Normobaric hyperoxia after stroke: a word of caution

KEYWORDS:

• Stroke
• hyperoxia
• outcome
• oxygen
• hyperoxemia

1. Introduction

Oxygen is one of the most abundant elements in the universe and is essential for life as it is the main component of water and is involved in mitochondrial respiration. In the medical setting, oxygen therapy was introduced during surgical procedures almost a century ago and is now widely used in hospitalized patients or in emergency medicine for different medical conditions. Oxygen therapy is the first-line intervention in acute hypoxemia. Indeed, up to 85% of intensive care unit (ICU) patients who are not treated with mechanical ventilation receive supplementary oxygen via different interfaces, including nasal cannulas and facial masks [1]. Nevertheless, target values for arterial oxygen tension (PaO₂) or hemoglobin saturations (SaO₂) should not be ‘as high as possible,’ which has long been considered optimal. Excessive oxygen therapy may expose patients to high PaO₂ levels, which are associated with potential harm.

2. Harmful effects of hypoxia and hyperoxia

2.1. In health

Although low oxygen concentrations may contribute to hypoxic cell damage, especially in acute conditions, providing oxygen therapy in hyperoxemic ranges can also alter cellular function and potentially induce organ dysfunction. Human beings may occasionally be exposed to hypoxia, for example at high altitude or in case of pulmonary disease, and will react to maintain adequate tissue oxygenation by increasing cardiac output and oxygen extraction rate [2]. By contrast, humans are never exposed to hyperoxia in the absence of oxygen therapy. In case of hyperoxia, the production of reactive oxygen species (ROS) can exceed the natural antioxidant capabilities and promote cell death by lipid and DNA peroxidation, which eventually induce apoptosis [3]. Moreover, hyperoxia is associated with an increased release of endogenous damage-associated molecular pattern molecules (DAMPs) that stimulate an inflammatory response, notably in the lungs, where it can cause tissue injury and nitrogen washout, leading to alveolar collapse and de-recruitment and promoting organ dysfunction. As a result of increased tissue oxygen concentrations, a reduction in nitric oxide (NO) promotes vasoconstriction that could protect cells from the harmful effects of high PaO₂ but at the same time reduce capillary density and promote heterogeneity of the microvascular flow [4].

2.2. In acute illness

Is excessive oxygen harmful also during acute illness? In ICU patients receiving mechanical ventilation, observational studies have reported a U-shaped relationship, with increased mortality rates at low and high PaO₂ [5]. Another recent study has shown conflicting effects on patient outcome when a ‘liberal’ oxygen therapy strategy (i.e. SaO₂ >96–97%) was compared to a ‘conservative’ strategy (i.e. SaO₂ 90–93%) [6]; however, the ‘liberal’ group was not specifically exposed to hyperoxia. More recently, a randomized clinical trial (RCT) assigned patients with sepsis to a factorial design according to oxygen (hyperoxia vs. normoxia) and fluid (hypertonic or isotonic saline) therapy; this study was stopped prematurely because of increased mortality rates in the hyperoxia arm (and in the hypertonic saline arm) [7].

Thus, this ‘double-edged sword’ character of oxygen therapy is particularly crucial in critically ill patients, in whom acute illness (e.g. major surgery, trauma, and life-threatening infection) can generate an ischemia/reperfusion injury, which may be worsened by supranormal oxygen concentrations, resulting in increased ROS production and worsening of cellular metabolism due to impaired mitochondrial function.

2.3. In acute brain injury

Although hyperoxia may be harmful in critically ill patients as seen earlier, it may have beneficial effects in patients with acute brain injury. Indeed, progressive dysfunction of cerebral metabolism is a well-recognized consequence of different forms of brain injury, including traumatic brain injury (TBI), anoxic injury, and subarachnoid hemorrhage (SAH) [8]. Moreover, in these conditions, increased intracranial pressure, altered systemic hemodynamics, and dysfunction of cerebral autoregulation and microcirculation can result in brain hypoxia [9]. Thus, the majority of cerebral glucose is metabolized anaerobically and converted into lactate, which cannot be easily metabolized by neurons because of the mitochondrial dysfunction induced by tissue hypoxia [10]. It would then be logical to increase oxygen concentrations and, by so doing, reduce tissue hypoxia and promote a shift towards aerobic metabolism, as the kinetics of mitochondrial redox enzymes are enhanced when brain tissue oxygen pressure (PbtO₂) values are high.

In patients with TBI, normobaric hyperoxia was associated with equivocal effects on tissue oxygenation in hypoperfused brain regions but was clearly associated with enhanced excitotoxicity [11]. Both hypoxemia and hyperoxemia on admission were independently associated with worse outcomes [12]. Nevertheless, in a small RCT in which 68 patients with TBI were included, 6-month outcome was better in those who received 80% oxygen during mechanical ventilation than in those who received 50% oxygen [13]. In post-anoxic brain injury following cardiac arrest, hyperoxemia (PaO₂ >300 mmHg) was associated with poor neurological outcome [14], although these findings were not unequivocal and no specific threshold for a ‘toxic’ PaO₂ was identified [15]. Finally, after SAH, exposure to hyperoxia during the early phase after hospital admission was associated with delayed cerebral ischemia and poor outcome [16], although this was not observed in patients with moderate levels of hyperoxemia (PaO₂ >150 mmHg) [17].

2.3.1. In acute ischemic stoke

Ischemic stroke is the most common form of acute brain injury and differs from TBI, SAH, or anoxic injury in several aspects, including the predominance of male sex, the relatively old age of the patients, the focal injury, and the pathophysiology, i.e. local ischemia due to the occlusion of a cerebral vessel. As in the experimental setting [18,19], observational clinical studies have shown controversial results with the use of hyperoxia in ischemic stroke. In a retrospective study including 2894 patients treated with mechanical ventilation for different forms of stroke (ischemic, hemorrhagic, and SAH), hyperoxemia, defined as PaO₂ >300 mmHg, was associated with increased in-hospital mortality [20]. In 2643 patients with acute ischemic stroke, Young et al. observed no association between the lowest or the highest PaO₂ in the first 24 h after ICU admission and mortality [21]. Nevertheless, it remains difficult to compare experimental and clinical data for several reasons: animal studies typically use only large-vessel occlusion while patients suffering from ischemic stroke have different forms of vascular injury, with different locations and variable collateral supply. Also, in animal studies, therapeutic interventions are started in the very early phase after the brain injury, whereas in clinical practice specific therapies (thrombolysis and endovascular procedures) may not be administered until a few hours after the onset of symptoms, depending on the speed of recognition of the disease and call for medical intervention. Finally, mechanical ventilation is usually applied in animal studies but is used less commonly in patients with stroke. Moreover, the few human studies that have evaluated the effects of hyperoxia on neurological recovery after ischemic stroke have several limitations. First, in some studies, only patients who did not receive adequate reperfusion therapy were included, which would limit the clinical effects of hyperoxia because of persistent vascular occlusion and the absence of possible local oxygen delivery. Second, only the first or the worst arterial blood gas analysis was taken into account in the final analysis instead of a mean value over the first 24–48 h, which could better reflect oxygen exposure.

Only a few prospective studies have specifically evaluated the effects of hyperoxia after ischemic stroke. Singhal et al. showed that high-flow oxygen therapy (45 L/min) was associated with a transient improvement in clinical deficits and magnetic resonance imaging abnormalities in selected patients (n = 9) with acute ischemic stroke not eligible for thrombolysis when compared to untreated patients (n = 7) [22]. In another study, the same group showed that the increase in lesion volume was smaller during high-flow oxygen therapy (n = 11) than when breathing ambient air (n = 8), at least in the very early phase of therapy [23]. In an RCT of 40 patients with ischemic stroke in the anterior circulation who were not eligible for thrombolysis, normobaric hyperoxia (oxygen flow of 10 L/min) did not improve outcome when compared to no oxygen administration [24]. The Normobaric Oxygen Therapy in Acute Ischemic Stroke Trial (NCT00414726), which evaluated high-flow oxygen (30–45 L/min for 8 h within the first 9 h after stroke onset), was prematurely halted because of the higher mortality rate in the high-flow oxygen group compared to  the control group. Most deaths were due to early withdrawal of life support in the high-flow oxygen group probably because of the lack of any clinical improvement despite hyperoxia therapy. Finally, a large RCT conducted in the United Kingdom (n = 8003) showed similar outcomes in patients receiving continuous oxygen administration for 72 h compared to those treated only during the night (i.e. when hypoxemia is more frequent) or those treated only when indicated [25]. Nevertheless, oxygen was administered at ‘low dose’ and measured oxygen saturations were very similar among groups.

Some authors still support the use of normobaric hyperoxia as a potential neuroprotective strategy in stroke, in particular in patients admitted very early, with large vessel occlusion (i.e. large penumbra area) and only in case of reperfusion therapy. This idea needs to be carefully validated as most data available today suggest harmful effects of hyperoxia in stroke. Moreover, several questions remain open. First, if normobaric hyperoxia is used, should oxygen be administered via a mask or using mechanical ventilation? Second, should oxygen therapy consist of a fixed high flow or should it be titrated according to arterial or – even better – cerebral oxygen concentrations (e.g. PbtO₂ or noninvasive cerebral oxygen saturation with near-infrared spectroscopy)? Third, which method should be used to assess the beneficial/harmful effects of hyperoxia: clinical evaluation? Neuroimaging? Or the biological assessment of ROS production? Fourth, as the pathophysiology of brain damage after ischemic stroke may be ongoing for 24–48 h after vascular occlusion, should we limit oxygen administration to a few hours or to longer periods? Fifth, other variables, such as hypocapnia, may also be important determinants of cerebral perfusion and therefore of poor outcome in this setting; this was not considered in the human studies evaluating hyperoxia after ischemic stroke. Finally, the mechanisms underlying the potential benefit from hyperoxia after ischemic stroke should be further clarified. Although high-flow oxygen can result in an early increase in cerebral blood flow velocity, this phenomenon is secondary to peroxynitrite formation and a dysregulation of endogenous NO availability, which could potentially impair cerebral autoregulation and tissue perfusion in case of systemic changes (e.g. reduction in blood pressure, which is typically applied in patients with ischemic stroke who are treated with thrombolysis).

3. Conclusions

The known adverse effects of hyperoxia and the trend towards a more conservative approach to oxygenation in critically ill patients deserve further clinical investigation to understand the role of high-flow oxygen therapy in patients with ischemic stroke. As with other areas of intensive care management, oxygen therapy should be individualized and both beneficial and detrimental effects should be assessed in all patients.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Additional information

Funding

This paper was not funded.