In Warning Letter 320-25-38 the FDA did not attack visual inspection as a technique. It objected to a cleaning programme in which visual inspection was the only evidence on the table — no acceptance criteria for carryover residues, no analytical data, and no procedures for the rest of the equipment train.
Visual Inspection Alone Was Not Enough: What FDA Expected to See
In Warning Letter 320-25-38 the FDA did not attack visual inspection as a technique. It objected to a cleaning programme where visual inspection was the only evidence on the table — no acceptance criteria for carryover residues, no analytical data, nothing covering the rest of the equipment train.
What happened
The FDA inspected Wuhu Nuowei Chemistry Co., Ltd. (FEI 3014438694), Jingde, Xuancheng, China, 2–6 September 2024. Warning Letter 320-25-38 — MARCS-CMS 697727 — is dated 4 February 2025 and published by CDER on 11 February 2025. It cites three deviations from CGMP for APIs: impurity test methods and specifications, quality-unit oversight of stability data, and cleaning.
The Warning Letter identifies the third as "Failure to establish adequate written procedures for cleaning equipment and its release for use in manufacture of API." On the substance: "You provided a protocol for cleaning the (b)(4) tank used to manufacture (b)(4), which does not include acceptance criteria for carryover residues. Instead, you relied on visual inspection to determine cleanliness of your equipment and could not provide any analytical or microbiological data to support the efficacy of your cleaning procedures." The equipment was shared, used for multiple APIs.
The firm responded with an updated tank protocol and report, supported by microbiological data. FDA still called the response inadequate — analytical testing of carryover residues had not been considered, and no cleaning procedures were provided for the other equipment used in API manufacture. Microbiological data is not worthless, but it does not demonstrate removal of a chemical residue.
The requested actions read like a cleaning lifecycle checklist: improvements "with special emphasis on incorporating conditions identified as worst case"; updated SOPs ensuring "an appropriate program is in place for verification and validation of cleaning procedures for products, processes, and equipment"; and "a risk evaluation of potential cross-contamination due to inadequate cleaning for APIs within expiry in distribution in the United States." FDA placed the firm on Import Alert 66-40 on 24 January 2025; the company stated it had ceased US-market production and deregistered its facility.
Why it matters
Cleaning findings are common; this one is worth reading for how narrowly it is scoped. The agency did not demand that visual inspection be abolished — it demanded that a release decision rest on stated acceptance criteria and on data that addresses the residue being controlled. Where a protocol says "visually clean" and nothing else, there is no limit to compare against, no way to challenge a result, and no defensible basis for releasing shared equipment into the next product.
It also frames cleaning as a release activity, not housekeeping: the citation heading is "cleaning equipment and its release for use in manufacture of API".
What is particularly interesting
First, visual inspection was not the accused — it was the sole evidence. ASTM E3263 does accept visual inspection as a basis for releasing equipment, but on conditions: a statistically determined visual detection limit, qualified inspectors, a justified hazard analysis, a validated and repeatable cleaning process, and — in section 4.5 — "a suitable safety margin between the VRL and MSSR." None of that scaffolding appears here. The letter shows what visual inspection looks like without the qualification that makes it defensible.
Second, the response package mattered as much as the original gap: microbiological data in reply to a carryover-residue finding answers a different question. The reply was judged inadequate not for lack of effort but for a mismatch between evidence and risk.
Third, the corrective action reached product already shipped — a cross-contamination risk evaluation for APIs still within expiry and in US distribution. Fixing the procedure going forward does not close the gap behind you.
How such a gap can realistically develop
This pattern rarely comes from deliberate corner-cutting; it develops through three quiet compression steps. Visual inspection starts as a sensible additional check on top of swab and rinse data, then becomes the primary criterion because it is fast and always passes. The protocol stays generic — "visually clean, no residue, no odour" — and the carryover limit for the worst-case product pair never reaches the equipment-specific procedure. Then, because qualification focuses on microbial control for a shared tank, the analytical side is never commissioned. Each step is locally reasonable; together they yield a programme that looks complete on paper and has never been challenged with a number.
Questions worth asking in our own programme
- Does our cleaning protocol state numerical acceptance criteria for carryover residues — per equipment train, for the worst-case product — or does it describe equipment as "visually clean"?
- If visual inspection is used as a release criterion, can we produce the qualified visual detection limit, inspector qualification records, and the demonstrated margin to the health-based exposure limit?
- When a cleaning finding arises, does our response package contain analytical residue data from the actual worst-case surfaces, or does it stop at microbiological swabs?
- Are cleaning procedures defined per equipment item — tanks, hoses, filters, dedicated as well as shared — rather than assumed to be covered by a single generic procedure?
- When a cleaning gap is identified, does our risk assessment extend to product already in distribution?
- Can our documentation show how worst-case conditions (product, soil, contact time, hardest-to-clean location) were selected and justified?
Wider context
Cleaning validation expectations have converged for years: health-based exposure limits as the toxicological anchor, worst-case justification as the design principle, analytically supported residue data as the evidence. Visual inspection has an established, standards-backed role inside that framework — a qualified method with a characterised detection limit, not a default. Read alongside Import Alert 66-40, the letter shows how one missing layer of evidence in a single unit operation can contribute to a facility-level outcome. The lesson is durable: whichever techniques a programme uses, the release decision should name the limit it was made against.
Sources
- U.S. Food and Drug Administration, Warning Letter 320-25-38, MARCS-CMS 697727, Wuhu Nuowei Chemistry Co., Ltd., 4 February 2025 (published 11 February 2025): https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/wuhu-nuowei-chemistry-co-ltd-697727-02042025
- FDA, Import Alert 66-40: https://www.accessdata.fda.gov/cms_ia/importalert_189.html
- ASTM E3263-22e1, Standard Practice for Qualification of Visual Inspection of Pharmaceutical Manufacturing Equipment and Medical Devices for Residues: https://store.astm.org/e3263-22e01.html
- Pharma Manufacturing, "FDA slaps two Chinese API manufacturers with warning letters over cGMP violations": https://www.pharmamanufacturing.com/industry-news/news/55267470/fda-slaps-two-chinese-api-manufacturers-with-warning-letters-over-cgmp-violations
- RAPS, "Drugmakers receive FDA warning letter for cGMP violations…", 11 February 2025: https://www.raps.org/resource/drugmakers-receive-fda-warning-letter-for-cgmp-vio.html
- Fierce Pharma, "FDA hands out warning letters to 2 Chinese API manufacturers": https://www.fiercepharma.com/manufacturing/fda-slaps-china-based-api-makers-separate-warning-letters-outlining-quality-testing
- The FDA Group Insider, "FDA Uncovers Critical Manufacturing Violations at Chinese API Makers": https://insider.thefdagroup.com/p/fda-uncovers-critical-manufacturing
Validation takeaway
Visual inspection remains a legitimate part of a cleaning validation programme, but it is defensible only as a qualified method with a characterised detection limit, trained inspectors, and a demonstrated margin to the health-based residue limit. Where a shared-equipment protocol states no acceptance criteria for carryover residues and offers no analytical data, the release decision has no limit to be made against — which is exactly what FDA cited here.
Questions worth asking in our own programme
- Does our cleaning protocol state numerical acceptance criteria for carryover residues — per equipment train, for the worst-case product — or does it describe equipment as "visually clean"?
- If visual inspection is used anywhere as a release criterion, can we produce the qualified visual detection limit, inspector qualification records, and the demonstrated margin between that limit and the health-based exposure limit?
- When a cleaning finding arises, does our response package contain analytical residue data generated on the actual worst-case surfaces, or does it stop at microbiological swabs?
- Are cleaning procedures defined per equipment item — tanks, hoses, filters, dedicated as well as shared — rather than assumed to be covered by a single generic procedure?
- When a cleaning gap is identified, does our risk assessment extend to product already in distribution, not only to future manufacture?
- Can our validation documentation show how worst-case conditions (product, soil, contact time, hardest-to-clean location) were selected and justified?
