Sterile Is Not the Same as Safe: The 2026 Glutathione Endotoxin Recalls

What was recalled, when, and by whom

Between 5 August and 9 September 2026, three Texas compounding pharmacies recalled injectable glutathione and two glutathione-containing blends because the bulk powder carried elevated bacterial endotoxin. Victory Medical Center Pharmacy in Austin went first, on 5 August: three lots of compounded glutathione shipped to Texas, Florida and New York, found by its own testing, with reports of fever, chills, headache, vomiting and blood-pressure changes. Optimal Balance Pharmacy in Houston followed on 19 August with one lot sent to fourteen states, also found by pharmacy testing, with a similar list plus symptoms consistent with an allergic reaction. Neither gives a patient count.

Centric Compounding, also in Houston, announced the third in a release dated 9 September; FDA's page lists the company announcement date as 11 September and posted it on 13 September. Six lots of three products, glutathione, a blend the notice spells Myer's Cocktail, and Tri-Immune Boost, shipped nationwide to patients and prescribers, all made with a glutathione ingredient containing elevated endotoxin. The company had received nine reports of fever, chills, chest pain, nausea or vomiting, headache and malaise. It names no supplier. And five of the six beyond-use dates had already passed when the recall was announced: most of the recalled vials had been used or had passed their beyond-use date before anyone was told.

On 27 August FDA said it knew of at least 30 patients with adverse events after intravenous glutathione, alone or with other ingredients, some hospitalised, with fever, chills, shock and sepsis-like symptoms it called consistent with excessive endotoxin. Every pharmacy then known to it had bought glutathione lot 229536 from Medisca Inc. of Plattsburgh, New York, labelled as dietary supplement grade, and Medisca had told FDA on 20 August that it had warned its customers not to use it in injectables. It referred to two Texas pharmacies that had recalled; Centric's notice came thirteen days later and names no supplier, so we do not write that Centric used that lot. CDC has since added Centric to the pharmacies whose product came from dietary-supplement-grade glutathione.

CDC's page, dated 15 September, counts over 25 people ill in at least four states, Texas, Washington, Florida and New York, at least three hospitalised and one in intensive care for shock and abnormal blood clotting. The two counts come from two agencies on two dates; we report both as given. Neither says how much endotoxin was in any vial.

Sterile is not the same as pyrogen-free

Endotoxin is lipopolysaccharide from the outer membrane of Gram-negative bacteria; the lipid A portion carries most of the toxicity. Bacteria shed it as they grow and release more when they die. It is a heat-stable, water-soluble molecule, and the steps that make a preparation sterile do not remove it.

USP says so in one sentence: pyrogens are organic compounds soluble in water and not removed by filtration or steam sterilisation. A sterilising filter with 0.22-micron pores stops bacteria; the endotoxin monomer is a 10 to 20 kilodalton molecule, far smaller, and passes through. Autoclaving kills spores and leaves endotoxin intact; USP notes that endotoxins resist dry heat better than the most heat-resistant spores, and the reference condition for destroying them is 250 degrees Celsius for 30 minutes, which no drug solution survives. Depyrogenation is done to empty glassware, never to the glutathione in it. That leaves the starting material as the only place endotoxin can be controlled, which is why USP requires an injectable made from nonsterile components to be sterilised within six hours: whatever grows in the bulk solution sheds endotoxin that the filter then leaves behind.

FDA treats the two attributes as separate. Its manufacturing guidance for outsourcing facilities lists sterility and a limit for bacterial endotoxins as distinct required specifications, and its insanitary-conditions guidance names compounding with components not verified for endotoxin as insanitary in itself. A vial can pass a sterility test, contain no living organism, and still make the person who receives it shake with fever. None of the three notices says whether a sterility test was run. It would not have mattered.

What endotoxin does to a person

The human response is well characterised, because purified endotoxin has been given to volunteers for a century as a model of inflammation. In a 1981 study the minimal pyrogenic dose was about 0.1 to 0.5 nanograms per kilogram; one endotoxin unit is roughly a tenth of a nanogram of reference material. A 1999 study giving 20 men up to 4 nanograms per kilogram found dose-related rises in symptoms, temperature, white cells and inflammatory cytokines.

At the top of that range the picture is a circulation, not a fever. A 1989 New England Journal of Medicine study gave 4 nanograms per kilogram to normal volunteers: three hours later heart rate was up 36 percent, cardiac index up 53 percent and systemic vascular resistance down 46 percent, the vasodilated state of early septic shock in people with no infection. Later work describes endotoxin as a specific agonist of Toll-like receptor 4, the innate immune sensor that starts the cascade, and that dose range as the one used to create a sepsis-like state on purpose.

The same happens whenever endotoxin gets into an injectable. In 1998 CDC investigated 20 patients at one Los Angeles hospital, with severe shaking chills, often with fever, a fast heart rate or a 20-point fall in systolic blood pressure, within three hours of intravenous gentamicin. Cultures were negative. The lots met their monograph limit; giving a whole day's dose in one hour had pushed exposure past 5 endotoxin units per kilogram. In a 1975 dialysis unit, 23 of 70 patients had chills, fever and hypotension, no infection was found, and endotoxin was in the dialysate and their blood.

Set the 2026 notices beside that: fever and chills in every one, headache, vomiting, body aches, blood-pressure changes, dizziness, chest pain, and one patient in intensive care for shock. FDA itself called the symptoms sepsis-like. Our reading of the mechanism is that this is innate immunity rather than allergy: it needs no prior sensitisation, is reproduced reliably in volunteer challenge studies, and in those studies and the gentamicin cluster developed over the first three hours, although the 2019 Letco patients were reported ill within minutes. Regulators do not draw that line sharply, the notices list anaphylactic shock among possible outcomes, and a person who is ill should not try to draw it either.

The limit, the test, and who has to run it

The limit is a rate. USP's bacterial endotoxins chapter sets it as K divided by M: K is 5 endotoxin units per kilogram of body weight per hour by any route other than intrathecal, 0.2 for intrathecal, and M is the maximum dose per kilogram given within an hour. For a 70 kilogram adult that is 350 endotoxin units in any one hour: our arithmetic on the rule, not a dose of anything. FDA's 2026 guidance adds that a product with several ingredients must meet the overall limit as a whole, the point for a cocktail.

The test is the Limulus amebocyte lysate assay, a clotting reaction from horseshoe crab blood, FDA's accepted finished-product test since the late 1970s. The question is whether anybody was required to run it, and when.

For a 503A pharmacy the answer is USP chapter 797, the state-enforced sterile compounding standard. It requires an endotoxin test on injectables made from nonsterile components only in defined cases: Category 2 preparations with a beyond-use date long enough to require sterility testing, and all Category 3 preparations. For Category 2 injectables with shorter dates it says they should be tested, a recommendation. And a preparation that requires sterility testing may be dispensed before results are back, provided the pharmacy has procedures to notify the prescriber and recall unused product if it fails. Our reading is that this is the lawful route by which an endotoxin-failing lot reaches patients: release first, recall on the result.

A 503B outsourcing facility is registered with FDA, inspected by FDA, must report adverse events and is subject to good manufacturing practice, under which sterility and an endotoxin limit are release specifications for every batch. CDC calls all three recalling firms pharmacies, and two carry the word in their names; section 503A is the pathway for a licensed pharmacy compounding for an identified patient on a prescription. None of the notices states which pathway the firm operated under, and we have not verified it.

The powder was labelled as a dietary supplement, again

FDA's 27 August statement contains the sentence that should sting: the agency previously communicated in 2019 about dietary ingredient glutathione in injectables and remains concerned about continued risks to patients. The 2019 event is this one with the dates changed.

In January 2019 FDA received a report of seven patients injected with compounded glutathione on the same day who developed nausea, vomiting, chills and body aches within minutes; one had low blood pressure and difficulty breathing and was hospitalised. A February report described a patient with sudden chills, fever and shaking, admitted for a possible bloodstream infection. Eight in all. The powder had been distributed by Letco Medical of Decatur, Alabama. Its label read Caution: Dietary Supplement and, beside that, For Manufacturing, Processing, Repackaging and Pharmaceutical Compounding; the manufacturer told FDA the ingredient was not intended for use in sterile drugs, and its original invoice said it was not for pharmaceutical use. FDA's testing found excessive bacterial endotoxin, some results as high as five times the appropriate limit.

In October 2024 the same failure arrived with a different molecule: compounders had used food-grade NAD+ from repackagers in intravenous products, and patients had severe chills, shaking, vomiting and fatigue consistent with excessive endotoxin. FDA wrote that food-grade ingredients are unsuitable for sterile compounding because of the high risk of contamination with microbes and endotoxins, and in 2026 that supplement-grade ingredients may contain impurities and contaminants, including endotoxins.

Grade is the whole story. A powder sold as a supplement ingredient is made and tested for something that will be swallowed, with the gut and liver in the way, and its certificate need say nothing about endotoxin. Injected, it bypasses those defences. Three warnings in seven years, two about the same tripeptide, are what happens when a supply chain built for capsules fills vials and the rule that would catch it is, for some preparations, only a recommendation.

What the three products are and what the evidence says

Glutathione is the tripeptide gamma-glutamyl-cysteinyl-glycine, the cell's main antioxidant thiol. No FDA-approved glutathione injection exists for any systemic use. In June 2022 FDA proposed not to add it to the 503A bulk list; its advisory committee narrowly disagreed, FDA has not decided, and glutathione is compounded meanwhile under interim discretion. The recall notice says the products were used to support immunity, fatigue and general wellness; a 2025 JAMA Internal Medicine survey of 255 intravenous-spa websites found glutathione offered by 54 percent, every site making benefit claims and two citing a source.

For those claims there is no controlled trial: a PubMed search this month for randomised trials of intravenous glutathione returns nothing on fatigue, immunity or wellness in healthy adults. The legitimate injectable use is narrow: double-blind trials of 50 to 151 patients between 1995 and 2002 gave a few supervised doses of pharmaceutical-grade glutathione around cisplatin or oxaliplatin and found less neurotoxicity on subjective scales. The one placebo-controlled study of intravenous glutathione for skin lightening analysed 32 women after eight treated women were withdrawn for deranged liver function tests and one for anaphylactic shock; the result was not significant, and a 2025 systematic review called intravenous use contraindicated.

Myers' cocktail is a name, not a formula. Its namesake, a Baltimore physician, died in 1984 leaving no written recipe, and the practitioner who reconstructed it from patients' recollections wrote in 2002 that it was not clear exactly what it consisted of. Published and marketed recipes differ from one another, and a clinical pharmacology society's 2023 position paper says there is no standardisation of the mixtures administered. The one randomised, double-blind trial gave 34 adults with fibromyalgia a course of infusions of the cocktail or a placebo solution and found no significant difference on any outcome. Centric's notice does not say what was in its version.

Tri-Immune Boost is an industry trade name for vitamin C, glutathione and zinc sulfate in one vial. A PubMed search for any trial of that combination by injection returns zero records. The Cochrane reviews of vitamin C and of zinc for colds, 11,306 and 8,526 participants, cover swallowed or nasal products, none injected, and are null for prevention in the general community and modest for duration. In the largest randomised trial of intravenous vitamin C, LOVIT, in 872 intensive care patients with sepsis on vasopressors, the vitamin C arm had more death or persistent organ dysfunction at 28 days, 44.5 against 38.5 percent. Centric's notice lists no ingredients for this product either, so even the trade formula is an assumption.

Why this channel keeps failing

Every product in this event shares a shape: made by a compounding pharmacy from bulk powder, sold for a claim no trial supports, through a clinic or by home delivery, and given where nobody would be watching two days later. FDA's compounding page says compounded drugs are not FDA-approved and that the agency does not verify their safety, effectiveness or quality before marketing. A clinical pharmacology society's position paper adds the part that matters: side effects develop in the days after an infusion, when there is no medical oversight.

The JAMA survey measured the setting: of 87 spas called by secret shoppers, 28 percent required a consultation with a licensed professional before treatment and 24 percent described any risk. FDA had said in October 2021 that it was increasingly aware of hydration clinics, medical spas and mobile infusion services compounding drugs that may not meet the conditions of section 503A, and that adverse events are underreported. Texas passed a law, effective 1 September 2025, defining elective intravenous therapy and restricting whom a physician may delegate it to.

The channel is not limited to vitamins. It supplies compounded semaglutide: in June 2026 FDA sent a warning letter to a mobile hydration business for marketing compounded semaglutide and tirzepatide as coming from an FDA-approved pharmacy, which FDA says no compounding facility is. BPC-157 shows the same weakness at the supply end: FDA's briefing for its July 2026 advisory committee noted that the nominator's certificate of analysis mentioned no endotoxin or bioburden test, and called the endotoxin test a critical quality attribute for anything meant to be injected. The glutathione recalls are what that concern looks like when it materialises.

What a reader can do with this

This is not medical advice, and the regulators' guidance is specific enough that we need not add to it. CDC tells clinicians not to administer injectable glutathione compounded from dietary-supplement-grade ingredients and to confirm with the pharmacy that ingredients are appropriate for injection; it tells patients to ask whether the product uses injection-appropriate ingredients and about sterilisation and endotoxin testing. All three notices say to stop using the affected vials: Centric's says to discard them, Victory's and Optimal Balance's say to quarantine them and contact the pharmacy to arrange retrieval.

The lot table at the top of this article is the complete list from the three notices. A lot number not on it means the vial is not on this list, not that it is safe; the notices do not say which supplier lots each pharmacy used. The checklist below turns the regulators' questions into something a person can ask across a counter.

A reaction looks like the clusters above: rigors, fever, headache, nausea or vomiting, body aches, a racing heart, dizziness or a fall in blood pressure, beginning within hours rather than days. CDC says severe reactions warrant emergency medical care and nothing more specific, so neither do we. What makes it easy to misread is that in every documented cluster it looked like sepsis and cultures came back negative. FDA says such events are underreported; any count that closes this one is likely to be smaller than the number of people it touched.