When a result is out of specification

A specification is a list of tests, methods and acceptance limits that a material must meet. An out-of-specification result, usually shortened to OOS, is any result that falls outside those limits. In regulated manufacturing, what happens next is governed by written procedure, and the central principle is simple: an OOS result is treated as real until an investigation shows otherwise.

The reference most often cited is the US Food and Drug Administration’s guidance on investigating out-of-specification test results for pharmaceutical production. EU GMP Chapter 6, on quality control, sets the same expectation: out-of-specification results must be investigated and the investigation recorded. The practices described below follow those documents.

Phase one: the laboratory investigation

The first step is to find out whether the result came from an error in the laboratory. It begins as soon as the result is seen, before any solution or sample is discarded. The analyst and a supervisor review the work together, typically checking:

1. Calculations, transcriptions and units.
2. Whether the correct method, method version and specification were used.
3. Instrument performance, including the system suitability results and calibration status.
4. Reference standard identity, preparation and expiry.
5. Sample preparation: weights, dilutions, glassware and the age of the solutions.
6. Whether the sample tested was the sample intended, by its label and log entry.

Retained solutions and preparations can be examined again. A prepared solution can be re-injected to test whether an injection failed. What cannot be done is to set the original result aside because it is unwelcome.

If a clear, documented laboratory error is found — a wrong standard concentration, an instrument fault confirmed by records, a calculation mistake — the original result is invalidated, the reason is written down and the test is repeated. The invalidated result stays in the record with its explanation. It is not deleted.

Phase two: the full-scale investigation

If the laboratory review finds no assignable cause, the investigation widens beyond the lab. It examines the manufacturing and handling history of the batch: production records, deviations, equipment, raw materials, storage and transport. It looks for other batches made under similar conditions that might share the problem. The quality unit, not the laboratory alone, owns this phase and its conclusion.

Retesting, resampling and averaging

Retesting means analyzing more of the original sample. It is permitted within a full investigation under rules set in advance: who decides, how many retests are run, and how their results will be judged. The guidance is direct about the practice these rules exist to prevent, which it calls testing into compliance: repeating a test until a passing result appears and reporting that one.

Resampling means taking a new sample from the batch. It is appropriate when the investigation shows the original sample was not representative or was compromised, not as a second attempt at a failing result.

Averaging has limits. For a test of the batch’s uniformity, individual results matter, and an average can hide a unit that fails. Even where averaging is part of the method, a failing individual result inside an average still has to be investigated.

Out of trend is a different signal

A result can be inside its specification and still be unexpected: a stability result that falls further in one interval than the trend predicts, or an impurity that rises in a batch where it was previously steady. These out-of-trend results are not failures, but they are early warnings, and many quality systems require them to be reviewed before they become OOS.

How it ends

Every OOS investigation ends with a written conclusion: the root cause if one was found, the evidence for it, the disposition of the batch, and the corrective and preventive actions taken. A batch that fails its specification on a valid result is rejected. A lab error that was found is followed by an action that makes it less likely to happen again, such as a revised procedure, retraining or an equipment repair.

For anyone reading a certificate, OOS handling is mostly invisible, because certificates report results that met specification. Its effect shows up elsewhere: in the existence of a written procedure, in investigation records an auditor can request, and in a laboratory’s willingness to show that a result was invalidated with a documented reason rather than quietly replaced.