Oxygen-Saturation Targets for Critically Ill Adults Receiving Mechanical Ventilation

Semler. N Engl J Med 2022; 387:1759-1769. DOI: 10.1056/NEJMoa2208415

Clinical Question

  • In patients receiving invasive mechanical ventilation, does a target oxygen saturation of 90%, 94%, or 98% increase days alive and free of mechanical ventilation at 28 days?


  • Despite the ubiquity of oxygen therapy in the critical care setting, the optimal oxygenation target remains elusive
  • Higher oxygen targets provide a degree of protection against hypoxaemia, but expose patients to the potential harms of a high fraction of inspired oxygen (FiO2)
  • Even a small survival difference could result in a significant reduction in mortality, given the vast numbers of patients treated with mechanical ventilation
  • A number of trials have compared restrictive and liberal oxygen therapy, with more recent trials showing equipoise between the two groups


  • Single centre, cluster-randomised, cluster-crossover trial
  • Non-blinded
  • Patients assigned to trial from emergency department (ED) or intensive care unit (ICU) at first receipt of invasive mechanical ventilation
  • Cluster design meant ED and ICU switched to one of three targets every 2 months, in a randomly generated sequence
  • The three oxygen targets were defined as
    • lower (SpO2 90%; range 88-92%)
    • intermediate (SpO2 94%; range 92-96%)
    • higher (SpO2 98%; range 96-100%)
  • The final 7 days of each period were considered a wash-out period, and were not included in the primary analysis
  • Primary outcome defined as days alive and ventilator free at 28 days
  • Sole secondary outcome defined as in-hospital death at 28 days
  • 2250 patients would provide 92% power at alpha level of 0.05 to detect absolute difference of 2 ventilator-free days between any two of the three groups


  • Single centre in Nashville USA
  • Data collection occurred from July 1st 2018 to August 31st 2021
  • There was a 2 month pause due to Covid-19 during April and May 2020


  • Inclusion:
    • All adult patients (>18 years of age)
    • Located in ED or ICU
    • Requiring mechanical ventilation
  • Exclusion:
    • Pregnant
    • Incarcerated
    • Patients with Covid-19 were managed on a separate ICU. Those with a laboratory confirmed diagnosis were not included in the primary analysis
  • 2987 patients enrolled
    • 446 (14.9%) excluded due to enrolment during washout period, or Covid-19 diagnosis
  • Baseline characteristics of lower vs intermediate vs higher target groups were broadly similar
    • Age: 57 vs 59 vs 59
    • Female: 44.7% vs 44.8% vs 46.8%
    • Comorbidities
      • COPD: 18.3% vs 20.4% vs 19.3%
      • Coronary artery disease: 17.9% vs 17.7% vs  20.4%
    • Acute illness
      • Cardiac arrest: 15.5% vs 11.6% vs 12.5%
      • Myocardial infarction: 16.8% vs 16.1% vs 16.6%
      • Sepsis: 34% vs 28.8% vs 32.4%
      • Received vasopressors: 19.8% vs 19.9% vs 17.5%
    • Median time from initiation of ventilation to enrolment: 0.0 vs 0.0 vs 0.0hr
    • Separation of SpO2 and FiO2 between trial groups occurred


  • Lower oxygen target
    • Target SpO2 90%; range 88-92%
  • Intermediate oxygen target
    •  Target SpO2 94%; range 92-96%


  • Higher oxygen target
    • Target SpO2 98%; range 96-100%

Management common to both groups

  • Only the changes in FiO2 were mandated by the trial protocol
  • FiO2 was titrated by respiratory therapists within 15 minutes of assignment to trial
  • All other aspects of patient care, including ventilatory parameters and PEEP, were left to the discretion of the treating clinician


  • Primary outcome: number of days alive and free of mechanical ventilation at 28 days – no significant difference
    • 20 days in lower; 21 days in intermediate and higher
    • Odds Ratios
      • Lower vs intermediate: 0.95 (95% CI 0.79-1.13)
      • Intermediate vs higher: 1.00 (95% CI 0.84-1.19)
      • Lower vs intermediate: 0.95 (95% CI 0.78–1.14)
  • Secondary outcome: No significant difference in secondary outcome or numerous prespecified exploratory outcomes
    • Comparing lower vs intermediate vs higher targets
    • No significant difference in 28 day mortality
      • 34.8% vs 34.0% vs 33.2%

Authors’ Conclusions

  • Among critically ill patients receiving mechanical ventilation, days alive and ventilator free did not differ among groups in which lower, intermediate and higher oxygenation targets were used


  • Very large numbers considering this was a single centre trial
  • Excellent separation between the three groups
    • Average sats achieved for each group 93%, 94% and 97% respectively
    • Average FiO2 administered for each group 0.31, 0.37 an 0.45 respectively
  • Use of SpO2 rather than PaO2 allowed for vast, highly granular data set
    • Nearly 8 million individual saturation readings
  • Excellent overall adherence to SpO2 targets
  • Cluster crossover format reduces selection bias and reduces time to enrolment
  • Sequential randomisation over the study period reduces risks of bias due to seasonality
  • Patient-centred primary outcome with face validity – that is, it is relevant to the intervention being tested
  • The choice of primary outcome allows a greater degree of discrimination between interventions than a simple mortality outcome


  • A few factors reduce the generalisibility of these results
    • Single centre trial
    • The ICU involved did not have surgical patients
    • The trial intervention was implemented by respiratory therapists, a role not present in some settings
    • Some unusual baseline characteristics for the patients
      • For example, 14% on home oxygen and 3% of patients chronically ventilated. This last group of patients could never achieve the primary outcome
  • Though separation was good, the target saturations for the lower group were not really met overall, increasing the risk of a beta error
    • This is likely due to reaching an FiO2 of 0.21, as occurred in around a 3rd of patients in the lower group
  • There is a chance that the use of different FiO2s may have introduced clinician bias, affecting their decision to attempt a spontaneous breathing trial and extubation
  • The primary outcome does not discern between dying and being ventilated for 28 days, a difference which would clearly matter to patients

The Bottom Line

  • In this trial of low, intermediate and high oxygen targets, no significant difference was found in days alive and ventilator free at 28 days
  • This excellent trial has challenged my prior inclination towards lower oxygen saturation targets and lower FiO2s. I remain confident in my belief that unnecessarily high FiO2s are harmful, but I will be less aggressive in reducing the FiO2 in the absence of hyperoxia
  • Those who use higher targets will be reassured. The chance that there is a large treatment benefit to any particular oxygen target is extremely low
  • I will continue to aim for the equivalent of an intermediate target of oxygenation, except in a research setting
  • I await the results of Mega-ROX with great anticipation

External Links


Summary author: Omar Jundi @bi0marker
Summary date: 01/12/22
Peer-review editor: @davidslessor

Image by Defence-Imagery from Pixabay



  • Dr Alex D Mattin

    Thanks for the review.
    Don’t forget to eagerly await the results of our very own UK ROX trial too, we are recruiting 18,000 patients UK wide

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.