Out-of-specification (OOS) results are among the most sensitive events a pharmaceutical Quality Control laboratory can face: an analytical value outside the registered limits calls into question the batch, the method and — if the investigation is weak — the entire quality system. Not by chance, incomplete or "convenient" OOS investigations have ranked for years among the most frequent findings in FDA Warning Letters and EMA inspections. In this article we look at how to build a solid, audit-proof OOS procedure: definitions, regulatory framework, Phase I and Phase II investigations, the rules on retesting, resampling and averaging, and the typical mistakes to avoid.
What out-of-specification (OOS) results are and why they matter
For the FDA, an out-of-specification result is any test result that falls outside the specifications or acceptance criteria established in drug applications, compendia or the manufacturer's internal specifications — including in-process laboratory tests. It helps to distinguish three different cases right in the procedure:
- OOS (Out of Specification): a result outside the approved specification limits. It always requires a formal investigation before any decision on the batch.
- OOT (Out of Trend): a result within specification but outside the expected historical trend. An early signal to be investigated proportionately.
- Atypical or aberrant result: anomalous versus expectations (e.g. inconsistent values between replicates), even if formally compliant.
The core principle is simple: an OOS result is never discarded and never "retested away". It is investigated, and the batch remains on decisional hold until the investigation is concluded and documented. The same logic applies to contract laboratories: the quality agreement must require immediate notification of the OOS to the client before any additional testing.
The regulatory framework: FDA, EU GMP and MHRA
The most operational reference remains the FDA guidance Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production, updated in May 2022 (Level 2 revision), which structures the investigation in two phases and sets the rules on retesting, resampling and the use of averages. The roots of this approach go back to the landmark United States v. Barr Laboratories ruling (1993), the first to censure the practice of "testing into compliance". In Europe, Chapter 6 of the EU GMP (EudraLex Volume 4, effective 1 October 2014) requires at clause 6.35 that out-of-specification results or significant atypical trends be investigated. The MHRA guidance on OOS/OOT investigations completes the picture and is widely used by inspectors as a benchmark for the quality of laboratory investigations.
If topics like OOS, deviations and audit readiness are part of your daily work, subscribe to The Pragmatic GMP, our free weekly newsletter: every week one GMP topic explained in a practical, no-nonsense way, ready to apply on the shop floor or in the lab.
Phase I: the laboratory investigation
Phase I serves to establish, before anything else, whether there is an assignable cause of laboratory error. It should start as soon as possible, ideally while solutions and glassware are still available, and involves the analyst and the supervisor:
- The analyst confirms the method was followed, checks system suitability, sample integrity, calculations and instrumentation. Solutions and preparations must not be discarded until the data have been evaluated.
- The supervisor conducts an objective, documented assessment: interviews the analyst, reviews raw data and the audit trail, verifies calculations, dilutions and injection sequences, and evaluates the method's historical performance.
An OOS result may be invalidated only if the investigation identifies and documents an assignable cause (e.g. a demonstrated dilution error, an evident instrument malfunction). "Analyst error" without evidence is not an assignable cause: it is the fastest way to collect a finding.
Phase II: the full-scale investigation
If Phase I does not identify a laboratory cause, the investigation extends to production and the process (Phase II), with QA involvement: review of the batch record, raw materials, utilities, any concurrent deviations and the product history. In parallel, additional laboratory testing may be foreseen — the area where inspectors dig deepest:
| Activity | What the procedure should define | Typical audit mistake |
|---|---|---|
| Retesting | Same sample unit; maximum number of retests predefined in the SOP; preferably a second qualified analyst; a clear investigation hypothesis. | Retesting "until it passes" (testing into compliance) with no predefined number or rationale. |
| Resampling | Only if the original sample was not representative or was compromised, with a documented scientific justification and the same qualified sampling plan. | Resampling to replace an "inconvenient" result obtained from a valid sample. |
| Averaging | Used only where the method provides for it; the average must never mask variability or hide individual OOS values, which must still be investigated. | Averaging a passing and an OOS value to obtain a "pass" outcome. |
| Statistical (outlier) tests | Predefined criteria in the procedure; for chemical tests an outlier test cannot be the sole basis to invalidate a result. | Invalidating a data point just because it is "statistically anomalous". |
Once the investigation is closed, the batch decision rests with QA (and the QP for release): all results, including invalidated ones, must remain traceable and the conclusion must be consistent with the evidence. A confirmed OOS must be handled as a deviation, with an impact assessment on other batches and adequate CAPAs.
How to structure the OOS procedure
For the SOP to hold up in an inspection, it should define at least: scope and definitions (OOS, OOT, atypical); responsibilities of analyst, supervisor, QA and QP; timelines for opening and closing each phase; minimum content of the Phase I checklist; criteria and limits for retesting and resampling; rules on averaging and outliers; documentation requirements and the link to the deviation/CAPA system; periodic OOS trending (by product, method, analyst) as input to the Product Quality Review. Periodic training of analysts and supervisors on the procedure — with practical cases — is what distinguishes a living SOP from shelfware.
GuideGxP recommendation
The best way to defend an OOS investigation in an audit is to show that the path was defined before the event: phases, responsibilities, maximum retest numbers and invalidation criteria written down in advance, not decided case by case. Reread your procedure with an inspector's eyes: if you find wording like "a retest may be considered" without criteria, rewrite it. And remember that every confirmed OOS enters the deviation system: to structure that flow properly, from classification to CAPA effectiveness checks, our guide GMP Deviations and CAPA: effective implementation and audit defensibility gives you the process, examples and ready-to-use checklists.
