Out-of-specification (OOS) and out-of-trend (OOT) results

An out-of-specification (OOS) result is a test result that falls outside the specification or acceptance criteria that apply to it. Those criteria may come from the marketing authorisation, a pharmacopoeial monograph or the company's own specification. OOS applies to release testing, in-process testing, raw material testing and stability testing.

An out-of-trend (OOT) result is still inside the specification, but it does not follow the pattern expected from earlier results. A stability assay that suddenly drops 3% between two time points, while every other batch drops 0.5%, is a typical OOT. It is an early warning that something may be changing before a true OOS happens.

Many SOPs also use the term atypical or aberrant result for data that look unusual for other reasons, for example an unexpected extra peak or poor replicate agreement, even when the reported value passes. These are handled under a similar, usually lighter, process defined in your SOP.

In the US, the CGMP rules require that any unexplained discrepancy or failure to meet specification is thoroughly investigated and documented. FDA's guidance on investigating OOS results, issued in 2006 and revised in 2022, explains how labs are expected to do this. The MHRA has published its own guidance on OOS and OOT investigations, which uses a similar phased model.

The core idea in both is simple. A failing result is assumed to be real until a documented, scientific investigation shows otherwise. You cannot keep testing until you get a passing result and then report that one — regulators call this “testing into compliance”.

The first rule of OOS: do not throw anything away. Keep sample solutions, standard solutions, glassware set-ups and the instrument as they are until your supervisor has looked at them.

As soon as an OOS result is seen, the analyst informs the supervisor. Together they carry out a quick, documented check for an obvious error. In the MHRA model this is Phase Ia; FDA calls the whole lab stage the Phase I laboratory investigation. The aim is to find a clear, provable mistake, not to look for a reason to reject the result.

• If a clear error is found and documented (for example a calculation mistake or wrong standard weight entry), the original result may be corrected or invalidated as defined in your SOP, with QA approval.
• If nothing obvious is found, the investigation moves to Phase Ib.

1. 1Analyst stops, records the result and tells the supervisor without delay.
2. 2Check calculations, dilution factors, units and transcription against the raw data.
3. 3Confirm the correct method version, sample, reference standard and reagents (with valid expiry) were used.
4. 4Check system suitability and instrument status: was it within calibration, and were there errors, leaks or power cuts during the run?
5. 5Look for documented events at the time, such as a spilled flask or a wrong volumetric pipette noted on the worksheet.
6. 6Review the chromatogram or spectrum: integration, peak identity, baseline problems.

In Phase Ib the lab forms a hypothesis about what might have gone wrong and tests it with a pre-planned, documented experiment. Examples are re-injecting the same sample solution to check for an injection fault, re-measuring the same preparation, or checking whether extraction was incomplete by extending sonication on retained solution. The plan, including what result would confirm or reject the hypothesis, is written and approved before the experiment is run.

If Phase Ib proves a laboratory root cause, the original result is invalidated and the test is repeated. If no lab error is confirmed, the OOS is treated as potentially real and Phase II begins. Phase II is a full-scale investigation led by QA that extends into manufacturing: batch records, process deviations, raw materials, equipment and people. It may include additional laboratory testing under a written protocol.

Some companies also describe a Phase III: assessing the impact on other batches, products and markets, and deciding on the batch. FDA describes the whole full-scale stage as Phase II; follow the terms in your own SOP.

A retest is a repeat analysis of the same original, homogeneous sample. A resample is a new sample taken from the batch. They are not interchangeable, and both need rules written in advance.

• Retests are done on the original sample material, usually by a second analyst where possible, as part of the investigation plan.
• The number of retests is fixed in the SOP or protocol before testing starts. It must not depend on whether the results pass.
• Passing retests do not simply overrule the original OOS. Without an assignable laboratory cause, all results are evaluated together and the batch decision is scientifically justified.
• Resampling is appropriate only when there is evidence that the original sample was not representative or was compromised, for example damaged during transport or taken incorrectly. It follows the same approved sampling procedure.
• Averaging is acceptable when the method itself defines the reportable result as an average, such as the mean of replicate injections or preparations. Individual values are still recorded and checked.
• Averaging is not acceptable when it hides variability that the test is designed to detect, as in content uniformity, or when an OOS value is averaged with passing retests to make the batch pass.
• Statistical outlier tests are generally not used to reject a chemical test result. FDA guidance allows them in limited cases, such as highly variable biological assays, when defined in advance.

Worked example: assay is 93.8% (limit 95.0–105.0%). Phase Ia finds nothing. The Phase Ib hypothesis is a dilution error; re-injection of the retained sample solution gives 93.9%, so the solution itself is low and the injection was fine. A wrong-flask hypothesis is then checked against the worksheet. If no lab cause is confirmed, the OOS stands and Phase II starts — retests alone cannot clear the batch.

A result may be invalidated only when the investigation identifies an assignable cause that is documented and scientifically sound. A suspicion is not enough. “Analyst error” without specific evidence is one of the most common weaknesses inspectors find in OOS files.

• Valid reasons: a proven calculation error, a documented instrument malfunction during the run, a confirmed wrong reagent or standard, a sample preparation error recorded at the time or confirmed experimentally.
• Not valid on their own: “probable” analyst error, a passing retest, an outlier test on chemical data, or the fact that previous batches all passed.
• Every invalidation is approved by QA and the original result is kept in the record with the reason.
• Labs should trend invalidated results. A high rate of lab-error invalidations points to a training, method or equipment problem.

Whether the cause is in the lab or in manufacturing, the investigation should end with a root cause, not just a symptom. Tools such as the “5 whys”, fishbone (Ishikawa) diagrams and fault trees help teams look beyond the immediate error to why it was possible.

Corrective and preventive action (CAPA) then fixes the cause and stops it happening again. A corrective action deals with the problem found; a preventive action reduces the chance of similar problems elsewhere. CAPA effectiveness should be checked after a defined period.

Stability data are the main place where OOT results are found, because each batch builds a series of results over time. The aim is to spot a result that does not fit the expected degradation pattern while there is still time to act.

There is no single regulatory method for OOT detection. Common approaches compare a new result with limits calculated from historical data: regression-based prediction limits for the batch, by-time-point limits from other batches at the same time point, or a check that a batch's slope is consistent with earlier batches. The approach, the data set used and the action limits should be defined in an SOP and kept up to date.

1. 1Confirm the result is correct with a Phase Ia-type lab check, as for an OOS.
2. 2Compare with previous time points for the same batch and with other batches at the same time point.
3. 3Check storage conditions: chamber excursions, sample pull dates and packaging.
4. 4Decide with QA whether the trend may lead to an OOS before expiry, and whether extra testing or an earlier time point is needed.
5. 5Record the OOT, the evaluation and any actions, and feed it into the product quality review.

An OOT result is not a failure, but ignoring it is. Many stability OOS results were visible as OOT trends months earlier.