Calibration Intervals: Risk, Performance History and Review

A calibration interval is a decision about the period over which available evidence supports an instrument’s intended use. It is not a universal validity period attached to a certificate. Reviewing it requires comparable data, operating conditions and the consequences of undetected deterioration: a series of passing results may conceal adjustments or replacements.

This article addresses interval review, rather than the design of an entire metrology programme. The worksheet and case are original GuideGxP proposals; the example’s numerical intervals and decisions are not recommendations for an equipment family.

1. Separate applicable constraints from internal choices

First identify any intervals prescribed by applicable provisions, methods, contractual commitments or approved procedures. An internal assessment cannot override a binding requirement. If a company procedure specifies a frequency, changing it requires the authorised revision process; changing a label alone does not update the quality system.

EU GMP Chapter 3, §3.41, calls for calibration and checks at defined intervals, using appropriate methods and adequate records. It does not assign an annual frequency to every instrument. NIST explains that it does not generally recommend one interval: actual use matters. Record the origin of each due date, distinguishing external requirements, manufacturer recommendations and justified internal decisions.

Also define the starting point: the calibration date, return-to-service date or another justified basis. The day the certificate arrives does not automatically replace the measurement date.

2. Establish an initial interval that can be evaluated

Describe the quantity, range, configuration, error and uncertainty requirements, workload and environment. Consider wear, transport, thermal cycling, storage and the ability to detect a problem before it affects a decision. Infrequently used equipment may still deteriorate in storage; operating hours do not cover every mechanism.

Use relevant recommendations and experience with genuinely comparable equipment, making unknowns explicit. Equipment new to the laboratory does not automatically inherit the history of a similar model used under different conditions. Plan which data to collect and when to review the initial assumption.

A risk score helps prioritise work but does not itself establish a metrologically supported duration. Record assumptions and the available margin against intended-use requirements. Avoid converting a “low criticality” classification into months through an unvalidated formula.

3. Make historical data comparable

For each event retain the date, elapsed time or usage, points, method, reference, uncertainty, conditions and configuration. Distinguish results before intervention, as-found, from results after adjustment or repair, as-left. To observe change over a period, normally compare the previous final state with the next initial state under relevant conditions and with uncertainty considered.

Comparing only two post-adjustment values may show that the service can bring the instrument close to the reference; it does not demonstrate stability between visits. A PASS without numerical results and the applied criterion provides less information than a complete series.

Mark replacements, software changes, impacts and abnormal conditions on the timeline. A sensor replacement creates a discontinuity requiring assessment, even if the asset number remains unchanged. Retain the earlier history: separate the periods and explain which comparisons remain legitimate. Do not combine different points, units or configurations to create an artificially longer series.

4. An original interval-review worksheet

An empty field does not mean “no risk”. The worksheet must connect identifiable records to the conclusion without becoming a total score that conceals missing evidence.

5. Select a method and manage exceptions

ILAC-G24/OIML D 10:2022 presents several review methods and asks laboratories to assess their suitability. It does not provide a universal schedule.

In the local procedure, specify how trends, variability and margins against requirements are assessed. If using a statistical model, state included data, assumptions and verified predictive capability. A line fitted to a few points, or a series without anomalies, does not prove future behaviour. Do not confuse statistical control limits with intended-use acceptance limits.

An extension needs evidence relevant to the proposed duration and conditions; shortening may be justified by a credible deterioration signal requiring preventive action. Intermediate checks add information only about characteristics actually covered. A zero check does not demonstrate span stability; an unstable reference may make the instrument appear unstable.

Separate routine review from extraordinary events: repair, impact, a change of use or an abnormal check may require an earlier assessment. An overdue calibration requires management of use status and any impact under the applicable procedure; it cannot be remedied by backdating an extension. Budget shortages or supplier delays are not evidence of stability, as NIST GMP 11 also explains.

6. Simulated case: three certificates, two configurations

A laboratory proposes extending a thermometer-chain interval from 12 to 18 months. Its last three certificates show conformity after servicing and values close to the reference. The summary appears stable, but the reports show adjustments at the first two visits and a sensor replacement eight months after the second. Complete initial results are missing for some visits.

The three PASS outcomes do not describe three equivalent periods of stability. The review separates the old and new configurations, retrieves available pre-intervention evidence and checks what the intermediate checks actually measured. Time on the new sensor is not added to time on the old one to support the extension.

The 18-month proposal is not approved on this evidence alone. The responsible person documents a plan for the new configuration, required checks and the next review date. This does not automatically establish that 12 months is adequate: the existing interval also needs justification. Any concerns about historical results follow a separate assessment; the final certificate alone does not make everything invalid or everything safe.

For the broader framework, read the calibration strategy; for the role of evidence, see measurement uncertainty. Related routes are in the Calibration and Metrology hub.

7. Verified sources and limitations

Accessed 1 October 2026. The worksheet and case are original; no numerical interval is treated as a general obligation.

1. European Commission, EU GMP Chapter 3: full text, effective 1 March 2015, §3.41; defined intervals and records.
2. ILAC-G24/OIML D 10:2022: full text, §§4–6; guidance rather than universal mandatory frequencies.
3. NIST, Recommended Calibration Interval: institutional page consulted; user responsibility and pre-/post-service information.
4. NIST GMP 11, May 2019: full document, particularly §2. Here GMP means Good Measurement Practice, not pharmaceutical GMP. NIST-context intervals and examples are not transferred to pharmaceutical laboratories.