Liberal or Conservative Oxygen Therapy
for Acute Respiratory Distress Syndrome

Barrot. N Engl J Med 2020; 382:999-1008. doi: 10.1056/NEJMoa1916431

Clinical Question

  • In patients with Acute Respiratory Distress Syndrome (ARDS), does conservative vs. liberal oxygen reduce death at 28 days?


  • The optimal amount of oxygen intensive care patients should receive is unknown. Hypoxaemia is dangerous but, conversely, hyperoxia can cause cellular damage via oxidative stress, absorption atelectasis, acute lung injury, central nervous system toxicity, reduced cardiac output, and cerebral and coronary vasoconstriction.
  • A number of trials have demonstrated the absence of benefit for normoxic patients receiving oxygen. AVOID demonstrated no benefit in giving oxygen to normoxic patients with STEMI undergoing PCI. DETO2X showed no difference in one year mortality between room air and oxygen therapy in normoxic patients with suspected acute myocardial infarction. The ICU-ROX study showed no difference in ventilator-free days for patients receiving conservative (SaO2 91-96%) oxygen therapy
  • A number of trials have also demonstrated potential harms associated with hyperoxia. OXYGEN-ICU previously found a striking mortality benefit in those critically ill patients given a conservative oxygen regimen (SpO2 94-98% or PaO2 70-100 mmHg), although it had several problems with methodology. The IOTA meta-analysis suggested that in a heterogeneous group of ICU patients hyperoxia increased mortality, although there was very heavy weighting in this meta-analysis on the OXYGEN-ICU trial
  • Long-standing and widely-adopted clinical guidelines (such as ARDSNet) have previously recommended accepting a lower oxygenation goal (PaO2 55-80 mmHg) in patients with ARDS


  • Multicentre, randomised control trial
  • Computer randomisation performed in blocks of four
  • Randomisation stratified by age, centre, & severity of respiratory failure
  • Non-blinded
  • Based on a power of 90%, a sample size of 850 patients would provide the ability to detect an absolute between groups difference of 9% mortality, assuming a death rate of 30% at day 28, and a significance level of 0.05
  • Intention-to-treat analysis
  • At pre-planned interim analysis, after 205 patients had been enrolled, the Independent Data & Safety Monitoring Board decided to stop the trial because of the potential increased risk of serious adverse events and futility


  • 13 ICUs in France
  • Patients enrolled from June 2016 to September 2018


  • Inclusion:
    • Adults
    • Intubated and receiving mechanical ventilation for <12 hours for ARDS (defined according to the Berlin definition), with a
    • PaO2:FiO2 ratio of 300 mmHg or less, at a positive end-expiratory pressure of 5cm of water or more, less than seven days after a known clinical insult, with bilateral opacities on chest imaging not fully explained by heart failure or fluid overload
  • Exclusion: The use of long-term oxygen therapy; Non-invasive ventilation at home; Cardiac arrest; Traumatic brain injury; Intracranial hypertension
  • 205 patients were enrolled. Four were subsequently excluded (two withdrew consent, one was legally incapacitated and unable to consent, and one was diagnosed with cardiogenic acute pulmonary oedema), leaving 201 for the analysis
  • Comparing baseline characteristics of conservative vs. liberal oxygen group
    • Age: 63 vs. 64
    • Male sex: 66% vs. 63%
    • Tidal volume (ml/kg predicted body weight): 6.0 vs. 6.2, p=0.02
    • Plateau pressure (cm of water): 19.8 vs. 20.8
    • PaO2: FiO2 – mmHg: 117 vs. 120
    • FiO2: 0.8 vs. 0.8
    • Use of catecholamines: 71% vs. 72%
    • SOFA score: 9.3 vs. 8.9
    • Main cause of ARDS
      • Pulmonary: 78.8% vs. 72.5%
      • Extra-pulmonary: 21.2% vs. 27.5%


  • Conservative oxygen therapy (n=99)
    • PaO2 target between 55-70 mmHg (7.3 -9.3 kPa)
    • Between ABGs, SpO2 was maintained between 88-92%
    • In 58 patients, a median of 2 ABGs, had a PaO2 <55mmHg
    • In 97 patients, a median of 6 ABGs, had a PaO2 >70mmHg


  • Liberal oxygen therapy (n=102)
    • PaO2 target between 90-105 mmHg (12-14 kPa)
    • Between ABGs, SpO2 was maintained at >=96%
    • In 98 patients, a median of 8 ABGs, had a PaO2 <90mmHg
    • In 98 patients, a median of 7 ABGs, had a PaO2 >105mmHg

Management common to both groups

  • Arterial blood bases measured 6 hourly; between ABGs, SpO2 used to determine oxygen therapy
  • If the PaO2 was not within the predefined range according to ABG analysis, the FiO2 was modified by 0.05 if the difference from the assigned target was <5mmHg, and modified by 0.10 if the difference was greater
  • If the SpO2 was not within the predefined range, the FiO2 was modified by 0.05 every 5 minutes until the assigned target was reached
  • Patients were ventilated using the volume assist-control mode, with a tidal volume of 6 ml/Kg predicted body weight
  • PEEP was adjusted according to the PaO2:FiO2
  • Neuromuscular blocking drugs were recommended for 48 hours in patients with a PaO2:FiO2 <150 mmHg. Prone positioning was recommended in patients with a PaO2:FiO2 <150 mmHg
  • In the case of interventions oxygenation was managed at the discretion of the responsible clinician. The study investigators recommended following the protocol, however, and returned to it as soon as possible


  • Primary outcome: Death from any cause at 28 days – No significant difference
    • Conservative group 34.3% vs. liberal group 26.5%
    • Difference 7.8% (95% C.I. -4.8 to 20.6%)
  • Secondary outcomes:
    • Time-adjusted difference in PaO2 between conservative vs. liberal group
      • -28.1 (95% C.I. -38.4 to -17.7)
    • Significantly greater in conservative group:
      • Day 90 mortality
        • 44.4% vs. 30.4% (difference 14%, 95% C.I. 0.7-27.2)
      • Mesenteric ischaemic events
        • 5% vs. 0%
      • Heart rate
    • Comparing conservative vs. liberal group, no significant difference in:
      • ICU mortality
        • 36.4% vs. 26.5%
      • Cardiac adverse events
        • Arrhythmia: 23.2% vs. 15.7%
        • New Onset AF: 21.2% vs. 12.7%
      • Patients receiving mechanical ventilation
        • Day 28: 16.1% vs. 15.1%
        • Day 90: 3.8% vs. 2.9%
      • Ventilator acquired pneumonia
        • 17.2% vs. 21.6%
      • Septicaemia
        • 11.1% vs. 18.6%
      • Neurologic adverse events
        • Seizure
          • 2% vs. 0%
        • Stroke
          • 4% vs. 1%
        • Delirium
          • 11.1% vs. 10.8%

Authors’ Conclusions

  • A conservative oxygenation strategy during the first seven days of mechanical ventilation did not reduce mortality at 28 days
  • These results suggest a clinically relevant excess of mortality in the conservative oxygen group, and an increased incidence of mesenteric ischaemia. The trial was stopped early because of this risk


  • Given oxygen is ubiquitously administered in ICU patients, and there is little known of the optimal dose, this trial asks an important question
  • The PaO2/SpO2 targets chosen for scrutiny are consistent with long-standing guidelines and widespread clinical practice and likely reflect a generally-accepted ‘lower limit’ of arterial oxygenation
  • Despite the trial stopping early, it was able to demonstrate a clinically important excess of mortality in patients receiving a conservative oxygen strategy, and an increase in mesenteric ischaemic events
  • Randomised controlled trial
  • Multi-centre


  • Clinicians were unblinded. This may bias the attending physician to make decisions regarding management which impacts the outcome
  • The trial was stopped early, with only 201 patients eligible for analysis of the originally calculated 850. Consequentially, it may be underpowered, and unable to demonstrate statistical significance for other clinically important differences between the intervention and control groups
  • The unplanned termination of a study is also known to increase the likelihood of effect overestimation
  • The use of a low range of PaO2 (55-70 mmHg) may have exposed patients to unsafe PaO2 It is possible that SpO2 values were not precise enough to avoid hypoxic events between blood gas samples
  • The conservative oxygen group received fewer interventions aimed at increasing their PaO2. In particular, there was a lower incidence of prone positioning in the conservative oxygen group, which has previously been shown in the PROSEVA trial to lead to survival benefit in ARDS. Retrospective analysis by the authors suggested that even when prone positioning was controlled for, the conservative oxygen group continued to demonstrate an excess of mortality
  • Because this study focused on ARDS specifically, it is possible that fewer patients were enrolled with the conditions previously found to benefit from a conservative oxygenation approach (e.g., acute hypoxic encephalopathy, as noted in ICU-ROX)

The Bottom Line

  • In patients with ARDS conservative vs. liberal oxygen therapy did not reduce 28 day mortality
  • There were concerning secondary outcome results with an increased 90 day mortality and rates of mesenteric ischaemia with the use of conservative oxygen therapy. There were also non-statistically significant increased rates of 28 day and ICU mortality, as well as cardiac and neurological adverse events with the use of conservative oxygen therapy
  • Continuous pulse oximetry may not be precise enough to avoid unrecognised hypoxic events in those patients who are most at risk
  • For those patients with ARDS, we will continue to avoid hyperoxia, and now adopt a lower SpO2 limit of 90% (particularly where this can be achieved with acceptable oxygen fractions) or more frequent ABG monitoring. Given the totality of the evidence currently available, targeting a PaO2 of 70-100mmHg appears reasonable

External Links


Summary author: Steven Alderson
Summary date: 27/08/2020
Peer-review editor: Luke Collett

Post-peer review editor: Dave Slessor

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.