Hydrogen Peroxide Is a Bottle of Oxygen. Sometimes It Arrives in the Wrong Place

Macro photograph of a tipped glass reagent bottle on a steel tray, with a rising column of oxygen micro-bubbles dissolving into a fine branching vessel pattern

The patient was thirty-one, with tuberculosis of the third lumbar vertebra and an abscess tracking alongside the spine. She went to theatre prone, the pus was evacuated, and then the cavity was washed out repeatedly with 200 millilitres of hydrogen peroxide at 3 per cent — the strength sold in any pharmacy. She arrested almost immediately. During resuscitation, transoesophageal echocardiography showed the right side of her heart filled with what the Chinese surgical team described as a diffuse “snowflake-like” gas embolus, and a CT of her head showed scattered pockets of air in both frontal lobes. Circulation came back. She was taken for hyperbaric oxygen therapy without delay. There was no improvement, and she died in intensive care on the seventh day. The case report published in 2024 in the International Journal of Surgery Case Reports is unusually blunt about why: peroxide had produced a volume of oxygen nobody had accounted for, in a cavity with no way to vent it.

That case is not a freak accident of chemistry. It is arithmetic.

A household bottle is a gas cylinder in disguise

Hydrogen peroxide decomposes on contact with catalase, the enzyme that sits in mucous membranes, liver, kidney, red cells and bone marrow. The reaction is unremarkable on paper — two molecules of peroxide give two of water and one of oxygen — but the volumes are not. A review published this week in The American Surgeon works it through: complete decomposition of a single millilitre of 3 per cent weight-per-volume solution yields roughly 9.9 millilitres of oxygen at 0°C and one atmosphere, or 11.2 millilitres at body temperature. A 200-millilitre wash therefore has the theoretical potential to liberate something close to two litres of gas. Concentrated solutions are worse. The Geneva case report traces a figure to the 1998 paper that first used hyperbaric oxygen to reverse a peroxide-induced stroke: 30 millilitres of 35 per cent peroxide can yield as much as 3.5 litres of oxygen.

In an open wound with a swab and a suction tip, that gas leaves. In a semi-closed cavity, in a wound being irrigated under pressure, in an abscess bed beneath the fascia, or in a stomach, it has only one route: into the venous plexus. From there it travels to the portal vein, to the right heart where it can lock the outflow tract, or — through a patent foramen ovale, a pulmonary shunt or the arterialised bed of an inflamed mucosal surface — into the systemic circulation and the brain.

The author of the new review, an independent author writing from Wyoming, is careful about the limits of the operative literature: published reports describe suspected pulmonary, demonstrated intracardiac and cerebral gas events, but they “establish neither incidence nor a safe exposure threshold”. The same review makes a second point that matters for anyone who defends the practice, namely that peroxide’s clinical benefits are setting-specific and must be distinguished from its laboratory antimicrobial effect — in other words, the fact that it kills bacteria in a dish is not evidence that irrigating a cavity with it helps a patient.

What poison centres actually record

Ingestion is the route with the largest denominators, and the benchmark dataset is a ten-year abstraction of the American National Poison Data System published in Annals of Emergency Medicine in 2017. Of 1,054 reported high-concentration peroxide ingestions, 294 met inclusion criteria. Forty-one of those patients, 13.9 per cent, showed evidence of an embolic event — seizure, altered mental status, hypoxia, haemodynamic instability, ECG change, radiographic stroke, focal deficit, raised troponin. Twenty, 6.8 per cent, either died or were left with ongoing disability when the poison-centre chart was closed. The authors also reported that outcomes were better after early hyperbaric oxygen, a finding that must be read for what it is: a retrospective chart abstraction, not a randomised comparison, with the usual possibility that the patients who reached a chamber quickly were also the ones who were less catastrophically ill.

The timing question was taken up in a 2023 systematic review in Clinical Toxicology, which pooled 85 published cases with 41 from that same poison-centre study to reach 126. Clinically significant embolic phenomena clustered within ten hours of exposure, though delayed events do occur and belong in any decision about how long to observe. The review’s most useful sentence is its most cautious one: it “remains equivocal whether hyperbaric oxygen therapy is beneficial in cases of primary portal venous gas without systemic involvement.” An earlier case series from Oregon, covering eleven patients with portal venous gas after concentrated ingestions, found that hyperbaric treatment cleared the gas completely in nine and almost completely in two, with ten patients discharged within a day — encouraging, again, and still uncontrolled.

Three settings, one mechanism

The Geneva case is the most instructive ingestion report of the past two years. Two men in their late seventies drank a sip from an unlabelled bottle they believed to be white wine. Laboratory analysis later identified nearly pure peroxide, above 90 per cent. Both developed portal venous gas. One became somnolent and then confused six hours after the dose; the other, who had only epigastric discomfort, developed a troponin rise and a small ECG change that his physicians labelled a type 2 non-ST-elevation myocardial infarction, possibly from a cardiac gas embolus. Both were treated approximately thirteen hours after ingestion on a long Comex 18 table, the European equivalent of US Navy Table 6. Follow-up CT nine hours later showed no gas anywhere in the portal system. Both went home after 48 hours. Both also sustained bilateral grade III ear barotrauma — a reminder that the treatment is not a spa session.

Surgery supplies the second setting, and the 2026 reports are sobering. A man in his twenties was undergoing debridement and external fixation of a femoral fracture from a blast injury. About forty-five minutes after induction, hydrogen peroxide was used to irrigate the semi-closed wound. End-tidal carbon dioxide fell, the heart rate climbed, and within moments he was in asystole. Six minutes of compressions and immediate intubation restored circulation, and he recovered without neurological deficit — a good outcome that depended on the arrest happening under a monitor with an anaesthetist present.

How common the practice is matters here, because a rare complication of an abandoned practice is a curiosity while a rare complication of a routine one is a public-health problem. A 2025 survey of 242 German neurosurgeons, published in Pharmaceuticals, found that 81 per cent use hydrogen peroxide in their operating lists, 62 per cent use it inside the skull, and more than half of those use it as deep as the intradural layer. Only 5 per cent could cite literature on how it damages neural tissue. The same paper measured that damage: human cortical slices irrigated with 1.5 and 3 per cent peroxide showed vacuolation to a mean depth of 544 and 859 micrometres, against 2,504 micrometres for bipolar cautery. Peroxide is, on that evidence, the gentler instrument locally. It is the gas phase that kills.

The third setting is the enema. The ACG Case Reports Journal added a portal venous gas embolism after a peroxide enema to the record in 2025, and the mechanism is identical — only the anatomy changes.

Not every case ends up in a chamber, and that variation is itself worth reporting. A 2026 report in the Journal of Stroke and Cerebrovascular Diseases describes an 82-year-old woman who drank 50 millilitres of 35 per cent peroxide, developed left-sided weakness, and turned out to have the scattered, non-territorial pattern of punctate cortical diffusion restriction that distinguishes air embolism from ordinary cardioembolism — a distinction that mattered, because she also had incidental atrial fibrillation and anticoagulating her would have been the wrong move given the corrosive gastritis found at endoscopy. She improved with oxygen and supportive care. A second 2026 report in BMJ Case Reports describes a middle-aged man with multiple cerebral infarcts from air emboli, also managed with high-concentration oxygen. Whether those patients would have done better in a chamber is exactly the question the case literature cannot answer.

Why the chamber, and what it cannot fix

Gas embolism is the first entry on the Undersea and Hyperbaric Medical Society’s list of accepted indications, and it is named explicitly in Medicare’s national coverage determination for hyperbaric oxygen therapy, which has been in force since April 2017. The physiology is taught in every introductory course: at pressure, a bubble obeys Boyle’s law and shrinks; breathing 100 per cent oxygen under pressure washes nitrogen out of the plasma and steepens the gradient driving gas out of the bubble; and the dissolved oxygen in plasma keeps ischaemic tissue alive while flow is restored.

What the chamber cannot do is undo a caustic injury. The mucosa burned by peroxide still needs endoscopy, proton-pump blockade, and time, and the 2017 poison-centre analysis found that routine endoscopy rarely changed management — grade 3 or 4 lesions appeared in only five of 294 patients. Nor does the chamber guarantee anything when the gas load is large. The Lanzhou patient received hyperbaric oxygen immediately and still progressed to brain death. That is the honest shape of the evidence: a mechanism that is well understood, a treatment that is widely accepted and physiologically rational, and an outcome literature made entirely of case reports and registries, with no trial and none on the horizon. This is also why the timing advice reads the way it does — treat as early as possible, and do not withhold treatment simply because hours have passed.

The regulatory frame, and the marketing temptation

The FDA’s August 2025 letter to health care providers remains the anchor for any discussion of chamber safety. It notes that hyperbaric devices are Class II, cleared through the 510(k) pathway under product code CBF, and that the agency is aware of “reports of fires that occurred with HBOT devices that resulted in serious injuries and deaths”, with the root cause not known. It directs providers to follow the manufacturer’s instructions, train and monitor staff, observe cleaning and maintenance intervals, and keep electrical and static-generating items and synthetic clothing out of the chamber. Those are not abstract warnings; they are the reason a chamber is a controlled environment rather than a room with a pressure vessel in it. The UHMS goes further in its own guidance, warning that low-pressure chambers sold to “wellness centres” and spas, often without physician supervision and in buildings that do not meet fire and pressure-vessel standards, are “putting those exposed at risk for serious injury and even death.”

Which brings the story back to the bottle. A peroxide solution effervesces on a wound, and that fizz is the entire marketing proposition — visible oxygen, doing something. The same catalase reaction, uncoupled from an open surface and a vent, is what put two litres of gas into a paravertebral abscess cavity in Lanzhou and what stopped a young man’s heart in an operating theatre. Neither hyperbaric oxygen nor any wellness chamber makes peroxide safer. What reduces the number is duller than that: irrigate with saline where saline will do, keep peroxide out of closed cavities and deep wounds, let the gas escape when it must be used, tell the team before it is poured, image the abdomen early in a concentrated ingestion, and observe for at least ten hours rather than sending an asymptomatic patient home at hour two. The chamber is a rescue, and rescues are what you arrange after prevention has already failed.

Sources

← Back to blog