Out of specification (OOS) is the term FDA and MHRA use for any analytical result that falls outside the registered specifications or acceptance criteria: in the dossier, in the pharmacopoeias or in the manufacturer's internal standards. OOS handling is one of the most heavily inspected processes in the Quality Control laboratory in the pharmaceutical industry, and one where mistakes cost the most: invalidating a result without an assignable cause, or "testing into compliance", is among the most frequent findings in FDA Warning Letters. In this guide we show how to build an audit-proof OOS process, comparing the 2022 FDA guidance and the MHRA phased approach.
What out of specification means: definitions and boundaries
Under the FDA guidance, an out of specification result includes any test result that falls outside the specifications or acceptance criteria established in drug applications, Drug Master Files, official compendia or by the manufacturer; the definition also covers in-process laboratory tests outside established specifications. It helps to distinguish three categories that SOPs often blur:
- OOS: a result outside the registered specification limits. It always requires a formal investigation.
- OOT (Out of Trend): a passing result that falls outside the expected historical trend. EU GMP (Chapter 6, §6.9) requires out-of-trend data to be addressed and investigated as well.
- Atypical result: an anomaly that breaches neither specification nor trend (e.g. an unusual chromatographic profile), to be assessed per procedure.
The underlying principle is identical for every authority: an OOS result cannot be discarded because it "looks wrong". Either an assignable laboratory cause is demonstrated with evidence, or the result stands and the investigation extends to the batch.
The regulatory framework: FDA, MHRA and EU GMP
There are three references to know. The first is the FDA guidance Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production, issued by CDER and updated in May 2022 (the "Level 2" revision): it structures the investigation into two stages, the laboratory investigation (Phase I) and the full-scale investigation (Phase II). The second is the MHRA guidance Out-of-specification investigations, which breaks the pathway into Phase Ia, Phase Ib, Phase II and Phase III, through to batch disposition. The third is Chapter 6 (Quality Control) of EudraLex Volume 4: §6.7 requires a procedure for investigating OOS and OOT results, §6.9 requires any out-of-trend or out-of-specification data to be addressed and investigated, and §6.35 extends the investigation duty to significant atypical trends.
If you deal with OOS results, deviations and laboratory investigations every week, subscribe to The Pragmatic GMP, GuideGxP's free weekly newsletter: real cases, operational checklists and regulatory updates explained with an audit-ready mindset.
Phase I: the laboratory investigation
Phase I answers a single question: is there an assignable laboratory cause that invalidates the result? The MHRA approach splits it in two. In Phase Ia the analyst and supervisor check together, before discarding the preparations, for obvious errors: calculations, dilutions, glassware, system suitability, sample integrity, correct execution of the method. In Phase Ib, if no obvious error emerges, a documented hypothesis-testing investigation follows: possible laboratory causes are formulated and verified with targeted checks, such as re-analysis of the same preparation or verification of the standard.
Three rules make this phase defensible in audit: the investigation starts immediately, while the solutions and the analytical sequence are still available; every observation is recorded contemporaneously, signed by analyst and supervisor; the original result may only be invalidated against a documented assignable cause, never against a passing retest.
Phase II: the full-scale investigation
If Phase I identifies no laboratory cause, the OOS result is presumed representative of the batch and the investigation becomes full-scale, involving QA and production. The key elements per FDA are:
- Production review: batch records, process deviations, raw materials, historical trends for the product and related batches.
- Retesting: performed on the same portion of the original sample where possible, by a second qualified analyst, with the number of replicates defined in advance in the SOP, never decided after the fact.
- Resampling: allowed only with a documented justification, for instance when the investigation shows the original sample was not representative or was compromised.
- Averaging: the mean can never be used to mask an OOS; out-of-specification replicates remain results to be investigated.
- Outlier tests: to be used with extreme caution, only if foreseen by the procedure and never as the sole criterion to invalidate a result in a chemical test.
FDA vs MHRA: the OOS investigation phases compared
| Stage | FDA approach (2022) | MHRA approach |
|---|---|---|
| Initial check | Phase I – laboratory investigation (obvious errors, raw data, system suitability) | Phase Ia – obvious errors, immediate check with the supervisor |
| Extended laboratory investigation | Phase I – in-depth assessment with checklist and assignable cause | Phase Ib – documented hypothesis testing |
| Batch investigation | Phase II – full-scale investigation (production + retesting/resampling) | Phase II – production and process investigation |
| Wider impact | Included in Phase II (other batches, related products, CAPA) | Phase III – impact on other batches and products, CAPA |
| Final decision | Batch disposition with QA | Formal batch disposition, with QP assessment |
The typical mistakes inspectors look for
- Testing into compliance: repeating the test until a result passes, then using only the passing one.
- Invalidation without an assignable cause: discarding the OOS based on a passing retest or unverified hypotheses.
- Undefined replicates: the number of retests decided during the investigation instead of written in the SOP.
- Improper averaging: averaging passing and failing results to "dilute" the OOS.
- Copy-paste investigations: generic root causes ("analyst error") with no evidence and no effective CAPA.
- Ignored OOT: atypical trends left uninvestigated, against §6.9 and §6.35 of EU GMP Chapter 6.
GuideGxP recommendation
Build your OOS SOP around the logic of the phases, not around forms: define up front who does what in Phase Ia/Ib, how many retest replicates are allowed, when resampling is justifiable and who signs the batch disposition. Formally link the OOS process to deviations, CAPA and trending (OOT), and measure effectiveness with two simple KPIs: the percentage of OOS closed with a documented assignable cause and closure times per phase. In an inspection, the first request will be precisely the OOS log of the last 12 months with the classification of outcomes.
The OOS process is one of Quality Assurance's key responsibilities: GuideGxP's operational guide to the Pharmaceutical QA Manager role covers investigations, deviations and batch disposition with ready-to-use, audit-defensible templates.
