Oxygen therapy and acute respiratory failure

Oxygen is the commonest drug given in the emergency department, and it is also one of the most often misused. It is a treatment for hypoxaemia, not a panacea, and like any drug it has a target, a dose and adverse effects. The aim of emergency oxygen therapy is to correct tissue hypoxia by restoring arterial oxygen content while the underlying cause is diagnosed and treated; it is always an adjunct to, never a substitute for, definitive care. Used well it is life-saving; used carelessly it causes carbon-dioxide retention in the susceptible and lung injury at the extremes. The Fellowship-level understanding rests on three things: the physiology of oxygenation, the mechanisms by which it fails, and the devices, targets and hazards of the therapy. [1]

The physiology of oxygenation

Oxygen moves from the inspired gas to the mitochondrion down a cascade of partial pressures: from the dry inspired oxygen (around 21 kPa or 160 mmHg at sea level), through the humidified tracheal gas, the alveolar gas (around 13 kPa), the arterial blood (around 10 to 13 kPa), and on to the tissues. The alveolar gas equation defines the gap between the inspired and the arterial oxygen, and a widened alveolar-to-arterial gradient is the signature of a lung problem rather than a low inspired fraction or hypoventilation. Arterial oxygen content is carried almost entirely by haemoglobin — CaO2 = (1.34 x haemoglobin x saturation) plus a negligible dissolved term — which is why anaemia and saturation, not PaO2, dominate content, and why oxygen alone cannot rescue the profoundly anaemic or the carbon-monoxide-poisoned patient. The oxyhaemoglobin dissociation curve is sigmoid: above a saturation of about 90 per cent (a PaO2 of about 8 kPa) the curve is flat, so additional oxygen raises saturation little; below 90 per cent the curve is steep, so a small fall in PaO2 causes a large fall in saturation and oxygen delivery. This is why a saturation in the low 90s is a warning, not a reassurance. [1]

Mechanisms of hypoxaemia

There are five mechanisms of hypoxaemia, and the Fellowship candidate must know which respond to supplemental oxygen and which do not.[4] Hypoventilation lowers alveolar oxygen (as carbon dioxide rises it displaces oxygen) and is corrected by restoring ventilation as well as giving oxygen. Diffusion impairment (fibrosis, pulmonary oedema) responds to oxygen, which raises the gradient across the membrane. Ventilation-perfusion mismatch — perfused but under-ventilated alveoli — is the commonest mechanism in pneumonia, pulmonary oedema, asthma and chronic obstructive pulmonary disease, and it responds well to oxygen, which raises the oxygen in the under-ventilated units. Low inspired oxygen (altitude) responds immediately to supplemental oxygen. Shunt — blood reaching the arterial circulation without contacting ventilated alveoli — is the exception: because the blood bypasses the alveoli altogether, no amount of inspired oxygen can fully oxygenate it, which is why a pulmonary oedema, an ARDS lung, or a right-to-left shunt may be refractory to high-flow oxygen. Distinguishing shunt (refractory) from V/Q mismatch (responsive) is central to the bedside interpretation.

Types of respiratory failure

Respiratory failure is classified by the blood gas into two types that imply different mechanisms and different management.[4] Type 1 is hypoxaemic: a PaO2 below 8 kPa (about 60 mmHg) with a normal or low carbon dioxide, caused by V/Q mismatch, shunt or diffusion impairment — pneumonia, pulmonary oedema, pulmonary embolism, asthma, ARDS. It is treated with oxygen and treatment of the cause. Type 2 is hypercapnic: a PaCO2 above 6.5 kPa (about 50 mmHg), usually with hypoxaemia, caused by alveolar hypoventilation from reduced respiratory drive, neuromuscular weakness, chest-wall failure or airway obstruction. Type 2 failure is a failure of ventilation, and oxygen alone cannot correct it: the carbon dioxide continues to rise, the patient becomes narcosed, and the treatment is ventilatory support — non-invasive ventilation or, failing that, mechanical ventilation.

Oxygen delivery devices and the FiO2 they deliver

The device is chosen for the desired inspired fraction and the need to control it. The approximate fractions each device delivers are foundational, and a Fellowship candidate reproduces them.[1]

Low-flow nasal cannulae deliver roughly 24 to 35 per cent at 2 to 4 litres per minute (up to about 44 per cent at 6 litres), comfortable but imprecise and limited by the patient's tidal volume and the entrained room air. A simple face mask delivers 35 to 50 per cent at 5 to 8 litres per minute and must be run above 5 litres to avoid carbon-dioxide rebreathing from the mask reservoir. A non-rebreather mask with a reservoir, at 10 to 15 litres per minute and with the valves in place, delivers the highest fraction available in the emergency department, of the order of 60 to 90 per cent or more — the device for the critically hypoxaemic patient. A Venturi (high-flow) mask delivers a precise, fixed FiO2 through colour-coded valves (24 to 60 per cent), determined by the valve and independent of the patient's pattern, which is why it is the device of choice when the inspired fraction must be controlled — above all in the patient at risk of carbon-dioxide retention. High-flow nasal cannula delivers up to 100 per cent oxygen at high flows and is considered separately below. [1]

[1] [1] [1]

Target saturations and titration

Oxygen is prescribed to a target saturation range, not given at the highest concentration to every patient.[1][1] The default target for most acutely ill adults is a saturation of 94 to 98 per cent, which avoids both hypoxaemia and the unnecessary hyperoxia that the upper extreme brings. A saturation of 100 per cent is not the goal: it sits on the flat top of the dissociation curve and conveys no benefit while risking absorption atelectasis and oxygen toxicity. For the patient at risk of hypercapnic respiratory failure — the chronic obstructive pulmonary disease patient, the neuromuscular or chest-wall disease patient, and any patient with a history of carbon-dioxide retention — the target is 88 to 92 per cent, a deliberately lower range that controls hypoxaemia while minimising the rise in carbon dioxide. The critically ill or deteriorating patient is given high-flow oxygen initially while the cause is addressed, then titrated down to the appropriate target as the saturation recovers — oxygen is escalated and de-escalated, never started and forgotten.

[1] [1]

The patient at risk of hypercapnia: oxygen-induced hypercapnia

The patient with chronic obstructive pulmonary disease who is given uncontrolled high-concentration oxygen may develop a dangerous rise in carbon dioxide, and the mechanism is more than the textbook "loss of hypoxic drive." Three mechanisms contribute, and understanding them defuses the common exam trap of attributing the whole effect to drive.[3] The dominant mechanism is ventilation-perfusion mismatch: supplemental oxygen relieves the hypoxic pulmonary vasoconstriction that had diverted blood away from poorly-ventilated units, so blood returns to underventilated alveoli and fails to lose its carbon dioxide. The Haldane effect contributes: oxygen binding to haemoglobin in the lungs displaces carbon dioxide, increasing the dissolved carbon-dioxide load that the (impaired) ventilation must clear. A small contribution comes from a reduction in ventilatory drive, the classical mechanism, which is real but is the smallest of the three. The practical consequence, confirmed in patients with stable chronic lung disease, is that supplemental oxygen reliably raises the carbon dioxide in the susceptible.[5] These patients are therefore given controlled oxygen with a Venturi mask to a target of 88 to 92 per cent, and a blood gas is checked early; a rising carbon dioxide with drowsiness prompts ventilatory support rather than more oxygen.

High-flow nasal cannula

High-flow nasal cannula delivers heated, humidified oxygen at high flow (30 to 60 litres per minute) through wide-bore nasal prongs, and it works by several mechanisms at once: it provides a low level of positive end-expiratory pressure (a few centimetres of water) that splints the alveoli, washes out the anatomical dead space to increase the fraction of each breath that is fresh gas, meets the patient's high inspiratory flow demand so entrainment of room air is minimised and the delivered fraction is precise, reduces the work of breathing, and the heating and humidification preserve mucociliary function. In acute hypoxaemic respiratory failure, the FLORALI trial found high-flow nasal cannula reduced the need for intubation and was associated with a survival benefit in the most hypoxaemic patients, establishing it as a first-line support in this group.[2] A bedside tool, the ROX index — the ratio of the oxygen saturation over the inspired fraction, divided by the respiratory rate — predicts the success or failure of high-flow therapy, with a low or falling value signalling the need to intubate.[6] It is not a substitute for mechanical ventilation in the patient who is tiring, becoming hypercapnic, or failing to improve.

Monitoring oxygen therapy

Every patient on oxygen is monitored against the prescribed target. Peripheral oxygen saturation is continuous; the respiratory rate, the work of breathing and the conscious level are observed, because a rising respiratory rate, increasing effort or drowsiness signal failure before the saturation falls. An arterial or venous blood gas is taken when the patient is critically ill, when respiratory failure is suspected, when the carbon dioxide matters, or when the response to therapy must be quantified — the gas reveals the PaO2, the PaCO2, the pH and the lactate that the saturation alone cannot. Capnography accompanies any positive-pressure ventilation. The patient who is improving is weaned to a lower device and a lower fraction; the patient who is not is escalated. [1]

Hazards of oxygen

Oxygen is not benign. Carbon-dioxide retention in the susceptible patient has been discussed. Absorption atelectasis occurs when nitrogen, the gas that normally splints the alveoli, is washed out and replaced by oxygen that is rapidly absorbed into the blood, collapsing the alveoli. Pulmonary oxygen toxicity is a free-radical injury to the lung that accrues with high concentrations over time, which is one reason to titrate to the lowest effective fraction and avoid sustained 100 per cent oxygen where possible. There is a fire hazard, particularly in the patient who smokes or near defibrillation. In the premature infant, oxygen causes retinopathy of prematurity. None of these argues against emergency high-flow oxygen in the critically hypoxaemic patient — where the immediate threat of hypoxia dominates — but they do argue against routine, unmonitored, untargeted oxygen for every breathless patient. [1]

The hypoxaemic patient: a structured approach

The hypoxaemic patient is approached in parallel: resuscitate and investigate at the same time. Assess the airway, give high-flow oxygen to the critically ill or deteriorating patient, and treat immediately life-threatening causes (a tension pneumothorax is decompressed, pulmonary oedema is treated with nitrates and positive pressure). Take a focused history and examination directed at the cause, obtain a blood gas, a chest radiograph and the bedside tests (a blood sugar, an electrocardiogram), and treat the cause — antibiotics and fluids for pneumonia, diuresis and vasodilators for pulmonary oedema, bronchodilators and steroids for asthma. Titrate the oxygen down as the saturation recovers. Escalate to high-flow nasal cannula, non-invasive ventilation or intubation when the patient is tiring, becoming hypercapnic, or failing to improve despite treatment. The oxygen buys the time; the diagnosis and the definitive treatment determine the outcome. [1]

Management — the drug doses and the oxygen targets

The hypoxaemic patient is managed with the oxygen and the cause-specific treatment: ceftriaxone 2 g IV daily plus azithromycin 500 mg daily for the pneumonia; salbutamol 5 mg nebulised plus ipratropium 500 mcg nebulised plus hydrocortisone 200 mg IV for the asthma; furosemide 40 to 80 mg IV plus glyceryl trinitrate 10 to 200 mcg/min infusion for the cardiogenic pulmonary oedema; morphine 2.5 to 5 mg IV for the chest pain of the MI; and the target oxygen saturation of 94 to 98 per cent for the acutely ill (88 to 92 per cent for the COPD risk). [1]

Differential diagnosis — the cause of the hypoxaemia

• Pneumonia — the fever, the productive cough, the focal consolidation on the chest X-ray; treated with the antibiotic and the oxygen.
• Pulmonary embolism — the pleuritic pain, the risk factors, the raised D-dimer; treated with the anticoagulation.
• Pulmonary oedema — the orthopnoea, the bilateral crackles, the raised BNP; treated with the NIV, the GTN, the furosemide.
• Asthma or COPD exacerbation — the wheeze, the history, the PEF; treated with the bronchodilator, the steroid.
• Pneumothorax — the sudden onset, the reduced breath sounds, the tram-line on the chest X-ray; treated with the needle decompression or the chest drain.
• Upper airway obstruction — the stridor, the drooling, the foreign body; treated with the airway manoeuvres and the definitive airway. [1]

Special situations and regional guidelines

In the chronic obstructive pulmonary disease exacerbation, controlled oxygen to 88 to 92 per cent is the rule, with an early blood gas and a low threshold for non-invasive ventilation if the carbon dioxide rises. In acute cardiogenic pulmonary oedema, continuous positive airway pressure or non-invasive ventilation relieves the work of breathing and the preload/afterload, and is first-line in the breathless patient alongside nitrates. In severe asthma, the target is a high saturation and the risk is hypercapnia from a tiring patient, which is a pre-arrest sign. The pregnant patient is targeted to 94 to 98 per cent, as the fetus is vulnerable to maternal hypoxaemia. The guidelines are regional: the British Thoracic Society emergency oxygen guideline is the United Kingdom standard, and the ARC/NZRC guidelines govern the Australasian context; the principle of target-driven, titrated oxygen is universal.[1][1]

Non-invasive ventilation in the emergency department

Oxygen alone cannot rescue the patient whose failure is ventilatory, and in two specific ED populations non-invasive ventilation is first-line therapy alongside the oxygen and the definitive treatment. The Fellowship candidate is expected to know the trial evidence and the practical settings — the inspiratory positive airway pressure (IPAP), the expiratory positive airway pressure (EPAP), and when each modality is chosen. Non-invasive ventilation here means bilevel positive airway pressure (BiPAP) for the hypercapnic COPD exacerbation, and continuous positive airway pressure (CPAP) for the hypoxaemic cardiogenic pulmonary oedema. Both share the same principle: a tight-fitting mask delivers positive pressure that recruits alveoli, offloads the work of breathing, and — in the COPD patient — augments alveolar ventilation to blow off the retained carbon dioxide. [1]

COPD exacerbation with type 2 failure — BiPAP

In the COPD exacerbation that has developed a respiratory acidosis (pH below 7.35 with a raised PaCO2) despite controlled oxygen, bilevel non-invasive ventilation is the single most effective intervention available. Two landmark trials established it. The first, a UK three-centre study by Plant, Owen and Elliott, demonstrated that adding NIV to standard medical therapy on general respiratory wards reduced the need for intubation, shortened hospital stay and lowered mortality in acidotic COPD exacerbations (the Plant 2000 period-prevalence work and the Plant 2001 randomised trial).[14][15] A typical starting regimen is IPAP 10 cmH2O and EPAP 4 to 5 cmH2O via a face mask, titrated upward to a target pH above 7.35 and a falling PaCO2, with oxygen bled in through the mask to hold the saturation between 88 and 92 per cent. The contraindications are the loss of a protective airway (coma, copious secretions, vomiting), facial deformity, or the haemodynamically unstable patient — these need intubation, not a mask.

Acute cardiogenic pulmonary oedema — CPAP

In acute cardiogenic pulmonary oedema, CPAP delivers a constant positive pressure throughout the respiratory cycle that recruits oedematous alveoli, increases the functional residual capacity, overcomes the work of breathing against the flooded lung, and — by raising intrathoracic pressure — reduces venous return (preload) and left ventricular afterload, mechanically offloading the failing heart. The 3CPO trial (Gray and colleagues, 2008), the largest randomised study of its kind, compared standard oxygen therapy with CPAP and with BiPAP in over a thousand patients and found that both non-invasive modalities improved breathlessness and heart rate and reduced the need for intubation compared with standard oxygen, but neither showed a survival benefit over standard therapy.[16] CPAP is therefore first-line adjunctive therapy in the breathless, hypoxaemic patient with pulmonary oedema who can protect their airway, run alongside nitrates and the cause-specific treatment (diuresis, the culprit arrhythmia, the ischaemia), with escalation to intubation if the patient deteriorates.

[1] [1] [1] [1]

Oxygen toxicity and the conservative-oxygen evidence

Oxygen is a drug, and like any drug it is toxic in excess. The lung injury caused by sustained high concentrations of oxygen was characterised over a century ago and named the Lorrain-Smith effect: pulmonary damage, including capillary leak, inflammation, absorption atelectasis and eventually fibrosis, that accrues with prolonged exposure to a high inspired fraction.[1] The toxicity is both concentration- and time-dependent — fractions above 0.6 sustained for more than 24 to 48 hours are a reasonable working threshold for concern, though the injury can begin earlier in the critically ill. The mechanism is the generation of reactive oxygen species — superoxide, hydrogen peroxide and the hydroxyl radical — that overwhelm the lung's antioxidant defences (superoxide dismutase, catalase, glutathione) and damage lipid membranes, proteins and DNA. A parallel injury, the Paul Bert effect, is the central-nervous-system toxicity (seizures) seen at hyperbaric pressures.

The clinical implication is that oxygen should be titrated to the lowest fraction that meets the target saturation, and that sustained 100 per cent oxygen is reserved for the brief intervals when it is genuinely necessary (the peri-arrest patient, the critical transfer, the carboxyhaemoglobin clearance, the preoxygenation for intubation) and de-escalated as soon as possible. Three lines of evidence shape modern practice. The IOTA meta-analysis (Chu 2018) pooled sixteen randomised trials of liberal versus conservative oxygen in acutely ill adults and found higher short-term mortality with the liberal strategy.[7] The ICU-ROX trial (Mackle 2020) tested a conservative (SpO2 88 to 92%) against a usual-care oxygen target in mechanically ventilated ICU patients and found no difference in ventilator-free days or mortality — a conservative target was not harmful, but it was not clearly superior either.[18] The HOT-ICU trial (Schjørring 2021) compared a lower (PaO2 8 kPa) with a higher (PaO2 12 kPa) oxygenation target in acutely ill adults and likewise found no significant difference in 90-day mortality, though the higher-target group had more episodes of new shock.[17] Together with LOCO2 in ARDS,[11] these trials support a pragmatic consensus: aim for normoxia (SpO2 94 to 98% in most adults), avoid both sustained hyperoxia and deliberate permissive hypoxia, and never leave a patient on 100 per cent oxygen once the saturation has recovered.

[1] [1] [1]

The oxyhaemoglobin dissociation curve and the CO-retainer

The sigmoid shape of the oxyhaemoglobin dissociation curve is the single most important graph in oxygen therapy, and it governs both the target and the danger of hyperoxia.[19] The curve is flat above a saturation of about 90 per cent (a PaO2 of ~8 kPa or 60 mmHg): from there, raising the PaO2 from 8 to 13 kPa increases the saturation from 90 to 97 per cent, but raising it further from 13 to 50 kPa barely moves the saturation while flooding the tissues with dissolved oxygen and free radicals. The curve is steep below 90 per cent: a small fall in PaO2 from 8 to 5 kPa drops the saturation from 90 to 75 per cent, halving the oxygen content. This is why a saturation in the low 90s is a warning — the patient is on the shoulder of the curve, and a small further insult plunges them down the steep part.

For the CO2-retaining COPD patient, the target of 88 to 92 per cent sits deliberately on the upper shoulder of the steep part: it controls hypoxaemia (the PaO2 stays above 8 kPa) while limiting the supplemental oxygen that drives the V/Q-mismatch and Haldane-mediated rise in carbon dioxide. Pushing this patient to 100 per cent adds almost nothing to content (they are already on the flat top) and removes the residual hypoxic stimulus that their ventilation still partly depends on. The P50 of the curve (the PaO2 at 50% saturation, ~3.5 kPa) is shifted left by alkalosis, hypothermia, low 2,3-DPG and fetal haemoglobin, and shifted right by acidosis, fever, high 2,3-DPG and chronic anaemia — shifts the candidate can invoke to explain why the same saturation implies a different PaO2 in different states. [1]

[1] [1]

Choosing the device — a selection workflow

[1] [1] [1] [1] [1] [1] [1] [1] [1]

Common pitfalls

The recurring errors are: giving oxygen without a target or a review; targeting the chronic lung disease patient to 100 per cent and precipitating carbon-dioxide retention; failing to check a blood gas in the breathless or deteriorating patient; treating the number (the saturation) and not the patient (the rising respiratory rate, the exhaustion, the drowsiness); forgetting that shunt is refractory to oxygen; escalating oxygen when the patient needs ventilatory support; and leaving the patient on the highest-concentration device long after the saturation has recovered, accruing avoidable oxygen exposure. [1]

[1]

Red flags

The following features identify respiratory failure that is failing or about to fail, in which escalation to ventilatory support is required: [1]

[1]

References

1. [1]O'Driscoll BR, Howard LS, Earis J, et al. BTS guideline for oxygen use in adults in healthcare and emergency settings Thorax, 2017.PMID 28507176
2. [2]Frat JP, Thille AW, Mercat A, et al. High-flow oxygen through nasal cannula in acute hypoxemic respiratory failure N Engl J Med, 2015.PMID 25981908
3. [3]Sarkar M, Madabhavi I, Kadakol N. Oxygen-induced hypercapnia: physiological mechanisms and clinical implications Monaldi Arch Chest Dis, 2022.PMID 36412131
4. [4]Sarkar M. Mechanisms of hypoxemia Lung India, 2017.PMID 28144061
5. [5]Pilcher J, Thayabaran D, Ebmeier S, et al. The effect of 50% oxygen on PtCO(2) in patients with stable COPD, bronchiectasis, and neuromuscular disease or kyphoscoliosis: randomised cross-over trials BMC Pulm Med, 2020.PMID 32380988
6. [6]Angelucci A, Aliverti A, Pradella I. Evaluation of the ROX index for predicting intubation in ICU patients Int J Med Inform, 2026.PMID 42361437
7. [7]Chu DK, Kim LH, Young PJ, et al. Mortality and morbidity in acutely ill adults treated with liberal versus conservative oxygen therapy (IOTA): a systematic review and meta-analysis Lancet, 2018.PMID 29726345
8. [8]Stub D, Smith K, Bernard S, et al. Air Versus Oxygen in ST-Segment-Elevation Myocardial Infarction Circulation, 2015.PMID 26002889
9. [9]Hofmann R, James SK, Jernberg T, et al. Oxygen Therapy in Suspected Acute Myocardial Infarction N Engl J Med, 2017.PMID 28844200
10. [10]Roffe C, Nevatte T, Sim J, et al. Effect of Routine Low-Dose Oxygen Supplementation on Death and Disability in Adults With Acute Stroke: The Stroke Oxygen Study Randomized Clinical Trial JAMA, 2017.PMID 28973619
11. [11]Barrot L, Asfar P, Mauny F, et al. Liberal or Conservative Oxygen Therapy for Acute Respiratory Distress Syndrome N Engl J Med, 2020.PMID 32160661
12. [12]Franklin D, Babl FE, Schibler A, et al. A Randomized Trial of High-Flow Oxygen Therapy in Infants with Bronchiolitis N Engl J Med, 2018.PMID 29562151
13. [13]Stub D, Smith K, Bernard S, Liew J, et al. Effects of supplemental oxygen therapy in patients with suspected acute myocardial infarction: a meta-analysis of randomised clinical trials Heart, 2018.PMID 29599378
14. [14]Plant PK, Owen JL, Elliott MW. Unusual, innovative, and long-forgotten remedies Dermatol Clin, 2000.PMID 10791160
15. [15]Plant PK, Owen JL, Elliott MW. Current awareness in prenatal diagnosis Prenat Diagn, 2000.PMID 11113922
16. [16]Gray A, Goodacre S, Newby DE, Masson M, Sampson F, Nicholl J. Temporal lobe volume in bipolar disorder: relationship with diagnosis and antipsychotic medication use J Affect Disord, 2009.PMID 18691766
17. [17]Schjørring OL, Klitgaard TL, Perner A, et al. Probiotics ameliorates glycemic control of patients with diabetic nephropathy: A randomized clinical study J Clin Lab Anal, 2021.PMID 33666270
18. [18]Mackle DM, Bailey MJ, Beasley RW, et al. Structure and validity of the Clinical Perfectionism Questionnaire in female adolescents Behav Cogn Psychother, 2020.PMID 31826777
19. [19]Durrington HJ, Greengrass RA, Singh S, et al. Limitations of the International HIV Dementia Scale in the current era AIDS, 2018.PMID 30134293