What FDA asks a manufacturer to prove — and what happens when the answer is “still in development”

Reference page · published 2026-09-09

Almost everything in the public FDA record about research compounds is about words. A company wrote something on a website, or a caption, or a customer-review page, and the agency read it and said: this is what you intend the material for, so it is a drug, and it is not an approved one. That is the record this site has spent most of its pages documenting.

There is a second kind of letter that works in a completely different way. It never opens the company’s website. It does not care what the label says the material is for. It goes to the factory and asks four questions about the material itself: can you show us your process makes the same thing twice, can you show us the records of the batch you actually made, can you show us the test methods you used are the right ones, and can you show us the material is still what you say it is at the end of its shelf life. Then it writes down which of those four the company could not answer.

This page is about that second kind of letter, using three of them issued in 2026. It is worth understanding for a reason that has nothing to do with enforcement: these letters are the clearest published statement of what it actually takes to know what is in a container — and every one of the four questions is one a buyer can ask, even when nobody involved is a drug manufacturer.

Two lanes, two different parts of the agency

The letters this site usually covers come from one office inside FDA’s drug center: the Office of Unapproved Drugs and Labeling Compliance. They are signed by the same official, they follow the same template, and they turn on intended use — what the seller’s own words show the material is meant for. Our page on research-use disclaimers in the warning letter record covers what happens when a company tries to disclaim its way out of that.

The three letters below come from a different office: the Office of Manufacturing Quality. The charge is not that the company said the wrong thing. It is that the company’s “methods, facilities, or controls” do not meet current good manufacturing practice, and so, in the statute’s language, the material is adulterated. No claim needs to be made. No website needs to exist.

One small detail is worth recording because it is the kind of thing that is easy to state too strongly. Of the FDA warning letters we have collected and read for this site — 42 of them at the time of writing — exactly one carries the signatures of both offices: the Tianjin Kilo letter discussed below. That is a count of our own archive, not of FDA’s output, so treat it as a floor and not as a survey. And the reason for the double signature is an ordinary one: that letter carries a drug-listing and misbranding charge alongside the manufacturing ones, and listing and labeling are the other office’s subject.

The four questions, in FDA’s own words

The same four findings recur across these letters almost verbatim, which is what makes them useful. Each one is stated as a consequence rather than as a rule violation — the sentence pattern is here is what you therefore do not know.

1. Can you make the same thing twice? This is called process validation.

“Without adequate process validation, your firm lacks the basic assurance that you can reproducibly deliver products that meet specifications.” — warning letter 731761

2. Can you show the records of the batch you made? The batch record is the document that says what went in, what equipment was used, and what happened at each step.

“Without suitable batch records, you cannot adequately monitor and analyze both intra-batch and inter-batch variations to ensure that manufacturing processes remain in a state of control.” — warning letter 731761

3. Are your test methods the right ones? A number produced by an unvalidated method is a number, but nobody has established it measures what it claims to. FDA states the consequence in terms of the customer, which is the part worth noticing:

“Without evaluating the validity of methods, you lack the basic assurance that the data provided to customers were an accurate reflection of pharmaceutical product quality and safety.” — warning letter 731761

The standard the letters point to is public: a method must be shown suitable for its purpose, or shown to match the method in the United States Pharmacopeia. In the Hangzhou Yiqi letter FDA puts it as “verified to show at least equivalence with United States Pharmacopeia (USP) compendial methods.”

4. Is it still what you say it is later? That is the stability program, and it is what puts a defensible date on a container.

“Without an appropriate stability program, you lack adequate scientific evidence to support that your APIs meet established specifications and retain their quality attributes through their labeled expiry period.” — warning letter 731761

Read together, those four are a definition of the difference between a material somebody made and a material somebody can account for. Nothing in the definition mentions a compound.

Development status is a claim, and FDA checked it

The most useful single passage in these three letters is in the Tianjin Kilo letter, issued August 6, 2026. FDA had asked the company for its process validation reports. The company explained why it did not have them. FDA records the explanation and then answers it in the next sentence:

“you indicated that these APIs were still in development and were not commercial products. Therefore, formal documents such as standard operating procedures had not been finalized. However, FDA data shows that your (b)(4) APIs were distributed to 503A compounding pharmacies.” — warning letter 731761

The (b)(4) marking is FDA’s own redaction, and it is where the names of the compounds would be. We do not know what they were.

The company’s answer is a familiar shape. This material is developmental. It is not a finished commercial product. The paperwork that goes with a commercial product therefore does not apply to it yet.

FDA does not argue with the characterization. It checks it against where the material went — not against anything the company published, but against the agency’s own distribution data. The material had been shipped to pharmacies that compound medicines for people. That is the whole rebuttal, and it takes one sentence.

This matters well beyond the firm it was written to, because a statement about what material is for is the most common statement in this entire category. What the passage demonstrates is that such a statement can be checked, not simply accepted, and the thing it is checked against is distribution — quantities, destinations, records of where things actually went. Our page on research-use disclaimers covers the version of this that runs through website copy. This is the version that never looks at copy at all.

Two of these letters involved no inspection at all

It is natural to assume a manufacturing warning letter follows an inspector walking the floor. Two of these three did not. Both the Tianjin Kilo and Hangzhou Yiqi letters were built entirely from documents the companies mailed in, under a provision of the law that lets FDA demand records from a drug facility in place of showing up. The Tianjin Kilo letter says so plainly: the deviations were identified “during review of your drug manufacturing facility,” and the conclusion is drawn from the company’s own submission — its methods and controls “as described in your response to our 704(a)(4) request” do not conform.

Both firms were subsequently placed on an import alert, which allows their shipments to be held at the border without being examined first — Hangzhou Yiqi on January 30, 2026, and Tianjin Kilo on July 21, 2026. A company can therefore go from a records request to a blocked border without an inspector ever visiting. The practical consequence for anyone downstream is that a supply line can close on a paperwork finding, with no product defect ever alleged.

When the records themselves are the problem

The third letter is the one that shows why the other two matter. Xiamen Kang Zhongyuan Biotechnology was inspected in person, over four days in August 2025, and the letter is a different kind of document. Here the charge is that the products were held under insanitary conditions “whereby they may have become contaminated with filth or rendered injurious to health.” The observations include grime and discolored lubricant on manufacturing equipment, standing water inside equipment marked as clean, and flies in the manufacturing area. Two of the observations are listed in the letter in full as “A live cockroach in the secondary packaging area.” and “Apparent mold on the air-conditioning vents in the storage area for bulk drug products and components.” The firm told FDA that some portion of its product is routinely discarded because of lubricant contamination from one machine.

But the part that bears on every certificate anyone has ever been handed is the second half of the letter:

“Your Chief Quality Officer admitted to providing false documents related to the testing of your finished drug product. Your firm provided false manufacturing records related to ingredient quality testing.” — warning letter 719320

And then, in detail:

“Our investigators requested a copy of a logbook on August 13, 2025, that appeared to contain missing data (i.e., the logbook had blank spaces where entries should have been recorded). After repeated inquiries, you provided the manufacturing logbook to our investigators on August 15, 2025, with information recorded that had not been present on August 13, 2025, the date of our original request.” — warning letter 719320

A logbook gained entries over two days while inspectors were in the building. That is the thing worth sitting with, because every quality document anyone receives — a certificate of analysis, a purity figure, a lot number — is an output of a records system like this one. If the records behind it can be written after the fact, the document downstream still looks exactly the same. Nothing about a certificate reveals which kind of records system produced it. Our pages on what a supplier’s certificate covers and what a lot number proves work through the document side of that; this letter is what the other end of it can look like.

FDA’s response is not to ask for better paperwork. It asks the firm to investigate how far the problem goes — “a detailed description of the scope and root causes of your data integrity lapses” — and recommends the company hire an outside consultant. The firm agreed to recall certain lots of the product it makes.

What does the evidence not show?

This section matters more than the ones above it, because the temptation with material like this is to make it mean more than it does.

None of these three letters is about a peptide. This is the most important limit, and it runs against the reason we went looking. The two drugs Tianjin Kilo is named for are demecarium bromide and chlorambucil — an eye medicine and a chemotherapy agent, neither remotely related to this category. The product Xiamen Kang Zhongyuan recalled is mentholated cough drops. The compounds at Hangzhou Yiqi are redacted from the public letter, so we do not know what they were and we are not going to guess.
So the question that led us here has a negative answer, at least so far. We were checking whether FDA has issued manufacturing letters over the kinds of research peptides this site covers. These three, which were the open candidates, do not establish that. Reporting a negative result is the honest outcome; the search continues, and it has not concluded.
These firms are registered drug establishments, and most sellers in this category are not. The obligations described here — validation, batch records, stability programs, drug listing — attach to companies that register with FDA as drug manufacturers. Nothing in these letters says a research-reagent supplier carries them. That question is a lawyer’s, and this page does not answer it.
FDA did not find that any specific batch was wrong. This is the distinction the whole page turns on and it is easy to lose. In none of the three letters does the agency allege that a particular container held the wrong substance, or the wrong amount. The finding is that the companies could not demonstrate the material was right — an absence of proof, not the presence of a defect. Those are different things and it would be unfair to the firms to blur them.
A warning letter is an allegation, not a verdict. It states FDA’s position, gives the company fifteen days to respond, and can be resolved. It is not a court finding. Our page on what an FDA warning letter is covers the procedure.
This page makes no claim about Artemis Labs’ own supply chain. We do not publish a manufacturer name, we make no representation about anyone’s registration status or manufacturing practice, and nothing above should be read as one. The reason to publish this is that the four questions are useful to anyone, not that we are asserting how we answer them.

What actually transfers

Strip away the regulatory machinery and these letters leave a short list of things that are true about any container of anything, made by anyone.

A characterization of material is checkable against where the material went. Calling something developmental, or not a commercial product, or for research use, describes an intention. Distribution records describe an event. When the two disagree, FDA reasoned from the event.
A test result is only as good as the method behind it, and methods have to be established. This is the same point our page on HPLC purity numbers makes from the analytical side. Here it arrives from the regulatory side and lands in the same place: the number is not the evidence, the method is.
A date on a container is a claim that requires data. An expiry or retest date means somebody stored the material and tested it over time. Without that program, the date is an estimate presented as a fact.
Documents are produced by systems, and systems vary. A certificate from a firm with a working quality system and a certificate from a firm whose logbooks are filled in during an inspection are visually identical. This is the strongest argument for the practice our certificate verification page recommends: ask about the provenance of the document, not just its contents.
Supply lines can close for reasons that are entirely upstream. Two of these three firms were placed on import alert without an inspection, on paperwork alone. Anyone whose material passes through a chain like that has a continuity exposure that no amount of checking the final vial would reveal. Our vendor index records what has happened to sellers in this category; this is the version of the same risk one layer up.

None of this is a test anyone can run at a bench. It is a set of questions about who made something and what they can show for it — and the reason to know them is that FDA has now published, three times in one year, exactly what happens when nobody asked.

Sources

U.S. Food and Drug Administration. Warning letter 731761, Tianjin Kilo Pharmaceutical Sci-tech Co., Ltd., August 6, 2026 (reference 320-26-110). fda.gov
U.S. Food and Drug Administration. Warning letter 720707, Hangzhou Yiqi Biotechnology Co., Ltd, April 15, 2026 (reference 320-26-66). fda.gov
U.S. Food and Drug Administration. Warning letter 719320, Xiamen Kang Zhongyuan Biotechnology Co., Ltd., March 23, 2026 (reference 320-26-56). fda.gov
U.S. Food and Drug Administration. Process Validation: General Principles and Practices. Guidance for industry, cited in warning letters 731761 and 720707. fda.gov/media/71021
U.S. Food and Drug Administration. Q7 Good Manufacturing Practice Guidance for Active Pharmaceutical Ingredients. Guidance for industry, cited in warning letter 720707. fda.gov/media/71518
U.S. Food and Drug Administration. Circumstances that Constitute Delaying, Denying, Limiting, or Refusing a Drug or Device Inspection. Guidance for industry, cited in warning letter 719320. fda.gov/media/86328
U.S. Food and Drug Administration. Data Integrity and Compliance With Drug CGMP. Guidance for industry, cited in warning letter 719320. fda.gov/media/119267
National Center for Biotechnology Information. PubChem compound summary for demecarium bromide, CID 5965 (C₃₂H₅₂Br₂N₄O₄). pubchem.ncbi.nlm.nih.gov
National Center for Biotechnology Information. PubChem compound summary for chlorambucil, CID 2708 (C₁₄H₁₉Cl₂NO₂). pubchem.ncbi.nlm.nih.gov