Article Text

Research paper
Hyperoxia may be related to delayed cerebral ischemia and poor outcome after subarachnoid haemorrhage
  1. Sang-Beom Jeon1,2,3,
  2. H Alex Choi2,3,
  3. Neeraj Badjatia3,4,
  4. J Michael Schmidt3,
  5. Hector Lantigua3,
  6. Jan Claassen3,
  7. E Sander Connolly5,
  8. Stephan A Mayer3,
  9. Kiwon Lee2,3
  1. 1Department of Neurology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea
  2. 2Departments of Neurology and Neurosurgery, The University of Texas Medical School at Houston, Houston, Texas, USA
  3. 3Department of Neurology, Columbia University College of Physicians and Surgeons, New York, New York, USA
  4. 4Department of Neurology, University of Maryland School of Medicine, Baltimore, Maryland, USA
  5. 5Departments of Neurosurgery, Columbia University College of Physicians and Surgeons, New York, New York, USA
  1. Correspondence to Dr Kiwon Lee, Departments of Neurology and Neurosurgery, The University of Texas Medical School at Houston, 6431 Fannin St., MSB 7.152, Houston, TX 77030, USA; Kiwon.Lee{at}


Objective To determine the association between exposure to hyperoxia and the risk of delayed cerebral ischaemia (DCI) after subarachnoid haemorrhage (SAH).

Methods We analysed data from a single centre, prospective, observational cohort database. Patient inclusion criteria were age ≥18 years, aneurysmal SAH, endotracheal intubation with mechanical ventilation, and arterial partial pressure of oxygen (PaO2) measurements. Hyperoxia was defined as the highest quartile of an area under the curve of PaO2, until the development of DCI (PaO2≥173 mm Hg). Poor outcome was defined as modified Rankin Scale 4–6 at 3 months after SAH.

Results Of 252 patients, there were no differences in baseline characteristics between the hyperoxia and control group. Ninety-seven (38.5%) patients developed DCI. The hyperoxia group had a higher incidence of DCI (p<0.001) and poor outcome (p=0.087). After adjusting for modified Fisher scale, rebleeding, global cerebral oedema, intracranial pressure crisis, pneumonia and sepsis, hyperoxia was independently associated with DCI (OR, 3.16; 95% CI 1.69 to 5.92; p<0.001). After adjusting for age, Hunt–Hess grade, aneurysm size, Acute Physiology and Chronic Health Evaluation II score, rebleeding, pneumonia and sepsis, hyperoxia was independently associated with poor outcome (OR, 2.30; 95% CI 1.03 to 5.12; p=0.042).

Conclusions In SAH patients, exposure to hyperoxia was associated with DCI. Our findings suggest that exposure to excess oxygen after SAH may represent a modifiable factor for morbidity and mortality in this population.


Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Supplemental oxygen is commonly administered to patients after subarachnoid haemorrhage (SAH) during the prehospital, emergency and critical care.1 During mechanical ventilation, hyperoxia is not uncommon in the emergency room and intensive care unit during the management of SAH. Recent studies in patients after cardiac arrest and traumatic brain injury have suggested that hyperoxia is associated with worse outcomes.2 ,3 Contrarily, others have suggested that hyperoxia may be of therapeutic benefit for patients with traumatic brain injury and ischaemic stroke.4–7 The effect of hyperoxia on outcomes in patients with SAH has not been studied.

At supra-physiologic levels, oxygen may have toxic effects on the human brain through production of superoxide anion, hydrogen peroxide and hydroxyl radical.8 The overproduction of these reactive oxygen species results in oxidative stress, a deleterious process which may lead to neuronal inflammation and necrosis as well as reduction of cerebral blood flow.9 ,10 After SAH, these processes may contribute to the development of delayed cerebral ischaemia (DCI).

We hypothesised that hyperoxia may increase the risk of DCI after SAH. We aimed to investigate the association between exposure to hyperoxia and the incidence of DCI and poor 3-month outcome.

Patients and methods


Columbia University SAH Outcomes Database Project is a single centre, prospective observational cohort study. Data fields include demographics, clinical, radiological and outcome data for all adult spontaneous SAH patients. All data are prospectively collected by a dedicated investigator and are adjudicated by the study team in weekly meetings.11 ,12 This retrospective analysis considered all SAH patients collected in our registry between August 1996 and April 2011. We included patients who (1) were 18 years of age or older; (2) had aneurysmal SAH confirmed by brain CT with angiography and/or CT angiography; (3) were endotracheally intubated with ventilators within 3 days after onset (excluding patients intubated solely for surgical interventions); (4) underwent arterial partial pressure of oxygen (PaO2) measurements ≥3 times (to be able to calculate area under the curve) within 14 days, including twice or more within 3 days. We excluded patients who expired within 14 days after symptom onset because (1) early death would lead to less chance of DCI; (2) it was uncertain that the cause of early death was related to DCI; (3) severe hypoxia could precede death.

This study was approved by the institutional review board of our centre. Informed consents for this study were obtained from patients or their family members.

Clinical, radiological and laboratory assessment

We obtained demographics (age, sex and ethnicity), risk factors or underlying comorbidities (hypertension, smoking, alcohol consumption, previous ischaemic or haemorrhagic stroke, cardiac disease, such as myocardial infarction and arrhythmia, and pulmonary disease, such as chronic obstructive pulmonary disease), clinical, radiological and laboratory findings on initial assessments (Glasgow coma scale (GCS), Hunt–Hess grade, modified Fisher scale,13 Hijdra scale,13 global cerebral oedema on initial CT,14 APACHE II (Acute Physiology and Chronic Health Evaluation II) score, PaO2, and arterial partial pressure of carbon dioxide), and acute complications of SAH on follow-up assessments (rebleeding,15 global cerebral oedema on follow-up CT, hydrocephalus, intracranial pressure crisis, meningitis, pneumonia or pulmonary oedema, and sepsis), DCI, and modified Rankin Scale (mRS).15

Definitions of hyperoxia and other variables

Oxygen burden was defined as the area under the curve of PaO2 divided by the observation period since the onset of SAH. The main objectives of this study were to investigate the effects of hyperoxia on DCI, which mostly occurs within the first 2 weeks. For this purpose, oxygen burden was calculated until the time point when each patient developed DCI. For patients who did not develop DCI, oxygen burden until postbleed day 6 was adopted for comparison. Six days were the median day from symptom onset to DCI in our study population. We further defined hyperoxia as the uppermost quartile of oxygen burden. The median oxygen burden was 146.5 mm Hg (IQR, 123–173 mm Hg) (figure 1). Thus, oxygen burden ≥173 mm Hg was considered as hyperoxia, as aforementioned. We chose these cutoff points, because (1) we investigated the ‘excess’ oxygen's effects and (2) there is no consensus on a specific cutoff point of hyperoxia in a human SAH population.

Figure 1

Distributions of oxygen burden.

DCI was defined as either a symptomatic vasospasm (ie, clinical deterioration, such as a new focal deficit and/or a decrease of the level of consciousness, in the presence of vasospasm determined by conventional angiography or CT angiography) or development of new cerebral infarction which is not explained by other causes such as hydrocephalus, seizures, intracranial pressure crisis and infection.16 This definition of DCI has been demonstrated to have a better predictive value for poor clinical outcomes than symptomatic, angiographic, or transcranial Doppler vasospasm.16 mRS was obtained by in-person interview or phone interview either with the patient or with a surrogate if the patient was unable to participate. Poor outcome was defined as mRS 4–6 at 3 months after the onset of SAH.16 We recorded the day of aneurysmal rupture as bleed day 0.

Clinical managements

The management of these patients followed guidelines set forth by the American Heart Association and has been described in detail previously.15 Endotracheal intubation was performed and mechanical ventilator was applied to the patients whose respiratory drive or reflexes were abnormal. All patients underwent neurological examinations each day. Patients underwent conventional angiography on admission, and between postbleed days 4 and 8 if clinically indicated. Transcranial Doppler was performed daily, and CT angiography was performed for clinical purposes. Patients who developed DCI were treated with hypertensive hypervolemic therapy. Endovascular treatment, including intra-arterial verapamil hydrochloride injection and/or angioplasty, was attempted on patients who developed DCI and did not respond to medical treatment. We did not increase fraction of inspired oxygen as a modality of prevention or treatment of DCI.15

Data analysis

We compared demographics, risk factors and comorbidities, findings on initial and follow-up assessments, and the exposure to hyperoxia with the development of DCI, and also with poor outcome. For univariate analysis, we used Pearson's χ2 test with the exact method, Student t test and the Mann–Whitney U test as appropriate. Multiple logistic regression analysis was performed after examining collinearity to estimate the independent contributions of variables to DCI and poor outcome. Variables with a p value <0.1 by univariate analysis were included in the multivariable model as candidate variables, and then removed by a backward stepwise selection procedure. We further performed all analysis using a forward selection procedure. A two-tailed p value of <0.05 was considered to indicate a significant difference in all statistical analysis. SPSS for Windows V.13.0 (SPSS, Chicago, Illinois) was used for all statistical analyses.


During the study period, 1140 patients with aneurysmal SAH were identified. Of those, 888 patients did not meet inclusion criteria and were excluded. Thus, the remaining 252 patients were included for the final analysis. Age was the only baseline characteristic that was different between 252 included and 888 excluded subjects, included patients being older (mean±SD, 56.7±13.9 years vs 53.7±14.7 years; p=0.005). SAH severity was more severe in included patients compared with excluded patients: GCS (median (IQR), 8 (6–10) vs 15 (12–15); p<0.001); Hunt–Hess grade (grade 1, 4% vs 29%; grade 2, 4% vs 22%; grade 3, 25% vs 30%; grade 4, 43% vs 9%; grade 5, 24% vs 10%; p<0.001); modified Fisher scale (group 0, 0% vs 4%; group 1, 10% vs 33%; group 2, 11% vs 9%; group 3, 41% vs 37%; group 4, 38% vs 17%; p<0.001); and APACH II score (mean±SD, 17.7±6.7 vs 10.6±7.2; p<0.001) (see online supplementary table S1).

Of the 252 included patients, 179 (71%) were women. A mean age was 56.7±13.9 years. The median time from symptom onset to the initial PaO2 was 16.4 h (IQR, 10.3–25.9 h). The median frequency of PaO2 measurements, within 14 days after symptom onset, was 23.0 (IQR, 15–31). Initial Hunt–Hess grade ranged from grade 1 to grade 5, with high grades predominant (table 1).

Table 1

Patient characteristics

Hyperoxia and baseline characteristics

Among demographics, risk factors, underlying comorbidities and initial findings, there was no variable significantly related to hyperoxia (see online supplementary table S2).

Hyperoxia and DCI

Table 1 shows demographics, risk factors or comorbidities, findings on initial assessments, findings on follow-up assessments and an exposure to hyperoxia, assorted by DCI. DCI was observed in 97/252 (38.5%) patients, and the median time interval from symptom onset to the occurrence of DCI was 6.0 days (mean±SD, 6.4±3.1 days; IQR, 4–8 days) (figures 2 and 3). Table 2 shows factors associated with DCI: highest grade of modified Fisher Scale (p=0.012), global cerebral oedema on follow-up CT (p=0.004), intracranial pressure crisis (p=0.009), pneumonia and/or pulmonary oedema (p=0.030), sepsis (p=0.046), and hyperoxia (p=0.001). Rebleeding was marginally associated with DCI (p=0.070). Increase in oxygen burden was related to high proportion of patients with DCI (figure 4). In multiple logistic regression analysis, modified Fisher Scale grade 3 (p=0.040) and grade 4 (p=0.005), global cerebral oedema on follow-up CT (p=0.004), pneumonia and/or pulmonary oedema (p=0.010) and hyperoxia (p<0.001) were independently associated with DCI.

Table 2

Factors associated with the delayed cerebral ischaemia

Figure 2

Average daily PaO2 on each postbleed day. PaO2 according to the development of (A) delayed cerebral ischaemia (DCI) and (B) poor outcome. Histogram shows the number of patients who developed DCI.

Figure 3

Oxygen burden according to severity of SAH. (A) In all grades of modified Fisher scale, oxygen burden was higher in patients with delayed cerebral ischaemia (DCI) than in those without DCI. (B) Oxygen burden was higher in patients with poor outcome than in those without poor outcome.

Figure 4

Relationship of oxygen doses to delayed cerebral ischaemia (DCI) and poor outcome. The proportion of patients with DCI and poor outcome increased in high oxygen burden.

Hyperoxia and poor outcome

mRS At 3 months was available in 202 (80.2%) patients (see online supplementary table S3). Poor outcome occurred in 108/202 (53.5%) patients. Table 3 shows the factors associated with poor outcome: older age (p<0.001), APACHE II score (p=0.001), higher Hunt–Hess grade (p<0.001), aneurysm size (p=0.006), rebleeding (p=0.011), global cerebral oedema on follow-up CT (p=0.001), pneumonia and/or pulmonary oedema (p=0.004), and sepsis (p=0.022). Exposure to hyperoxia was marginally associated with poor outcome (p=0.087). Increase in oxygen burden was related to high proportion of patients with poor outcome (figure 4). In multiple logistic regression analysis, older age (p=0.001), higher Hunt–Hess grade (p=0.012), aneurysm size (p=0.024), rebleeding (p=0.035) and hyperoxia (p=0.042) independently predicted poor outcome.

Table 3

Factors associated with poor outcome


Our main finding is that exposure to hyperoxia was associated with an increased risk for DCI and poor outcome after SAH. As expected, given our selection of high-grade SAH patients, we had a high rate of DCI (38.5%) and poor functional outcome (53.5%). Exposure to hyperoxia was related to approximately three times the risk of DCI and twice the risk of poor 3-months outcome. This relationship was independent of the SAH severity and other comorbidities.

Currently, no guidelines for oxygen therapy in SAH exist. In this study, during the first week after SAH, three quarters of patients were exposed to an average PaO2≥123 mm Hg, and a quarter of patients were exposed to an average PaO2≥173 mm Hg. As seen in our cohort, a recent study has shown that the risk of being exposed to hyperoxia is common in critically ill patients, and such condition was not addressed with any adjustments in ventilator settings.17

Hyperoxia causes oxidative stress by generation of reactive oxygen species and is known to cause lung injury.8 ,18–20 Detrimental cerebral effects of hyperoxia have been reported. Oxidative stress may cause brain injury through unavailability of nitric oxide, inflammation and neuronal necrosis.21 ,22 Hyperoxia decreases the availability of nitric oxide through reduced basal production, oxidative quenching and decreased release from the S-nitrosohemoglobin binding.21 This contributes to vasoconstriction and decrease in cerebral blood flow.10 ,23–25 Additionally, mitochondria-derived reactive oxygen species can cause neuronal inflammation.26 ,27 Hyperoxic reperfusion after global cerebral ischaemia has been reported to promote neuronal inflammation and long-term neuronal loss.22 At high pressure, hyperoxia has caused paralysis due to neuronal necrosis in rats.28 Furthermore, oxidative stress has a role in the formation and rupture of cerebral aneurysm through inflammation, activation of matrix metalloproteinases, and lipid peroxidation.29

Recently, hyperoxia has been shown to be associated with poor outcomes in other acute brain injury populations. A large multicentre cohort study found that exposure to hyperoxia was common following resuscitation from cardiac arrest, and associated with increase in hospital mortality.2 Retrospective studies in patients with traumatic brain injury have also found an association between worse functional outcomes and higher mortality in patients exposed to hyperoxia.3 ,30 In both studies, oxidative stress due to reactive oxygen species has been postulated as the pathomechanism of the aggravation of initial brain injury.22 Conversely, studies have found that hyperoxia may be of therapeutic benefit for patients with brain injury. A pilot study showed that high-flow oxygen therapy via facemask transiently improved neurological deficits and MRI abnormalities compared with room-air therapy in patients with acute ischaemic stroke.6 A magnetic resonance spectroscopy study also found that normobaric oxygen therapy decreased lactate level and preserved N-acetyl-aspartate peak in patients with acute ischaemic stroke.7 Other studies showed that hyperoxia was related to improvement of cerebral metabolism and reduction of intracranial pressure in patients with traumatic brain injury.4 ,5 At present, the role of hyperoxia remains controversial in patients with acute brain injury. Future clinical trials are warranted to address this issue.

DCI is an important complication after SAH, increasing the mortality and morbidity for survivors.16 ,31 Complex pathophysiology is related to the development of DCI, including vasospasm, neurovascular inflammation and microcirculatory dysfunction.32 Narrowing of the lumen of large or small vessels may cause ischaemia distal to the involved sites. Effects of hyperoxia on the vasoconstriction and reduction in cerebral blood flow may contribute to DCI.10 ,23 Oxygen-free radicals may also be involved in neurovascular inflammation through activation of microglia and astrocytes.22 ,26 ,27 Furthermore, hyperoxia can impair cerebral microcirculation by disturbing fibrinolysis and activation of the coagulation system.33 Vasoconstriction, inflammation and impaired cerebral microcirculation have all been postulated to be important processes for the development of DCI.

Our study has limitations. First, this was an observational study, and findings should be interpreted cautiously. Despite our findings of hyperoxia's association with DCI, hyperoxia may be a surrogate marker rather than a cause in the development of DCI. In a retrospective study, potential interactions among variables may not be completely excluded by multivariate analysis. Second, we chose to include only patients who were intubated and had PaO2 examined. Patients included in our study were older and had more severe symptoms than those who were excluded. This is a potential source of selection bias and it is unclear if our findings can be generalised to the low-grade SAH patient. Third, due to the design, we were also not able to obtain detailed information on ventilator settings, such as modes and fraction of inspired oxygen, shunting and metabolic status. However, regardless of the ventilator settings, we believe that the PaO2 levels were the most critical values regarding the development of DCI because the effect of ventilator settings, shunting and metabolic status would most likely work through PaO2 levels. Additionally, measurements of PaO2 were inconsistent across patients in terms of time points and interval. These limitations, due to retrospective nature, should be addressed in future prospective studies. Fourth, a consensus on the definition of hyperoxia in SAH is not available in terms of a level and duration of PaO2. A recently published large multicentre cohort study defined hyperoxia as PaO2 of 300 mm Hg or greater in the first (within 24 h after admission) arterial blood-gas analysis.2 However, their definition of hyperoxia was based on ‘consensus of all authors’, because the precise PaO2 level associated with maximal risk is unknown.2 Moreover, the previous study's population was cardiac arrest victims, whose secondary brain injury would occur mostly within 24 h. On the contrary, in patients with SAH, secondary brain injury in the form of DCI occurs typically between 3 and 14 days after SAH.16 Another study of patients with traumatic brain injury averaged PaO2 level during the first 24 h of hospital admission and defined hyperoxia as PaO2 levels of 200 mm Hg or greater.3 This definition of hyperoxia, however, was arbitrarily based on the authors’ clinical experiences of rarely seeing PaO2 values greater than 200 mm Hg. Moreover, to our best knowledge, there are no laboratory data to define hyperoxia in SAH population. Thus, we defined hyperoxia as the highest quartile of area under the curve of PaO2 based on our patients’ data. Although hyperoxic PaO2 levels (≥173 mm Hg) of our patients were moderate compared with PaO2 levels (≥300 and ≥200 mm Hg) of the previous studies, we believe that continuous exposure to excess oxygen could also be harmful. When our patients were exposed to higher oxygen doses, the risks of developing DCI and poor outcome further increased. We favour avoiding hyperoxia, but a cutoff value of 173 mm Hg based on our retrospective study cannot be recommended. Further studies remain to be published to address the cutoff value.

In conclusion, our study suggests that exposure to hyperoxia may be associated with DCI and poor 3- month outcome in SAH patients requiring mechanical ventilation. Exposure to excess oxygen after SAH may represent a modifiable factor for morbidity and mortality in this population. Further studies are necessary to validate our findings and explore its clinical implications.


We are grateful to Younsuk Lee, MD, PhD, Institute of Statistical Research and Department of Anesthesiology, Dongguk University College of Medicine, Goyang, Gyeonggi-do, Republic of Korea, for statistical assistance.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors S-BJ, HAC and KL contributed to the article by participating in the concept and design. S-BJ, HAC, NB, JMS, HL, JC, ESC, SAM and KL: acquisition of data. S-BJ, HAC, NB and KL: analysis and interpretation of data. S-BJ, HAC and KL: drafting of the manuscript. S-BJ, HAC, NB, JMS, HL, JC, ESC, SAM and KL: critical revision of the manuscript for intellectual content.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval The institutional review board of New York Presbyterian Hospital.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Linked Articles