ISO 17025 method validation and verification: which applies, and what the file must show

ISO 17025 method validation or verification: which one applies?

The first question on every method in your intended scope is simple: did someone else design and publish it, or did your laboratory write it? The answer decides the route and the size of the file.

For a published standard method, the method's authors have already done the design work. Your job is to verify it: to prove that your analysts, your instruments and your environment reproduce what the method promises. For an in-house method, nobody has done that design work except you, so you validate it: you establish, with evidence, that the method does what the customer's decision needs.

Grey areas are where laboratories get caught. A published method with a changed sample size, an extra clean-up step, a different instrument technique or a matrix the method never mentions is no longer the method its authors tested. Sterling's working rule is to treat any change you would have to explain to an assessor as a reason to move the method from the verification route to the validation route, and to record that decision. Check the exact triggers against clause 7.2.2 in your licensed copy of the standard before you finalise the method list.

The wider picture of clause 7, and where methods sit among the other process requirements, is covered in ISO/IEC 17025 requirements explained.

Selecting methods before you prove them

Proof work is expensive, so settle four questions before anyone runs a replicate.

• What does the customer need to decide? A pass or fail against a limit, a trend over time and a calibration value with a stated uncertainty each place different demands on the method.
• Which method does the customer, the regulator or the tender name? If a specific published method is written into the requirement, a laboratory alternative may not be accepted, however good it is.
• Is the edition you hold current? Keep a method register that shows the edition in use, where it came from and when it was last checked against the publisher.
• Can you resource it? Equipment, reference materials, environmental control and trained staff all need to exist before verification can start.

Agreeing the method with the customer belongs in the contract review. If the method changes later, the customer should hear about it before the report does.

What a verification file should show

Verification is lighter than validation, but it is not a signature on a form: the assessor wants evidence of your own performance against the method.

A short plan

State the method and edition, the scope you will claim (items, matrices or ranges), the characteristics you will check and the acceptance criteria, taken from the method itself where it gives them and from the customer's need where it does not.

Evidence from your own bench

Run the checks with the analysts who will do the routine work, on the equipment listed in the equipment register, using reference materials or reference standards whose traceability you can show. Sterling typically builds the plan around the performance the method claims, such as its precision and its agreement with a known value, and the range you intend to report; confirm the characteristics that apply to your discipline against the method and the accreditation body's guidance.

A conclusion someone owns

Compare the results with the criteria, record any limitation (a narrower range, a matrix excluded), and have the technically responsible person sign that the method is fit to go on the scope. That signature is also where analyst authorisation for the method begins.

What a validation file should show

A validation file tells the whole story of an in-house method, from the requirement to the decision. Assessors often read it line by line.

• The requirement. What the result is for, what quantity is measured, over what range and in which items, and how good the result must be for the customer's decision.
• The method as it stands. A controlled, written procedure, frozen at the version being validated, so the evidence relates to the method you will actually run.
• The plan. Which performance characteristics you will determine, how, how many runs and by whom, and the criteria for acceptance, all set before the data exist.
• The data. Raw results, calculations and instrument records that let someone reconstruct every conclusion, including runs that failed and what was done about them.
• The comparison. Each characteristic against its criterion, in a table, with no conclusion stronger than the data support.
• The statement. A dated decision that the method is fit for its intended use, with any limits, and the name of the person authorised to make it.

The same data usually feed the uncertainty budget, so plan the two together; the method for that budget is set out in Measurement uncertainty under ISO/IEC 17025. Traceability of the reference standards and equipment used in the work belongs to Equipment, calibration and metrological traceability under ISO/IEC 17025.

Worked example: one laboratory, four methods

An illustrative case: a testing laboratory prepares its first application with four methods. The table shows how each is routed and what its file needs. The details are invented for explanation.

Only after each file closes do the methods move on to routine quality control and proficiency testing and interlaboratory comparison, which then show that the performance holds over time.

Mistakes assessors find in method files

• Criteria written after the results. Acceptance limits chosen to fit the data prove nothing. Set them in the plan and date the plan.
• Verification done by the wrong people. A supervisor runs the checks and three other analysts do the routine work. Each person who reports results needs evidence of their own competence on the method.
• The file describes a different method. The procedure was edited after validation and nobody reassessed it. Link every file to a controlled version.
• Scope wider than the evidence. The application claims a range or matrix the data never covered. Claim what you proved; extend later.
• Old methods left unproven. A method in use for years is assumed to be fine. Long use is not evidence; build the file.

How assessors test these files in person is covered in The ISO/IEC 17025 assessment, and how your own auditors should examine them in ISO/IEC 17025 internal audit: the technical clauses.

Method validation under ISO/IEC 17025: questions buyers ask