Proficiency Testing: Coverage, Scores and What Follows a Bad Result

Internal quality control tells a laboratory it is consistent. Only comparison with other laboratories tells it whether it is right. That is what proficiency testing does, and it is among the first things an accreditation assessor asks to see.

This guide covers how participation is planned across an accreditation schedule, how PT items should be handled, what the scores mean, and what a laboratory is obliged to do when one comes back bad — which is more than investigate the item.

What proficiency testing is

A provider distributes a sample of known value to many laboratories, collects their results, and evaluates each participant against pre-established criteria. You find out not just whether your result was close to the assigned value, but where you sat relative to everyone else measuring the same thing.

Clause 7.7.2 of ISO 17025 requires laboratories to monitor performance by comparison with other laboratories where available and appropriate, through proficiency testing or other interlaboratory comparison. Accreditation bodies almost always add their own minimum participation requirements on top.

Planning proficiency testing coverage, not participation

The common mistake is to build a proficiency testing plan from the schemes that happen to be available. Build it from your accreditation schedule instead, and treat any entry with no matching scheme as a gap requiring an alternative.

Plan across a defined cycle — four years is typical, subject to what your accreditation body requires — and record for each schedule entry:

• the scheme and provider, and whether the provider is accredited to ISO/IEC 17043;
• the matrix and concentration level, which should reflect where your customers’ decisions are made;
• the planned frequency and the date of the next round;
• where no scheme exists, the alternative comparison arranged instead.

Book early. Scheme calendars are fixed months ahead and rounds fill. For a laboratory seeking first accreditation, a missed round is one of the few things that can delay the whole programme by a year — the documents can be written in a fortnight, but you cannot compress a distribution schedule.

Handle the item as routine work

This is where laboratories quietly undermine the exercise. A PT item must be run the way customer samples are run: same method, same personnel, same equipment, same conditions.

That means no extra replicates, no assigning it to the most experienced analyst, no running it three times and submitting the middle value, and no calling another laboratory to compare notes before the deadline.

The temptation is understandable — the score is visible to your accreditation body. But a result obtained under exceptional conditions measures something the laboratory does not actually do, which makes it worthless as a control and misleading as assurance. Record the PT item as a technical record like any other.

Reading the scores

Most schemes report a z-score: how far your result sat from the assigned value, in units of the target standard deviation.

Two things laboratories get wrong here.

A questionable result is not a pass. It is a warning, and two questionable results in the same direction are a bias you have now been told about twice.

Read across rounds, not round by round. Four consecutive satisfactory scores of +1.4, +1.6, +1.5 and +1.7 are all individually fine and collectively describe a systematic positive bias. Each round in isolation passes; the sequence does not. Plot them.

When a result comes back unsatisfactory

Treat it as nonconforming work and investigate properly. In rough order of how often each turns out to be the cause:

1. Transcription and units. Reported in mg/L when the scheme wanted µg/L, a decimal place lost, the wrong field on the form. Unglamorous and extremely common — check it first, however trivial it feels.
2. Calibration and traceability. Standard preparation, an expired reference material, an instrument drifting since its last check. A persistent bias here usually shows up in the measurement uncertainty budget too.
3. Method performance. Recovery, matrix effects, an interference the validation did not cover.
4. The analyst. Technique, or a step performed differently from the written method.
5. The assigned value itself. Occasionally the scheme is wrong. Challenge it with the provider and record that you did — but only after eliminating the four above, and never as a first resort.

The step most often skipped

Investigating the PT item is the easy half. The obligation that makes proficiency testing a control rather than an examination is this: assess the customer results you produced in the same period, by the same method.

If your PT result was 29% high, the customer results run that week were plausibly 29% high too. Identify them, evaluate whether they are affected, and where they are, notify and reissue. A file containing a thorough investigation of the PT sample and no assessment of customer work is one an assessor will raise a finding against.

Where no scheme exists

Some measurands have no commercial scheme, particularly in niche calibration and specialised testing. Options, in descending order of strength:

• an interlaboratory comparison arranged directly with another accredited laboratory, with the design and evaluation documented in advance;
• measurement of a certified reference material not used in your calibration;
• comparison against a different method, ideally based on different physical principles;
• measurement of a retained item previously measured elsewhere.

Record the absence of a scheme and the alternative chosen. A schedule entry with neither is a gap, and it is a gap the assessor will find because they will work down your schedule.

Choosing a proficiency testing provider

Prefer a provider accredited to ISO/IEC 17043, which covers the competence of proficiency testing providers themselves. Where none exists for your measurand, record the basis on which you selected the provider you used.

Either way, the provider is an external provider under clause 6.6: approved, monitored, and reviewed like any other supplier whose failure could affect your results.

What proficiency testing records to keep

Retain the participation plan and its coverage against schedule, the scheme reports, your own raw data and technical records for the PT item, the evaluation of each result including satisfactory ones, and the investigations and actions for anything questionable or worse.

Proficiency testing performance and coverage are inputs to management review, and they are usually among the first three things an assessor asks for — alongside the calibration schedule and your handling of an out-of-tolerance instrument.

Our ISO 17025 Toolkit includes a proficiency testing procedure and a plan and results register that tracks coverage against the accreditation schedule, alongside internal quality control and the nonconforming work process an unsatisfactory result feeds into.

Frequently asked questions

How often should a laboratory participate in proficiency testing?

There is no single answer in the standard. Your accreditation body will set minimum expectations, commonly framed as coverage of the schedule across a four-year cycle with at least annual participation in major areas. Frequency should also reflect the consequence of an undetected error.

What is the difference between proficiency testing and an interlaboratory comparison?

Proficiency testing is a formal scheme in which a provider evaluates participant performance against pre-established criteria. An interlaboratory comparison is any organised measurement of the same or similar items by two or more laboratories, and may be arranged privately without formal scoring.

Does a poor proficiency testing z-score mean losing accreditation?

Not on its own. Accreditation bodies expect occasional unsatisfactory results — what they examine is the response: whether it was investigated to a cause, whether customer results were assessed, and whether the corrective action worked. Repeated unsatisfactory results for the same measurand, with no effective action, is a different matter.

Can a laboratory decline to participate?

Where no suitable scheme exists, yes — but you must record that and put an alternative comparison in place. Declining on grounds of cost or inconvenience where a scheme is available will not survive assessment.