Critical Care, Critical Care Alert

Critical Care Alert: Conservative Oxygen Therapy in Mechanically Ventilated Patients - the ICU-ROX Study

Critical Care Alert

ICU-ROX Investigators and the Australian and New Zealand Intensive Care Society Clinical Trials Group, et al. Conservative Oxygen Therapy during Mechanical Ventilation in the ICU. N Engl J Med. 2019; epub ahead of print Oct. 14.

To evaluate if conservative oxygen therapy would lead to more ventilator-free days than usual oxygen therapy in adults on mechanical ventilation and to establish a regimen for doing so.


There have been a series of trials looking at the impact of conservative oxygen therapy on mortality. The OXYGEN-ICU trial found that patients on a conservative oxygen regimen (lowest possible FiO2 to target SpO2 94-98%) showed a dramatic difference in mortality among ICU patients (11.6% compared to 20.2%). A similar reduction in harm or lack of benefit of excess oxygen was also seen in the trials HYPERS2S (looking at patients with sepsis), SO2S (stroke patients) AVOID (STEMI patients), and DETO2X (AMI patients).  Enough RCT trials have taken place that the IOTA trial completed meta-analysis on 25 RCTs looking at the impact of conservative oxygen therapy compared to liberal oxygenation of a variety of patient presentations (stroke, MI, critically ill, emergency surgery, sepsis) receiving oxygen via room air, nasal cannula, facemask or ET tube. The study ultimately found a significant reduction in mortality in patients receiving the conservative oxygen therapy.

The ICU-ROX study was designed as the largest RCT of mechanically ventilated ICU patients with the hope of examining this question and establishing a more research driven regimen for conservative oxygen delivery.

965 adult patients (> 18 yo) who had received less than 2 hours of mechanical ventilation and were expected to continue to need mechanical ventilation beyond the day of recruitment.

Patients were excluded if hyperoxia was indicated for their medical management (carbon monoxide poisoning), if avoidance of hyperoxia was indicated (COPD, chronic hypercapnic respiratory failure), if they were pregnant, death was deemed inevitable, or if they were chronically ventilated patients.

Randomized control trial across 21 ICUs in New Zealand and Australia that was single-blinded (blinding of the outcome assessors but not the clinicians providing the treatment)

Conservative oxygen therapy: Maintain SpO2 at > 90% and < 97%

  • If the SpO2 rose above 96% the FiO2 was lowered until the SpO2 is < 97% or the FiO2 was 21%
  • If the SpO2 was between or equal to 91-96% the FiO2 was lowered until FiO2 was 21% or the SpO2 fell below the target range
  • If the SpO2 fell below 91%, the FiO2 was returned to the last that achieved the target SpO2 of 91-96%

Liberal oxygen therapy: No upper limit to SpO2 with a targeted lower limit of > 90%. The use of FiO2 < 30% was discouraged.



  • Number of ventilator-free days up through day 28


  • Death by day 90 and day 180
  • Length of survival
  • Cognitive function
  • Health-related quality of life at day 180
  • Proportion of patients who had been employed at baseline but were then unemployed at day 180

There was no significant difference for any of the outcomes: number of ventilator free days, mortality at days 90 or 180, quality of life metrics, cognitive function, nor employment at day 180.

There was also no difference between in the number of adverse events (2 events of hypoxia in the conservative group, and 1 stroke in the liberal group).

The authors had planned to do a series of subgroup analyses. Interestingly, the analysis on the subgroup with hypoxic-ischemic encephalopathy found statistically significant improvement in mortality (43% conservative group v. 59% liberal group), ventilator free days, and cognitive function.


  • This is the largest trial on mechanically ventilated ICU patients that is asking this very clinically relevant question of conservative v liberal oxygen therapy
  • Randomized controlled trial with intention to treat analysis and good follow up
  • The control and intervention group had a clear difference in oxygen exposure (measured in PaO2, SpO2 and FiO2 differences between groups)


  • Non-blinded clinicians
  • This trial was done as much of the conversation and initial studies were starting to point toward more conservative oxygen therapy. While there was a difference between the conservative and liberal groups, the relatively low oxygen exposure in the liberal group suggests that group may have had a more conservative oxygen therapy than liberal groups in prior RCTs, potentially blunting the difference in outcomes.

Even though there was no significant difference between the liberal and conservative oxygen groups in this study, there is a large body of evidence pointing to the benefit of conservative oxygen therapy. While unfortunately there is still not a standard regimen for conservative oxygen therapy, this study is not likely to change practice much from the current standard of trying to limit excess oxygen exposure in mechanically ventilated patients.

Interestingly, the authors of this study have proposed a larger study - MEGAROX, with 40,000 patients - with increased subgroup analyses, so stay tuned. Given the huge number of patients, there are likely to be interesting result and much conversation.

Related Articles

Critical Care Device Series: Transvenous Pacemaker

Temporary transvenous pacing (TTVP) utilizes central venous access to pass an electrode into the right ventricle. TTVPs are one of the most infrequently performed procedures by emergency physicians;

Aortic Occlusion: A Rare Presentation of Back Pain

Back pain is a common chief complaint in the emergency department. However, not all back pain is simple lumbago, and as emergency physicians, it is crucial to be aware of more insidious causes of back