In Carbon Monoxide Poisoning, the Pressure of the First Session May Matter More Than the Number That Follow

Stylised illustration of a hyperbaric chamber control console: a large circular pressure gauge with a red needle against a teal therapeutic-range arc, a bank of smaller dials, brass valve handles and a condensation-flecked porthole, lit in cool teal with amber accents

Carbon monoxide poisoning is the emergency that built modern hyperbaric medicine, and it remains the clearest example of the therapy working on a physiological problem rather than a marketing promise. It is also the indication where the prescription is least standardized. Clinicians who agree that a patient should be treated routinely disagree about how hard to pressurize and how many times to repeat the exposure — a disagreement that has survived four decades because the randomized trials were built to answer a different question.

A retrospective published in the October 2026 issue of Clinical Toxicology takes that unanswered question head-on, and its answer is blunt: the pressure of the first session appears to matter, and the number of sessions that follow appears not to.

Thirty years of records from one centre

The study comes from the hyperbaric referral service at Rambam Health Care Campus in Haifa, Israel, with analysis by Dr. Itay Gur and colleagues at the University of Haifa. They reviewed every patient treated for carbon monoxide poisoning at the facility over roughly thirty years, then narrowed the group to 312 people who completed their prescribed course and had at least a year of medical records available.

Delayed neuropsychiatric sequelae — the cognitive, motor and psychiatric symptoms that can surface days to weeks after an apparently recovered poisoning — were defined as any new such symptom appearing between two and forty days after exposure. Four treatment regimens were compared: 2.0 ATA for 90 minutes in a single session; the same protocol three times; 2.8 ATA for 90 minutes once; and 2.8 ATA once followed by two sessions at 2.0 ATA.

The sequelae rates tracked the pressure, not the session count. They were 36 percent after one session at 2.0 ATA and 37 percent after three sessions at 2.0 ATA. After a first session at 2.8 ATA the figure fell to 20 percent, and to 19 percent when the higher-pressure session was followed by two conventional ones.

Pooled, patients whose first exposure was at the lower pressure developed delayed sequelae 36.2 percent of the time, against 19.3 percent for those started at 2.8 ATA — a difference the authors report at P = 0.0013. A multivariable logistic regression adjusting for age, carboxyhemoglobin level and presenting symptoms put the higher-pressure protocol at an adjusted odds ratio of 0.45, with a confidence interval of 0.26 to 0.77, meaning roughly a 55 percent reduction in the odds of developing sequelae. Increasing the number of sessions from one to three was not associated with any significant benefit, either within the 2.0 ATA group or the 2.8 ATA group. Seizures on admission were the strongest predictor of a poor outcome, with an adjusted odds ratio of 4.27.

Why the first hour would matter at all

The result is easiest to dismiss as an artifact of sicker patients being treated differently, but the physiology offers a coherent explanation. Hyperbaric oxygen hastens the clearance of carboxyhemoglobin — the mechanism most people know — but that clearance is only part of the story.

As Dr. Lindell Weaver of Intermountain Healthcare and the University of Utah describes in the carbon monoxide chapter of the 15th edition of the UHMS Hyperbaric Indications Manual, hyperbaric oxygen also modulates the inflammatory cascade set off by poisoning: it transiently inhibits lipid peroxidation, impairs leukocyte adhesion to injured microvasculature, improves mitochondrial function, and reduces brain inflammation caused by CO-induced adduct formation on myelin basic protein. Normobaric oxygen does not produce those effects.

If the durable injury is inflammatory rather than simply hypoxic, then the therapeutic lever is not total oxygen delivered over a course — it is whether the initial insult is interrupted early enough and deeply enough. That is the Rambam authors’ reading, and it fits the pattern in their data: the higher first pressure helped even in high-risk patients who presented with seizures, while the second and third sessions added nothing measurable.

The randomized question nobody answered

This is where the new study has to be read carefully. It is not a controlled comparison; it is a single-centre retrospective in which treatment choices were made by clinicians over three decades, and the requirement that patients complete their prescribed course and have a full year of records filters the sample toward those who stayed in the system.

The randomized evidence on the therapy’s neurological benefit, meanwhile, remains unresolved. The 2011 Cochrane review by Nicholas Buckley and colleagues pooled six trials with 1,361 participants and found no statistically significant benefit for hyperbaric oxygen over normobaric oxygen on neurological deficits, with a pooled odds ratio of 0.78 and a confidence interval spanning 1.12. The reviewers were careful to note that the trials were too heterogeneous and too flawed to call the question either way: two “positive” trials stopped early for benefit, one showed an apparent change in its primary outcome during analysis, and one positive result may have reflected a failure to adjust for multiple testing. Their conclusion was that existing randomized trials do not establish whether hyperbaric oxygen reduces adverse neurologic outcomes in poisoned patients.

Crucially, those trials compared hyperbaric oxygen against normobaric oxygen. They did not compare one pressure against another. The dose question the Haifa group examined was never the subject of a randomized trial, which is why a retrospective of this size is worth reading — and why it cannot be the last word.

Guidance already disagrees with itself

The absence of a settled protocol shows up in the guidelines. The UHMS indications chapter takes a broad position: hyperbaric oxygen “should be considered for all cases of acute symptomatic carbon monoxide poisoning,” and it is specifically indicated where poisoning is complicated by cyanide exposure, as often happens with smoke inhalation.

The 2025 position statement of the Polish Medical Society’s clinical toxicology section draws a narrower line. It treats immediate normobaric 100 percent oxygen as the cornerstone of care, continued until carboxyhemoglobin falls to roughly 3 percent and for no less than six hours, and reserves hyperbaric oxygen for selected patients — chiefly those with persistent neurological or cardiac symptoms, or metabolic acidosis, despite normobaric oxygen, regardless of their carboxyhemoglobin level. Pregnant patients are the exception; there the statement says hyperbaric treatment is always indicated, and the first session should ideally occur within six hours.

Both positions are compatible with a dosing study like this one, and neither is settled by it. What the guidelines do agree on is that the diagnosis is harder than it looks: a 2025 review in Diagnostics notes that carboxyhemoglobin and pulse CO-oximetry both have significant limitations, and that prognosis worsens with delayed treatment.

A pressure ceiling is not free

There is a reason protocols drift toward the lower end of the pressure range even where higher pressures look better, and it is toxicity. Oxygen is a drug with a narrow therapeutic index at pressure, and the complications of hyperbaric treatment — central nervous system and pulmonary oxygen toxicity, barotrauma of the ears, sinuses and lungs — are reviewed in a 2025 primer for emergency physicians in the Journal of Emergency Medicine. Raising the initial pressure to 2.8 ATA buys a larger margin against delayed neurological injury and spends some of the margin against oxygen toxicity, which is exactly the kind of trade a retrospective cannot price.

The regulatory picture is separate. Medicare’s national coverage determination lists acute carbon monoxide intoxication among the conditions covered when the therapy is delivered in a chamber, and it says nothing about the pressure or session count — that has always been left to clinicians. On the device side, the FDA’s August 25, 2025 letter to health care providers reminds facilities that hyperbaric oxygen therapy devices are Class II devices cleared through the 510(k) pathway under Product Code CBF, and that the agency is aware of fires involving these devices that caused serious injuries and deaths. The letter’s recommendations — following manufacturer instructions, maintaining fire prevention, training and monitoring staff, controlling prohibited electrical and static-generating items — apply to every session, at every pressure.

Carbon monoxide poisoning kills hundreds of Americans and sends more than 100,000 to emergency departments each year, according to the CDC, and a therapy that reduces the chance of lasting brain injury has obvious value. But this study does not tell anyone to demand 2.8 ATA. It tells them that the dose question is real, that a randomized comparison of pressures is long overdue, and that the familiar reflex of adding sessions is not supported by the best data anyone has yet assembled.

Sources

← Back to blog