Chemically Pure, Biologically Contaminated: What Endotoxin Testing Measures

On 13 September 2026 the FDA published a nationwide recall by Centric Compounding of six lots of injectable glutathione, Myer’s Cocktail and Tri-Immune Boost, after the products were found to have been compounded with a glutathione active ingredient containing elevated endotoxin levels.¹ It is the third such recall since August, and all three trace back to the same upstream material. The chemistry of that material was not in question. Glutathione is a tripeptide and a well-characterised one. What failed was a property no purity assay measures. Condor Research supplies peptides as reference materials for laboratory research use only; this article discusses pharmaceutical quality control and contains no human use guidance.

Three recalls, one ingredient

Centric Compounding, of Houston, announced its recall on 9 September and the FDA posted it on 13 September. Six lots of 30 mL multidose vials were recalled to the consumer level, distributed nationwide by home delivery and to prescribers. The company reported nine cases of fever, chills, chest pain, nausea and vomiting, headache or malaise, and stated the risk plainly: elevated endotoxin in an injectable product can cause fever, hypotension, inflammatory reactions, anaphylactic shock and death.¹

Two Texas pharmacies had already recalled glutathione 200 mg/mL for the same reason: Victory Medical Center Pharmacy of Austin on 5 August, and Optimal Balance Pharmacy of Houston on 19 August.²³ Three separate pharmacies, three separate recalls, one shared cause.

The FDA and CDC have described the common thread. At least 30 patients experienced adverse events after intravenous glutathione, including fever, chills, pain, dizziness, signs of shock and sepsis-like symptoms, some requiring hospitalisation, consistent with exposure to excessive endotoxin.⁴ The pharmacies were different. The glutathione was not: all of it came from one distributor, Medisca, and was labelled dietary supplement grade, which is not an appropriate ingredient for making injectable drugs. Medisca told the agency on 20 August 2026 that it had warned its United States customers.⁴

The contamination did not originate in any of the three pharmacies. It arrived with the raw material and was faithfully carried into everything made from it.

This has happened before, with the same substance and the same root cause. In 2019 the FDA warned compounders not to use L-glutathione distributed by Letco Medical after seven patients in one clinic developed symptoms within minutes of infusion. FDA testing found excessive bacterial endotoxin, some samples as high as five times the appropriate limit. The powder was labelled “Caution: Dietary Supplement”.⁵

What an endotoxin is, and why purity does not see it

An endotoxin is not a contaminant in the sense most people picture. It is lipopolysaccharide, a structural component of the outer membrane of Gram-negative bacteria, and it is released when those bacteria die. Killing the organisms is therefore not the same as removing the hazard: sterilisation can leave a product with no viable bacteria and a full load of the molecules their membranes left behind. Reviews of sterilisation methods treat endotoxin removal as a separate problem from microbial kill for exactly this reason.⁸ LPS is also markedly more robust than the organisms that produced it, which is why a separate test exists rather than an inference from sterility.

That is the whole point of the bacterial endotoxins test in USP General Chapter <85>, the chapter the FDA cited in its 2019 glutathione notice, and which sits at the centre of the agency’s guidance on pyrogen and endotoxins testing, reissued in a second edition in March 2026.⁵⁶ Limits are not fixed per vial; they are calculated from the maximum dose intended for administration, which is why the same material can be acceptable for one route and unacceptable for another.

Now place that next to a purity figure. HPLC separates a sample by chemical behaviour and reports the main peak as a proportion of total peak area. Lipopolysaccharide does not appear as a peak competing with a peptide in a typical reversed-phase run with UV detection, and would not be counted against the main peak even if it did. A vial can honestly report 99% chromatographic purity and carry an endotoxin burden far above any injectable limit. The two numbers are not in tension; they are answers to unrelated questions.

How often this shows up in peptide material

Glutathione recalls are a compounding story, and compounding is not the gray peptide market. But the two meet at the same question, and there is now a figure for the peptide side. Mendias and Awan analysed 6,441 samples across fourteen compounds, including BPC-157, TB-500, GHK-Cu, ipamorelin, CJC-1295, semaglutide, tirzepatide and retatrutide, from a large independent testing dataset of preparations sold directly to consumers. Between 41.6% and 71.1% failed basic quality criteria depending on the framework applied, and measurable endotoxin contamination was present in 15% of samples.⁷

That figure belongs next to how the gray market actually works, and next to the COA section where endotoxin and sterility are reported, which is the one most buyers skip.

The preprint is not peer-reviewed and its sample is self-selected, which almost certainly flatters the market: these are preparations someone chose to have tested. One in seven carried detectable endotoxin anyway.

The lesson is about materials, not paperwork

The sentence worth taking from the 2019 FDA notice is the one the agency wrote for compounders and that applies with equal force to anyone buying research material: quality should be built into production, and testing alone should not be relied on to ensure drug quality.⁵

Three pharmacies in 2026 did nothing visibly wrong at their own bench. They bought a chemically correct substance whose label said, accurately, that it was a dietary ingredient, and made injectable preparations from it. A certificate of analysis for finished product would not have flagged the problem unless someone had specifically ordered an endotoxin test, because chemically the material was what it claimed to be. The failure was a grade decision made one step upstream, and it propagated intact.

For laboratory work the same logic holds in a lower key. If an experimental system is sensitive to inflammatory signalling, endotoxin in the material is not a background nuisance; it is an active variable, and LPS is one of the most potent immunological stimuli known. A cytokine response attributed to a test compound, in material never assessed for endotoxin, may be a response to the packaging its bacteria left behind. That is a reproducibility problem before it is a safety one, and the questions that address it are specific: which grade the starting material was, whether an endotoxin test was run, on which batch, and to what limit.

The companion to this article is our report on the Australian case in which a vial sold as retatrutide contained no retatrutide, which covers the identity half of the same problem. On chemical impurities specifically, see the evidence that TFA is not an inert counterion and what HPLC and mass spectrometry can establish about a sample. The regulatory backdrop is in what the FDA is about to debate on peptide impurities and in what the July 2026 compounding vote actually decided.

How this was checked. The Centric Compounding recall was read on the FDA recall page, including lot numbers and dates, on 21 September 2026. The August recalls and the common-source finding were taken from the FDA and CDC pages describing the glutathione investigation, and the 2019 precedent from the FDA’s own compounding notice, which is where the USP <85> citation and the quoted principle originate. Endotoxin figures for gray-market peptides are from a preprint and are reported as such. Descriptions of what each test measures are general pharmacopoeial practice and are stated without reference to any supplier’s programme. Version 1.0, first published 21 September 2026.

Condor Research supplies peptides as characterised reference materials for laboratory research use only: not for human or veterinary use, not for diagnostic or therapeutic application, and not for any food or cosmetic purpose. None of the recalled products discussed here are supplied by Condor Research, and nothing in this article describes, recommends or implies the preparation or administration of any substance for human use.

Condor Research · Scientific desk
Atrio Sciences s.r.o., IČO 57 669 651, Nitra (SK) · info@condorresearch.com