Can a Cleaning Validation Demonstrate Reproducibility If Every Run Samples Somewhere Different?

What happened

Warning Letter 320-26-43 (MARCS-CMS 719225), dated February 6, 2026, identifies Cosmetic Manufacturers (Aust) Pty Ltd., FEI 3007187387, at 14 Kingston Drive, Helensvale, Queensland, Australia. The inspection ran from March 31 to April 8, 2025.

Despite a name that reads like a cosmetics business, the site is a contract manufacturer of finished drug products for the U.S. market, and the letter is written under 21 CFR parts 210 and 211. FDA cites four CGMP violations: inadequate batch release testing (211.165(a)), inadequate laboratory controls (211.160(b)), inadequate written production and process control procedures (211.100(a)), and equipment not of appropriate design and suitably located to facilitate cleaning and maintenance (211.63).

The cleaning criticism sits inside that fourth citation. FDA states the firm "[has] also not established whether your cleaning procedures consistently reduce chemical and microbiological residues to acceptable levels", then:

> "The cleaning validation protocol was not designed to demonstrate reproducibility of the cleaning process. For example, the cleaning validation protocol required (b)(4) swab samples to be taken from different locations for each of the (b)(4) qualification runs."

That observation is the heart of the letter. The counts are redacted as (b)(4), but the defect is stated plainly: the sampling locations changed run to run.

A related critique targets equipment equivalence. FDA found the firm "did not adequately justify the decision to use a (b)(4) tank in lieu of the (b)(4) tank used to manufacture the (b)(4) drug products", adding that "there is no discussion of the (b)(4) tanks' material(s) of construction and interior features (e.g., drains, ports, (b)(4))" — exactly the attributes that determine whether two tanks present equivalent surfaces.

FDA also deemed the response inadequate: the firm did "not commit to adequately justify your various (b)(4) methodologies and to revalidate your cleaning process". All drugs and drug products offered for import from the firm went on Import Alert 66-40 on January 8, 2026; the firm stated it has ceased drug production for the U.S. market.

Why it matters

Reproducibility is the entire claim of a validation exercise. Three cleaning runs are meaningful only if they are replicates of the same measurement — same product, equipment, procedure, sampling sites, analytical method — so that variation between runs reflects the process and not the study design.

When each run samples different locations, that comparison collapses. Run 1 does not measure what Run 2 measured: there is no replicate, no estimate of run-to-run variability, and no basis to conclude the process is consistent. The protocol can still generate a pile of passing results, but the dataset cannot support the sentence cleaning validation exists to support. FDA's phrasing — "not designed to demonstrate reproducibility" — says the study was incapable of answering the question it was built to answer.

What is particularly interesting

First, this is not an execution error. Nobody failed to swab; the protocol instructed them to swab elsewhere each time. It is a design defect, and design defects do not appear in a deviation log — they surface only when someone asks whether the runs are comparable.

Second, FDA's remediation list confirms the intended design. The letter asks for improvements "with special emphasis on incorporating conditions identified as worst case", including "swabbing locations for areas that are most difficult to clean" and "maximum hold times before cleaning". Worst-case locations are only a worst case if held constant across runs; sample the hardest-to-clean site in one run and a benign one in the next, and the study never tests the worst case.

Third, the tank critique shows how quickly an equivalence argument fails without a technical basis. Materials of construction, interior finish, drains, ports and dead legs are precisely the attributes that govern cleanability; substituting a vessel without documenting them turns an equivalence claim into an assumption.

How such a gap can realistically develop

Usually through coverage thinking: twenty potential sites exist but only nine swabs fit per run, so sites are rotated to "cover the whole equipment item" across the campaign. The intent is thoroughness; the effect is to destroy the replicate structure — and nothing in a swab result flags it, because every individual result is in specification.

It develops further when sites come from a historical list rather than a documented worst-case rationale. Without a written basis — geometry, flow, dead legs, drains, seams, manual-clean areas, prior deviations — site selection becomes habit, and rotation looks defensible because the list is long. Grouping and bracketing compound it: group equipment by assumption rather than documented equivalence, and the sampling map loses its anchor. It then survives review because protocol and report are internally consistent; only a reviewer who asks "which sites were sampled in each run, and are they the same?" will catch it.

Questions

• For each validated cleaning process, are the swab locations identical across all qualification runs, and can we show that in a site-by-run matrix?
• Is worst-case location selection documented with a technical rationale — geometry, drains, ports, dead legs, seams, manual-clean areas — or inherited from a historical list?
• Where equipment was substituted or bracketed, is there a written equivalence argument covering materials of construction and interior features?
• Does the design let run-to-run variability be calculated, or does each run measure different sites?
• Are maximum hold times before cleaning fixed and reproduced across runs as part of the worst-case set?
• When a protocol is revised, does change management force reassessment of design and acceptance criteria, not only procedure wording?

Wider context

FDA evaluates cleaning validation as study design, not as a document set. The remediation language — worst-case drugs, worst-case swabbing locations, maximum hold times before cleaning, equipment change management, interim controls — is a recurring template across warning letters to contract manufacturers. The expectation is consistent: define the worst case, hold it constant, and prove the process performs the same way every time.

It is also a reminder about responsibility: regardless of its agreements with customers, a contract manufacturer bears responsibility for the quality of the drugs it produces. For a firm that has ceased U.S. drug production and sits on Import Alert 66-40, the point is moot; for anyone still manufacturing, the sampling-location question is worth answering before an investigator asks it.

Sources

• FDA Warning Letter 320-26-43, Cosmetic Manufacturers Pty Ltd., MARCS-CMS 719225, February 6, 2026 — primary source.
• FDA Import Alert 66-40 (Red List entry dated January 8, 2026).
• RAPS, "FDA warns contract manufacturer for GMP violations".
• Happi, "FDA Issues Warning Letter to Cosmetic Manufacturers Pty Ltd. in Australia".