Guide

Computer system validation (CSV) in pharma

Computer system validation (CSV) is the documented evidence that a computer system used in a GxP process does what it's intended to do, consistently, and keeps its data accurate and secure for as long as that data matters. In pharma it covers everything from an MES and a LIMS to an eQMS, an ERP batch release step and a spreadsheet that calculates assay results. If a system creates, changes, stores or uses GxP data, or controls a GxP process, it falls in scope.

Last checked: 12 November 2026 · By Sachin Bhandari

I've spent more than 25 years in pharma quality IT, and the best definition I've found in practice is the one an inspector applies: can you show me, from your records, that this system is fit for what you use it for, and that it has stayed that way since you went live?

Why regulators require it

Medicines are made, tested and released on the strength of data, and most of that data now lives in software. If a LIMS miscalculates a result, or an MES lets an operator skip a step without a record, a batch can go out that shouldn't. If an audit trail can be switched off, nobody can show what really happened.

Validation shows 3 things. The system works for its intended use before patients depend on it. Data integrity controls are in place, so records can be trusted. And the system stays in a validated state through changes, upgrades and supplier releases.

The regulations and guidance

21 CFR Part 11

FDA's rule on electronic records and electronic signatures, published in March 1997. It requires validation of systems that hold regulated electronic records, secure computer-generated time-stamped audit trails, access controls, and electronic signatures linked to their records. FDA's September 2003 guidance on Part 11 scope and application set out a narrower reading of the rule and a risk-based approach to validation and audit trails. See 21 CFR Part 11 in the glossary.

21 CFR 211.68

Part of the US drug CGMP regulations. It requires that automatic, mechanical and electronic equipment, including computers, is routinely calibrated, inspected or checked under a written programme, that only authorised people change master production and control records, that inputs and outputs are checked for accuracy, and that a backup of entered data is kept. This is the core US basis for validating computer systems in drug manufacturing.

EU GMP Annex 11

The European annex on computer systems, part of EudraLex Volume 4, in its current form since 2011 and mirrored in the PIC/S GMP guide. It covers risk management, supplier assessment, validation, data, audit trails, change control, periodic review, security and business continuity. A much longer revised draft went to public consultation from 7 July to 7 October 2025, alongside a revised Chapter 4 on documentation and a new draft Annex 22 on artificial intelligence. On 12 November 2026 these are still drafts and the 2011 Annex 11 remains in force.

GAMP 5 Second Edition

ISPE's GAMP 5 guide, Second Edition, published in July 2022. It's industry guidance, and it's the practical reference most pharma companies build their CSV procedures on. The Second Edition puts critical thinking at the centre, discusses FDA's computer software assurance approach, supports iterative and incremental development, and adds appendices on cloud and on AI and machine learning. I was one of its contributors.

PIC/S guidance

PIC/S PI 011-3, "Good Practices for Computerised Systems in Regulated GxP Environments" (25 September 2007), is still a widely used inspector reference. PIC/S PI 041-1, "Good Practices for Data Management and Integrity in Regulated GMP/GDP Environments" (1 July 2021), sets out what inspectors expect on data integrity, including for computer systems.

FDA computer software assurance guidance

FDA's CDRH and CBER issued "Computer Software Assurance for Production and Quality System Software" as a draft in September 2022. It was finalised September 2025, updated February 2026 to align with the QMSR, and is now titled "Computer Software Assurance for Production and Quality Management System Software". Its stated scope is medical device production and quality management system software, and pharma applies the same principles through GAMP 5 Second Edition. The guidance contains no percentage targets for effort reduction.

The CSV lifecycle and its deliverables

GAMP 5 describes a lifecycle from concept to retirement. Here's how it runs in practice, with the documents you'd usually produce. Scale them to the system's risk and complexity.

PhaseWhat happensTypical deliverables
Concept and planningDecide the system is needed and define its GxP impactGxP assessment, validation plan
RequirementsDefine intended use and what the system must doUser requirements specification (URS)
Supplier selectionAssess the supplier's quality system and evidenceSupplier assessment or audit report
Specification and configurationDefine how the system will meet the URSFunctional and configuration specifications, design specification for custom code
Risk assessmentIdentify functions that affect patient safety, product quality and data integrityRisk assessment, including Part 11 and data integrity risks
VerificationTest that the system works as specified, in proportion to riskIQ, OQ and PQ protocols or test records, traceability matrix
Reporting and releaseConfirm the system is fit for intended useValidation summary report, release approval
OperationKeep it in a validated stateChange control records, incident records, access reviews, audit trail reviews, backup and restore checks, periodic review
RetirementRetire the system and keep the data availableRetirement plan, data migration or archive verification
PlanGxP assessment, validation planURSintended useFunctional andconfig specsDesign spec(custom code)PQworks in real useOQworks as specifiedIQinstalled rightConfigure, buildA traceability matrix links eachrequirement to its risk and its testReport and releasevalidation summary reportOperation (the longest phase)Change control, incidents, access andaudit trail reviews, backup and restoretests, periodic reviewRetirementarchive or migrate, verify the data
Figure 1. The lifecycle drawn as a V. Each specification on the left is verified by the test level opposite it, and operation is where the system spends most of its life.

The traceability matrix links each requirement to its risk and its test. It's the document inspectors use most to check whether your testing covered what mattered. IQ, OQ and PQ are explained in the glossary.

GAMP software categories

GAMP sorts software into categories that guide how much validation effort a system needs.

Category 5Custom applicationscode written for you, macros,complex spreadsheetsCategory 4Configured productseQMS, LIMS, ERP, MESCategory 3Non-configured productsused as suppliedCategory 1Infrastructure softwareoperating systems, databasesMorelifecycle controlMoresupplier reliance
Figure 2. Lifecycle control and supplier reliance move in opposite directions as you climb the categories.

GAMP 5 has no Category 2. Real systems often mix categories, so assess the parts and avoid labelling a whole system with one number.

Who does what

The process owner owns the business process the system supports and answers for intended use. The system owner owns the system's availability, support and maintenance, usually from IT. Quality assurance approves the validation approach, the key deliverables and the release, and stays independent of the delivery team. Subject matter experts write requirements, assess risk and run tests. The supplier provides the product, its documentation and often test evidence you can use once you've done a supplier assessment.

Write these roles into your validation plan by name. When an inspector asks who decided a function was low risk, you want a name and a record.

Common inspection findings

Across 300+ inspections, zero critical findings, the same themes have come up again and again, and they match what you read in published observations and warning letters across the industry:

Most of these are about operation, after go-live. A clean validation package doesn't protect you if the system drifts in use.

Where CSA fits

Computer software assurance is the risk-based way of doing CSV. You still validate for intended use. What changes is the depth: you put scripted testing and full records on high-risk functions, and use lighter assurance, such as unscripted testing and system-generated evidence, on the rest. GAMP 5 Second Edition supports this approach for pharma. My method for moving a team from traditional CSV to CSA is called DRIVE. The comparison is on CSV vs CSA, the step-by-step move is in how to move from CSV to CSA, and my computer system validation consulting page covers how I help with a specific system or programme.

Where AI fits

AI and machine learning systems are computer systems, so the same principles apply: intended use, risk, evidence and control in operation. They add problems traditional CSV wasn't built for: the quality and independence of training and test data, acceptance criteria for model performance, explainability, and drift over time.

The draft EU GMP Annex 22 is the first GMP text written for AI. In its consultation form it applies to AI models used in critical GMP applications, meaning those with direct impact on patient safety, product quality or data integrity, and it covers only static models with deterministic output. It says dynamic models, probabilistic models and generative AI should not be used in critical GMP applications. It covers intended use, acceptance criteria, test data and its independence, test execution, explainability, confidence and operation. It's a draft and its wording may change; my Annex 22 tracker records each step. ISPE published the "GAMP Guide: Artificial Intelligence" in July 2025, and my book, Validating AI in GxP Environments: A Practitioner's Handbook, covers how to extend a CSV programme to AI systems. The short version is in my guide to AI validation in GxP.

Frequently asked questions

What is CSV in pharma?

CSV is the documented proof that a computer system used in a GxP process works as intended, protects its data, and stays in control through its life. It's required under 21 CFR 211.68, 21 CFR Part 11 and EU GMP Annex 11.

Which systems need validation?

Any system that creates, changes, stores or uses GxP data, or controls a GxP process. That includes MES, LIMS, chromatography data systems, eQMS, ERP functions used for batch release, and spreadsheets used for GxP calculations. A GxP assessment at the start decides scope.

What is the difference between CSV and CSA?

CSA is a risk-based way of carrying out CSV. It puts testing effort on the functions that affect patient safety, product quality and data integrity, and accepts lighter evidence elsewhere. The regulatory obligation is the same.

Is GAMP 5 a regulation?

No. GAMP 5 is industry guidance from ISPE. Regulators know it well and most pharma companies base their CSV procedures on it, but the legal requirements sit in regulations such as 21 CFR Part 11, 21 CFR 211.68 and EU GMP Annex 11.

How often should a validated system be reviewed?

Annex 11 requires periodic evaluation to confirm systems stay in a validated state, and the frequency is yours to set by risk. In my experience high-impact systems are commonly reviewed every 1 to 2 years and lower-impact ones less often, with the rationale written into your procedure.

Do cloud and SaaS systems need validation?

Yes. The supplier hosts and tests the platform, but you remain responsible for intended use. You assess the supplier, use their evidence where it's sound, test your configuration and high-risk functions, and control their releases through change management.

Ground your team in CSV and CSA

If your team needs a grounding in CSV and CSA built around your own procedures, I run it in-house. If you want to talk about a specific system or an upcoming inspection, book a 30-minute call.

Sources

Last checked: 12 November 2026. I checked the sources listed below on that date.

  1. eCFR, 21 CFR Part 11
  2. eCFR, 21 CFR 211.68
  3. FDA, Part 11, Electronic Records; Electronic Signatures: Scope and Application (September 2003)
  4. European Commission, EudraLex Volume 4
  5. European Commission, consultation on Chapter 4, Annex 11 and new Annex 22
  6. European Commission, draft Annex 22 consultation document
  7. ISPE, GAMP 5 Second Edition
  8. ISPE, announcement of the GAMP Guide: Artificial Intelligence (29 July 2025)
  9. PIC/S, PI 011-3 (25 September 2007)
  10. PIC/S, PI 041-1 (1 July 2021)
  11. FDA, Computer Software Assurance for Production and Quality Management System Software

Related guides and pages