What does 21 CFR 211.192 require?

It requires that any unexplained discrepancy or batch failure be thoroughly investigated, that the investigation extend to other batches and products that may have been associated, that the investigation be recorded in writing, and that it include conclusions and follow-up.

Does 211.192 apply to out-of-specification results?

Yes. An OOS result is an unexplained discrepancy, so the same requirement applies: a thorough investigation, an assessment of whether other batches were affected, a written record, and conclusions with follow-up.

What is the most common gap in a 211.192 investigation record?

The other-batches assessment. The regulation asks the investigation to extend to other batches and products that may have been associated with the failure, and records frequently omit it or assert it in a single sentence with no reasoning.

What 21 CFR 211.192 requires in an investigation

Resources/What 21 CFR 211.192 requires in an investigation

What the regulation says

The text is short. It requires that any unexplained discrepancy or the failure of a batch or its components to meet any of its specifications be thoroughly investigated, whether or not the batch has already been distributed. The investigation must extend to other batches of the same drug product and other drug products that may have been associated with the failure. It must be recorded in writing, and it must include the conclusions and follow-up.

That is four obligations in one paragraph: thorough, extended, written, concluded.

The regulation does not name a form, a template, a timeline, or a root-cause methodology. Those come from your own procedures. What it fixes is the content the record has to contain — which is why an investigation can follow your SOP exactly and still be thin against 211.192.

The four things, and what each looks like on the page

1. Thoroughly investigated

Thorough is not a length. A thorough investigation shows what was considered and ruled out, not only what was concluded. A record that names one cause and stops has not shown its work — a reader cannot tell whether alternatives were dismissed for a reason or never raised.

Thin

Investigation determined the cause to be a calibration drift on Balance B-14.

Shows its work

Balance B-14 calibration, sample handling, analyst technique and reagent lot were each assessed. Reagent lot and technique were excluded (same lot and analyst produced conforming results on the preceding four samples). Calibration records show drift exceeding tolerance at the verification on 12 March.

2. Extended to other batches

This is the requirement most often missing, and the one an investigator asks about first — because it is the question that decides whether product already released is affected. The record needs to say which other batches were considered, on what basis, and what was concluded. A single sentence asserting no impact is an answer without a reason.

Assertion

No other batches were impacted.

Assessment

Balance B-14 was used for batches 24-118 through 24-127 between the 12 March verification and the 2 April failure. Each was reviewed against the drift magnitude; 24-121 and 24-124 fell within the affected weight range and were placed on hold pending re-assay. The remaining eight were outside it, per the calculation in Attachment C.

Note what the second version gives a reader: a boundary, a method, a number of batches, and a place to check. None of that requires the conclusion to be different.

3. Recorded in writing

The obligation is the record, not the meeting. Discussion that happened but was never written is, to a reader, discussion that did not happen. This is where investigations lose their reasoning: the assessment was genuinely done, in a room, and the record captured only its output.

4. Conclusions and follow-up

A conclusion states what happened and why it was allowed to. Follow-up states what changes and how anyone will know it worked. An investigation that ends at a conclusion has satisfied half of the fourth requirement.

The swap test. Read the conclusion, then read the event title. If you could exchange one for the other and lose no information, the conclusion is restating the event rather than explaining it. That is the circular root cause pattern, and it is the single most common reason a 211.192 record reads as thin.

The four gaps, in the order they get found

No other-batch reasoning. Present as a sentence, absent as an assessment.
A conclusion that restates the event. "Root cause: analyst did not follow the sampling plan" for an event titled "Sampling plan not followed."
Follow-up with no verification. An action is named; nothing says how anyone will know it worked, or when that gets checked.
A procedure cited without a revision. "Performed per SOP-220" leaves the reader unable to tell whether the version in force on the event date is the version described.

Where this sits in Dry Run. The quality-event criteria pack checks the mechanical parts of this — whether a procedure is cited with a revision, whether scope is stated, whether a conclusion restates its event, whether follow-up names a verification. Each criterion carries its regulatory reference as fixed text written by a person; the model is instructed not to produce citations, because a citation is a checkable fact and generating one fresh per run means getting it right by coincidence.

What it cannot judge is whether your science is correct or your batch boundary is right. Those stay with the reviewer, and a screening that returns REVIEW on them is telling you the document did not contain enough to decide — not that the point is fine.

See it on sample data. The demo screens a canned deviation-style record and flags where an investigation is thin — including scope and follow-up.

See the demo → All resources