“May be retested” — when your acceptance criteria are wide enough to absorb the failure
483 response filed 20 October 2025 · Citations 21 CFR 211.42(c)(10), 211.113(b)
A sterile injectables site in Germany ran decontamination validation cycles for three years and passed nearly all of them. Biological indicators grew in almost every cycle — as many as eight survivors at a time, at various locations inside the equipment. The cycles passed anyway, because the acceptance criteria permitted it.
Issue 05 dealt with corrective actions that could not work. This one deals with something quieter: a quality system operating exactly as written. When a limit is wide enough to absorb the failure it was designed to detect, the record shows a state of control the process does not have — and every review downstream inherits the error rather than catching it.
The mechanism appears twice, in two unrelated systems. Both times the firm defended the result instead of the limit.
What FDA found
FDA inspected the firm’s German site unannounced in September 2025. The findings concern the isolator and RABS lines used to aseptically fill sterile products for the U.S. market.
The first is the decontamination cycle. Positive biological indicators — “upwards of 8 recoveries” — appeared in nearly all validation cycles run between 2022 and 2025, at various locations within each line. Only two were judged unacceptable. FDA is precise about why: “your acceptance criteria were inappropriate in that they allowed routine recovery of multiple BI survivors.” Those cycles did not pass because they worked. They passed because the criterion had room in it.
The second is seal integrity testing. The firm’s own translated procedure permitted a seal that failed pressure decay testing to be retested, with replacement required only after a defined number of failures — and retesting was often deferred. One seal failed on 1 July 2025 and did not pass until 19 July; the production schedule showed batches manufactured during those eighteen days.
- 2022 – 2025
- Biological indicator growth recovered in nearly every decontamination validation cycle. Two cycles rejected.
- Jun 2023 – Sep 2025
- At least 47 microbial recoveries at ISO 5 cleaning use points; 14 exceed the action limit.
- 1 – 19 Jul 2025
- A seal fails leak testing and is not retested for eighteen days. Batches are made in the interval.
- 18 – 26 Sep 2025
- FDA inspects the Halle/Westphalia facility unannounced.
- 3 Mar 2026
- Warning letter. Every response found inadequate.
What the firm said
The response was substantial and technically engaged, which is worth saying plainly. The firm committed to procedural updates, added quality oversight, a rule against proceeding to production “following an invalid or failed test,” lower allowable BI growth, and an investigation for any growth.
“[T]here has been no impact to any released batches or the overall control of sterility assurance at the site.” Firm’s 483 response, as quoted by FDA
It attributed the 2025 validation growth to a “rogue” phenomenon in one biological indicator lot, and the leak test failures to criteria wrongly programmed into the testing device. FDA also recorded the firm’s position that in its operators’ experience pressure decay failures generally do not indicate a non-integral seal, and that they relied on visual inspection instead.
Why FDA rejected it
The limit was the finding, not the result
The firm addressed the failures. It did not address the criterion that had been classifying them as passes since 2022. Lowering allowable BI growth going forward leaves three years of accepted cycles standing on a limit now conceded to be wrong.
“Recurrent failures are indicative of poor cycle robustness and require further investigation.” FDA, Warning Letter 320-26-49
A retest allowance is a written decision to ship on an unconfirmed result
That clause was not an oversight; it was approved, translated and followed. Its effect was that a seal could fail, stay in service, and be confirmed weeks later — or not at all. Eighteen days of production ran on a result the firm had not yet obtained, backed by visual checks FDA dismissed in a line: “Visual inspections, on their own, are insufficient.”
“Rogue” is a claim about one lot, and the data covered several
FDA’s objection is arithmetic rather than scientific: the explanation “fails to recognize the multiple positive BIs that occurred with other BI lots in the same year.”
Absence of observations was read as evidence
Operators can only report failures their method is capable of resolving. A sub-visible breach produces no observation, and no observation was being treated as a passing one.
The pattern worth carrying forward
Nothing here was concealed. The BI recoveries were recorded, the leak test failures were recorded, the retest allowance was written into an approved procedure. Everything FDA needed was already in the firm’s own files — which is why it could say, in good faith, that sterility assurance remained under control. Its indicators said so.
That is what makes this structural rather than careless. A criterion set too wide does not produce alarming data; it produces reassuring data. It converts a signal into a pass, and the pass propagates — into batch release, into annual product review, into the management review concluding the site performs well. Nobody has to ignore anything. The system reports success accurately against the criterion. The criterion is what is wrong, and nothing inside the system can notice, because the criterion is what noticing has been defined as.
- When did this acceptance criterion last reject something? If the honest answer is “it hasn’t,” you cannot yet distinguish a robust process from a permissive limit. A criterion that has never fired has never been tested.
- Does any approved procedure of yours permit a retest before rejection or replacement? Read that clause as an inspector will — as a documented decision to keep operating on an unconfirmed result. Then ask how long the interval runs in practice, and whether anyone has measured it.
- If this test failed today, what would actually stop? Trace it to the batch, the line, the release decision. If production continues while the result is adjudicated, the test is not a control. It is a record.
A defensible response would have inverted the analysis: not explaining the 2025 growth, but asking what the criterion had absorbed since 2022, and stating plainly which batches were made under conditions never actually qualified.
A limit that has never rejected anything is not a limit. It is a description of what you have been doing.
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