Intended for healthcare professionals


Oxygen therapy for medical patients

BMJ 2018; 363 doi: (Published 24 October 2018) Cite this as: BMJ 2018;363:k4436

Linked practice

Oxygen therapy for acutely ill medical patients: a clinical practice guideline

This article has a correction. Please see:

  1. Daniel Horner, professor1 2,
  2. Ronan O’Driscoll, consultant respiratory physician2
  1. 1Royal College of Emergency Medicine, London, UK
  2. 2Salford Royal NHS Foundation Trust, Salford, UK
  1. Correspondence to: Daniel Horner danielhorner{at}

New lower targets will change practice, but evaluation must continue

Oxygen has long been a friend to the medical profession.1 Even old friendships need reappraisal, however, in light of new information. Oxygen therapy in acute medical illness is a friendship past its due date for review.

In a linked paper, Siemieniuk and colleagues publish a set of rapid recommendations2 highlighting the increasingly well known risks of hyperoxia and proposing some radical new targets for future care, based on recently published evidence.3 They propose a target oxygen saturation range of 90-94% for most patients with acute medical illness and no more than 90-92% for those with acute stroke or myocardial infarction. The authors make a strong recommendation against letting saturation rise above 96% in any patient with a medical problem who is receiving supplemental oxygen.

These target ranges are a step change downwards from the 2008 British Thoracic Society guideline (target range 94-98%) and the 2015 Thoracic Society of Australia and New Zealand guideline (target range 92-96%).45 Although technically simple to deliver, these recommendations will require a shift in culture—away from oxygen as a friend, to treating oxygen as any other clinical intervention, requiring consideration and careful supervision.

Despite the increasing evidence on the harms of hyperoxia over the past decade, serial audits published by the British Thoracic Society demonstrate limited progress with oxygen prescribing and the use of target ranges.6 In order to deliver effective change, we must first understand the rationale and barriers, alongside creating a coalition of urgency.

Oxygen is an inexpensive, painless, and easily applied therapy. It continues to be recommended as a routine part of the primary management for most medically ill patients, endorsed internationally across paediatric, adult, and trauma life support courses by the European Resuscitation Council.7 Oxygen provides a supposedly supraphysiological “buffer” for the patient during work-up and initial management.

But do we love it too much? Previous articles have drawn attention to the common medical maxim that “if some is good, more is better.”8 This is rarely true. Perhaps the use of oxygen as a buffer is more about treating our own anxiety than helping patients.9 We do not routinely tell patients what we are hoping to achieve with supplemental oxygen, or what the potential harms are.1011

Oxygen toxicity seems to develop through several potential pathways, described by J B Downs over 15 years ago in a seminar on the fallacies of oxygen therapy.8 Harms include intrapulmonary shunting from absorption atelectasis and reversal of hypoxic pulmonary vasoconstriction. It has also been proposed that toxicity in hyperoxia occurs through free radical formation and oxidative damage throughout multiple organ systems, compounded by reduced blood flow because of systemic vasoconstriction.

There are other, more indirect harms. The buffer provided by hyperoxia during diagnosis and assessment can provide false reassurance, obscure the primary pathology, and delay the recognition of clinical deterioration.8 Also, many clinicians would be surprised to learn that hypercapnia is four times more common than pure hypoxia in arterial blood gas samples.12 Lastly, the epidemiological evidence is increasingly compelling. Systematic review data and observational cohort studies of oxygen use in cardiac arrest, intensive care, and acute medical illness are reaching the same conclusions; hyperoxia is consistently associated with a higher mortality than normoxia, and often a higher mortality than hypoxia.3131415

How can we effect change? Evidence based recommendations such as those by Siemieniuk and colleagues are a good start and build on previous guidelines. The strength of the evidence, inclusion of patient preferences, and clear practical guidance should hopefully cut the usual lag time in knowledge translation and encourage early adoption. The move towards “aim for normality, or tolerance of near-normality” also ties into a large body of recent critical care work.161718 The message is becoming more and more familiar, and harder to ignore.

As we strive for culture change on the wards, however, we must recognise the limitations of the evidence. The largest meta-analysis is formed principally from studies of oxygen for patients with stroke, myocardial infarction, critical illness, and cardiac arrest; generalisability outside of these groups is not guaranteed.3 Multiple clear indications for oxygen therapy remain, such as those highlighted in box 3 within the linked article, and emergency periprocedural use. We must balance the need for tighter control of oxygen therapy during acute illness with a determination to avoid previous mistakes encountered with other strict variable controls, including tight blood glucose limits.19 It is easy to envisage unintended hypoxia because of a militant but unclear message on the harms of oxygen therapy.

Although the case for an upper saturation limit of 96% is strong, the case for a target range of 90-94% for most patients is weaker. The authors are clear on the need for further assessment of lower saturation limits through clinical study, which we suggest should occur through large scale randomised controlled trials. More evidence is also needed regarding the clinical impact of lower saturation targets on surgical and obstetric patients, with particular regard to wound healing, and on patients with brain injury.

However, this work should not delay change that is already supported by a compelling body of evidence. This will be hard. Sometimes the oldest friendships are the hardest to change. But they are often better for it.


  • This article has been updated to clarify that the recommendations on oxygen saturation levels apply only to patients receiving oxygen.

  • Competing interests: We have read and understood BMJ policy on declaration of interests and declare the following: none

  • Provenance and peer review: Commissioned; not peer reviewed.


Log in

Log in through your institution


* For online subscription