Goldilocks in the ICU: Oxygenation targets for mechanical ventilation

Like all medical therapies, we have learned that treatment with oxygen comes at a cost. The medical literature is replete with the detriments of “prophylactic” oxygen and adverse effects of hyperoxia in the treatment of myocardial infarction, acute stroke, cardiac arrest and septic shock. The DETO2X-SWEDEHEART showed no survival benefit for higher PaO2 in patients with myocardial infarction.1 The Oxygen-ICU study showed increased mortality for critically ill patients receiving conventional oxygen (SpO2 97% to 100%) versus conservative oxygen therapy (SpO2 94 to 98%).2 In patients resuscitated following cardiac arrest, there is a dose-dependent relationship between elevated PaO2 and in-hospital death.3 Moreover, hyperoxia after resuscitation from cardiac arrest is independently associated with neurologic disability at hospital discharge as measured by modified Rankin Scale in a prospective trial.4 Finally, in patients with septic shock, the HYPER2S trial demonstrated that ventilation with 100% fraction of inspired oxygen (FiO2) inducing elevated arterial oxygen partial pressure was associated with increased mortality and fewer ventilator-free days.5

Based on this data suggesting harm associated with high arterial oxygen tension, the authors of the multicenter ICU-ROX trial (Intensive Care Unit Randomized Trial Comparing Two Approaches to Oxygen Therapy) investigated whether avoidance of hyperoxia, i.e. conservative oxygen therapy with a target SpO2 90% to 97%, would increase the number of ventilator-free days as compared to usual oxygen therapy targeting SpO2 > 90% with no upper limit.6 They enrolled all adults expected to receive mechanical ventilation in the ICU beyond the day after trial recruitment within 2 hours of institution of mechanical ventilatory support. Patients were randomized to receive conservative oxygen therapy with upper limit SpO2 97% versus usual oxygen therapy with no upper SpO2 limit. The lower limit for SpO2 in both groups was 90%, although this could be adjusted by the treating clinician. The primary outcome was ventilator-free days defined as the total number of calendar days with unassisted breathing within 28 days of randomization. Secondary outcomes included death from any cause between days 90 and 180 and total duration of survival. 


A total of 965 patients were enrolled with 484 patients randomized to receive conservative oxygen and 481 patients randomized to receive usual oxygen therapy. The conservative oxygen group spent less time with an SpO2 >97% (median 27 hours vs. 49 hours in the usual oxygen group). There was no observed difference in the primary outcome of ventilator-free days (median 21.3 days in the conservative oxygen group and 26.2 days in the usual oxygen group). Mortality in patients receiving conservative oxygen therapy was not significantly different than those receiving usual oxygen therapy with no SpO2 upper limit, but there was a signal for increased mortality with higher oxygenation target (35.7% vs. 34.5%, respectively). In a prespecified subgroup of critically ill patients with hypoxic-ischemic encephalopathy, mortality at 180 days was higher in the usual oxygen therapy group, which was considered a hypothesis-generating finding suggesting possible harm for higher oxygenation targets. 


Unlike the, single-center Oxygen-ICU trial where conservative oxygen therapy was associated with a greater number of ventilator-free days and lower risk of death, the ICU-ROX trial found no difference in ventilator-free days for conservative oxygen therapy in the ICU, although there was a signal of harm in the subgroup of patients with hypoxic-ischemic encephalopathy. One difference between the two studies is that the usual care group in the Oxygen-ICU study targeted an SpO2 of 97-100% whereas a max SpO2 of 97% was targeted in the ICU-ROX usual care group. Therefore, the authors acknowledge that limiting the peak SpO2 to 97% or under in the usual care group could have muted the treatment effect. Despite this variance, similar PaO2 levels were achieved in both trials.


What about patients with acute respiratory distress syndrome (ARDS) who have the most severe oxygenation impairment? Based on ARDSnet guidelines, a PaO2 between 55 and 80 mm Hg has typically been targeted; however, this has not been prospectively validated. In the LOCO2 trial (Liberal Oxygenation versus Conservative Oxygenation in ARDS), Barrot et al. hypothesized that targeting a PaO2 of 55 to 70 mm Hg (SpO2 88-92%) would improve 28-day morality in ARDS patients compared to a liberal target of PaO2 of 90-105 mm Hg (SpO2 > 96%).7 Patients were included if they received mechanical ventilation for less than 12 hours and met the Berlin definition for ARDS with P/F ratio of 300 mm Hg or less. Both groups received lung protective ventilation with tidal volumes of 6 mL per kg predicted body weight and neuromuscular blockade and prone positioning was recommended in patients with P/F < 150 mm Hg. After enrolling 205 patients in 13 ICUs in France, the trial was stopped early for safety and lack of anticipated difference in the primary outcome. The mortality at 28 days was 34.3% in the conservative oxygen group and 26.5% in the liberal oxygen group (difference of 7.8%, 95% CI -4.8 to 20.6). However, at day 90, mortality was significantly higher in the conservative-oxygen group than in the liberal oxygen group (44.4% vs 30.4%, respectively). The lower probability of survival in the conservative oxygenation group remained after adjusting for age and severity of illness. Five patients in the conservative oxygenation group developed mesenteric ischemia versus none in the liberal oxygenation group. Interestingly, these findings were consistent with subgroup analysis of patients in the Oxygen-ICU where the lowest mortality was observed with patients who had a median PaO2 between 87 to 93 mm Hg as compared to a lower PaO2 54 to 81 mm Hg. 


These landmarks trials demonstrate that targets for oxygen therapy may parallel glycemic control: not too high (SpO2 > 97%) and not too low (SpO2 88% to 92%), particular in patients with severe ARDS that may be subject to hypoxemia. So, what is just right? The optimal oxygenation target for mechanical ventilation is not definitive and probably depends on the disease process and degree of oxygen impairment based on P/F ratio. Perhaps what we have learned from the Oxygen-ICU, ICU-ROX and LOCO2 trials, is that both hypoeroxia (SpO2 >97%) and hypoxemia (SpO2 <90%) should be avoided, and a target SpO2 of 92% to 96% (PaO2 ~60s-90s) for Goldilocks may be “just right.” The end to this fairytale remains to be told.    



1. DETO2X-SWEDEHEART. Oxygen therapy in suspected acute myocardial infarction. New Engl J Med2017;377(13):1240-1249. 


2. Oxygen-ICUEffect of conservative vs conventional oxygen therapy on mortality among patients in an ICU. JAMA2016;316(15):1583-1589. 


3. Kilgannon JH, et al. Relationship between supranormal oxygen tension and outcome are resuscitation from cardiac arrest. Circulation 2011;123:2717-2722. 


4. Roberts BW, et al. Association between early hyperoxia exposure after resuscitation from cardiac arrest and neurological disability. Circulation 2018;137:2114-2124.


5. HYPERS2S. Hyperoxia and hypertonic saline in patients with septic shock: a two-by-two factorial multicenter, randomized, clinical trial. Lancet Respir Med 2017;5(3):180-190. 


6. ICU-ROX. Conservative oxygen therapy during mechanical ventilation in the ICU. New Engl J Med 2020;382(11):989-997. 


7. LOCO2Liberal or Conservative Oxygen Therapy for Acute Respiratory Distress Syndrome. New Engl J Med2020;382(11):999-1007.