OOS Investigations: Why You Cannot Just Retest

An out-of-specification (OOS) result is a test result outside the specification or acceptance criteria. What happens next is one of the most scrutinised procedures in a quality system, because it is where the temptation to test until you get the answer you want is strongest.

FDA's guidance on investigating OOS results is the reference, and the structure it describes is two-phase.

Phase I: the laboratory investigation

Before anything else, the analyst and supervisor determine whether the result came from a laboratory error.

What this permits:

• Reviewing the data, the calculation, the instrument, the standards and the method.
• Identifying an assignable cause: a documented, specific laboratory error such as a dilution mistake, an equipment malfunction, or a standard prepared incorrectly.

What it does not permit:

• Invalidating a result because it is unexpected. "The result does not match the historical trend" is not an assignable cause.
• Retesting to see whether the first result repeats. That is not an investigation, and it is the single most common finding in this area.

If Phase I finds a documented assignable cause, the result may be invalidated and the test repeated. The original result and the investigation stay in the record permanently.

Phase II: the full-scale investigation

If no laboratory error is found, the investigation moves outward: to manufacturing, the process, the batch record, and other batches.

This phase decides batch disposition, and it asks whether the OOS reflects something real about the material.

Retesting in Phase II is permitted, but only under a predefined procedure stating in advance how many retests, by whom, and how the results will be interpreted. Deciding the retest plan after seeing the first retest is what "testing into compliance" means.

Outlier tests may be used for biological assays where the procedure allows, and may not be used to invalidate a chemical assay result.

Why this attracts so much attention

An OOS investigation is where data integrity and product quality intersect. An investigation that concludes "laboratory error, retested, passed" without a documented assignable cause is the pattern inspectors are trained to look for, and it appears repeatedly in warning letters.

The record has to show all results, including the original OOS, every retest, and the reasoning. A batch released on a passing retest whose original failure is not visible in the record is the finding.