Twelve FDA warning letters issued between 3 December 2024 and 24 July 2026 cite cleaning and cross-contamination control. Read together rather than case by case, they stop looking like twelve incidents and start looking like one audit checklist: worst-case selection, hold times, recovery, MACO justification, visual-only strategies, sampling locations, change control and lifecycle management — in that order of frequency.
Twelve warning letters, one checklist
The cleaning-related enforcement record between December 2024 and July 2026 is unusually consistent. Twelve FDA warning letters issued by CDER to twelve different firms — API manufacturers, finished-dose contractors and OTC producers across India, China, Canada, Australia and the United States — cite cleaning, cross-contamination control or equipment status as a significant deviation. Read individually, each is a company story. Read together, they are one document.
The recurring request from FDA is almost verbatim across the set. Six elements appear again and again: identify and evaluate the worst-case drug on the shared train; establish and justify maximum hold times before cleaning and, where relevant, clean hold times; demonstrate that the sampling and analytical procedure can actually retrieve and detect the residue; justify the carry-over limit rather than defaulting to a number; stop treating visual inspection as the sole acceptance criterion; and justify the sampling locations — including the locations that are hardest to clean. Around them sit two systemic findings that are cited just as often but discussed less: no re-evaluation when products or equipment change, and no lifecycle management of the validated cleaning state.
This synthesis covers all twelve cases without re-telling any of them; individual case analyses are published separately in this series. What follows is what the pattern says about the state of cleaning validation practice — and what a programme owner can use as a self-audit before an inspector does it for them.
1. Worst case is asked for in nine of twelve letters, in nearly identical words
Nine of the twelve letters — from Micro Orgo Chem onward — contain the same enumerated list of conditions FDA expects a cleaning validation to address: drug products with higher toxicities, higher potencies, lower solubility in cleaning solvents, characteristics that make them difficult to clean, swabbing locations in areas most difficult to clean, and maximum hold times before cleaning. The wording is stable enough across 2024, 2025 and 2026 letters that it functions as a standing expectation, not a case-specific demand. Brands International, Wuhu Nuowei, Granules India, MediNatura, Cohance Lifesciences, Cosmetic Manufacturers, Chemco, Harbin Jixianglong and Dabur all received some or all of that list.
The important part is where the worst-case determination was absent or stale, not merely incomplete. Cohance's 2018 cleaning validation study did not use the drug product the firm had itself identified as worst case. ABBE Laboratories' 2009 kettle study contained no worst-case determination at all and did not even record which drug product had been cleaned during the study; the firm then began manufacturing End Zit Acne Drying Lotion — its own most-difficult-to-clean product because of the titanium dioxide active — on the same shared kettle without a new study. Granules India argued low risk using retrospective MACO calculations; FDA rejected the argument and stated that the batch-size denominators used were inconsistent with the size of the relevant vessels and that the limits were not adequately justified.
Why it matters. Worst-case selection is the load-bearing assumption of every cleaning validation: it defines which product, which residue and which limit the study actually covers. If the selection is missing, unrecorded or overtaken by a new product, the study describes equipment and products that no longer exist. What to check. Take the cleaning validation report for every non-dedicated train and answer four questions: which product was declared worst case, on what documented criteria, when was that determination last re-evaluated against the current product list, and would a new product introduced since then displace it? ABBE and Cohance show that both answers — "no determination" and "stale determination" — are cited findings.
2. Hold times are required as data, not as convention
Hold times appear in FDA's remediation requests with the same regularity as worst case, but with less visible practice to support them. In every letter where the enumerated list appears, "maximum hold times before cleaning" is included — the dirty hold time. ABBE's version extends to maximum dirty and clean hold times. MediNatura and Cosmetic Manufacturers received the same request in the context of campaign production, where carry-over between a prior batch and the next is a routine, not exceptional, exposure.
The underlying finding is usually documentation, not arithmetic. MediNatura lacked equipment usage logs and cleaning documentation to demonstrate that equipment was adequately cleaned and to prevent carry-over contamination from a prior batch during a production campaign. Micro Orgo Chem was unable to demonstrate that rinse and swab samples had been collected and tested as its own procedure required. Where the logbook itself is unreliable, no hold-time limit can be shown to have been respected — and in two 2026 letters the logbook was worse than unreliable: Dabur's letter cites a falsified equipment usage logbook among the quality-unit findings, and Granules India's inspection ended with torn CGMP records in at least 15 plastic waste bags.
Why it matters. A hold time is an operational control that must be enforceable on the floor; a number in a protocol that the plant cannot demonstrate is an open finding. What to check. Confirm that each validated cleaning process states a maximum time-before-cleaning and — where the process uses one — a maximum clean hold time, that the limit was derived from data rather than convention, and that equipment usage and cleaning records are complete enough to show compliance between the two. Then test one campaign retroactively and see whether the records alone would allow an inspector to reconstruct it.
3. Recovery and detectability: the control has to be able to see the risk
Four letters address the sampling and analytical side directly, and the language is unusually blunt. Granules India: swabs taken from areas after HEPA filters detected residues of multiple previously manufactured drug products plus TNTC microbial contamination, with no documented cleaning procedures for the duct sections between the HEPA filter and the cleaning point. FDA rejected the firm's reliance on limited reserve-sample testing and stated in the letter that cross-contamination is not uniform, that testing of limited reserve samples alone cannot ensure products are contaminant-free, and that while MACO is helpful in validated cleaning scenarios, it cannot apply where cleaning has not been performed or exposure is inconsistent. That letter also requests limits that take recovery study results into account. Cohance Lifesciences produced the sharpest formulation of the principle: visible residues on capsule filling machine PD/CFM-03 that had been documented as clean and had undergone changeover cleaning, analytical testing confirming multiple APIs above the firm's allowable limit, and FDA's statement that the firm's cleaning verification "lacks the ability to detect cross-contamination from carryover." FDA additionally asked Cohance to establish the adequacy of its analytical methodology and to investigate unknown peaks found in reserve samples. Harbin Jixianglong's study did not contain the swab sample testing results its own cleaning procedure required. Cosmetic Manufacturers' protocol required swab samples from different locations in each qualification run, which FDA stated was not designed to demonstrate reproducibility of the cleaning process.
Why it matters. Recovery and method capability are what connect a reported result to a limit. Where the method cannot find the residue at the carry-over limit, or where sampling varies between runs, an acceptable result is not evidence of a clean surface. What to check. For each residue method, does the recovery-corrected limit of quantification sit below the carry-over limit, and is the recovery study documented with the surfaces and worst-case residue? Are sampling locations fixed and justified per equipment item — with a defensible explanation for any variation between runs — and would the analytical method detect the lowest-limit product in the train? Cohance shows that visible residue and passing documentation can coexist; only an independent measurement separates them.
4. MACO is a calculation, not a control — and the limit has to be justified
Three letters turn specifically on carry-over limits. Granules India used retrospective MACO calculations on equipment that had, in FDA's finding, not been reliably cleaned; the letters state that the calculation cannot substitute for the cleaning step. Harbin Jixianglong's cleaning procedure required a maximum allowed carryover limit based on API toxicity — but no limit was established in the cleaning validation study for equipment used to manufacture peptide APIs including semaglutide and tirzepatide, and no justification was provided for the swab sampling locations. Wuhu Nuowei's tank cleaning protocol contained no acceptance criteria for carryover residues at all. Chemco's letter records white residue in Tank #68 that the quality unit had verified as clean in the tank cleaning logbook, and a statement during inspection that no cleaning validation had been performed for non-dedicated equipment.
Why it matters. A limit is a claim about acceptable patient exposure, and it has to be traceable to a toxicological basis plus the size of the equipment actually used. Granules India shows both failure modes at once: a limit derived from denominators that did not match the vessels, and a limit applied to a step that was not performed. What to check. Can you trace each carry-over limit to its toxicological or health-based basis, to the batch size and surface area actually in the train, and to the recovery of the method used to test against it? And is the limit applied only to cleaning steps with demonstrated execution — not to campaigns and ducts where cleaning was partial or undocumented?
5. Visual-only strategies fail, but visual inspection is not the problem
Four letters address visual-only or visually-dependent cleaning control. Wuhu Nuowei's cleaning protocol for the tank "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 firm's updated protocol with microbiological data was still found inadequate because analytical testing for potential carryover residues had not been considered. Dabur India's Unit/Line cleaning validation "relied exclusively on visual inspection", with FDA listing the missing elements explicitly: scientifically determined maximum allowable carryover limits, quantitative residue testing, and direct surface sampling of difficult-to-clean areas — no cleaning validation had been performed although multiple different OTC products were manufactured on the equipment. Cosmetic Manufacturers' request included swabbing locations for the hardest-to-clean areas. In the background of the same letters sits ASTM E3263-22e1, the standard for qualifying visual inspection for residues, whose scope requires a statistically valid visual detection limit, inspector qualification, an acceptable hazard analysis, a validated and repeatable cleaning process and a suitable safety margin between the visual residue limit and the maximum safe surface residue.
Why it matters. The failure is not using eyes; it is using eyes as the only control, without the qualification that makes a visual criterion defensible and without knowing where the limit of visual detection sits relative to the carry-over limit. What to check. If visual inspection is an acceptance criterion on a non-dedicated line, is there a qualified visual detection limit for the residue and surface in question, qualified inspectors, and a documented comparison against the lowest carry-over limit in the train? Dabur's letter shows what happens when that comparison is absent: the firm's remediation — validated methods, carryover limits, direct surface swabbing, revalidation and line dedication — had to be demanded by FDA rather than adopted voluntarily.
6. Sampling locations are part of the evidence, not a sampling detail
Sampling design appears in five letters. Cosmetic Manufacturers required swab samples from different locations in each of the qualification runs, which FDA stated was not designed to demonstrate reproducibility. Micro Orgo Chem's requested worst-case evaluation includes the most difficult-to-clean swabbing locations, as does the FDA list reproduced in the Brands International, Granules India, Cohance and Dabur letters. Harbin Jixianglong's study provided no justification for its swab sampling locations. Wuhu Nuowei's and Micro Orgo Chem's letters sit in the same territory from the other direction: FDA asked for risk-based identification of where cross-contamination actually resides, including duct sections between the HEPA filter and the cleaning point and other equipment for which no cleaning procedure had been provided.
Why it matters. A validation run only covers what it samples. Rotating locations across runs trades evidence of reproducibility for nominal coverage, and it is the reproducibility — three consistent runs — that the validation claim rests on. What to check. Are sampling locations fixed per equipment item, mapped to the equipment drawing, and chosen by a documented risk review of flow, geometry, dead legs, seals, flanges and drainage? If locations change between runs, is that a deliberate, justified widening of coverage in addition to — not instead of — consistent worst-case points? And for enclosed sections such as ducts, is there a defined sampling or rinse strategy at all, or does that portion of the equipment simply exit the validated boundary?
7. Change control is the single most repeated systemic gap
Six letters describe a validated cleaning state that was invalidated by change without being re-evaluated or revalidated. MediNatura's own response acknowledged "that the cleaning validation is not reflective of your current operations, because you continued to purchase new equipment and add new drug products to your operations", and FDA asked for a described change-management step before introducing new manufacturing equipment or a new product. Cohance did not re-evaluate its cleaning validation programme when new drug products were introduced onto shared equipment. ABBE's validation was from 2009 and no new study was performed when a new product entered the shared kettle. Wuhu Nuowei had no cleaning procedures for other equipment used in API manufacture. Cosmetic Manufacturers substituted a different tank without adequate justification — no discussion of materials of construction and interior features such as drains and ports, and no commitment to revalidate. Chemco's and Dabur's letters likewise request change-management provisions and updated SOPs. FDA's standing request across the set — improved SOPs for the verification and validation of cleaning procedures for products, processes and equipment — is the procedural counterpart.
Why it matters. Every one of these findings could have been detected internally, prospectively, in a change-control step that asked a single question: does this change invalidate a cleaning validation assumption? In none of the letters did that happen. What to check. Does a new product or a new equipment item trigger a documented cleaning-impact assessment before it enters a shared train — covering worst-case selection, limit derivation, sampling locations, hold times and the analytical method? And has the current product and equipment list been compared against the last cleaning validation study, retroactively, with the differences closed?
8. Lifecycle: equipment released as clean, and a state nobody maintains
The most direct expression of the lifecycle gap is equipment status. Aspen Biopharma Labs' inspector observed manufacturing equipment labelled "Cleaned" containing liquid with a floating substance or residue on product-contact surfaces, with no cleaning validation studies to support the cleaning procedures — a need the firm had already identified in an internal deviation report. Chemco: white residue in Tank #68, verified as clean in the logbook, on equipment for which no cleaning validation had been performed. Cohance: visible stains and residues on equipment documented as clean, with analytical results above the allowable limit. Micro Orgo Chem: equipment that appeared dirty and not appropriately maintained, with FDA noting that inadequate cleaning and maintenance of non-dedicated equipment may lead to cross-contamination between drug products. ABBE stopped documenting cleaning activities and equipment status around 2020, contrary to its own procedure ABBE-010. MediNatura's last official validation study dated from 2007 under previous ownership, and FDA asked for a summary of vulnerabilities specifically in the process for "lifecycle management of equipment cleaning". The remedy FDA requested in several letters is the same systemic instrument: an independent, comprehensive retrospective assessment of cleaning effectiveness across every piece of equipment used for more than one product, a CAPA plan built on it, and — in Cohance's case — an independent consultant review of CAPA effectiveness, alongside the recommendation to engage a qualified CGMP consultant.
Why it matters. A validated state decays through changes nobody records and documents nobody updates, and the equipment status label is where the decay becomes visible. What to check. Does each equipment item have a current, documented cleaning status backed by records, and is the validated state maintained under a named owner with periodic review against the current portfolio? If a retrospective assessment were requested tomorrow, could you produce it — scope, residues, equipment list, and a conclusion on products released to the market?
The synthesis: an audit checklist, not twelve stories
The pattern across the twelve letters is not a rise in cleaning violations discovered; it is a regulator asking the same structural questions regardless of product type, region or facility size. Worst-case selection, hold times, recovery and detectability, justified limits, defensible visual inspection, fixed and justified sampling locations, change-triggered re-evaluation and lifecycle management of the validated state form a single chain. Break one link and the others stop protecting anything: a perfect MACO calculation applied to a cleaning step that was not performed (Granules India), a worst case superseded by a new product (ABBE, Cohance), a visual check used where the residue limit sits far below the limit of detection (Dabur, Wuhu Nuowei), or a status label that no longer reflects the tank (Aspen, Chemco).
Two secondary observations are worth carrying forward. First, FDA's responses to remediation plans are consistently about completeness and interim control: responses were repeatedly found inadequate for lacking an impact assessment of distributed product within expiry, for lacking interim cross-contamination controls during the remediation period (Dabur), for limited retrospective testing that was statistically insufficient (Dabur), or for lacking a systemic CAPA addressing the mismatch between procedure and study (Harbin Jixianglong). A reply to a 483 that does not cover released product and interim controls is unlikely to close the finding. Second, procedural maturity did not prevent cleaning findings: letters in this set also cite missing annual product reviews, unvalidated methods, data integrity issues and falsified or torn records — but none of those, on their own, is a cleaning violation. The cleaning findings stand alone, and they are specific enough to be tested before publication of a validation report rather than after an inspection.
Sources
- US FDA Warning Letter 320-25-21, Micro Orgo Chem, MARCS-CMS 686458 (3 December 2024) — https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/micro-orgo-chem-686458-12032024
- US FDA Warning Letter 320-25-25, Brands International Corporation, MARCS-CMS 689983 (17 December 2024) — https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/brands-international-corporation-689983-12172024
- US FDA Warning Letter 320-25-38, Wuhu Nuowei Chemistry Co., Ltd., MARCS-CMS 697727 (4 February 2025) — https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/wuhu-nuowei-chemistry-co-ltd-697727-02042025
- US FDA Warning Letter 320-25-48, Granules India Limited, MARCS-CMS 697115 (26 February 2025) — https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/granules-india-limited-697115-02262025
- US FDA Warning Letter 320-25-50, Aspen Biopharma Labs Private Limited, MARCS-CMS 698665 (5 March 2025) — https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/aspen-biopharma-labs-private-limited-698665-03052025
- US FDA Warning Letter 320-25-98, ABBE Laboratories, Inc., MARCS-CMS 708309 (6 August 2025) — https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/abbe-laboratories-inc-708309-08062025
- US FDA Warning Letter 320-26-27, MediNatura New Mexico, Inc., MARCS-CMS 716025 (10 December 2025) — https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/medinatura-new-mexico-inc-716025-12102025
- US FDA Warning Letter 320-26-40, Cohance Lifesciences Limited, MARCS-CMS 718812 (30 January 2026) — https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/cohance-lifesciences-limited-718812-01302026
- US FDA Warning Letter 320-26-43, Cosmetic Manufacturers (Aust) Pty Ltd., MARCS-CMS 719225 (6 February 2026) — https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/cosmetic-manufacturers-pty-ltd-719225-02062026
- US FDA Warning Letter 320-26-60, Chemco Corporation, MARCS-CMS 721592 (7 April 2026) — https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/chemco-corporation-721592-04072026
- US FDA Warning Letter 320-26-73, Harbin Jixianglong Biotech Co., Ltd., MARCS-CMS 723330 (1 May 2026) — https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/harbin-jixianglong-biotech-co-ltd-723330-05012026
- US FDA Warning Letter 320-26-105, Dabur India Limited, MARCS-CMS 728828 (24 July 2026) — https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/dabur-india-limited-728828-07242026
- 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
- EU GMP Annex 15, Qualification and Validation (visual inspection as part of, not a substitute for, acceptance criteria)
- US FDA, Guide to Inspections of Validation of Cleaning Processes (1993/1998 inspection guide) — https://www.fda.gov/media/71023/download
Validation takeaway
Across twelve FDA warning letters issued between 3 December 2024 and 24 July 2026, cleaning findings are not random and not product-specific: they converge on eight recurring themes — worst-case selection, hold times, recovery and detectability, justified carry-over limits, visual-only strategies, sampling locations, change-triggered re-evaluation and lifecycle management of the validated state. The recurring failure is not the mathematics but the correspondence between the validation document and the plant: a limit applied to a cleaning step that was not performed (Granules India), a worst case superseded by a product or equipment change (ABBE, Cohance, MediNatura), a visual check standing in for measurement (Wuhu Nuowei, Dabur), or an equipment status label that no longer matches the tank (Aspen Biopharma, Chemco). Because FDA asks the same questions in nearly identical words regardless of firm, region or dosage form, this set functions as a usable pre-inspection self-audit: verify worst-case currency, hold-time data, method detectability against the lowest carry-over limit, fixed and justified sampling locations, a change-control gate before new products or equipment enter a shared train, and a maintained validated state with an owner and records. Remediation plans that omit interim controls and an impact assessment of distributed product within expiry have been found inadequate repeatedly.
Questions worth asking in our own programme
- For each non-dedicated equipment train: which product is documented as worst case, on what criteria, and when was that determination last re-evaluated against the current product list? ABBE Laboratories and Cohance both received findings for a missing or stale worst-case selection.
- Does every validated cleaning process state a maximum hold time before cleaning — and a clean hold time where one applies — and do the equipment usage and cleaning records allow an inspector to reconstruct any campaign retroactively?
- Is the recovery-corrected limit of quantification of each residue method below the carry-over limit of the lowest-limit product in the train, and where it is not, what control closes the gap?
- Can we trace each carry-over limit to its toxicological basis and to the batch size and surface area actually used — and is that limit applied only to cleaning steps with documented execution?
- If visual inspection is an acceptance criterion anywhere on a non-dedicated line, do we have a qualified visual detection limit, qualified inspectors, and a documented comparison against the lowest carry-over limit in the train?
- Are swab sampling locations fixed per equipment item, mapped to the drawing and justified by a documented risk review — and if they vary between validation runs, can we defend that variation as additional coverage rather than as a substitute for reproducibility?
- Does a new product or a new equipment item trigger a documented cleaning-impact assessment before it enters a shared train, covering worst-case selection, limit derivation, sampling locations, hold times and the analytical method?
- Does each equipment item have a current documented cleaning status backed by records, with a named owner who periodically reviews the validated state against the current portfolio?
- If FDA requested a comprehensive independent retrospective cleaning-effectiveness assessment tomorrow, could we produce the scope, the residue list, the equipment list and a conclusion on products already released to the U.S. market?
- Do our responses to cleaning findings address interim cross-contamination controls for the remediation period and an impact assessment of distributed product within expiry — the two omissions FDA cited most often when finding a response inadequate?
- Have we compared the current product and equipment list against the last cleaning validation study, closed the differences prospectively, and recorded the outcome in change control rather than in a retrospective file?
