Why this poison is interesting

Everybody learns that carbon monoxide binds haemoglobin more avidly than oxygen does. That is correct — 230 to 300 times more avidly¹ — and if it were the whole mechanism, carbon monoxide poisoning would behave like acute anaemia. It does not. An anaemic patient does not develop a movement disorder three weeks later, and a patient with a haemoglobin of 70 g/L is usually walking around.

Three things follow from carbon monoxide doing more than occupying a binding site, and each one contradicts something that is widely taught.

It does not merely reduce carrying capacity — it also prevents unloading. Binding one haem group shifts the oxyhaemoglobin dissociation curve leftward, so the oxygen that is still carried is held more tightly.¹ A patient with 40% carboxyhaemoglobin is worse off than a patient with a haemoglobin halved, because the remaining haemoglobin is worse at its job. Established
It leaves the blood before it leaves the tissue. Carbon monoxide also binds skeletal and myocardial myoglobin, and elimination times in tissue and blood differ, so tissue injury can develop with a delay.¹ The number you measure clears first. Established
The injury continues after the gas has gone. At cellular level carbon monoxide produces neutrophil activation, lymphocyte proliferation, mitochondrial dysfunction and lipid peroxidation, with oxidative stress, inflammation and apoptosis comparable to a reperfusion injury.¹ That is a process, not an exposure — and it explains a syndrome that appears after recovery. The source hedges this one ("leads—among others—to") and the underlying evidence is largely animal and in vitro. Inferred

The result is a poisoning in which the measurement is the least informative thing available, and where the most consequential clinical question — whether to use hyperbaric oxygen — has been addressed by six randomised trials that do not agree.

A poison is a drug whose kinetics have escaped its pharmacology.

The toxic principle

The poison is carbon monoxide itself. There is no metabolism, no activation, no toxic metabolite and nothing to block — which puts it in the small group of poisons on this site where the Metabolism and the metabolites section is genuinely empty. What replaces it is a set of binding targets.

Haemoglobin: Carbon monoxide diffuses rapidly through the alveolar membrane and binds with an affinity 230–300 times that of oxygen, preferably to the iron ion in haem.¹ This reduces oxygen-carrying capacity in proportion to the fraction occupied. Established
Haemoglobin, again — the allosteric effect: Conformational changes lead to a leftward shift of the oxyhaemoglobin dissociation curve, to reduced oxygen transport capacity, and to reduced oxygen release into peripheral tissue.¹ This second effect is the one usually left out, and it is why carboxyhaemoglobin is worse than equivalent anaemia. Established
Myoglobin: Within tissue, carbon monoxide also binds other haem-containing proteins such as skeletal and myocardial myoglobin, and since elimination times in tissue and blood differ, tissue injury can also develop with a delay.¹ Established
Mitochondria: Carbon monoxide also binds the haem groups of the mitochondrial respiratory chain, and the review lists mitochondrial dysfunction among its cellular effects.¹ This is the mechanism it shares with cyanide and, at one remove, with methanol's formate — see methanol. Inferred

Toxicokinetics

Carbon monoxide — a gas, so the kinetics are respiratory rather than metabolic
Parameter	Therapeutic	In overdose	Why it changes
Absorption	Diffuses rapidly through the alveolar membrane¹. Uptake depends on inspired concentration, minute ventilation and duration of exposure	Same mechanism; the variable is exposure. An exposure period longer than 24 hours is a risk factor for late neuronal sequelae (OR 2.0, 95% CI 1.0–3.8, P=0.046)¹	Minute ventilation is part of the dose. A person who is exerting themselves, or a child with a higher respiratory rate per kilogram, takes up more from the same atmosphere — which is why members of the same household are affected unequally and why the least affected person's story is the least reliable guide.
Protein binding	Not binding in the pharmacological sense — it is bound to haem iron on haemoglobin, myoglobin and respiratory-chain cytochromes¹	Unchanged	This is the crucial structural point. Carbon monoxide is not distributed to tissue as free drug; it is held on proteins in several compartments with different release kinetics, which is why blood and tissue do not clear together.¹
Volume of distribution	Not meaningfully definable — distribution is determined by where haem proteins are, not by lipid or water partitioning	Unchanged	The muscle and myocardial myoglobin pool is the reason the concept fails usefully here: a substantial reservoir sits outside the blood, releases slowly, and is invisible to the assay.
Half-life	Carboxyhaemoglobin half-life: indoor air 320 min; 100% normobaric oxygen 74 min; 100% hyperbaric oxygen at 3.0 bar 20 min¹	Same figures. Note these are blood half-lives	The single most useful set of numbers on this page, and the most commonly over-read. They describe the clearance of the marker, not of the injury — tissue elimination is slower,¹ and the inflammatory process the exposure has started is not described by any of them. High-flow oxygen shortens the blood half-life more than fourfold, which is why it is given immediately and to everybody symptomatic.
Elimination route	Pulmonary, entirely. Exhaled unchanged; a trace is produced endogenously from haem catabolism	Unchanged	There is nothing hepatic or renal to fail. Elimination is a function of inspired oxygen partial pressure and minute ventilation, which makes oxygen simultaneously the antidote and the elimination-enhancement technique — a combination no other poison in Band A offers.
Order of kinetics	First order, and the clearance constant is set by inspired oxygen	First order	Nothing saturates. The therapeutic lever is not the drug's kinetics but the competing substrate's concentration — pure mass action at the binding site.
Dialysability	—	Not applicable. No EXTRIP recommendation exists, and none is needed¹	The lung is already the most efficient extracorporeal circuit available for this poison. Removing carboxyhaemoglobin by any other route would be an elaborate way of doing something a facemask does better.

Metabolism and the metabolites

There is none. Carbon monoxide is not metabolised, is not activated, and generates no toxic product — it is inhaled, it binds, and it is exhaled unchanged. This section is kept because the section spine is fixed across the library, and its emptiness is informative: every therapeutic strategy that works on the other Band A poisons is unavailable here. There is no enzyme to inhibit as in the toxic alcohols, no cofactor to replace as in paracetamol, no pathway to divert, and no elimination to alkalinise.

What replaces a metabolic pathway is a binding cascade, and it is worth drawing because the three targets fail at different times.

Carbon monoxide — three targets, three timescales

Inhaled carbon monoxideDiffuses rapidly across the alveolar membrane¹
Binding to haemoglobin haem ironCarboxyhaemoglobinAffinity 230–300× that of oxygen¹. This is the only compartment you can measure
Binding to myoglobinCarboxymyoglobinSkeletal and myocardial. Clears more slowly than blood, so tissue injury is delayed¹
Binding to respiratory-chain haem proteinsImpaired oxidative phosphorylationMitochondrial dysfunction¹ — the mechanism shared with cyanide
Reduced delivery AND reduced unloading AND reduced utilisationThe leftward shift of the dissociation curve means the retained oxygen is also less available¹
Neutrophil activation, lymphocyte proliferation, lipid peroxidation¹Continues after the gas has been cleared
Oxidative and inflammatory injury comparable to reperfusion injuryThe substrate of the delayed neurological syndrome¹

Elimination and accumulation

Elimination is entirely pulmonary and entirely governed by the partial pressure of inspired oxygen, which is what makes the half-life figures so striking: 320 minutes on room air, 74 minutes on 100% normobaric oxygen, 20 minutes on hyperbaric oxygen at 3.0 bar.¹ Oxygen is the antidote and the elimination-enhancement therapy — one intervention doing what bicarbonate and dialysis do separately in salicylate poisoning.

Why the carboxyhaemoglobin concentration does not grade the patient

It falls fast, and it falls faster once treatment starts. A level taken after an ambulance journey on high-flow oxygen has already been treated — the 74-minute half-life applies from the moment the mask goes on.¹
It measures one of three compartments, and not the one where the injury is occurring.¹
It does not capture duration. A brief exposure to a high concentration and a night-long exposure to a low one can produce similar levels and very different outcomes — and exposure longer than 24 hours is an independent risk factor for late sequelae (OR 2.0).¹
One reviewed guideline recommends against routine _non-invasive_ carboxyhaemoglobin measurement, alongside ECG and cardiac enzymes as the diagnostic evaluation.¹ Note the limit of that: the same review recommends measuring COHb on a blood gas for verification, and uses a COHb below 3% with resolved symptoms as the endpoint for stopping oxygen.¹ The level is not useless — it is a poor severity grade. Established

Target organs — and why those

Brain

TargetNeurons, and the white matter between them

Why hereThe brain has the highest oxygen consumption per gram of any organ and almost no anaerobic reserve, so a simultaneous failure of oxygen delivery, unloading and mitochondrial utilisation is expressed there first. The classical imaging finding is bilateral globus pallidus injury — the same logic as the putamen in methanol poisoning and zone 3 in paracetamol: a watershed territory with high metabolic demand fails when a diffuse metabolic ceiling drops. Delayed white-matter demyelination is the substrate of the late syndrome. Inferred

At the bedsideHeadache, nausea, dizziness and confusion acutely — a presentation indistinguishable from viral illness, which is why whole households are misdiagnosed together. Cerebellar dysfunction before treatment was strongly associated with cognitive sequelae (OR 5.71, 95% CI 1.69–19.31, p=0.005) in the Weaver trial.²

Heart

TargetMyocardial myoglobin and the cardiac mitochondrial chain

Why hereMyocardium binds carbon monoxide directly through myoglobin¹ as well as suffering reduced delivery, and it has a high extraction ratio at baseline — so it cannot compensate for reduced supply by extracting more. In a patient with coronary disease the two insults are additive. Because tissue and blood elimination differ, myocardial injury can appear after the blood level has fallen.¹ Established

At the bedsideThis is why an ECG and cardiac enzymes appear in the reviewed guidelines as the routine diagnostic evaluation, while routine non-invasive carboxyhaemoglobin measurement does not.¹ Myocardial injury is also an independent predictor of long-term mortality.

Brain again — the delayed syndrome

TargetAn inflammatory and oxidative process rather than the gas

Why hereNeutrophil activation, lymphocyte proliferation, mitochondrial dysfunction and lipid peroxidation, producing oxidative stress, inflammation and apoptosis comparable to reperfusion injury.¹ This is why the second illness is separated from the first by a well interval: it is not continued exposure but a continuing process, and immune-mediated demyelination is the usual proposed mechanism. Inferred

At the bedsideCognitive impairment, personality change, parkinsonism and incontinence appearing days to weeks after apparent recovery. Age over 36 is a risk factor for late neuronal sequelae (OR 2.6, 95% CI 1.3–4.9), as is exposure longer than 24 hours (OR 2.0).¹

The fetus

TargetFetal haemoglobin

Why hereFetal haemoglobin binds carbon monoxide with even greater avidity than adult haemoglobin, and fetal carboxyhaemoglobin lags and then exceeds the maternal level, clearing more slowly. The fetus is therefore both more exposed and slower to recover than the mother whose level is being measured. This is standard teaching and is not sourced to any reference on this page — flagged rather than dropped, because the clinical implication is real and the gap should be visible. Inferred

At the bedsidePregnancy is one of the two situations in which the reviewed guidance considers hyperbaric treatment sensible despite the weak overall evidence — "the decision in favour of HBOT seems sensible in severe CO intoxication or in pregnant women."¹

Timeline of effects

Carbon monoxide — two illnesses with a well interval between them

Time

What you seeWhat is happening

During exposureAcute
What you seeHeadache, nausea, vomiting, dizziness, malaise, breathlessness on exertion. Indistinguishable from influenza or gastroenteritis, and often affecting several people in one household. Then confusion, ataxia, collapse, seizures.
What is happeningCarboxyhaemoglobin rises; the dissociation curve shifts left, so both carriage and release fall.¹ Myoglobin and mitochondrial binding are already occurring in parallel.¹ The pulse oximeter reads normal throughout, and so does the PaO₂.
Minutes to hours after removalApparent recovery
What you seeSymptoms improve rapidly once the patient is in clean air and faster still on high-flow oxygen. Many feel well within an hour or two.
What is happeningCarboxyhaemoglobin falls with a half-life of 320 minutes on room air but only 74 minutes on 100% oxygen.¹ The marker is clearing. Tissue-bound carbon monoxide is clearing more slowly,¹ and the oxidative–inflammatory process is not clearing at all.
Hours to 2 daysCardiac
What you seeChest pain, arrhythmia, troponin rise, decompensated heart failure — sometimes after the carboxyhaemoglobin has normalised.
What is happeningMyocardial myoglobin releases its bound carbon monoxide more slowly than blood does,¹ and the myocardium has no extraction reserve. Pre-existing coronary disease converts a metabolic insult into ischaemia.
2 days – 6 weeksThe well interval
What you seeNothing. The patient has been discharged, feels recovered and has often returned to work. There is no test to perform and no sign to find.
What is happeningThe inflammatory and oxidative cascade — neutrophil activation, lymphocyte proliferation, lipid peroxidation, apoptosis, comparable to reperfusion injury — continues in the absence of any carbon monoxide at all.¹ Demyelination develops. This is the only latent phase in Band A that persists after the poison has left the body. Inferred
3 days – 6 weeks onwardsDelayed neurological sequelae
What you seeCognitive impairment, memory and attention deficits, personality change, parkinsonism, apraxia, incontinence. Frequently attributed to something else, because the exposure is weeks in the past.
What is happeningEstablished demyelinating and inflammatory injury. Risk factors: age over 36 (OR 2.6, 95% CI 1.3–4.9) and exposure longer than 24 hours (OR 2.0, 95% CI 1.0–3.8).¹ Cognitive sequelae occurred in 46.1% of the normobaric-oxygen arm of the Weaver trial at six weeks.² Long-term risks including cardiovascular events and raised mortality are also described.¹

Name the reason for this latent phase

Carbon monoxide — an inflammatory process continuing after the poison itself has gone

The other 26 kinds of latent phase in this library

Amphetamines and MDMA — a hormone acting normally on a kidney behaving normally, while the patient supplies the water
Antipsychotics — a physical object in the stomach — extended-release quetiapine forming a pharmacobezoar
Arsenic — a tissue declaring on its own timetable rather than the poison's — the arsenic is excreted within days, but the nail that was growing while it circulated does not show its white transverse line for several weeks
Arsine and stibine — a red cell mass haemolysing faster than a kidney can cope with — the exposure is over, the haemolysis is silent until the urine changes colour, and the renal failure that follows is the cause of death
Beta-blockers — a repolarisation lesion waiting for an ectopic beat to fall inside it — sotalol prolongs the QT and then, for hours, nothing happens
Calcium-channel blockers — a tablet that has not yet dissolved
Chlorine and ammonia — paraquat's cause at its shortest — the time an injured alveolar epithelium takes to leak
Cocaine — a package that has not yet failed — the calcium-channel blocker's cause moved outside the tablet, where wrapping integrity cannot be measured
Digoxin — distribution of the drug to its target — which that page argues is not a latent phase in the paracetamol sense at all
Ethylene glycol — paracetamol's cause with a different enzyme — glycolate accumulating behind glycolate oxidase while the patient looks merely drunk
GHB and GBL — not the poisoning but its withdrawal — a half-life under an hour means admission for any reason interrupts round-the-clock redosing
Hydrofluoric acid — an ion diffusing far enough to reach a nerve ending — and the thinner the solution, the further it travels before anybody feels it
Iron — a true remission — the corrosive injury settling while absorbed iron moves into the cells it will poison, so the patient improves on their way to the dangerous phase
Ketamine — nitrous oxide's cause on a longer clock — urothelial damage accumulating over months to years of repeated exposure
Lead — bone giving lead back — a store with a half-life of 10 to 30 years releasing its contents when pregnancy, lactation, menopause or osteoporosis resorbs it, so the blood concentration rises after the exposure has ended
Lithium — transport across cell membranes
Mercury — distribution on two clocks — tissue concentrations peaking within 24 hours everywhere except the brain, which is not reached until 2 to 3 days, and which then cannot let the poison out again
Methaemoglobin inducers — paracetamol's cause at its shortest — one to four hours manufacturing the toxic species
Methanol — paracetamol's cause again — formate accumulating behind a folate-dependent disposal step that primates perform poorly
Nitrous oxide — damage accumulating to a threshold
Opioids — an antidote wearing off before the poison does — renarcotisation, the only gap in this set that treatment creates rather than reveals
Organophosphate insecticides — a fat store emptying — and, separately, a second and unexplained lesion declaring itself at a neuromuscular junction the first phase had already left
Paracetamol — time spent manufacturing a toxic metabolite
Paraquat — the body responding to an injury that is already complete
Sodium-channel blockade — a gap that cannot be shortened
Thallium — hair on its own clock rather than the poison's — sensory symptoms come first and the alopecia that makes the diagnosis obvious follows them, well after the interval in which treatment is recommended

The process was started by the poison and does not require it to continue. That is why nothing measured on the day predicts it well, and why follow-up rather than a level is what the mechanism argues for.

What the mechanism predicts at the bedside

Why oxygen is given immediately and before any level is available

Because the therapeutic lever is mass action at a binding site, and it works from the moment the mask is applied — reducing the carboxyhaemoglobin half-life from 320 minutes to 74.¹ Waiting for a number delays the only intervention with unambiguous evidence behind it. The reviewed guidance is unequivocal: "All patients with symptomatic carbon monoxide poisoning should be treated with 100% oxygen as soon as possible."¹ Established

Why a normal saturation and a normal blood gas do not exclude it

Because a pulse oximeter cannot distinguish carboxyhaemoglobin from oxyhaemoglobin, and because PaO₂ measures dissolved oxygen, which is genuinely normal. Both instruments are working correctly and answering a question that is not the one being asked. Only co-oximetry measures carboxyhaemoglobin. Established

Why the diagnosis is made from the story

The acute syndrome is non-specific to the point of invisibility, the routine monitors are falsely reassuring, and the level falls fast during transport. What remains discriminating is the exposure history and the epidemiology: several people in one household unwell together, symptoms that improve away from home and recur on return, a pet affected, a faulty appliance, a generator or barbecue used indoors, a shisha session. The clinical picture cannot make this diagnosis; the circumstances can.

What the hyperbaric evidence actually shows

This is the most contested question in Band A, and it deserves to be laid out rather than resolved. Two well-conducted randomised trials point in opposite directions, and both are in the same Cochrane review.

The trial in favour

Weaver and colleagues randomised patients with symptomatic acute carbon monoxide poisoning to three hyperbaric-oxygen sessions within 24 hours or one normobaric-oxygen treatment plus two sham sessions, double-blind.²

Cognitive sequelae at six weeks: 19/76 (25.0%) with hyperbaric oxygen versus 35/76 (46.1%) with normobaric oxygen, p = 0.007 — adjusted odds ratio 0.45 (95% CI 0.22–0.92), p = 0.03.²
Cognitive sequelae were also less frequent at 12 months on intention-to-treat analysis (p = 0.04).²
The trial was stopped after the third of four scheduled interim analyses.²
Cerebellar dysfunction before treatment predicted cognitive sequelae (OR 5.71, 95% CI 1.69–19.31) and was more frequent in the normobaric arm (15% versus 4%, p = 0.03) — a baseline imbalance in the strongest prognostic variable.²

The evidence against

The Cochrane review identified seven randomised trials and analysed six involving 1,361 participants. Two found benefit at one month; four did not. Pooled random-effects meta-analysis did not suggest a significant benefit: OR 0.78 (95% CI 0.54–1.12) — with significant methodological and statistical heterogeneity, and design or analysis flaws evident in all trials.³
The review notes that one positive trial's conclusions may have been influenced by failure to adjust for multiple hypothesis testing, and that interpretation of the other is hampered by a high risk of bias introduced during the analysis, including an apparent change in the primary outcome.³ That second criticism appears to be aimed at Weaver, and it is the same objection the German review makes independently — that the stated objective was delayed neurological deficit while what was reported was persistent deficit.¹
And Cochrane is equally hard on the negative trials, which this page previously did not say. It found that three negative trials had low power to detect a benefit — two excluded severely poisoned patients and one had very poor follow-up — and that a further trial finished years earlier had never reported its final analysis.³ Its own pooled estimate is qualified as needing to be interpreted cautiously given the heterogeneity, and its conclusion is that the trials do not establish whether hyperbaric oxygen helps — not that it does not.³
Annane and colleagues randomised 385 patients; hyperbaric oxygen at 2 atm conferred no cognitive benefit over normobaric oxygen, and repeated hyperbaric treatment tended towards worse outcomes.¹
A methodological criticism of the Weaver trial: its stated objective was delayed neurological deficit, but what it actually showed was rates of persistent neurological deficit¹ — a different outcome from the one the trial set out to measure.

The reviewed guidance's own summary is the fairest available: "The evidence for the benefit advantage of hyperbaric oxygen is weak in view of the heterogeneity of the available studies. The decision in favour of HBOT seems sensible in severe CO intoxication or in pregnant women."¹ And: "no clear, generally accepted recommendation exists for what should be done."¹

Why follow-up matters more than the acute number

Because the injury that determines quality of life develops after discharge, over weeks, in a patient who has been reassured. Nothing measured on the day predicts it reliably; the risk factors that are known — age over 36, exposure longer than 24 hours,¹ and cerebellar dysfunction at presentation² — are historical and clinical rather than biochemical. A patient told they have recovered and given no follow-up is a patient whose delayed syndrome will be attributed to depression, dementia or stress. Inferred

The antidote, from the poison's side

The antidote is oxygen, and it is the only one in this library that is also the physiological substrate the poison displaced. That gives it an unusual set of properties, all of which follow from mass action at a haem binding site.

It is competitive, so it works on concentration alone. Raising the inspired oxygen partial pressure shifts the equilibrium at the binding site — no receptor, enzyme or transporter is involved. This is why it is effective immediately and why no dose-finding is required. Established
It is simultaneously the elimination-enhancement therapy. The carboxyhaemoglobin half-life falls from 320 minutes to 74 on 100% normobaric oxygen and to 20 at 3.0 bar hyperbaric.¹ No other Band A antidote also removes the poison. Established
It does not address the second illness. Oxygen displaces carbon monoxide; it does not stop the inflammatory and oxidative cascade the exposure has already initiated.¹ The mechanistic hope for hyperbaric oxygen has always been that it does something about this — and the trial evidence does not settle whether it does. Inferred
It is free of the trade-offs that limit every other antidote here. No withdrawal as with naloxone, no seizures as with flumazenil, no anaphylactoid reactions as with acetylcysteine. The argument about carbon monoxide is never whether to give oxygen; it is only ever at what pressure.

Critical appraisal

Cherry-red skin is traditional teaching, and there is published work showing it cannot be relied on. Traditional teaching Findlay studied skin colour by reflectance spectrophotometry in ten fatal cases of carbon monoxide poisoning and set out the circumstances that make the cherry-red colour difficult to identify: low carbon monoxide concentration, skin pigmentation, wash-out of a previously high concentration, and deep venous dilatation with superficial vasoconstriction producing the impression of cyanosis.⁴ If the sign is unreliable in fatal poisoning studied instrumentally, it cannot be a bedside test in survivors. It is taught because it is memorable, and it belongs in the same category as the Wood's lamp in ethylene glycol poisoning.
The hyperbaric question is genuinely unresolved and is often presented as settled in both directions. Inferred A well-conducted double-blind trial found a halving of cognitive sequelae at six weeks;² a Cochrane meta-analysis of six trials and 1,361 participants found a pooled OR of 0.78 with a confidence interval crossing one, significant heterogeneity, and design or analysis flaws in every trial including the positive ones;³ and a 385-patient trial found no benefit with a suggestion of harm from repeated treatment.¹ Anyone quoting only one of these is quoting selectively. The most defensible summary is the reviewed guidance's own: the evidence is weak, and the decision seems sensible in severe poisoning or in pregnancy.¹
The Weaver trial has two specific weaknesses worth knowing, and neither makes it wrong. Inferred It was stopped early at the third of four interim analyses, which tends to overestimate effect size; and cerebellar dysfunction — the strongest prognostic variable in the trial, OR 5.71 — was imbalanced against the control arm at baseline (15% versus 4%, p=0.03).² The adjusted analysis addresses the second, and the confidence interval (0.22–0.92) is not comfortable. A separate critique holds that the trial reported persistent rather than delayed neurological deficit, which was its stated objective.¹
The 230–300-fold affinity figure is a range for a reason and is often quoted as a single number. Established for the range itself.¹ Reported values vary with measurement conditions and with haemoglobin variant, and the precision implied by quoting "240 times" is not in the source.
The mitochondrial contribution is the least demonstrated part of the mechanism. Inferred Mitochondrial dysfunction is listed among carbon monoxide's cellular effects,¹ and direct binding to respiratory-chain haem proteins is chemically plausible and widely accepted. How much of the human illness it accounts for, relative to impaired delivery and unloading, has not been separated — which is the same limitation flagged on the paracetamol page about mitochondrial adducts.
Delayed neurological sequelae have no agreed definition, which makes every incidence figure hard to compare. Inferred Trials have used cognitive test batteries, clinical syndromes and patient-reported outcomes, at intervals from one month to twelve. The 46.1% figure in the Weaver control arm² and the incidences in other trials are not measuring the same thing, which is a large part of the heterogeneity Cochrane identified.³
Carboxyhaemoglobin measurement persists in practice despite guidance against its routine non-invasive use.¹ It is easy to obtain, feels objective, and answers no clinical question that changes management. Its legitimate roles are confirming a suspected diagnosis and, in the same review, marking the endpoint of oxygen therapy — but not grading how ill the patient is, which is what it is most often asked to do.

References

1
Eichhorn L, Thudium M, Jüttner B. The diagnosis and treatment of carbon monoxide poisoning. Dtsch Arztebl Int 2018;115(51–52):863–70. PMC6381775 Open access. Source of the 230–300-fold affinity, the leftward shift of the oxyhaemoglobin dissociation curve, the myoglobin binding and the differing tissue and blood elimination times, the cellular effects (neutrophil activation, lymphocyte proliferation, mitochondrial dysfunction, lipid peroxidation) and the reperfusion-injury comparison, the carboxyhaemoglobin half-lives of 320 / 74 / 20 minutes, the late-sequelae risk factors (age over 36, OR 2.6; exposure over 24 h, OR 2.0), the Annane trial summary, the criticism of the Weaver trial's stated outcome, the guideline recommendations on 100% oxygen and against routine non-invasive COHb measurement, and the quoted conclusions on hyperbaric oxygen and on cyanide co-poisoning. Verified 1 Sep 2026 from the full text.
2
Weaver LK, Hopkins RO, Chan KJ, et al. Hyperbaric oxygen for acute carbon monoxide poisoning. N Engl J Med 2002;347(14):1057–67. PubMed 12362006 Double-blind randomised trial. Source of the 19/76 (25.0%) versus 35/76 (46.1%) cognitive-sequelae comparison at six weeks with p=0.007 and adjusted OR 0.45 (95% CI 0.22–0.92), the 12-month intention-to-treat result, the cerebellar-dysfunction odds ratio of 5.71 (95% CI 1.69–19.31) and its baseline imbalance (15% versus 4%, p=0.03), and the early stopping at the third of four interim analyses. Verified 1 Sep 2026 from the abstract.
3
Buckley NA, Juurlink DN, Isbister G, Bennett MH, Lavonas EJ. Hyperbaric oxygen for carbon monoxide poisoning. Cochrane Database Syst Rev 2011;(4):CD002041. PubMed 21491385 Source of the six-trial / 1,361-participant pooled analysis, the odds ratio of 0.78 (95% CI 0.54 to 1.12), the statements on significant methodological and statistical heterogeneity and on design or analysis flaws in all trials, and the specific criticisms of the two positive trials. Verified 1 Sep 2026 from the abstract.
4
Findlay GH. Carbon monoxide poisoning: optics and histology of skin and blood. Br J Dermatol 1988;119(1):45–51. PubMed 3408663 Reflectance spectrophotometry of skin colour in ten fatal cases. Source of the circumstances that make the cherry-red colour difficult to identify, including deep venous dilatation with superficial vasoconstriction producing the impression of cyanosis. This is the citation for the doubt behind the traditional-teaching badge. Verified 1 Sep 2026.
5
TOXBASE — carbon monoxide. National Poisons Information Service. toxbase.org (NHS login required. NPIS: 0344 892 0111. The authoritative UK source for oxygen therapy, hyperbaric referral criteria, cardiac assessment and follow-up. Login-gated, so not quoted here.)

Carbon monoxide

At a glance