Abstract

The beneficial effects of oxygen are widely known, but the potentially harmful effects of high oxygenation concentrations in blood and tissues have been less widely discussed. Providing supplementary oxygen can increase oxygen delivery in hypoxaemic patients, thus supporting cell function and metabolism and limiting organ dysfunction, but, in patients who are not hypoxaemic, supplemental oxygen will increase oxygen concentrations into nonphysiological hyperoxaemic ranges and may be associated with harmful effects. Here, we discuss the potentially harmful effects of hyperoxaemia in various groups of critically ill patients, including postcardiac arrest, traumatic brain injury or stroke, and sepsis. In all these groups, there is evidence that hyperoxia can be harmful and that oxygen prescription should be individualized according to repeated assessment of ongoing oxygen requirements.

1. Introduction

Oxygen is the third most abundant element in the universe and essential for life, but it was only officially “discovered” in the early 1770s separately by the British-born theologian, Joseph Priestly, and the Swedish apothecary, Carl Scheele [1, 2]. It took another few years for its role in respiration to be identified by the French chemist, Antoine-Laurent Lavoisier, who also gave it its name [1, 2]. Introduced into anaesthetic practice in the 1930s, oxygen is now one of the most widely used “drugs” in hospitalized patients. In a point-prevalence study conducted in 40 intensive care units (ICUs) in Australia and New Zealand in 2012, 59% of patients were receiving mechanical ventilation; among those not receiving mechanical ventilation, 86% were receiving oxygen via nasal cannulas, facial masks, or noninvasive ventilation [3]. However, although oxygen therapy clearly has important benefits in many patients, we have become increasingly aware of the potential harmful effects of high oxygenation concentrations in blood and tissues (Figure 1). In a retrospective study comparing mortality rates and PaO2 levels in mechanically ventilated ICU patients, de Jonge et al. reported a U-shaped relationship with increased mortality rates at low and high PaO2 [4]. The potential risks of hyperoxia, with a focus on recent clinical evidence in specific groups of critically ill patients (Table 1), will be the emphasis of this short narrative review.

2. Effects of Hyperoxia

Adequate cellular oxygenation is essential for normal cell function, and a low SaO2 is life-threatening, especially in acute conditions. Providing supplementary oxygen will increase oxygen delivery in hypoxaemic patients, thus supporting cell function and metabolism and limiting organ dysfunction. However, in patients who are not hypoxaemic, supplemental oxygen will increase oxygen concentrations into hyperoxaemic ranges. Although human beings may be exposed to hypoxia, for example, when at altitude or as a result of pulmonary disease, we are never exposed to hyperoxia, so that supplying extra oxygen to individuals who are not hypoxaemic is always a “nonphysiological” event.

Hyperoxia is associated with multiple effects in different organ systems. It can directly damage tissues via the production of reactive oxygen species (ROS) in excess of physiological antioxidant defence capabilities [5], leading to increased cell death by apoptosis and increased release of endogenous damage-associated molecular pattern molecules (DAMPs) that stimulate an inflammatory response, notably in the lungs [6] and vasoconstriction, likely as a result of reduced nitric oxide levels [7]. Orbegozo Cortés et al. recently reported that normobaric hyperoxia in healthy volunteers was associated with reduced capillary perfusion as assessed using sublingual side-stream dark field (SDF) video-microscopy [8]. It has been suggested that these vasoconstrictive effects may provide a means of protecting cells from the harmful effects of high PaO2 [9].

2.1. After Cardiac Arrest or Myocardial Infarction

Given the associated vasoconstriction and increased ROS release, hyperoxia may be particularly harmful after cardiac arrest [10]. Experimental and observational data have given conflicting results regarding the effects of hyperoxia in this setting [10]. In a retrospective analysis of data from 6326 postcardiac arrest patients admitted to ICUs in 120 US hospitals between 2001 and 2005, patients with hyperoxia (defined as PaO2 of ≥300 mmHg) on arrival in the ICU had higher mortality rates than those with normoxia or hypoxia; in multivariable analysis, hyperoxia exposure was an independent predictor of in-hospital death (odds ratio 1.8 [95% CI 1.5–2.2], ) [11]. In an analysis of a registry database, severe hyperoxia (as identified by a PaO2 > 300 mmHg) was associated with increased mortality in postcardiac arrest patients, whereas moderate or “probable” hyperoxia (PaO2 101–299 mmHg) was not [12]. In a retrospective cohort study, patients managed according to a conservative oxygen approach after cardiac arrest, targeting an pulse oximetry oxygen saturation (SpO2) of 88–92% had shorter lengths of ICU stay, although there were no differences in neurological outcomes [13]. In a multicentre trial conducted in 441 patients with ST-elevation myocardial infarction, 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. Patients treated with oxygen had 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 [14]. These results do not support the use of routine supplemental oxygen after cardiac arrest or myocardial infarction. A randomized multicentre study is ongoing in Sweden aiming to randomize 6,600 patients with suspected acute myocardial infarction and SpO2 ≥ 90% to either 6 L/min of supplemental oxygen for 6 to 12 hours or room air [15].

2.2. In Traumatic Brain Injury and Stroke

Reduced cerebral oxygenation after brain injury is associated with impaired mitochondrial function and reduced metabolic rate and may be associated with an increased risk of secondary brain damage [16, 17]. Treating such patients with hyperoxia may, therefore, be expected to have beneficial effects on outcomes. However, clinical studies have given conflicting results. In a retrospective study of more than 3000 patients with traumatic brain injury (TBI), hypoxaemia, and extreme hyperoxaemia (PaO2 > 487 mmHg) on admission were both associated with worse outcomes; a PaO2 value of 110–487 mmHg was considered optimal in this study [18]. Similar findings were reported by a more recent retrospective study in 1547 patients with TBI, with both low and high admission PaO2 levels independently associated with worse outcomes [19]. In a long-term outcomes study after TBI, although there was a significant association between hyperoxaemia and a decreased risk of 6-month mortality in univariate analysis, in multivariable analysis, hyperoxaemia was not independently associated with outcome [20]. In a small randomized trial, 68 patients with severe TBI received either 80% or 50% oxygen via mechanical ventilation in the first 6 hours after the TBI. Patients in the hyperoxia group had better outcomes at 6 months as assessed using the Glasgow Outcome Scale than patients in the normoxia group [21]. A planned larger study to compare treatment with an FiO2 of 0.4 or 0.7 in patients with TBI was terminated because of slow recruitment (NCT01201291). Interestingly, in a prospective study of 30 patients monitored with a brain tissue oxygen sensor, Vilalta et al. reported that a hyperoxia challenge was associated with improved cerebral metabolism only in patients with reduced metabolism at baseline [22].

Evidence from studies in patients with stroke is also conflicting. Lang et al. reported no effect on 3-month neurological outcomes or mortality of moderate hyperoxaemia during the first 24 hours after ICU admission in patients after subarachnoid haemorrhage [31], and Young et al. similarly reported no association between worst PaO2 in the first 24 hours after ICU admission and mortality in patients with acute ischaemic stroke [36]. However, other observational studies have reported detrimental effects on short and longer term outcomes in different groups of stroke patients [28, 30]. A study randomizing patients to room air or supplemental oxygen administered at 30–45 L/min for 8 hours was terminated early because of more deaths in the oxygen group (NCT00414726). In a pilot study comparing oxygen supplementation for 72 h via nasal cannulae with room air in 289 patients with acute stroke, there was a small improvement in neurological recovery at one week [37], but there were no significant differences between the groups at 6 months [38], findings supported by the larger Stroke Oxygen Study in more than 8000 patients [39].

2.3. In Sepsis and Septic Shock

The use of hyperoxia in patients with sepsis is also controversial [40]. Sepsis is already associated with increased formation of ROS, believed to play a role in the tissue damage and organ dysfunction seen during sepsis. Hyperoxaemia is known to stimulate release of ROS and could therefore further worsen organ function in these patients. In a rat caecal ligation and puncture model, hyperoxia was associated with increased inflammatory cytokine release and organ dysfunction compared to normoxia [41]. However, in other animal models of sepsis, hyperoxia has been associated with improved haemodynamics and anti-inflammatory effects [42, 43]. And in a sheep model of sepsis, we showed that hyperoxia was associated with better haemodynamics and organ function compared to normoxia (unpublished data). In experimental human endotoxaemia, hyperoxia had no effect on levels of inflammatory mediators [44], and in a small observational study of patients with suspected sepsis in the emergency department, there were no significant differences in mortality rates between hyperoxic and normoxic patients [32]. A clinical trial in patients with sepsis randomized to hyperoxia or normoxia and hypertonic or isotonic saline in a 2 × 2 factorial design was stopped prematurely because of increased mortality rates in the hyperoxia and hypertonic saline arms (NCT01722422). Two randomized studies, one comparing supplemental oxygen titrated to different PaO2 targets (105–135 mmHg versus 60–90 mmHg, O2-ICU study, NCT02321072) and one comparing supplemental oxygen at 15 L/min to no supplemental oxygen (NCT02378545), are currently ongoing and should provide final answers as to whether or not patients with sepsis may benefit from hyperoxia.

2.4. In Mixed ICU Patients

Use of liberal oxygen therapy is frequent in critically ill patients [45] and severe hyperoxaemia (PaO2 > 200 mmHg) is associated with higher mortality rates [35]. Interestingly, in three ICUs in the Netherlands, more than 70% of ICU patients had PaO2 levels that were higher than the upper limits identified by the ICU clinicians treating them [46]. Several studies have now compared so-called conservative oxygen strategies targeting lower PaO2 or SpO2 values with conventional oxygen administration. Panwar et al. compared target SpO2 values of 88–92% and ≥96% in 103 ICU patients and reported no significant differences between the groups in terms of organ function or ICU and 90-day mortality [33]. In the Oxygen-ICU study, which was terminated early, 434 patients were randomized to receive supplemental oxygen to maintain PaO2 at 70–100 mmHg (SpO2 94–98%) or to be managed conventionally allowing PaO2 to reach 150 mmHg (SpO2 97–100%) [34]. Patients in the conservative group had lower ICU mortality than those in the conventional group (relative risk 0.57 [95% CI 0.37–0.9]; ).

3. Conclusions

For many years, the known risks of hypoxia and less known adverse effects of hyperoxia have led to many patients receiving liberal oxygenation to avoid hypoxaemia at all costs. Although good quality data remain limited, results from the latest clinical studies seem to suggest that hyperoxaemia may be associated with worse outcomes in some critically ill patients (Table 1). The trend is therefore moving towards a more conservative approach to oxygenation aimed at maintaining SpO2 targets at 95–97%, although the optimal PaO2 level has not yet been defined and will likely change during the course of a patient’s illness. Indeed, there may be a time window during which patients may benefit from higher oxygen levels [47]. Further well-designed randomized controlled trials in carefully selected groups of patients may help provide some definitive answers to these questions. As with other areas of intensive care management, oxygen therapy should be individualized. Patients who are hypoxaemic clearly need to receive supplemental oxygen, but ongoing requirements need to be reassessed on a regular basis to limit any risks associated with hyperoxia.

Competing Interests

The authors declare that there are no competing interests regarding the publication of this paper.