Lower or Higher Oxygen Targets for Acute Hypoxaemic Respiratory Failure

O.L. Schjørring, N Engl J Med 2021 DOI: 10.1056/NEJMoa2032510

Clinical Question

  • In patients with acute hypoxaemic respiratory failure, does a lower vs higher oxygenation target reduce mortality at 90 days?


  • Patients admitted to the ICU with hypoxaemic respiratory failure often receive supplemental oxygen with a high fraction of inspired oxygen
  • Hyperoxia is potentially harmful, with sequelae such as resorption atelectasis, lung injury and oxidative stress. It remains unclear what the optimal oxygenation target should be that balances the risks of hypoxia and hyperoxia
  • Clinical practice guidelines give no clear recommendation for an optimal oxygenation target in hypoxaemic patients in ICU, aside from the setting of ARDS where a conservative range (PaO2 55 – 80mmHg) has been adopted based on historical data from the ARDSnet study
  • There have been a number of trials performed recently to address this issue with conflicting results. While the OXYGEN-ICU study demonstrated that lower oxygenation targets (SpO2 94-98%/PaO2 70 – 100mmHg) yield a mortality benefit, the study had a number of methodological issues. ICU-ROX did not demonstrate any difference in ventilator-free days between conservative and usual oxygen therapy. LOCO2 demonstrated no difference in 28-day mortality between liberal or conservative oxygenation targets in patients with ARDS. However, it did signal a potential for harm with conservative targets from its analysis of secondary outcomes


  • Randomized controlled trial
  • International, multi-centre
  • Computer-generated allocation concealment with permuted block-randomization
  • Single-blinded: outcome-assessors were blinded, clinicians were not
  • Based on a power of 90%, a sample size of 2928 patients would provide the ability to detect an absolute between group mortality difference of 5% at 90 days. This corresponds to a 20% mortality difference, assuming a death rate of 25% at day 90, with a two-sided alpha level of 0.05


  • 35 ICUs in Denmark, Switzerland, Finland, the Netherlands, Norway, United Kingdom and Iceland
  • Patients enrolled from June 20, 2017 to August 3, 2020


  • Inclusion:
    • Adults acutely admitted to the ICU with hypoxic respiratory failure
    • Receiving supplemental O2, with a flow of at least 10L via an open system OR at least an FiO2 of 0.5 via a closed-system (including NIV and CPAP)
    • Expected to receive supplemental O2 for the next 24 hours in the ICU
    • Arterial line placed
  • Exclusion:
    • Not randomized within 12 hours of ICU admission
    • Chronic mechanical ventilation for any reason
    • Home O2 therapy
    • Previous bleomycin treatment
    • Organ transplant within current hospital admission
    • Withdrawal of active therapies or death deemed imminent
    • Pregnancy
    • CO poisoning, CNpoisoning, methaemoglobinaemia, paraquat poisoning
    • Need for hyperbaric O2
    • Sickle cell disease
  • 2928 patients enrolled; 2888 (98.6%) included in primary analysis
  • Comparing baseline characteristics of intervention vs. control group
    • Age: 70 vs 70
    • Male: 63.7% vs 64.9%
    • Pneumonia the main acute illness: 57.7% vs 57.4%
    • ARDS: 12.3% vs 13.4%
    • Invasive ventilation: 57.4% vs 59.7%
    • NIV and CPAP: 13.7% vs 12.1%
    • Median tidal volume and PEEP: 499ml vs 499ml AND 8cmH2O vs 7cmH2O
    • Median P/F ratio: 118.6 vs 117.5
    • Median SOFA score: 9 vs 9
    • Use of vasoactive agents: 55.1% vs 54.3%


  • Lower oxygen target
    • Oxygen therapy targeting a PaO2 of 60mmHg (8kPa)


  • Higher oxygen targey
    • Oxygen therapy targeting a PaO2 of 90mmHg (12kPa)

Management common to both groups

  • Oxygen supplementation devices and ventilator settings chosen by the clinicians
  • All other ICU interventions determined by clinicians
  • Schedule of ABGs not mandated, but assumed 4 measurements per day
  • SpO2 correlated to PaO2 after each ABG
  • Episodic weaning of FiO2, based on the results of ABG analysis. FiO2 also weaned based on continuous pulse oximetry, with ABG correlation


  • Primary outcome: 90-day allcause mortality – no significant difference
    • 42.9% vs 42.4% (Risk ratio 1.02, 95% C.I. 0.94-1.11, p = 0.64)
  • Secondary outcomes – no significant differences
    • Median percentage days alive without life support: 87.8 vs 84.4 (p=0.1)
    • Median percentage days alive after hospital discharge: 55.6 vs 50.0 (p=0.67)
    • Serious Adverse Events (SAE)
      • Number of patients with ≥1 SAE: 36.1% vs 38.1% (p=0.24)
      • Shock: 33.9% vs 35.8%
      • Myocardial ischaemia: 1.0% vs 1.6%
      • Ischaemic stroke 1.3% vs 1.6%
      • Intestinal ischaemia: 2.2% vs 2.0%

Authors’ Conclusions

  • Targeting a lower oxygenation level of 60mmHg did not result in lower 90-day mortality in comparison to targeting Higher oxygenation levels of 90mmHg
  • Targeting a conservative oxygenation target does not result in increased adverse events, in contrast to the results of the LOCO2 study


  • High quality trial: international, multi-centre, outcome-assessor blinded, pragmatic trial design with a high follow-up rate
  • Important clinical question addressed in the largest trial to date of liberal vs conservative oxygen regimens in the ICU
  • Oxygenation targets for intervention and control groups fall within the commonly accepted range for ICU patients
  • Clear separation of FiO2 administered and PaO2 achieved between groups, (although this was not maintained towards the end of the trial period)
  • The lower-oxygenation therapy did not suffer hypoxia from targeting a lower PaO2


  •  Clinicians and investigators were not blinded which may introduce bias into their usual practice
  • 1264 patients were screened but not recruited, including 712 because consent could not be arranged. This cohort may have influenced the results
  • Median age of the population was high, a large proportion had medical illnesses and 20% had COPD. A high proportion of patients had ARDS, high P:F ratios and the overall mortality in this population was higher than expected. The results of this study may not apply to low-risk populations with hypoxaemic respiratory failure
  • Although a pa02 of 60mmHg was targeted, the intervention group had a median pa02 of 70.8 (IQR 66.6-76.5), compared to a median of 93.3 (IQR 87.1-98.7) in the control arm
  • Between-group differences in values for PaO2, FiO2 and SpO2 diminished over the observation period, as patients often left the ICU between day 14 and 21
  • This study has not yet answered the question of the effect of different oxygenation targets in key subgroups of patients, including traumatic brain injury and hypoxic-ischaemic encephalopathy (sub-group analysis pending)

The Bottom Line

  • While this trial does not change my practice, it is reassuring to see that targeting a PaO2 of 60mmHg (and delivering a Pa02 of ~70mmHg) for patients with hypoxaemic respiratory failure does not result in harm
  • I will continue to use the minimal FiO2 possible to maintain my patients’ SpO2>90%, whilst awaiting further trials such as MEGA-ROX
  • This trial adds to the current evidence base, which suggests that oxygenation targets aiming between PaO2 of 60 – 90mmHg may balance the risks of hypoxia and hyperoxia

External Links


Summary author: Aniket Nadkarni @aniketnadkarni
Summary date: 24/02/21
Peer-review editor: Luke Collett @luke_col

Image by: Bessi from Pixabay


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.