“Contamination occurred during handling” — the OOS reflex FDA keeps rejecting
483 response filed 6 November 2025 · Citations 21 CFR 211.192, 211.113(a), 211.100(a), 211.166(a)
A finished-product sample failed a microbiological limit. The lab could not identify the organism. The firm concluded the failure came from handling, resampled, got a passing result, and released the lot. Every step of that is the standard reflex — and FDA rejected all of it.
Issues 01 and 02 were about how firms answer a 483. This one is about the investigation that produces the answer. The failure here is upstream of the response: an out-of-specification result invalidated without ever establishing why the original result was wrong.
It is the most common laboratory error in the industry, and it almost never survives an inspection.
What FDA found
The firm — a contract manufacturer of topical prescription and OTC products — hit failing results in two separate places, and disposed of both the same way.
On the microbiology side, a finished lot failed total yeast and mould count. The contract testing laboratory could not identify the fungal species. With no organism identified and no laboratory error found, the firm nonetheless concluded — in FDA’s words, “without evidence” — that contamination had occurred “during handling, transportation and/or testing.” It resampled, the resample passed, and the lot was released.
On the stability side, the same move appeared twice more. The firm invalidated out-of-specification subpotent assay results, relying on resample data “with no identified laboratory error or rationale.”
A recurring phrase runs through the letter: the firm labelled wider storage conditions “at the instruction of your customer,” used a customer’s validation study, and stopped a validation “at the direction of your customer.” The lab was deferring — to the product owner, and to the more convenient of two results.
What the firm said
The response was constructive, and it conceded the central point more candidly than most. The firm committed to stop using the rework treatment on failing batches, to place reworked lots on real-time stability, and to reopen the complaints it had closed. On the customer-deference thread it went further, stating plainly that it “can no longer rely on written justifications from product owners to deviate from the general [C]GMP requirements.”
That is a firm that understood what had gone wrong. It still fell short — because acknowledging the practice is not the same as bounding its consequences.
Why FDA rejected it
A passing resample is not a root cause
This is the whole matter in one line. Invalidating a failing result requires conclusively identifying a laboratory error that explains it. A second sample that happens to pass explains nothing — microbiological contamination is not uniformly distributed, so a clean resample is as consistent with a real failure as with a false one. “Probably handling” is a hypothesis. FDA needed it proven, and it was not.
No organism, no error, no basis
The contract lab could not identify the fungal species. That absence is not neutral — it removes the only evidence that could have supported a laboratory-error conclusion. With nothing identified, the default is that the result stands and production is investigated, not that the result is set aside.
The scope was never bounded
FDA’s repeated demand is the same one from Issue 01: a retrospective, independent review of every invalidated failing/OOS result on the market and within expiry, each assessed for whether laboratory error was conclusively or only inconclusively established. The response invalidated individual results without ever asking how many others rested on the same reasoning.
“At the customer’s instruction” is not a defence
FDA’s position on contract manufacturing is categorical: “You are responsible for the quality of drugs you produce as a contract facility regardless of agreements in place with product owners.” A quality agreement allocates work. It does not transfer the obligation to reject a bad batch.
The pattern worth carrying forward
The OOS reflex is seductive because each step feels reasonable in isolation. A result fails. You suspect the lab, because labs do make errors. You resample. It passes, which feels like confirmation. You release. But the logic runs backwards: you began with the conclusion you wanted — the batch is fine — and treated a passing resample as permission to reach it. The passing result was never evidence of laboratory error; it was just a second draw from a batch whose contamination may not be evenly spread.
The discipline that separates a defensible invalidation from an indefensible one is simple to state and hard to hold to: you may invalidate a result only when you can name what went wrong in the lab, not when you have found a way to make the problem disappear.
- Can you name the specific laboratory error, with evidence? Not “possible” or “likely” — identified. A named cause with supporting data invalidates a result. A plausible story does not.
- Are you relying on a passing resample as your justification? If the passing retest is the evidence, you have no evidence. A resample can support a conclusion already established by root cause; it cannot be the root cause.
- If no laboratory cause is found, have you investigated production? An unexplained OOS is not a closed OOS. When the lab is clean, the batch record, equipment, and process are where the investigation goes next — not into the release file.
A defensible response here would have reopened every invalidation built on resampling, established for each whether a real laboratory error existed, and moved the ones that failed that test into a full production investigation. The firm was moving in that direction. It had not yet gone back to bound what the practice had already released.
An OOS result is a question the batch is asking. Resampling until it passes doesn’t answer the question — it just stops listening.
Veritas Quality Consultants works with pharmaceutical, medical device, and biologics manufacturers on Form 483 responses, root cause analysis, and CAPA development — built to FDA expectations inside the 15-day window.
Request a consultation