By the PharmaTradz BD Team. Published October 2026.
In pharma, an OOS (out of specification) result is any test result that falls outside the specification or acceptance criteria set for a material or product, whether those limits come from the registration dossier, a pharmacopoeia or the manufacturer's own standards. Every OOS result must be investigated before the batch can be judged. The investigation asks one question: is the result wrong, or is the batch wrong?
What counts as an OOS result
A specification is the list of tests, methods and limits a material must meet: for example, assay 95.0% to 105.0% of the labelled amount, a named impurity not more than 0.2%, water not more than 0.5%. A result outside any of those limits is OOS. This applies to APIs, excipients, in-process materials and finished products.
In the United States, the legal basis is 21 CFR 211.192, which requires that the failure of a batch or any of its components to meet any of its specifications be "thoroughly investigated, whether or not the batch has already been distributed", and that the investigation extend to other batches and products that may be affected. For APIs, ICH Q7 (section 11.15) says any OOS result should be investigated and documented according to a procedure, and that any resampling or retesting should follow a documented procedure. In Europe, EU GMP Part I, Chapter 6 requires a written procedure for investigating out of specification and out of trend results.
OOS vs OOT vs OOE
Only OOS means a limit has been broken; the other two are early warnings.
| Term | What it means | Typical example | What should happen |
|---|---|---|---|
| OOS (out of specification) | Result outside the approved specification or acceptance criteria | Assay of 93.8% against a limit of 95.0% to 105.0% | Formal investigation is mandatory; the batch cannot be released while it is open |
| OOT (out of trend) | Result still within specification but out of line with the expected pattern, such as earlier batches or earlier stability time points | An impurity that has sat at 0.05% for every batch suddenly reads 0.15% against a limit of 0.2% | Investigate; EU GMP Chapter 6 says out of trend data should be addressed and investigated |
| OOE (out of expectation) | Industry term for an atypical or unexpected result that is within specification and not a trend issue; MHRA's guidance speaks of atypical, aberrant or anomalous results | An unknown peak in a chromatogram, or duplicate results that disagree more than usual | Assess under the laboratory's procedure; it may point to a method or sample problem |
OOT matters most in stability testing. A product that is drifting towards its limit may fail before its expiry date even though every result so far has passed. FDA's guidance notes that much of its approach may also be useful for examining out of trend results.
The FDA guidance and its two phases
The key US document is FDA's guidance Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production. It was first issued in October 2006 and revised in May 2022. The 2022 document is labelled "Revision 1" on its cover, while FDA's website lists it as a "Level 2 revision", which is why you will see both descriptions. The revision was mostly editorial, with added text on averaging and outlier tests. It covers chemistry-based laboratory testing; it is not intended for biological assays or process analytical technology.
The guidance says an investigation should be "thorough, timely, unbiased, well-documented, and scientifically sound". It runs in two phases.
Phase I: laboratory investigation. The aim is to find out whether the result was caused by a laboratory error. The analyst keeps the test solutions and tells the supervisor, who reviews the work with no preconceived idea of the cause: the raw data such as chromatograms, the calculations, instrument calibration and system suitability, the reference standard and reagents, and the method's past performance. A result may be invalidated only when the investigation clearly establishes a laboratory root cause. FDA notes that laboratory errors should be relatively rare.
Phase II: full-scale investigation. If Phase I finds no laboratory error, the result appears accurate and the investigation widens. The quality unit leads it, with production, process development, maintenance and engineering involved as needed. It reviews the batch record, materials, equipment and any previous similar failures, looks at other batches that may be affected, and aims to find the root cause and put corrective and preventive action in place. Additional laboratory testing, meaning retesting and in some cases resampling, also sits within Phase II.
Retesting and resampling rules
Retesting means testing a further portion of the same original sample, the same homogeneous material that gave the OOS result. FDA's guidance expects the maximum number of retests to be fixed in advance in a written procedure, based on sound science, and not adjusted according to the results obtained. A second analyst, at least as experienced as the first, often performs the retest. Any extra testing should follow a protocol approved by the quality unit.
Resampling means taking a new sample from the batch. It is appropriate only when the investigation shows the original sample was not representative or was prepared incorrectly, and it must use the same validated methods. If the sampling method itself was inadequate, a new method has to be developed, documented and approved by the quality unit.
Crucially, if no laboratory error is found, FDA says there is "no scientific basis for invalidating initial OOS results in favor of passing retest results". All results, passing and failing, are reported and considered.
Why testing into compliance and averaging are not allowed
Testing into compliance means retesting a batch again and again until a passing result appears, then reporting the pass. FDA's guidance calls this "unscientific and objectionable under CGMP". A borderline batch will eventually pass by chance, which tells you nothing about it.
Averaging hides failures in a similar way. Here is a simple worked example. A tablet's assay limit is 95.0% to 105.0%. The first result is 93.8%. Phase I finds no error, so two retests are run on the original sample and give 97.0% and 97.4%. The average of all three is 96.1%, which looks like a pass. It is not. Averaging the original OOS result with retest results is inappropriate because it hides variability, and a spread from 93.8% to 97.4% is itself a warning about the batch or the method.
The guidance does allow averaging in narrow cases: when the method itself specifies replicate measurements and a reportable result, such as replicate injections with set variability limits, or for microbiological assays. Averages are inappropriate for content uniformity and blend uniformity testing, where unit-to-unit variability is exactly what is being measured, so individual results are reported. Statistical outlier tests should not be used to invalidate a chemical assay result, and they do not apply to content uniformity or dissolution at all.
ICH Q7 adds a related rule for APIs: an out of specification batch should not be blended with other batches to make it meet specification.
The MHRA OOS guidance
The UK regulator, the MHRA, publishes its own out of specification guidance, first issued in 2013 and reviewed in 2017. It is set out as a stepwise flow chart, and it splits the process more finely than FDA. Phase Ia deals with obvious errors such as a calculation slip, a spillage or a power cut. Phase Ib is the analyst-and-supervisor review of data, equipment and method using a checklist. Phase II follows when no laboratory cause is found, and starts with a manufacturing investigation, with any further testing driven by a written, approved hypothesis plan. Phase III applies once a batch is rejected and covers the impact on other batches and corrective and preventive actions.
The MHRA's message is that investigators should move from "I think this is the cause" to "I know this is the cause". Its inspectors have said the most common failing is invalidating OOS results on assumptions with no real evidence. That links OOS directly to data integrity and the principles behind ALCOA+: every original result must be kept, traceable and reported.
How OOS links to batch release and CAPA
The investigation ends in a batch decision made by the quality unit (in the EU and UK, the Qualified Person certifies the batch). An initial OOS result does not automatically mean the batch fails. There are three broad outcomes:
- Invalidated: a laboratory cause is proven and documented, and a valid result replaces the original. The original data stays on file.
- Confirmed: the batch is rejected, as 21 CFR 211.165(f) requires, and the investigation becomes a batch failure investigation covering other associated batches and products.
- Inconclusive: no cause is found and the result is not confirmed. The OOS result must still get full consideration, and FDA says the quality unit should always err on the side of caution.
If the batch is already on the market, reporting duties follow. For US application products, a Field Alert Report is due within three working days unless the OOS is invalidated in that time. EU GMP Chapter 6 says confirmed OOS results or significant negative trends affecting marketed batches should be reported to the competent authorities.
Every confirmed OOS should feed the company's CAPA system. If the root cause was a weak method, poor training or a process flaw, the fix must stop it happening on the next batch. FDA notes that frequent OOS results may point to a flaw in product or process design.
What buyers should check
A certificate of analysis shows the final results, not the history behind them. To see how a supplier handles failures, ask:
- Was there any OOS or OOT result on this batch? If yes, ask for the investigation summary, the root cause and the final disposition. FDA's guidance recommends that OOS investigation reports be provided to the customer.
- Does the CoA show every result? FDA's guidance says that when no assignable cause is found, all test results should be reported to the customer on the CoA. For content uniformity, ask for the individual results, not just an average.
- Can we see the OOS procedure? Check it defines Phase I and Phase II, fixes the number of retests in advance, and bans averaging an OOS result with retest results.
- How many OOS results were invalidated as lab error last year? A high rate is a red flag for testing into compliance.
- How are OOT results trended, especially in stability studies?
- Which CAPAs came out of recent confirmed OOS results, and were they checked for effectiveness?
During an audit, read two or three closed OOS investigations end to end, looking for raw data, a tested hypothesis, a clear conclusion and no unexplained retests. Our guide to verifying a pharmaceutical supplier covers the wider checks.
How PharmaTradz can help
When you source an API or finished product, tell us what quality evidence you need, such as CoAs with individual results, stability data and a summary of OOS history. Our documentation and quality pack gathers these documents up front, or you can send us an RFQ directly. For other terms, see our pharma abbreviations A-Z.
Frequently Asked Questions(FAQs)
What is OOS in pharma?
OOS stands for out of specification. It is any test result that falls outside the approved specification or acceptance criteria for a material or product, and it must be formally investigated before the batch can be released or rejected.
What is the difference between OOS and OOT?
An OOS result breaks a specification limit. An OOT (out of trend) result is still within specification but does not fit the expected pattern from earlier batches or earlier stability time points, so it is an early warning that should also be investigated.
What are the phases of an OOS investigation?
FDA's guidance uses two phases: Phase I, a laboratory investigation to look for analytical error, and Phase II, a full-scale investigation of production and the batch if no laboratory error is found. The MHRA's guidance splits this further into Phases Ia, Ib, II and III.
Can you retest after an OOS result?
Yes, but only on the original sample, under a written procedure that fixes the maximum number of retests in advance. If no laboratory error is found, a passing retest cannot simply replace the original OOS result, and repeated testing until a pass appears is not allowed.
Can an OOS result be averaged with retest results?
No. FDA's guidance says averaging the original OOS result with retest or resample results is not appropriate because it hides variability, and averages are also inappropriate for content uniformity testing.
