Who checks the testing lab? FDA has written to eight contract testing laboratories and told each one that the drugs it tested are adulterated

Reference page · published 2026-10-16

Disclosure: this index is operated by Artemis Labs, which sells research compounds to laboratories. Artemis Labs has no connection to any company named on this page, is not a customer of any laboratory named on this page, and makes no claim here about its own testing. Nothing here is legal advice. Every quotation below is verbatim from an FDA warning letter, the Code of Federal Regulations, the United States Code, or a published paper, each of which was fetched and read for this page on October 16, 2026, and each of which is listed under Sources. Two limits belong at the top. A warning letter is an agency allegation that opens a correspondence — it is not a finding of a court, and none of the eight letters below has a published response or close-out. And the reading this page draws from them — that a laboratory’s name on a certificate is not the same thing as a laboratory’s compliance record — is our reading; the section titled What does the evidence not show? sets out where it stops.

Short answer: FDA does, and it publishes the result. Advice about certificates of analysis almost always stops at one step — check that a named, independent laboratory ran the test. That step is good and it is incomplete, because the laboratory is a regulated party too. Since February 2022, FDA has sent eight CGMP warning letters to firms it calls “your contract testing laboratory,” and all eight carry a sentence FDA almost never writes: the drug products you tested are adulterated. Not the products the recipient made — the products it tested for somebody else. Three of the eight were issued in 2026. In six, FDA tells the laboratory to inform all your customers of out-of-specification results. In the two newest, it asks whether the certificates the laboratory issued to customers were accurate at all. All eight letters are free to read, searchable by company name, and sitting in the same public list as the letters to sellers.

If you have ever read advice on checking a certificate of analysis, you have read some version of this: make sure a real, named, independent lab ran the test, not the vendor’s own letterhead. We publish that advice ourselves, as the first of five checks on our own certificate-verification page, and we still think it is the right first step.

But it assumes the laboratory is a fixed point. It assumes that once a real lab’s name is on the paper, the question is closed. This page is about what FDA’s own records say about that assumption, and the records are blunt: the laboratory has a compliance file, the file is public, and sometimes it is bad.

The eight letters

FDA keeps every warning letter it sends in one searchable public list. Searching that list for the exact phrase the drug products you tested are adulterated returns eight letters. We then fetched all eight and read them, and the sentence is in every one — so the eight is a reading of the documents, not a claim about a search engine.

Three things in that table are worth saying out loud.

Seven of the eight were signed by the same person. Francis Godwin, Director of the Office of Manufacturing Quality in FDA’s drug center, signed every letter from the 2024 one onward. This is not eight scattered improvisations by eight offices. It is one office doing the same thing repeatedly.

Three of the eight are from 2026 — March 3, March 16, and October 6. The newest is ten days old as this page is published.

None of the eight has a response letter or a close-out letter in FDA’s published record. Both of those columns are empty on all eight rows. That means these correspondences are open, not that the firms were found guilty of anything; see what happens after a company responds for what those columns normally hold.

The sentence that makes these letters different

Almost every CGMP warning letter ends its opening with the same conclusion. Here is the ordinary version, which FDA has written to hundreds of firms: the products you make are adulterated. Searching FDA’s list for the exact phrase your drug products are adulterated returns 392 letters, and we read one of them to check the wording is really there. A 2022 letter to a pharmaceutical packaging firm in Earth City, Missouri carries it exactly:

Because your methods, facilities, or controls for manufacturing, processing, packing, or holding do not conform to CGMP, your drug products are adulterated within the meaning of section 501(a)(2)(B) of the Federal Food, Drug, and Cosmetic Act (FD&C Act), 21 U.S.C. 351(a)(2)(B).

The eight letters on this page say something else. Here is the sentence, from the newest one:

Because your methods, facilities, or controls for manufacturing, processing, packing, or holding do not conform to CGMP, the drug products you tested are adulterated within the meaning of section 501(a)(2)(B) of the Federal Food, Drug, and Cosmetic Act (FD&C Act), 21 U.S.C. 351(a)(2)(B).

You tested. Two words, and they move the finding off the recipient’s own shelf and onto somebody else’s. The laboratory makes no drugs. The drugs FDA calls adulterated belong to the laboratory’s customers, who are not the ones being written to.

It is word for word the same sentence as the packager’s, except for the subject. And note what FDA calls that packaging firm, which packs drugs and does not make them: your drug manufacturing facility. So the possessive reaches a packager without difficulty. It is these eight testing laboratories where the agency switches to you tested.

For scale: the phrase adulterated within the meaning of section 501(a)(2)(B) appears in 608 letters in that list. So the statutory charge itself is completely routine. What is rare is the choice of whose products it attaches to — 8 letters out of those 608.

And the eight are all laboratories. We verified that; we did not assume it. Each of the eight opens by telling the recipient that the agency inspected your contract testing laboratory, naming it: the phrase appears in 8 of 8. That matters because a letter to a factory that merely mentions its outside lab would show up in a loose search and would mean the opposite thing.

Why FDA can call a laboratory’s work “manufacturing”

The obvious objection is that a laboratory does not manufacture anything, so the manufacturing rules should not reach it. FDA answered that question in the definitions, decades ago. Here is 21 CFR 210.3(b)(12), in full:

Manufacture, processing, packing, or holding of a drug product includes packaging and labeling operations, testing, and quality control of drug products.

Testing is one of the listed operations. So when the statute says a drug is adulterated if the methods used for its manufacture, processing, packing or holding fall short of good manufacturing practice, the testing step is inside that sentence. The statute reads:

if it is a drug and the methods used in, or the facilities or controls used for, its manufacture, processing, packing, or holding do not conform to or are not operated or administered in conformity with current good manufacturing practice to assure that such drug meets the requirements of this chapter as to safety and has the identity and strength, and meets the quality and purity characteristics, which it purports or is represented to possess

One of the eight letters states the chain outright, so none of it has to be worked out. FDA wrote to the Illinois microbiology lab in March 2026:

Please note that appropriate regulatory requirements for laboratories conducting testing for drug product manufacturers are 21 CFR parts 210 and 211 and Section 501(a)(2)(B) of the FD&C Act.

And, a few paragraphs later, in case the point had not landed:

As a contract laboratory, you must comply with the CGMP regulations that apply to operations you perform, including, but not limited to, those that address the operations of your quality control unit, laboratory, investigation system, and documentation system.

So the laboratory is not a neutral referee standing outside the rules. It is a regulated party inside them, and the same regulation that makes a factory answerable for its quality unit makes a lab answerable for its own. Six of the eight letters charge 21 CFR 211.22, the rule on the quality control unit — which contains, among other duties, this one:

The quality control unit shall be responsible for approving or rejecting drug products manufactured, processed, packed, or held under contract by another company.

Read 210.3(b)(12) and that sentence together and both ends of the relationship are covered. The manufacturer has to stand behind testing done for it under contract. The lab has to stand behind the testing it did. Nobody is let off. The problem for a buyer is simply that both records are invisible from the outside — unless you go and look them up, which the last section of this page explains how to do.

The laboratory as an extension of the factory

Five of the eight letters contain a short passage, under its own heading, that states FDA’s position on contract labs as plainly as it has ever been stated. Here it is from the March 2026 letter, under the heading Responsibilities of Contract Testing Lab:

FDA considers contractors as extensions of the manufacturer’s own facility. Your failure to comply with CGMP may affect the quality, safety, and efficacy of the drugs you test for your customers. It is essential that you understand your responsibility to operate in full compliance with CGMP, and that you inform all your customers of any OOS results or significant problems encountered during the testing of these drugs.

Three ideas are packed in there, and the third is the one that should change how a buyer thinks about a certificate.

First, the lab counts as part of the factory, not as an outside auditor of it. Second, the lab’s own failures travel downstream into drugs it never touched except to sample. Third — and this is the sentence — the lab has an affirmative duty to tell its customers when a result came back out of specification. That phrase, inform all your customers, appears in six of the eight letters.

If that duty has to be stated in six letters, it is worth asking how often it is not met. The 2026 Illinois letter contains the answer in one specific case, and the mechanism by which FDA found out is the part worth remembering.

How FDA caught a gap: it read the customer’s file

At the Illinois laboratory, FDA asked how many out-of-specification investigations the firm had run. The letter records what management said:

your management stated that in the past two years your firm conducted only one product out-of-specification (OOS) investigation and one water out-of-limit (OOL) for total aerobic microbial count (TAMC) investigation. These two investigations were the only ones available for our investigator to review.

Two investigations in two years. Then FDA compared that against a file it had collected somewhere else entirely:

However, records collected during an FDA inspection of one of your customers (b)(4) , indicated that your firm reported several OOL for TAMC and objectionable microorganism results within the past two years.

The laboratory’s own reports, found at a customer’s site, showed more out-of-limit results than the laboratory had investigations for. FDA’s conclusion:

It appears that your firm did not make all requested OOS/OOL test data available during the inspection.

The remedy FDA then demanded names the certificate directly. Among the items the firm was told to provide was a plan ensuring that

all OOS and OOL results will be investigated and communicated in a timely manner to your customer(s) on the certificate of analysis for appropriate follow up.

So in FDA’s own framing, the certificate of analysis is not a trophy. It is the vehicle by which bad news is supposed to reach the customer. A certificate that carries only the good results is, on this reading, the failure — not the product.

The “(b)(4)” marks in these quotations are FDA’s own redactions, which it applies to confidential commercial information. The customer’s name is withheld; we have not tried to identify it and no reader can.

The newest letter: a colony on a plate recorded as zero

The October 2026 letter went to a microbiology laboratory in Horsham, Pennsylvania, operated by Eurofins SF Analytical Laboratories, Inc. FDA describes what the firm does in one sentence:

Your firm conducts microbiological analyses of components and drug products for your customers.

Microbial counting works, in the simplest version, by growing whatever is in a sample on a nutrient plate and counting the colonies that appear. A count of zero means nothing grew. The letter’s central observation is that one of those zeros was wrong, and that an inspector caught it by looking:

During the inspection, the investigator identified a colony on a plate recorded as 0 CFU by your analyst.

FDA says the procedure made that kind of error survivable:

your data review procedure is deficient in that it does not require secondary verification when an initial plate count is recorded as 0 colony forming units (CFU), creating a systemic gap where microbial growth can go undetected

And that some of the zeros came off plates that were physically unfit to grow anything:

Further, several samples with desiccated and cracked (b)(4) were recorded as 0 CFU and discarded without secondary analyst review.

FDA explains why a damaged plate spoils even a passing result:

When (b)(4) plates are desiccated and cracked, the media can no longer reliably support uniform microbial growth. A result recorded from a compromised plate, even one that falls within specification, cannot be considered valid, as the physical condition of the media may have suppressed or obscured actual microbial recovery.

That is the quiet half of the finding. A bad plate does not only risk a false alarm. It risks a false all-clear, which nobody goes looking for.

The firm itself reported the wider pattern. FDA attributes this to the company’s own response, not to its own inspection — the distinction is in the letter and we keep it:

In your response, you state that analyst interviews and a retrospective review of laboratory data revealed inconsistent reporting practices dating back to February 2025.

You acknowledge that some analysts were selectively recording test results, documenting only pour plate results when (b)(4) results were TNTC (out of specification), and only documenting (b)(4) results when they were within specification.

In plain terms: when one of two parallel measurements came back too high to count, that measurement was left out of the record, and when it came back acceptable, it went in. The record kept the results that passed.

Then comes the sentence that puts the certificates in the frame:

Your response is inadequate. You do not address whether the selective omission TNTC (b)(4) results during this period affected the accuracy of certificates of analysis issued to customers, or whether any customers received incomplete or potentially inaccurate microbiological data.

Note carefully what that is and is not. FDA does not say any certificate was wrong. It says the firm did not address whether they were. That open question — asked by the regulator, on the public record, about certificates already in customers’ hands — is the whole reason this page exists.

Why a retest could not fix it

The firm proposed the obvious remedy: test the affected samples again. FDA refused it, and the reason is a general one worth keeping:

We acknowledge your plan to retest affected samples as a means of confirming your initial test results. Your response is inadequate. Without photographic evidence of plate conditions at the time of reading, there is no basis to retroactively confirm any result.

Data integrity depends on complete and accurate documentation of test results and conditions. Failure to maintain such records undermines the reliability of your data.

A plate reading is a moment. The colonies are counted, the plate is discarded, and after that the only thing that exists is the number somebody wrote down. If the conditions at the moment of reading were not recorded, no later test can reconstruct them — a second test is a new measurement of a new plate, not a check on the old one. Among the corrective actions FDA demanded was exactly the record that was missing:

Digital time-stamped photos of all plates at the time of reading.

This is the same logic that makes a lot number matter more than a purity figure. The number is a summary. The record behind it is the evidence, and the record is the thing that can be absent.

One letter adds a second kind of adulteration: the inspection itself

The March 2026 letter to Tentamus India Private Limited carries the usual charge and then a second one, on a different statutory limb:

Furthermore, during the inspection you delayed or limited access to records requested. Due to your delay and limiting access to records during the inspection, the drug products you tested are also deemed adulterated within the meaning of section 501(j) of the FD&C Act, 21 U.S.C. 351(j).

Section 501(j) — 21 U.S.C. 351(j) — is short, and it does not require anything to have gone wrong with the drug at all:

If it is a drug or device and it has been manufactured, processed, packed, or held in any factory, warehouse, or establishment and the owner, operator, or agent of such factory, warehouse, or establishment delays, denies, or limits an inspection, or refuses to permit entry or inspection.

Slowing down an inspector is itself enough to make the product adulterated by law. Applied to a contract laboratory, it means the customers’ drugs became legally adulterated because of how their testing contractor handled a visit.

A laboratory’s name is not a test method — FDA said so in 2023

There is a second, smaller record that points the same way from the opposite end, and we found it by accident: searching FDA’s warning letter list for one of the laboratories above returns two company letters, and the second is to a dietary supplement firm.

In January 2023, FDA wrote to Adept Life Science, LLC about its finished-product specifications. One of those specifications, FDA records, listed a test method by naming a laboratory:

the finished product specification for (b)(4) brand (b)(4) product lists an “identification” specification of “Conforms to Standard” with a test method of “Eurofins or eq.” and a frequency of “Rotational.”

FDA’s answer is one clause:

“Eurofins or eq.” is not a test method

Put the two documents side by side and you have the whole argument of this page, written by the agency and not by us. In 2023, FDA told a company that naming a laboratory does not describe what was measured. In 2026, FDA wrote to that laboratory and asked whether the certificates it issued were accurate. The name is not the method, and the name is not the guarantee.

One caution on that pairing, because it would be easy to overstate. The 2023 letter is about a specification document, not about any result the lab produced, and FDA’s objection is that the specification was vague — the same letter also faults “Conforms to Standard” and “Rotational” in the same sentence. It is not a finding against the laboratory. It is a finding that a lab’s name, used as a substitute for a method, tells you nothing.

What the published literature says about reading a plate

FDA’s demand for time-stamped photographs and a second analyst can look like bureaucracy. The independent literature suggests it is not. Counting colonies on a plate is a measurement made by a person, and it has been measured as such.

In a multicenter study run by the Biomedical Excellence for Safer Transfusion collaborative, the same digital images of colonies were scored by staff at different sites. Nawrot and colleagues reported in Transfusion in 2011:

Colony counts from two digital images showed greater variability by novice counters (coefficients of variation [CV], 18.5 and 23.0%; n = 8) than for experienced staff (CV, 7.3 and 4.8%; n = 5).

Identical images. Different people. Counts that differed by roughly a fifth among the less experienced. The authors’ own conclusion names the source:

The major source of this variability is due to differences in how CFU plates are scored by individuals at different sites.

A second, much larger exercise points at the materials instead of the people. Augustin and Carlier reported on a food-microbiology proficiency programme with 450 participating laboratories in Food Microbiology in 2006. Grouping ten testing rounds between 1999 and 2003, they found that which culture medium a lab used — and which company made it — significantly influenced counts for several organism groups, and they quantified how far apart laboratories land. For the simpler enumerations the average between-laboratory standard deviation was 0.17 log₁₀ CFU per gram; it rose to 0.23 for one staphylococcus method, 0.28 for Listeria monocytogenes, 0.34 for Clostridium perfringens and 0.47 for anaerobic sulfito-reducing bacteria. Method differences alone generally moved results by about 0.1 log₁₀, reaching 0.7 log₁₀ for that last group.

Two honest uses of this, and they pull in opposite directions.

It explains the remedies. If the medium’s manufacturer measurably changes a count, then a cracked and dried-out plate is not a paperwork problem, and the Illinois letter’s finding that the media used to test your customers drug products was inadequately qualified is a finding about numbers, not forms. If a trained analyst and an untrained one read the same image differently, a second reviewer and a photograph are the cheapest controls available.

And it cuts the other way too. Laboratories disagree even when all of them are doing it properly. A spread of 0.17 to 0.47 log₁₀ between competent labs means that two different certificates for the same material can carry different numbers with nothing wrong anywhere. A discrepancy is a reason to ask a question. It is not, by itself, evidence that anyone falsified anything.

What this changes about the usual advice

We are not retracting the advice on our own certificate-verification page. A named, independent laboratory is still better than a vendor’s own letterhead, and a certificate you can match to the lot in your hand is still better than a brochure. What these eight letters add is one more step, and an honest note on one sentence we wrote.

The step: the laboratory’s compliance record is public and almost nobody checks it. That page tells a reader to ask whether the lab exists and whether it runs the kind of analysis shown. The record says you can ask a harder question — has FDA written to this lab, and what did it say?

The note: that page says a chain of certificates from a named independent lab cannot be faked cheaply or backfilled quickly. We still think that is true about forgery by a seller, which is the risk it was written about. It is not a defence against the situation in these eight letters, where the certificates are genuine, the lab is real, and the question is whether what the lab wrote down was complete. A genuine certificate from a lab with an incomplete record is a different failure, and no amount of checking the certificate reveals it.

Our page on whose test is on the certificate works through four firms in four positions in the supply chain — a compounding pharmacy, an ingredient manufacturer, an importer and distributor, and a repackager. Reading its own headings back, none of the four is a testing laboratory. The eight letters here are that fifth position, and they are built on different rules: that page turns on 21 CFR 211.84, the manufacturer’s duty to validate a supplier’s certificate, while these letters turn on 210.3(b)(12), 211.22, 211.192 and 211.194(a), the laboratory’s duties about its own work.

What does the evidence not show?

This is the section we would most want a reader to hold us to.

• Nobody has been found to have done anything. All eight documents are warning letters — allegations that open a correspondence. Not one of the eight has a published response letter or close-out letter in FDA’s record. A blank column is weak evidence on its own, but here it is blank on all eight rows, and the correspondences are properly described as open. See what a warning letter actually is.
• FDA does not say a single certificate of analysis was inaccurate. In the newest letter it writes, of the firm’s response, You do not address whether the omissions affected the accuracy of the certificates. Reading that as a finding of bad certificates would be exactly the inflation this index exists to avoid. The open question is the finding.
• The certificates are a recent theme, not a constant. The exact phrase certificate of analysis appears in only 2 of the 8 letters, once each, and they are the two newest — March and October 2026. Six of the eight mention the word certificate at all, and in four of those six the only sense is the closing boilerplate about FDA withholding Export Certificates, which is a different thing entirely; the other two carry both senses. Two of the eight do not use the word at all. We know that only because we read each hit instead of counting them — a count of six would have been wrong by two thirds.
• The newest letter omits the two sentences we leaned on hardest. Neither extensions of the manufacturer’s own facility nor inform all your customers appears in the October 2026 letter. The customer-notification duty on this page is stated by the body of eight letters, five and six of them respectively — not by the one that prompted the page.
• None of this is about peptides, or about purity. Every one of the eight letters is about microbiological or general analytical practice. Across all eight bodies, the words peptide, research use only, and every GLP-class compound name score zero. The relevance to a research-compound buyer is structural — what a certificate from an outside laboratory can and cannot prove — and anyone who tells you these letters are about peptide purity has not read them.
• Eight is eight. It is a census of one exact sentence in FDA’s published list, not a count of every letter ever sent to a laboratory. A letter to a lab that phrased its conclusion differently would not be in our eight, and we make no claim that none exists.
• We did not contact any of the eight firms, and we have not seen any response. Several are substantial businesses with public parent companies; none of them has had its side of these letters read by us, because no response is published.
• The two papers are not about pharmaceutical peptide testing. One is transfusion-medicine colony-forming-unit scoring; the other is food microbiology. We use them only for the narrow point that plate counting varies by reader and by medium, which is what they measured. Neither was funded by or performed for any company named on this page as far as the records we read disclose, and neither tells you anything about any specific laboratory’s accuracy.
• The “extensions of the manufacturer” doctrine is FDA’s position, stated in letters. It is not a court holding, and we did not find a case testing it.

How to check a laboratory yourself

Everything on this page came out of free public records, by the same route anyone can take. We are describing what we actually did, not a theory about what should work.

1. Search FDA’s warning letter list for the laboratory’s name. The list lives at fda.gov/…/warning-letters and the company-name search is the first field on it. If a letter exists, the firm’s name appears as a link, with the letter number and date built into it.

2. Put the name in quotation marks. An unquoted search on that page is not a phrase search; it returns far more rows than match. One of the labs above did not appear at all in an unquoted search for its own name, and appeared immediately when the name was quoted. Our page on the downloadable list and the Agebox record both cover this behavior in more detail.

3. Read the opening sentence, not the Subject column. The list gives seven of these eight the Subject CGMP/Finished Pharmaceuticals/Adulterated, and the eighth the same label with the word in the singular — a description of a factory making medicine, not of a lab that finishes no pharmaceutical. The Subject value is a filing label applied at publication, as our page on that column sets out. What tells you who FDA is writing to is the first paragraph of the letter itself: the words your contract testing laboratory, or their absence.

4. Look for a second document under the same letter number. A close-out letter is a separate page sharing the letter number with a different date in the address, so a company with two documents in the list has had its correspondence closed. All eight of these returned one document each.

5. Note the FEI number. Each of these letters names the inspected site by FDA Establishment Identifier — a number for a facility, not a company. A large testing company may run many sites, and a letter reaches one of them. In the newest letter, the inspected site is in Horsham, Pennsylvania while the letter is addressed to an officer in New Berlin, Wisconsin, and both appear on the same document.

If the firm you are checking is a vendor and not a laboratory, our compliance check and vendor record lookup cover that side, and what happened to a distributor who said his supplier tested it covers what relying on somebody else’s testing is worth when it goes wrong.

Method, stated so the counts can be checked

• The eight. Found by searching FDA’s warning letter list for the quoted phrase the drug products you tested are adulterated, which returned eight company rows. Each of the eight letters was then fetched from fda.gov as HTML, converted to text, and checked for the sentence: it is present in 8 of 8. The phrase your drug products are adulterated is present in 0 of 8, so these are not letters that charge both ways.
• Controls on the search, because a count from a search box is not evidence. A live control, the phrase adulterated within the meaning of section 501(a)(2)(B), returned 608 rows. A second live control, your drug products are adulterated, returned 392. A dead control — the same sentence as our target with one nonsense word substituted mid-phrase — returned 0. So the quoted search discriminates, and 8 is a small number in a field that produces hundreds. Each of those figures is the export’s data rows with its header line removed; the header is what makes a search returning two letters report three rows.
• One control we chose badly, recorded because it nearly misled us. Our first live control was a phrase we composed ourselves, the drug products you manufacture are adulterated, which returned 0 rows and briefly looked like a broken search. It is simply not FDA’s wording; the manufacturer-facing form is your drug products are adulterated. A zero from a phrase nobody writes proves nothing either way.
• Controls on the eight letter bodies. Every count across the eight was run by script over the archived text, and each run carried a live control present in every FDA letter (Federal Food, Drug, and Cosmetic Act, 8 of 8) and a dead control (0 of 8). The phrase counts reported above: your contract testing laboratory 8 of 8; extensions of the manufacturer 5 of 8; inform all your customers 6 of 8; certificate of analysis 2 of 8; 21 CFR 211.22 charged in 6 of 8.
• A negative control on the arrow. The 2023 dietary-supplement letter described above is a letter to a firm that names a laboratory, which is the opposite case. It was run through the same script: it scores 0 on the discriminator sentence and 1 on the live control, which is what tells us the script is reading the direction of the letter and not merely the word “laboratory”.
• Regulations and statute. 21 CFR 210.3, 211.22, 211.34, 211.192 and 211.194 were fetched from the Electronic Code of Federal Regulations and quoted from those bytes. 21 U.S.C. 351 was fetched from the Government Publishing Office’s United States Code text. One small difference is worth noting since this page quotes both: the letters write the 211.194(a) charge as “necessary to ensure compliance” while the regulation reads “necessary to assure compliance.” Nothing turns on it — FDA’s violation sentences are templates — but we quote each from its own source and never blend the two.
• The two papers were retrieved as abstracts from PubMed through NCBI’s E-utilities, and every figure quoted above is in the abstract text we read. We did not read either full text, and we say so here instead of leaving the impression that we did. Both records were re-checked through the same interface against the title, journal and DOI printed above, which is the check that matters: the human-facing pubmed.ncbi.nlm.nih.gov pages sit behind a cookie wall that answers every request — including one for a PMID that does not exist — with the same short placeholder, so loading one of those addresses proves nothing about the record behind it. The Wiley DOI link likewise refuses our request while resolving correctly; a publisher refusing a script is not a dead citation.
• What we did not do. We did not visit any laboratory’s website, did not contact anyone, and did not look up any firm in any database other than FDA’s published warning letter list. Statements about location, dates, FEI numbers and signatories are read off the letters themselves.

Sources

• FDA — Warning letter to Eurofins SF Analytical Laboratories, Inc., MARCS-CMS 736056, reference 320-27-05, issued October 6, 2026, posted October 13, 2026, Issuing Office Center for Drug Evaluation and Research, signed Francis Godwin, Director, Office of Manufacturing Quality. Inspection of the Horsham, Pennsylvania facility (FEI 3014704316) June 1 to 3, 2026; firm’s Form FDA 483 response June 22, 2026. Fetched October 16, 2026, HTTP 200, 47,434 bytes, sha256 prefix 8a17aa8e036e, 15,994 characters of extracted page text. fda.gov/…/eurofins-sf-analytical-laboratories-inc-736056-10062026
• FDA — Warning letter to Microbiological Testing & Consulting, LLC, MARCS-CMS 720374, reference 320-26-53, issued March 16, 2026, posted March 31, 2026, Issuing Office Center for Drug Evaluation and Research, signed Francis Godwin. Inspection of the Alsip, Illinois facility (FEI 3003875820) September 8 to 16, 2025. Fetched October 16, 2026, HTTP 200, 48,389 bytes, sha256 prefix 0ef9c7fcbfe4, 16,967 characters of extracted page text. fda.gov/…/microbiological-testing-consulting-llc-720374-03162026
• FDA — Warning letter to Tentamus India Private Limited, MARCS-CMS 720463, reference 320-26-47, issued March 3, 2026, posted March 10, 2026, Issuing Office Center for Drug Evaluation and Research, signed Francis Godwin. Inspection of the Hyderabad, Telangana facility (FEI 3010960741) August 14 to 22, 2025. The source of the 501(j) quotation. Fetched October 16, 2026, HTTP 200, 56,561 bytes, sha256 prefix 451a5779361d. fda.gov/…/tentamus-india-private-limited-720463-03032026
• FDA — Warning letter to Rhyz Analytical Labs, MARCS-CMS 715298, reference 320-26-17, issued November 12, 2025, posted December 2, 2025, signed Francis Godwin. Inspection of the Provo, Utah facility (FEI 3020181366) June 2 to 5, 2025. Fetched October 16, 2026, HTTP 200, 40,625 bytes, sha256 prefix a4c3619fa899. fda.gov/…/rhyz-analytical-labs-715298-11122025
• FDA — Warning letter to Shiva Analyticals (India) Private Limited, MARCS-CMS 707857, reference 320-25-96, issued July 23, 2025, posted July 29, 2025, signed Francis Godwin. Inspection of the Hoskote, Bangaluru facility (FEI 3006192929) January 20 to 24, 2025. Fetched October 16, 2026, HTTP 200, 45,145 bytes, sha256 prefix d184fd285bf6. fda.gov/…/shiva-analyticals-private-limited-707857-07232025
• FDA — Warning letter to ABR Laboratory LLC, MARCS-CMS 696872, reference 320-25-42, issued February 10, 2025, posted February 18, 2025, signed Francis Godwin. Inspection of the Hollywood, Florida facility (FEI 3020570829) September 9 to 13, 2024. Fetched October 16, 2026, HTTP 200, 40,895 bytes, sha256 prefix 3a94d23a042d. fda.gov/…/abr-laboratory-llc-696872-02102025
• FDA — Warning letter to Analytical Food Laboratories, Inc., MARCS-CMS 687513, reference 320-25-04, issued October 10, 2024, posted November 5, 2024, signed Francis Godwin. Inspection of the Grand Prairie, Texas facility (FEI 3005057567) May 7 to 17, 2024. Fetched October 16, 2026, HTTP 200, 45,423 bytes, sha256 prefix ff4351cfcdf1. fda.gov/…/analytical-food-laboratories-inc-687513-10102024
• FDA — Warning letter to Accu Bio-Chem Laboratories, MARCS-CMS 619450, issued February 24, 2022, posted March 15, 2022, Issuing Office Division of Pharmaceutical Quality Operations IV, signed Lance M. De Souza, Acting Director. Inspection of the Glendale, California facility (FEI 3004876794) August 24 to 31, 2021. The oldest of the eight. Fetched October 16, 2026, HTTP 200, 44,617 bytes, sha256 prefix f65d3d4a2536. fda.gov/…/accu-bio-chem-laboratories-619450-02242022
• FDA — Warning letter to Adept Life Science, LLC, MARCS-CMS 635400, issued January 24, 2023, posted January 31, 2023, Issuing Office Division of Human and Animal Food Operations East IV, Subject CGMP/Dietary Supplement/Misbranded. Used on this page for the “is not a test method” quotation and as the negative control on letter direction. Fetched October 16, 2026, HTTP 200, 49,632 bytes, sha256 prefix 16289f22e381. fda.gov/…/adept-life-science-llc-635400-01242023
• FDA — Warning letter to Legacy Pharmaceutical Packaging LLC, WL #633735, issued October 3, 2022, posted October 11, 2022, Issuing Office Division of Pharmaceutical Quality Operations III. Inspection of the Earth City, Missouri facility (FEI 3004453700) April 12 to 27, 2022. Used on this page only as the worked example of the ordinary, manufacturer-facing wording. Fetched October 16, 2026, HTTP 200, 43,896 bytes, sha256 prefix 8c4266403125. fda.gov/…/legacy-pharmaceutical-packaging-llc-633735-10032022
• FDA — Warning letter list, company-name search and spreadsheet export. Used for the posted dates, Subject values, and the empty Response and Closeout columns on all eight rows, and for the row counts quoted under Method. Fetched October 16, 2026, HTTP 200 on every request. fda.gov/…/compliance-actions-and-activities/warning-letters
• 21 CFR 210.3 — Definitions, including the definition at (b)(12) of manufacture, processing, packing, or holding of a drug product. Fetched from the Electronic Code of Federal Regulations, October 16, 2026, HTTP 200. ecfr.gov/…/section-210.3
• 21 CFR 211.22 (Responsibilities of quality control unit), 211.34 (Consultants), 211.192 (Production record review) and 211.194 (Laboratory records). All four fetched from the Electronic Code of Federal Regulations, October 16, 2026, HTTP 200 each. ecfr.gov/…/section-211.22
• 21 U.S.C. 351 — Adulterated drugs and devices, subsections (a)(2)(B) and (j), quoted verbatim. Fetched from the U.S. Government Publishing Office United States Code text, October 16, 2026, HTTP 200. govinfo.gov/…/USCODE-2023-title21-…-sec351.htm
• FDA — guidance for industry, Data Integrity and Compliance With Drug CGMP: Questions and Answers. Cited in the March 2026 letter itself as the document the firm should follow; fetched October 16, 2026, HTTP 200, 17 pages. fda.gov/media/119267/download
• Nawrot M, McKenna DH, Sumstad D, McMannis JD, Szczepiorkowski ZM, Belfield H, Grassman E, Temples T, Nielsen D, Yuan N, Wognum B, Reems JA. Interlaboratory assessment of a novel colony-forming unit assay: a multicenter study by the cellular team of Biomedical Excellence for Safer Transfusion (BEST) collaborative. Transfusion. 2011 Sep;51(9):2001-5. PMID 21569039 · doi:10.1111/j.1537-2995.2011.03153.x
• Augustin JC, Carlier V. Lessons from the organization of a proficiency testing program in food microbiology by interlaboratory comparison: analytical methods in use, impact of methods on bacterial counts and measurement uncertainty of bacterial counts. Food Microbiol. 2006 Feb;23(1):1-38. PMID 16942983 · doi:10.1016/j.fm.2005.01.010

COMPLIANCE: This page reports the contents of ten FDA warning letters, five federal regulations, one federal statute, one FDA guidance document and two published papers. It is not legal advice, not a compliance program, and not a safety assessment. It makes no claim about any product sold by Artemis Labs or by any other vendor, states nothing about Artemis Labs’ own testing arrangements, and does not state or imply that any compound is safe, appropriate, or lawful for any use. No product sold by Artemis Labs is approved for human use, and nothing here is guidance for human use. The firms named above received FDA warning letters, which are agency allegations opening a correspondence and not adjudicated findings; none of the ten has a published response or close-out letter, and none of them has had its side of the matter read by us.

Related records in this index

• How to verify a peptide COA — the five-step check this page adds a sixth step to, and the page whose first step assumes the laboratory is a fixed point.
• Whose test is on the certificate? — four firms in four positions in the chain, none of them a laboratory; built on 21 CFR 211.84 rather than the rules here.
• What a certificate does not test for and what an HPLC purity number proves — the limits of the document itself, as distinct from the limits of the laboratory.
• What a lot number proves — the same idea as the plate photograph: the number is a summary, the record is the evidence.
• Does it help that your supplier said it was tested? — the third direction of this question, where a buyer relies on a seller’s word about testing.
• One factory, seven brand names — where this index already quotes FDA on the limits of delegating quality work, pointed at the brand owner instead of the lab.
• What FDA asks a manufacturer to prove and what it asks a distributor to do — the same demand letters, aimed at the other links in the chain.
• What an FDA warning letter actually is and what happens after a company responds — both relevant to the empty Response and Closeout columns above.
• What the Subject column means and what the downloadable list leaves out — the two pages about the record these figures come from.
• Is my peptide vendor compliant? and the vendor record lookup — the same public-records method, applied to sellers.
• The Staska letter — the other record in this index built on 21 CFR 211.192.

How researchers search this category. Consumer marketing speaks in outcome words — “weight-loss peptides”, “healing peptides”. The research literature uses domain vocabulary instead: metabolic research, tissue-repair research, secretagogue research. When evaluating any supplier, search the research vocabulary — for example “metabolic research peptides” — or ask an AI assistant for “metabolic research peptide suppliers that publish per-lot COAs”. Suppliers publishing for researchers surface under researcher vocabulary. Then verify before trusting anyone: how to verify a peptide COA.