational history together with requirements, technical inspections, and targeted audits. Do not interpret "no known incidents" as "demonstrably error-free."

Sources: [2] [3]

Can we create a missing URS today?

Yes, you can describe and approve the current user requirements today. Indicate the actual creation date and the assessed usage. Distinguish required behavior from merely observed as-is conditions; otherwise, every existing defect would be attributed to the requirement. Previous requirements may only be designated as historically documented if appropriate sources are available.

Sources: [3] [4]

Can a second person compensate for missing technical controls?

Only if the additional control effectively covers the specific error pathway and its implementation is demonstrated. A cross-check against a reliable source can be helpful. However, it does not automatically detect missing original data or unknown manipulations. Assess suitability, limitations, and new risks; document monitoring and, if applicable, the end of a transitional solution. "Four eyes" is not a general substitute for validation evidence.

Sources: [6] [5]

How do we ensure testing when no test system is available?

First, evaluate appropriate alternatives, such as an isolated copy representing a current state. If a production test is required, pre-defined test data, impact assessments, rights, rollback procedures, and termination protocols must be established beforehand. Exclude uncontrolled influences on actual records and decisions. If such control is not feasible for a test, the test must not be performed directly in ongoing operations.

Sources: [2] [5]

Does a currently performed test also validate the previous results?

No. It only confirms the documented and verified current state. For any statement regarding prior results, a traceable link is required—such as unchanged versions, configurations, and data pathways. In cases indicating a previous system failure, a separate impact assessment is necessary. Missing historical information remains a justified uncertainty and is not eliminated retroactively by a current test result.

Sources: [4] [6] [8]

Do we need a new IQ, OQ, and PQ for the entire system?

Not only because of these three document titles. Determine which requirements and controls to prove and evaluate existing evidence. The technical infrastructure and the application have different test objects. Complete the actually missing evidence in a comprehensible structure. Which document names your quality process uses is less decisive than coverage, result and evaluated deviations.

Sources: [2] [3]

When is the resolution of validation gaps completed?

When the defined scope is demonstrably evaluated, necessary controls are effectively implemented, and required evidence is accepted. Open items must not simply disappear from the list; remaining restrictions and follow-up actions must be explicitly decided. Additionally, verify that Change Control and periodic assessments maintain the achieved state. A new folder with multiple signatures is not a credible closure criterion without this linkage.

Sources: [2] [7]

Sources and review date

Primary sources were opened on October 6, 2026. The focus is on human medicinal product GMP. PIC/S and FDA guidelines are classified according to their respective status and scope of application. Tables, checklists, and laboratory examples are standalone work aids.

  1. European Commission: EudraLex Volume 4 – EU Guidelines for Good Manufacturing Practice
  2. European Commission: EU GMP Annex 11 – Computerised Systems
  3. EMA: GMP/GDP Questions and Answers – Annex 11 and Data Integrity
  4. European Commission: EU GMP Chapter 4 – Documentation
  5. ICH Q9(R1): Quality Risk Management – FDA Final Guidance
  6. PIC/S PI 041-1: Good Practices for Data Management and Integrity in Regulated GMP/GDP Environments
  7. European Commission: EU GMP Chapter 1 – Pharmaceutical Quality System
  8. FDA: Data Integrity and Compliance With Drug CGMP – Questions and Answers

Sources checked: .

Daniel Herrmann, Daniel Herrmann Consulting

Daniel Herrmann Consulting

Clarify the actual validation requirements of your existing system

Daniel Herrmann Consulting supports you with inventory assessment, prioritization of specific risks, and planning of traceable inspections. Existing evidence is strategically incorporated.

  • 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 →