1.  Provision of life support (mechanical ventilation) is required for most critically ill patients who are admitted to an intensive care unit (ICU)  

2.  Delivery of supplemental oxygen to ICU patients who require mechanical ventilation often exposes them to a high fraction of inspired oxygen (FIO 2 ) and higher than normal arterial oxygen partial pressure (PaO 2 ).  

3. The human body is adapted to breathe air (21% oxygen) and it is biologically plausible that exposure to higher amounts of oxygen might be harmful .  

4. In the Intensive Care Unit Randomised Trial Comparing Two Approaches to Oxygen therapy (ICU-ROX) trial, conservative oxygen therapy targets neither increased nor decreased days alive and free from invasive mechanical ventilation (ventilator-free days) compared with usual (relatively liberal) oxygen therapy targets in mechanically ventilated adults anticipated to be ventilated in ICU beyond the calendar day after randomisation. 

5. Importantly though, the ICU-ROX trial findings did not exclude clinically important effects of the oxygen regimens tested on mortality.  Based on the distribution of data in the ICU-ROX trial there is a 46% chance that conservative oxygen therapy targets increase absolute mortality by more than 1.5 percentage points and a 19.3% chance that conservative oxygen therapy targets decrease absolute mortality by more than 1.5 percentage points compared with liberal oxygen therapy.   

6.  Given the number of patients who receive mechanical ventilation in ICU every year, an absolute effect on mortality of 1.5 percentage points would have profound global public health importance.  For every 100,000 patients treated, such a difference would equate to 1,500 lives saved or lost. 

7.  Accordingly, we are conducting a definitive phase 3 trial to test the two-sided hypothesis that, compared with liberal oxygen therapy targets, conservative oxygen therapy targets reduce mortality by 1.5 percentage points in adult ICU patients who are ventilated in ICU following an emergency admission or who are emergently intubated in the ICU. 

8.  This 40,000 participant trial will be conducted in multiple countries and includes innovative trial design features:  

i.  Linkage to identify enrolled patients in national ICU registries so that outcome data do not need to be collected specifically for the trial;  

ii.  Response adaptive randomisation giving trial participants an increased chance of being assigned to the oxygen regimen associated with the lowest mortality risk while the trial is ongoing. 

9. In the event that a zero percentage point absolute mortality difference between treatment groups is observed in our trial, 95% CIs would be expected to exclude the possibility of an absolute increase or decrease in mortality of well under one percentage point.  In this situation, in the absence of heterogeneity of treatment effect, we submit that our trial would effectively exclude the possibility of a clinically important effect of conservative oxygen therapy on in-hospital mortality in this patient population. 

10. Because we consider that there is a distinct possibility that conservative oxygen therapy will be best for patients with some diagnoses while liberal oxygen will be best for patients with other diagnoses (i.e. that there will be heterogeneity of treatment effect), we are conducting a number of parallel nested trials within the overall 40,000 participant trial sample.  Each of these nested trials will evaluate a pre-specified hypothesis in a specific cohort of critically ill patients and is accompanied by an appropriate power calculation.  

Hence, determining the effects of liberal versus conservative oxygen level targets in patient on invasive mechanical ventilation in general patients and some specific subpopulation is urgently indicated.