Normal Oxygenation Versus Hyperoxia in the Intensive Care Unit (ICU) (OXYGEN-ICU)

 Giradis. JAMA 2016; published on-line October 2016. doi:10.1001/jama.2016.11993

Clinical Question

  • In critically ill adults, does conservative oxygen therapy compared to liberal oxygen therapy reduce mortality?


  • Randomised controlled trial
  • Computerised random number generator
  • Allocation concealment maintained with use of sequentially numbered, closed, opaque envelopes
  • Non-blinded
  • Sample size calculation: 660 patients required to detect and absolute difference in mortality of 6% from a baseline of 23%, with a false negative rate of 20% and a false positive rate of 5%
    • Study stopped after unplanned interim analysis following earthquake that led to reduction in study hospital beds and low recruitment
  • Modified intention-to-treat analysis
    • Patients excluded if withdrew consent (n=2), lack of data during ICU stay/did not receive at least 1 arterial blood gas analysis per day (n=9), ICU stay <72 hours (n=35)
  • Intention-to-treat analysis also performed on all randomised patients excluding those that withdrew consent


  • Single medical-surgical ICU of University Hospital, Italy
  • Data collected March 2010 – October 2012


  • Inclusion criteria:
    • Aged ≥18
    • Admitted to ICU with expected length of stay of ≥72 hours
  • Exclusion criteria:
    • Pregnancy
    • ICU readmission
    • Decision to withhold life sustaining treatment
    • Immunosupression or neutropenia
    • ARDS with P/F ratio <150
    • Acute decompensation of COPD
    • Enrollment in another study
  • 480 patients randomised, of whom 434 analysed in modified intention-to-treat population


  • Conservative oxygen therapy: Target SpO2 94%-98% (n=236)
    • Lowest possible FiO2 to maintain PaO2 of 70-100mmHg
    • During intubation, airway suction, and hospital transfer patients only received supplemental oxygen if SpO2 <94%


  • Standard oxygen therapy: Target SpO2 97%-100% (n=244)
    • FiO2 of at least 0.4, allowing PaO2 of up to 150mmHg
    • Patients received FiO2 of 1.0 during intubation, airway suction, hospital transfer

Comparing baseline characteristics in conservative vs. conventional group:

  • Age (median): 63 vs. 65
  • Surgical admission: 64% vs. 61%
  • Respiratory failure: 56% vs. 59.%
  • Mechanical ventilation: 66% vs. 68%
  • Shock: 31% vs. 33%
  • Liver failure: 19% vs. 21%
  • Renal failure: 15% vs. 16%
  • Infection: 38% vs. 40%
  • Simplified Acute Physiology Score II (median): 37 vs. 39

Comparing oxygen control in conservative vs. control group (for modified intention to treat population):

  •  Median FiO2 was significantly lower (p<0.001)
    •  0.36 (IQR 0.30-0.40) vs. 0.39 (IQR 0.35-0.42)
  • Median PaO2 was significantly lower (p<0.001)
    • 87mmHg (IQR 79-97) vs. 102mmHg (IQR 88-116)


  • Primary outcome: ICU mortality – significantly lower in conservative oxygen group
    • 11.6% vs. 20.2% (Absolute risk reduction [ARR] 8.6%, 95% C.I. 1.7%-15%, p=0.01)
    • Number needed to treat 12
    • Fragility index 3 patients
  • Secondary outcomes: comparing conservative vs. conventional group
    • New organ failure during ICU stay
      • Any new organ failure: no significant difference
        • 19% vs. 25.7%, p=0.09
      • Respiratory failure: no significant difference
        • 6.5% vs. 6.4%
      • Shock: significantly lower in conservative group
        • 3.7% vs. 10.6%, p=0.006
        • Fragility index 4 patients
      • Liver failure: significantly lower in conservative group
        • 1.9% vs. 6.4%, p=0.02
        • Fragility index 1 patient
      • Renal failure: no significant difference
        • 12% vs. 9.6%, p=0.42
    • New infections during ICU stay
      • All new infections: no significant difference
        • 18.1% vs. 22.9%
      • Bacteraemia: significantly lower in conservative group
        • 5.1% vs. 10.1%, p=0.049
        • Fragility index 0 patients
  • Post-hoc analysis
    • Hospital mortality – significantly lower in conservative oxygen group
      • 24.2% vs. 33.9% (ARR 9.9%, 95% C.I. 1.3%-18.2%, p=0.03)
    • Mechanical ventilation free hours (median) – significantly higher in conservative oxygen group
      • 72 vs. 48, p=0.02
  • Analysis of intention-to-treat population yielded results similar to those of modified intention-to-treat population with regard to primary and secondary outcomes

Authors’ Conclusions

  •  For critically ill patients with an ICU length of stay of >72 hours, a conservative vs. a conventional protocol for oxygen therapy resulted in a lower ICU mortality


  • Randomised control trial
  • Reporting of intention-to-treat analysis as well as modified intention-to-treat analysis
  • Clear difference achieved between median PaO2 for conservative and conventional groups


  • Single centre
  • Non-blinded
  • Unplanned early termination of study – this is known to increase the likelihood of effect overestimation
  • Baseline imbalances in study population, favouring the conservative oxygen therapy group
  •  The results had a low fragility index, meaning that if the outcomes had been different for a few patients, then the results reported would have no longer been statistically significant

The Bottom Line

  • This single centre under powered study found that for critically ill patients a conservative, compared with a conventional, oxygen strategy resulted in a dramatic mortality benefit. As the authors state, a further multi-centre study is needed to confirm these findings. I hope that future trials also investigate the use of permissive hypoxia, which this trial did not investigate. Whilst awaiting for these trials I will avoid hyperoxia.

External Links


Summary author: David Slessor
Summary date: 13.10.2016.
Peer-review editor: Duncan Chambler


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.