Oxygen

Why this drug is interesting

Oxygen is given to more poisoned patients than every other antidote on this site combined, and it is almost never prescribed as one. For carbon monoxide it is not supportive care — it is the specific treatment, working by exactly the mechanism a chelator or a receptor antagonist works by, except that the competing ligand happens to be the molecule the body already runs on.

It is also the antidote with the longest-running and least resolved controversy attached to it, and one of the few where the argument is not about whether to give the drug but about what pressure to give it at.

Pharmacology

Why carbon monoxide is dangerous

Carbon monoxide binds haemoglobin with an affinity roughly 200–250 times that of oxygen, forming carboxyhaemoglobin. Three things follow, and only the first is widely taught:

• Reduced oxygen-carrying capacity — the obvious effect, proportional to the carboxyhaemoglobin fraction.
• A leftward shift of the oxyhaemoglobin dissociation curve — the remaining haemoglobin holds its oxygen more tightly and releases less to the tissues. This is why carbon monoxide is worse than an equivalent anaemia.
• Direct cellular toxicity — binding to myoglobin and to cytochrome c oxidase, impairing mitochondrial respiration in the same way cyanide does. This is thought to underlie the delayed neurological sequelae, and it is not reflected in the carboxyhaemoglobin level at all.

How oxygen reverses it

Binding is competitive and reversible. Raising the partial pressure of oxygen drives carbon monoxide off haemoglobin by mass action and dissolves additional oxygen in plasma, which supports tissue delivery independently of haemoglobin. The rate of elimination therefore depends directly on inspired oxygen partial pressure:

Why the monitoring lies to you

• Arterial blood gas PaO₂ is also normal, because dissolved oxygen is unaffected. The oxygen is in the plasma; it is the haemoglobin that is unavailable.
• You need co-oximetry to measure carboxyhaemoglobin. Most blood gas analysers do it; ask for it explicitly.
• Venous samples are adequate for carboxyhaemoglobin — an arterial puncture is not required to make the diagnosis.
• Cherry-red skin is a post-mortem finding, not a clinical sign to wait for.

Treatment

Normobaric oxygen

High-flow oxygen through a tight-fitting non-rebreathe reservoir mask, started immediately on suspicion and continued regardless of the SpO₂ reading, is the treatment. In the intubated patient, 100% inspired oxygen.

There is no titrating to saturations here: the pulse oximeter is measuring the wrong thing, and the therapeutic target is a partial pressure gradient, not a saturation. Continue until carboxyhaemoglobin has fallen and symptoms have resolved — conventionally for a minimum of several hours, and longer in pregnancy.

Alongside

• Consider concurrent cyanide poisoning in any enclosed-space fire — soot, altered mental status and a lactate ≥10 mmol/L. See hydroxocobalamin, and note that sodium nitrite is the wrong drug here precisely because it makes methaemoglobin in a patient whose haemoglobin is already compromised.
• Do not give activated charcoal — the route is inhalational.
• Look for the source. Faulty boiler, blocked flue, generator or barbecue indoors, shisha. Other occupants and pets may be affected, and the patient may be returning to the same house.
• Consider deliberate exposure and the mental health assessment that follows.
• ECG and troponin — myocardial injury is common and independently predicts worse long-term outcome.

The hyperbaric question

Hyperbaric oxygen shortens the carboxyhaemoglobin half-life dramatically and is biologically plausible for the mitochondrial and inflammatory components of injury. The question has always been whether it reduces delayed neurological sequelae, and the honest answer is that forty years of trials have not settled it.

What the Cochrane review found

The 2011 Cochrane review identified seven randomised trials, excluded one that did not evaluate clinical outcomes, and analysed six involving 1,361 participants. Two found a benefit at one month; four did not.¹

Existing randomised trials do not establish whether the administration of HBO to patients with carbon monoxide poisoning reduces the incidence of adverse neurologic outcomes.

Juurlink, Buckley and colleagues, Cochrane Database of Systematic Reviews, 2011¹

In practice

UK hyperbaric chambers are few, geographically concentrated and often coastal. Referral means transferring a poisoned, sometimes unstable patient — with the risk that entails, and with the transfer time itself eroding the theoretical advantage over the high-flow oxygen the patient is already receiving. The criteria in circulation (loss of consciousness, neurological signs, myocardial ischaemia, pregnancy, high carboxyhaemoglobin) are consensus, not trial-derived.

Safety

In acute poisoning, high-flow oxygen is close to risk-free and the theoretical hazards should not delay it.

Hypercapnic respiratory failure

The familiar concern in COPD. In carbon monoxide poisoning it does not justify withholding high-flow oxygen — the patient is dying of impaired oxygen delivery, not of a saturation target. Monitor the gases and support ventilation if needed

Oxygen toxicity

Pulmonary and CNS toxicity are functions of pressure and duration. Not relevant to hours of normobaric therapy; relevant to hyperbaric treatment, where seizures are a recognised complication

Barotrauma

A hyperbaric issue — middle ear, sinus and pulmonary. Pneumothorax is a contraindication to compression until drained

Fire

An oxygen-enriched atmosphere is a genuine hazard, particularly relevant in the patient who has just been in a fire and may wish to smoke

Paraquat

The one poisoning in which oxygen is actively harmful — it potentiates redox cycling and pulmonary fibrosis. Oxygen is withheld unless hypoxia is life-threatening

Critical appraisal

Oxygen for carbon monoxide poisoning is the rare case where the mechanism is unambiguous, the effect is measurable in real time, and the drug is free and universally available. Nothing about normobaric treatment is controversial and nothing should delay it.

The hyperbaric literature, by contrast, is a cautionary study in how a plausible mechanism plus an available technology plus a devoted specialty produces four decades of inconclusive trials. Two positive trials stopped early for benefit; four negative trials underpowered or excluding the sickest patients; one completed trial never reported at all. The Cochrane authors' verdict — that the trials do not establish the answer — is not a fudge; it is the accurate description of a literature nobody has been willing to fix.

One structural observation to close on. The delayed neuropsychiatric sequelae that hyperbaric oxygen is meant to prevent are the part of carbon monoxide poisoning that emergency departments never see — they appear weeks later, to somebody else. That is a large part of why the question remains open: the specialty that makes the decision is not the specialty that observes the outcome.

References

1. 1
   Buckley NA, Juurlink DN, Isbister G, Bennett MH, Lavonas EJ. Hyperbaric oxygen for carbon monoxide poisoning. Cochrane Database of Systematic Reviews 2011;(4):CD002041. PubMed 21491385 · PMC7066484. Six trials, 1,361 participants. Source of the pooled odds ratio 0.78 (95% CI 0.54–1.12), the two-positive/four-negative split, the flaws in all trials, the early stopping of both positive trials, the low power of the negative trials, the unreported completed trial and the authors' conclusion. Abstract verified 4 September 2026.
2. 2
   Juurlink DN, Buckley NA, Stanbrook MB, Isbister GK, Bennett M, McGuigan MA. Hyperbaric oxygen for carbon monoxide poisoning. Cochrane Database of Systematic Reviews 2005;(1):CD002041. PubMed 15674890. The earlier version, superseded by the 2011 update above; reached the same conclusion.
3. 3
   National Poisons Information Service. TOXBASE · NPIS 0344 892 0111. The operational UK source for carbon monoxide management, hyperbaric referral criteria and duration of oxygen therapy. A search of the electronic Medicines Compendium for "oxygen" and "medicinal oxygen" returned no products on 4 September 2026, so no UK SmPC is cited on this page; the carboxyhaemoglobin half-life figures are conventional toxicological values, not label-derived.