Proficiency testing: the complete guide

What proficiency testing is

Three terms are often used as if they meant the same thing. They describe related activities, and one question separates them.

Interlaboratory comparison (ILC). Two or more laboratories test the same or similar items under conditions agreed beforehand, and the results are evaluated. It is the umbrella term. A study of how reproducible a new method is counts as an ILC too, but it judges the method, not the laboratories.

Proficiency testing (PT). An interlaboratory comparison used to judge each participant’s performance against criteria set before the round. The scheme plan states in advance how the assigned value will be set, how results will be scored and which scores count as satisfactory. ISO/IEC 17043:2023, the standard for PT providers, defines both terms.

External quality assessment (EQA). The usual name in medical laboratories, and the one ISO 15189 uses. An EQA programme is proficiency testing, often with education added, such as case commentaries and method reviews.

PT and internal quality control do different jobs

Internal quality control (IQC) means running control materials alongside routine samples, in every run or at set intervals, and judging them against targets and limits the laboratory has set itself. Its question is narrow and urgent: can the results of this run be released? Because it runs so often, IQC catches rising imprecision, a sudden shift or a slow drift, usually on the day it starts.

But IQC compares today with the laboratory’s own yesterday. If yesterday was already wrong, IQC has nothing to object to. It cannot see a bias that was present when the targets were set, a bias shared with the laboratory’s own calibrator, or a shift that was absorbed when targets were reset after a lot change. It never sees a result typed wrongly into a report, a unit mix-up or a wrong interpretation.

PT covers those gaps. It brings an assigned value from outside the laboratory, compares the laboratory with peers who use the same method, and tests the whole chain from receipt of the item to the reported result. What PT cannot do is release a run: it samples a few days a year, so it catches a short-lived fault late or not at all.

Types of PT scheme

Schemes differ in several independent ways, so one scheme usually carries several labels at once.

What participants report

• Quantitative: a number with a unit, scored against an assigned value. Lead in drinking water in µg/L, or the compressive strength of a concrete cube in N/mm².
• Qualitative: a category or an identity. Whether Listeria monocytogenes is detected in a cheese sample, or which animal species is present in a meat product.
• Interpretive: a judgement, assessed against expert criteria. A comment on a blood film, or whether a set of water results meets a drinking-water standard.

How the items travel

• Simultaneous: portions of one homogeneous batch go to every participant at about the same time. Most testing schemes work this way.
• Sequential: one item travels from laboratory to laboratory, and a reference laboratory checks that it has not changed along the way. This suits stable measuring artefacts such as a reference thermometer.

How often, and special designs

• Single occasion or continuous: a one-off exercise, for example when a new contaminant comes under regulation, or rounds at regular intervals, such as four or six a year.
• Split-sample: one sample divided between two or a few laboratories, often by a customer comparing the laboratories it uses.
• Data transformation: participants receive data rather than material, such as raw readings or a chromatogram, and report what they calculate or conclude. It tests calculation and interpretation, not measurement.

One PT round, end to end

Rounds differ in detail, but most follow the same eight steps.

1. Plan. The provider fixes the measurands, the test items and their levels, how the assigned value (x_pt) and the standard deviation for proficiency assessment (σ_pt) will be set, how results will be scored, and the timetable.
2. Prepare and check the items. The provider shows that the items are sufficiently homogeneous after packaging and stable for the life of the round.
3. Dispatch the items, with instructions on storage, preparation, units and the closing date.
4. Test. Each participant tests the items as it would test routine samples.
5. Submit results by the closing date.
6. Evaluate. The provider sets x_pt and σ_pt as planned and scores every result.
7. Report. Each participant receives its own scores, identified by a code, together with summary statistics for the round.
8. Review. The laboratory checks its report, investigates anything unusual and records what it did.

The provider’s report is not the last step. The review inside the laboratory is where PT turns into improvement: a questionable or unsatisfactory score starts an investigation, and the record of that investigation is evidence in its own right.

How results are scored

For quantitative results the most common score is the z-score: the difference between the participant’s result x and the assigned value, divided by σ_pt.

z = (x − x_pt) / σ_pt

A |z| of 2.0 or less is usually satisfactory, a |z| between 2.0 and 3.0 questionable, and a |z| of 3.0 or more unsatisfactory. The two choices behind the score matter as much as the formula.

• The assigned value, x_pt. It can come from how the item was made (formulation), from a reference material with a certificate, from a reference laboratory, or from the consensus of the participants’ own results. A consensus needs robust statistics so that a few wild results cannot drag it; the best-known method is Algorithm A from ISO 13528.
• σ_pt. It sets how far from the assigned value a result may be and still count as satisfactory. It can be fixed in advance from what the results are needed for, taken from earlier rounds, or estimated from the round’s own data. Each choice changes what a satisfactory score means.

Qualitative results are scored differently. A result either matches the expected answer or it does not, and the expected answer comes from how the item was prepared or from the consensus of the participants.

Why laboratories take part

What the standards ask of laboratories

ISO/IEC 17025:2017, clause 7.7.2, asks laboratories to monitor their performance by comparing their results with other laboratories’ results whenever a suitable comparison exists. The monitoring is planned and reviewed, and it can use proficiency testing, other interlaboratory comparisons, or both.

That condition is not an exemption. If a suitable scheme exists, laboratories are expected to use it. If none exists, or none fits the matrix, range or method, they compare themselves with other laboratories in another way and record why.

ISO 15189:2022, clause 7.3.7.3, asks medical laboratories to take part in EQA programmes suited to their examinations and to the way they interpret results. In plain words, it also asks them to handle EQA samples as far as possible like patient samples, tested by the staff who normally do the work; not to send EQA samples to another laboratory for confirmation before submitting results; to use an alternative approach, with objective evidence, when no suitable programme exists; and to review EQA results with the staff concerned and act when criteria are not met.

ILAC P9 is the policy accreditation bodies apply to PT participation. For a laboratory it means a documented participation plan covering its scope of accreditation over the accreditation cycle, with the level and frequency of participation set by risk rather than by one fixed rule. Participation goes up when results drive regulatory, legal or clinical decisions, when staff are new to a test, when a method is new or has changed, when reference materials are scarce, and when recent scores were questionable or unsatisfactory.

When no scheme exists

Start by searching provider catalogues and PT scheme directories, such as the EqasOnline PT directory, and ask your accreditation body. Record the date of the search and what it found. A scheme with a different matrix or range may still be the best one available; decide, and record why.

If nothing fits, choose an alternative that compares the laboratory with something outside its own system, and fix the criteria before the comparison, not after it.

• An interlaboratory exchange with one or a few other laboratories. Agree the criteria first and check the partner’s competence.
• Split samples with another laboratory, using real samples. The partner may share your method, and its bias.
• A reference material with a certificate, tested as an unknown. Match the matrix as closely as possible, and do not use the material you calibrate with.
• Previously tested material, compared with its earlier assigned value. Check that it is still stable.

Re-testing your own retained samples is useful internal monitoring, but it shows consistency, not accuracy: a bias present both times stays hidden. Record and review the results of any alternative the same way as PT results, and search for a scheme again each year.

Taking part fairly

A PT score is evidence about the laboratory’s routine process. If the item gets special treatment, the score describes a process the laboratory never runs, and the problem it should have revealed stays hidden. Fair participation also protects other participants: a consensus assigned value assumes independent results, so copied or adjusted results distort everyone’s scores. Four rules cover most situations.

1. Test it like a routine sample. The staff who normally do the work, the routine method and equipment, and the routine number of determinations. If the laboratory normally reports a single result, it reports a single result.
2. Never refer a PT item to another laboratory before the closing date, even when a routine sample would be sent there. If an item cannot be tested, tell the provider and record why. No result is honest; a borrowed result is falsification.
3. Do not collude. Do not share, compare or discuss results with other participants until the provider has issued the report. That includes sister laboratories in the same group.
4. Report what was measured, mistakes included. Repeat a test only when the routine procedure calls for it, and never nudge a result towards an expected value.

The provider’s role under ISO/IEC 17043:2023

ISO/IEC 17043 sets out what makes a PT provider competent. In outline, a provider is expected to:

• be impartial, and keep participants’ identities and results confidential;
• have competent staff, suitable facilities and a management system;
• plan each scheme in writing before the round;
• prepare test items and show that they are sufficiently homogeneous and stable;
• use a statistical design suited to the scheme, usually drawing on ISO 13528;
• give participants clear instructions, evaluate their results and issue reports;
• handle complaints and appeals.

A provider can use competent external suppliers for some of the work and stays responsible for it: for example producing items, homogeneity and stability testing, measuring an assigned value in a reference laboratory, and transport. Planning the scheme, evaluating performance and authorising the report stay with the provider.

For a laboratory choosing a scheme, this gives a practical test. Look for a provider whose ISO/IEC 17043 accreditation includes that scheme in its scope, and read the scheme plan before enrolling. For a provider, most of the work in a round is the record-keeping around these duties: who enrolled, what was sent and when, whether the items passed their checks, which results arrived, how they were scored and which report version each participant received.

Where to go next

• Looking for a scheme? Search the PT directory by field, analyte and country.
• Want the statistics? Read Algorithm A (ISO 13528) explained, step by step, with a fully worked example.
• Want to go deeper? The free EqasOnline Academy course covers PT from first principles to investigating a poor score.