An OOS result opens a lab check, not a recall

211.192 sets the duty to investigate; it doesn't sort the result

21 CFR 211.192 requires a thorough investigation of any unexplained discrepancy, or any failure of a batch or its components to meet a specification, whether or not the batch in question has already been distributed. The investigation has to extend to other batches of the same product, and other products, that the specific failure could implicate, and a written record has to capture the investigation's conclusion and whatever follow-up it leads to. The regulation states the duty in general terms and never actually distinguishes between a result that reflects a real problem with the batch and one that reflects an error made in producing the number — that sorting job belongs to guidance, not the rule itself.

Phase I asks the narrower question first: is the number real

FDA's guidance, Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production, first issued in October 2006 and revised in May 2022, opens with what it calls the laboratory's own Phase I investigation: a review of analyst technique, instrument performance and calibration, calculation, and sample handling, looking for an assignable, documented cause specific to the test itself. Only a confirmed lab error found this way lets a firm invalidate the original result and test again under a sampling plan set before the failure occurred — the guidance is explicit that retesting a sample repeatedly until a passing number turns up is unscientific, and is a symptom of skipping the investigation rather than a substitute for running one.

Phase II opens only once Phase I can't close the question

When no assignable cause turns up in the lab, the investigation widens into Phase II: a full-scale review that reaches past the bench into manufacturing — batch production records, in-process data, equipment and environmental logs, and the performance of related batches — run jointly by quality and production rather than by the lab alone. The two-phase structure the current guidance sets out traces back in large part to a 1993 federal court decision, United States v. Barr Laboratories, which held that an out-of-specification result doesn't automatically mean a batch has failed and required an investigation to confirm or refute it before FDA could treat the result as conclusive. The holding didn't excuse a weak investigation; it just refused to let FDA skip to a conclusion either.

A confirmed OOS is a decision about the batch, not just a closed file

If Phase II turns up no explanation in the process either, the original result stands, and the batch is treated as having failed to meet its specification — which is the same fork this blog has already drawn on the device side, where a complaint record has to decide whether what it describes is actually a nonconformance before it becomes a CAPA. The written record 211.192 requires has to state that conclusion and the disposition decision it leads to, not just the steps the investigation walked through — an investigation that documents every test run but never states what the batch's fate was has recorded an activity log, not the written conclusion the regulation actually asks for.

Where this meets the file

An OOS investigation worksheet built around the Phase I/Phase II split — the lab's own assignable-cause check kept as a distinct, time-stamped section before any production-side investigation opens — is previewed in the launch catalog alongside this blog's other quality-investigation templates. If your site runs this investigation in a different order, the shelf takes that correction directly.