Harmful Effects of Hyperoxia in Postcardiac Arrest, Sepsis, Traumatic Brain Injury, or Stroke: The Importance of Individualized Oxygen Therapy in Critically Ill Patients

Table 1

Some recent clinical studies on the risks of hyperoxia after cardiac arrest or myocardial infarction, in traumatic brain injury, stroke, sepsis, and mixed ICU patients.

References	Study design	Hyperoxia measurements	Main finding
After cardiac arrest or myocardial infarction
Kilgannon et al. 2010 [11].	Retrospective cohort study, 120 hospitals, 6326 patients (nontraumatic cardiac arrest)	First PaO2 in the first 24 hours.; Hyperoxia: PaO2 ≥ 300 mmHg	Hyperoxia was associated with an increased hospital mortality compared with either hypoxia or normoxia (OR 1.8 [1.5–2.2])
Bellomo et al. 2011 [23]	Retrospective cohort study, 125 ICUs, 12108 patients (nontraumatic cardiac arrest)	Worst PaO2 in first 24 h.; Hyperoxia: PaO2 ≥ 300 mmHg; Normoxia: PaO2 60–300 mmHg	Hyperoxia group had a higher hospital mortality than normoxia (OR 1.2 [1.1–1.6])
Kilgannon et al. 2011 [24]	Retrospective cohort study, 120 hospitals, 4459 patients (nontraumatic cardiac arrest)	Highest PaO2 in the first 24 hours	A 100 mmHg increase in PaO2 was associated with a 24% increase in mortality risk (OR 1.24 [1.18 to 1.31])
Ranchord et al. 2012 [25]	Pilot randomized controlled trial, single-centre, 136 patients with STEMI	Patients randomized to receive high-concentration (6 L/min) or titrated oxygen (to achieve oxygen saturation 93%–96%) for 6 hours after presentation	No differences in number of deaths in the two groups (relative risk 0.5, 95% CI 0.05–5.4, )
Janz et al. 2012 [26]	Retrospective analysis of a prospective cohort study, single-centre 170 patients (cardiac arrest treated with mild therapeutic hypothermia)	Highest PaO2 in first 24 h.	Increased hospital mortality for every 100 mmHg increase in PaO2 (OR 1.49 [1.03, 2.14])
Lee et al. 2014 [27]	Retrospective cohort study, single-centre, 213 patients (cardiac arrest treated with therapeutic hypothermia)	Average PaO2 between ROSC and the end of rewarming.; Hyperoxia: PaO2 > 157 mmHg; Normoxia: PaO2 117–135 mmHg	V-shaped association between PaO2 and poor neurologic outcome at hospital discharge (OR 6.47 [1.68, 24.91])
Stub et al. 2015 [14]	Prospective, randomized, controlled trial, 9 hospitals, 441 patients with STEMI	Patients with an SpO2 > 94% were randomized to receive 8 L/min of oxygen or no supplemental oxygen from arrival of paramedics until transfer to the cardiac care unit	An increased rate of recurrent myocardial infarction, an increase in the frequency of cardiac arrhythmias, and an increase in myocardial infarct size at 6 months on magnetic resonance imaging in the supplement group
Elmer et al. 2015 [12]	Retrospective analysis of a high-resolution database, single-centre, 184 patients postcardiac arrest	Mean hourly exposure in first 24 h.; Normoxia: PaO2 60–100 mmHg; Moderate hyperoxia: PaO2 101–299 mmHg; Severe hyperoxia: PaO2 300 mmHg	Severe hyperoxia was associated with decreased survival (OR 0.83 [0.69–0.99] per hour exposure); moderate hyperoxia was not associated with survival but with improved SOFA score 24 h (OR 0.92 [0.87–0.98])
Eastwood et al. 2016 [13]	Retrospective before-after nested cohort study, single-centre, 50 patients postcardiac arrest	Conservative oxygenation: SpO2 88–92%	Conservative group had a shorter ICU length of stay; no difference in the proportion of survivors discharged from hospital with good neurological outcome compared to conventional group
In traumatic brain injury (TBI) and stroke
Davis et al. 2009 [18]	Retrospective cohort study, 5 trauma centres, 3420 moderate-to-severe patients	Extreme hyperoxia: first PaO2 > 487 mmHg	A PaO2 value of 110–487 mmHg was considered optimal. Extreme hyperoxia had an independent association with decreased survival (OR 0.50 [0.36, 0.71]) compared to optimal range
Brenner et al. 2012 [19]	Retrospective study, single-centre, 1547 severe TBI patients	Mean PaO2 in first 24 h hospital admission: Hyperoxia: PaO2 > 200 mmHg; Normoxia: PaO2 100–200 mmHg	Both low and high PaO2 had increased mortality.; Patients with hyperoxia had higher hospital mortality (OR 1.50 [1.15–1.97]) and lower discharge GCS scores at discharge (OR 1.52 [1.18–1.96])
Raj et al. 2013 [20]	Retrospective nested cohort analysis, 5 hospitals, 1116 ventilated moderate-to-severe TBI patients	Worst PaO2 in first 24 h ICU admission: Hyperoxia: PaO2 > 100 mmHg; Normoxia: PaO2 75–100 mmHg	Hyperoxia had no independent relationship with in-hospital mortality (OR 0.94 [0.65–1.36]) and 6-month mortality (OR 0.88 [0.63–1.22])
Rincon et al. 2014 [28]	Retrospective cohort, 84 ICUs, 2894 stroke patients	PaO2 in the first 24 hours.; Hyperoxia: PaO2 ≥ 300 mmHg; Normoxia: PaO2 60–300 mmHg	Hyperoxia was independently associated with in-hospital mortality (OR 1.22 [1.04–1.48])
Rincon et al. 2014 [29]	Retrospective cohort study, 61 hospitals, 1212 ventilated TBI patients	Hyperoxia: PaO2 > 300 mmHg; Normoxia: PaO2 60–300 mmHg	Hyperoxia was associated with a higher in-hosptial case fatality (OR 1.5 [1.02–2.4])
Jeon et al. 2014 [30]	Prospective, observational cohort database analysis, single-centre, 252 patients (subarachnoid haemorrhage)	PaO2 AUC by observation time until delayed cerebral ischemia (DCI). Hyperoxia: PaO2 ≥ 173 mmHg (upper quartile)	Hyperoxia group had a higher incidence of DCI (OR 3.16 [1.69 to 5.92]) and poor outcome (modified Rankin Scale 4–6 at 3 months after subarachnoid haemorrhage) (OR 2.30 [1.03 to 5.12])
Quintard et al. 2015 [17]	Retrospective analysis of a database, single-centre, 36 severe TBI patients	Hyperoxia: PaO2 > 150 mmHg	Hyperoxia was associated with increased cerebral microdialysis glutamate, indicating cerebral excitotoxicity
Lang et al. 2016 [31]	Retrospective analysis using 2 databases, 432 ventilated patients (subarachnoid haemorrhage)	Time-weighted average PaO2 during the first 24 hours; Low PaO2 < 97.5 mmHg; Intermediate PaO2 97.5–150 mmHg; High PaO2 >150 mmHg	Patients with an unfavorable outcome had significantly higher PaO2, but high PaO2 has no effect on 3-month neurological outcomes (OR 1.09 [0.61–1.97]) or mortality (OR 0.73 [0.38–1.40])
In sepsis
Stolmeijer et al. 2014 [32]	Prospective pilot study, 83 sepsis patients in emergency department, single-centre	PaO2 after 5 min of a VentiMask 40% with 10 L O2/min.; Hyperoxia: PaO2 > 100 mmHg	Of the hyperoxic patients, 8% died in hospital versus 6% with normoxia
In mixed ICU patients
de Jonge et al. 2008 [4]	Retrospective observational study, 50 ICUs, 36307 ventilated patients	Worst PaO2 in first 24 h.; Hyperoxia: PaO2 ≥ 123 mmHg (upper quintile) compared with PaO2 between 67 and 80 mmHg	In-hospital mortality was linearly related to FiO2 value and had a U-shaped relationship with PaO2. Hyperoxia had a higher mortality (OR 1.23 [1.13–1.34])
Panwar et al. 2016 [33]	Pilot randomized controlled trial, 4 ICUs, 103 patients	Conservative oxygenation: SpO2 88–92%; Liberal oxygenation: SpO2 ≥ 96%	No significant differences in measures of new organ dysfunction, or ICU or 90-day mortality (OR 0.77 [0.40–1.50])
Girardis et al. 2016 [34]	Open-label randomized trial, single-centre, 434 patients	Conservative oxygenation: PaO2 70–100 mmHg (SpO2 94–98%); Conventional oxygenation: PaO2 > 150 mmHg (SpO2 97–100%)	Patients in the conservative group had lower ICU mortality (RR 0.57 [0.37–0.9]) and fewer episodes of shock, liver failure, and bacteraemia
Helmerhorst et al. 2017 [35]	Observational cohort study, 3 ICUs, 14441 ventilated patients	First PaO2 at ICU admission; Mild hyperoxia: PaO2 120–200 mmHg; Severe hyperoxia: PaO2 > 200 mmHg	Severe hyperoxia was associated with higher mortality rates and fewer ventilator-free days in comparison to both mild hyperoxia and normoxia; Time spent in hyperoxia had a linear and positive relationship with hospital mortality

OR: odds ratio; SOFA: sequential organ failure assessment; ROSC: return of spontaneous circulation; DIC: delayed cerebral ischemia; AUC: area under the curve; STEMI: ST-segment elevated myocardial infarction.