key: cord-0017931-u7nq0by4
authors: Demiselle, Julien; Calzia, Enrico; Hartmann, Clair; Messerer, David Alexander Christian; Asfar, Pierre; Radermacher, Peter; Datzmann, Thomas
title: Target arterial PO(2) according to the underlying pathology: a mini-review of the available data in mechanically ventilated patients
date: 2021-06-02
journal: Ann Intensive Care
DOI: 10.1186/s13613-021-00872-y
sha: 9a0edb96c5211b4669b12e10237b755b745eaddc
doc_id: 17931
cord_uid: u7nq0by4

There is an ongoing discussion whether hyperoxia, i.e. ventilation with high inspiratory O(2) concentrations (F(I)O(2)), and the consecutive hyperoxaemia, i.e. supraphysiological arterial O(2) tensions (PaO(2)), have a place during the acute management of circulatory shock. This concept is based on experimental evidence that hyperoxaemia may contribute to the compensation of the imbalance between O(2) supply and requirements. However, despite still being common practice, its use is limited due to possible oxygen toxicity resulting from the increased formation of reactive oxygen species (ROS) limits, especially under conditions of ischaemia/reperfusion. Several studies have reported that there is a U-shaped relation between PaO(2) and mortality/morbidity in ICU patients. Interestingly, these mostly retrospective studies found that the lowest mortality coincided with PaO(2) ~ 150 mmHg during the first 24 h of ICU stay, i.e. supraphysiological PaO(2) levels. Most of the recent large-scale retrospective analyses studied general ICU populations, but there are major differences according to the underlying pathology studied as well as whether medical or surgical patients are concerned. Therefore, as far as possible from the data reported, we focus on the need of mechanical ventilation as well as the distinction between the absence or presence of circulatory shock. There seems to be no ideal target PaO(2) except for avoiding prolonged exposure (> 24 h) to either hypoxaemia (PaO(2) < 55–60 mmHg) or supraphysiological (PaO(2) > 100 mmHg). Moreover, the need for mechanical ventilation, absence or presence of circulatory shock and/or the aetiology of tissue dysoxia, i.e. whether it is mainly due to impaired macro- and/or microcirculatory O(2) transport and/or disturbed cellular O(2) utilization, may determine whether any degree of hyperoxaemia causes deleterious side effects.

Oxygen (O 2 ) is not only the final electron acceptor within the respiratory chain, but also one of the strongest oxidizing molecules [1, 2] . Approximately 1-3% of mitochondrial O 2 consumption is directed towards the production of "reactive oxygen species" (ROS), i.e. the more ATP produced, the more ROS are generated [3, 4] . ROS formation is directly related to the O 2 concentration [5] , so that hyperoxia (i.e. breathing inspiratory O 2 concentrations Open Access *Correspondence: peter.radermacher@uni-ulm.de 2 Institut Für Anästhesiologische Pathophysiologie Und Verfahrensentwicklung, Universitätsklinikum, Helmholtzstrasse 8-1, 89081 Ulm, Germany Full list of author information is available at the end of the article Demiselle et al. Ann. Intensive Care (2021) 11:88 F I O 2 > 0.21) and the consecutive hyperoxaemia (i.e. arterial PO 2 > 100 mmHg) result in a dose-dependent increase of ROS formation [6, 7] . It is noteworthy in this context that the definition of hyperox(aem)ia may vary as well: while some authors use a threshold PaO 2 value of 150 mmHg [8, 9] , others refer to hyperox(aem)ia as PaO 2 ≥ 300 mmHg [10] [11] [12] [13] [14] . The aggravated ROS formation resulting from increased O 2 concentrations is particularly pronounced during ischaemia/reperfusion (I/R) and/or hypoxia/re-oxygenation [5] . Nevertheless, the dichotomy for O 2 holds also true for ROS formation, in that despite their toxic potential ROS are vital players in host defence systems and as signalling molecules [15] .

In line with the potential toxicity of increased ROS formation resulting from supplemental O 2 administration, a meta-analysis of more than 16,000 patients previously concluded that patients with "liberal" oxygenation (defined as transcutaneous, pulse oximetry haemoglobin-O 2 saturation [SpO 2 ] median/range 96/94-99%) "had a dose-dependent increased risk of … mortality", yet found "no significant difference in disability, hospital-acquired pneumonia, or length of hospital stay" [16] . While providing robust data on a large general ICU population, only 8 out 25 studies analysed in this systematic review, however, had included patients undergoing mechanical ventilation. Moreover, the putative impact of the presence or absence of circulatory shock was not discussed at all. Finally, due to the scarcity of the literature available, a conclusive evaluation of some conditions normally necessitating ICU treatment could not be provided either, e.g. trauma-and-haemorrhage and/or traumatic brain injury [16] . Nevertheless, based on this metaanalysis, an expert panel concluded on a "strong recommendation" that when supplemental O 2 therapy is used, SpO 2 > 96% should be avoided in all in-hospital as well as prehospital medical patients. Explicitly, uncomplicated, elective surgical patients were not included because the expert panel had not reviewed the issue of peri-operative hyperoxia and surgical site infection, respectively [17] .

On the other hand, more recently, in a population of mechanically ventilated ICU patients, a "conservative" oxygenation target (defined as SpO 2 < 97%) yielded no benefit when compared to a "usual" oxygenation group [18] , and even suggested that "usual (liberal) oxygen therapy might be preferred" in the subgroup of patients with sepsis [19] . Therefore, given the U-shaped relation between mortality/morbidity and PaO 2 [20] , this mini-review will discuss the questions i) whether there is an "ideal" PaO 2 for ICU patients, and in particular; ii) if present, whether it possibly differs according to the underlying pathology. We will only discuss the available clinical data, since the vast majority of experimental models lack standard ICU care, which necessarily limits their translational value. Moreover, as far as possible from the data reported, we will focus on data from mechanically ventilated vs. spontaneously breathing patients. Finally, we will try to evaluate the possible impact of the presence or absence of circulatory shock, a condition where "administration of oxygen should be started immediately to increase oxygen delivery" [21] . This review will discuss pertinent clinical studies on general ICU populations and in the emergency department (ED), acute respiratory distress syndrome (ARDS), sepsis and septic shock, trauma and haemorrhage, traumatic brain injury, cardiopulmonary resuscitation (CPR) and post-cardiac management, and peri-operative hyperoxia. Last, the role of SpO 2 vs. PaO 2 measurements for the monitoring of oxygenation will be addressed.

A retrospective multicentre analysis of patients with an ICU stay > 24 h showed that the "hyperoxaemia dose", defined as the time integral of supraphysiological PaO 2 (> 100 mmHg) was associated with mortality in ICU patients. Interestingly, however, no dose-response relationship could be established [22] . In line with these findings, a retrospective analysis of more than 25,000 mechanically ventilated ICU patients found that the proportion of time spent at 95 ≤ SpO 2 ≤ 99% was associated with the lowest odds ratio for mortality, while both SpO 2 ≤ 94 and = 100% coincided with increased mortality [23] . Similarly, an observational study analysing large ICU data bases of a total of 35,287 patients identified 94 ≤ SpO 2 ≤ 98% as the optimal range with respect to survival [24] . In fact, targeting this interval, i.e. "conservative" oxygen therapy (PaO 2 = 70-100 mmHg or SpO 2 = 94-98%) vs. "standard" treatment (PaO 2 ≤ 150 mmHg or SpO 2 ≤ 98%) in a general ICU population of 434 patients with an expected length of stay of ≥ 72 h, had previously been associated with significantly reduced mortality (11.6 vs. 20.2%), de novo occurrence of shock (3.7 vs. 10.6%), liver failure (1.9 vs. 6.4%), and bacteraemia (5.1 vs. 10.1%). However, while the originally planned sample size had been 660 patients, the study was stopped early due to difficulties in enrolment after inclusion of 480 patients [25] . Moreover, only 2/3 of the patients needed mechanical ventilation, and only 30% of the patients presented with shock at inclusion [25] .

However, another retrospective single-centre study in general ICU patients mechanically ventilated for at 7 days did not show any association between in-hospital mortality and time-weighted PaO 2 > 120 mmHg [26] . Furthermore, other authors had even reported nadir mortality in a general ICU population at a mean PaO 2 over the total ICU length of stay ≈ 120-150 mmHg, while exposure to PaO 2 > 200 mmHg was indeed associated with increased mortality [27] .

An observational cohort study analysed the outcome of patients and exposed to ED hyperoxia. The study included 688 out of a total of 3525 patients already mechanically ventilated in the ED, who were all normoxic (60 ≤ PaO 2 ≤ 120 mmHg) on day 1 of their ICU stay. While ED normoxia was present in 50.9% of the patients, ED hyperoxia as defined as a PaO 2 > 120 mmHg occurred in 43.6%. Hospital mortality at day 28 was higher in patients with ED hyperoxia (29.7%) than in those with ED normoxia (19.4%) [28] . Interestingly, survival curves of patients with ED hyperoxia and normoxia, respectively, cleaved at day 4-5 only of hospital stay, i.e. several days after normoxia had resumed.

In conclusion, in general ED and ICU populations, so far, the available clinical data do not suggest that there is any ideal target PaO 2 except for avoiding prolonged exposure (> 24 h) to either hypoxemia or supraphysiological (PaO 2 > 100 mmHg).

The current standard of care of patients with ARDS recommends targeting for arterial oxygenation defined as PaO 2 = 65-75 mmHg and/or arterial haemoglobin O 2 saturation (SaO 2 ) = 90-95%, respectively [29] . In a secondary analysis of 2005 patients of the LUNG SAFE study, Madotto et al. reported an incidence of "hyperoxaemia" (PaO 2 > 100 mm Hg) on day 1 or "sustained hyperoxaemia" (i.e. on day 1 and 2) in 30 and 12% of the patients, respectively. In 66% of these hyperoxaemic patients, "excess O 2 use", i.e. an inspiratory O 2 concentration F I O 2 ≥ 0.6 and PaO 2 > 100 mmHg, was present. The authors concluded that despite being frequently present, hyperoxaemia was mostly only transient. However, neither hyperoxaemia nor excess O 2 use had any effect on patient outcome [30] .

The LOCO2 trial [31] compared "conservative O 2 " therapy for 7 days (target PaO 2 = 55-70 mm Hg and/or SaO 2 as measured by pulse oximetry [SpO 2 ] = 88-92%) to a "liberal O 2 " therapy arm (target PaO 2 = 90-105 mm Hg and/or SpO 2 ≥ 96%). Albeit the patients in the liberal O 2 arm showed significantly higher PaO 2 values than those in the conservative O 2 group without major intergroup overlap, about half of the patients of the conservative O 2 group presented with values beyond the upper target threshold of 70 mmHg. Nevertheless, during the 7 observation days, the conservative O 2 group needed less controlled ventilation, lower PEEP levels, and prone positioning. In contrast, heart rate was higher in this group. However, the conservative O 2 approach did not increase survival at 28 days, and the trial was prematurely stopped because of safety concerns after enrolment of [31] .

Finally, in a total of 2,928 patients with acute hypoxemic respiratory failure, the most recent ICU-HOT trial compared a "lower" (PaO 2 ≈ 60 mmHg) vs. a "higher" (PaO 2 ≈ 90 mmHg) oxygenation target until 90 days after randomization [32] . More than half of the patients needed mechanical ventilation and/or vasopressor support (58 and 54%, respectively). Interestingly, in both study groups, the median/interquartile range PaO 2 values were slightly higher than the target levels (71/67-77 vs. 93/87-99 mmHg, respectively). The two intervention groups did not show any difference with respect to mortality and morbidity.

In conclusion, in patients with ARDS, so far, the available clinical data do not suggest that there is any ideal target PaO 2 except for avoiding prolonged exposure (> 24 h) to either hypoxemia, in particular with respect to longterm neuropsychological sequelae [33, 34] , or supraphysiological (PaO 2 > 100 mmHg).

Hyperoxaemia is associated with systemic vasoconstriction, i.e. might theoretically counteract vasodilationrelated arterial hypotension [5] , and has antibacterial potential [35] . However, in a retrospective analysis of 141 patients (out of a total enrolment of 503 patients) with ventilator-associated pneumonia (VAP), the number of days with hyperoxaemia (as defined by a PaO 2 > 120 mmHg) was an independent risk factor for the occurrence of VAP, rather than suggesting any antibiotic property. It should be noted, however, that these patients also showed several other risk factors of VAP, e.g. more use of proton pump inhibitors, more frequent circulatory shock, more transfusion of packed red blood cells, and more frequent sedation [36] .

The prospective, randomized, controlled HYPER2S trial [37] compared standard therapy with mechanical ventilation with 100% O 2 during the first 24 h after diagnosis of septic shock. Despite a significantly lower SOFA score at day 7, the trial was prematurely stopped due to increased mortality in the treatment arm at days 28 and 90 (43 vs. 36 and 48 vs. 42%, respectively). Due to the premature stop of the trial, this increase in mortality did not reach statistical significance (p = 0.12 and 0.16, respectively) [37] . The detrimental effect of hyperoxia was further mirrored by a higher incidence of serious adverse events (85 vs. 76%, p = 0.02), in particular of ICU-acquired weakness (11 vs. 6%, p = 0.06). Of note, a post hoc analysis according to the Sepsis-3 criteria showed that the hyperoxia-related increase in mortality at day 28 was only present in patients with hyperlactataemia > 2 mmol/L (57% vs. 44%, p = 0.054), while no effect was present in patients with normal lactataemia (25 vs. 23%, p = 0.68) [38] . This finding suggests that the putative hyperoxaemia-related increase in tissue O 2 availability may have led to excess ROS production and consecutive oxidative damage because of a sepsis-induced impaired cellular O 2 extraction.

The above-mentioned "conservative" oxygenation target in the ICU-ROX study was not beneficial either in patients with sepsis: a post hoc analysis of the 251 patients fulfilling these criteria showed no statistically significant inter-group difference when compared to a "usual" oxygenation. Point estimates of treatment effects even consistently favoured the latter [19] . Hence, similar to ARDS, in patients with sepsis/septic shock, there seems to be no ideal target PaO 2 except for avoiding prolonged exposure (> 24 h) to either hypoxemia or supraphysiological (PaO 2 > 100 mmHg).

The blood loss-related reduction of O 2 transport capacity and the fall in cardiac output during trauma and haemorrhage cause a tissue O 2 debt, the rapid repayment of which determines outcome [39] . There is experimental evidence that during haemorrhage, the hyperoxia-related rise in PaO 2 can improve microcirculatory and tissue O 2 availability [40] . Moreover, lung-protective ventilation with an F I O 2 = 1.0 even attenuated organ dysfunction in resuscitated, long-term large animal experiments [41, 42] . Nevertheless, clinical data remain equivocal. In 471 consecutive mechanically ventilated patients with a median injury score (ISS) of 29, Russell et al. reported that there was no association between mortality and maximum PaO 2 in the first 24 h. This was true both for the overall analysis as well as in the subgroup with head trauma (n = 266) [43] .

More recently, Baekgaard et al. reported on 5,912 patients of a French trauma registry (median ISS 16), 32% of whom presented with traumatic brain injury (TBI). Univariate analysis showed higher mortality (12 vs. 9%, p < 0.0001) in patients with hyperoxaemia as defined by PaO 2 > 150 mmHg upon admission (43% of the population). However, propensity score matching for gender, age, prehospital heart rate and systolic blood pressure, temperature, haemoglobin and arterial lactate concentration, use of mechanical ventilation, presence of traumatic brain, initial Glasgow Coma Scale (GCS) score, ISS, American Society of Anesthesiologists physical health class > I, and presence of haemorrhagic shock yielded just the opposite result, i.e. hyperoxaemia was associated with even lower mortality [8] . Most recently, a large retrospective, propensity-matched, data analysis of a total of 864,340 trauma patients (median ISS = 9) investigated the possible association between supplemental O 2 administration in the ED and in-hospital mortality and development of ARDS. In all three patient categories as predefined according to SpO 2 < 94%, 94 ≤ SpO 2 ≤ 97%, and SpO 2 > 97%, respectively, supplemental O 2 was associated with a significantly increased odds ratio of both mortality and incidence of ARDS, no matter the presence or absence of TBI [44] . Unfortunately, the authors did not provide information on the presence of circulatory shock nor the need for mechanical ventilation. Finally, a post hoc analysis of the Focused Outcomes Research in Emergency Care in Acute Respiratory Distress Syndrome, Sepsis and Trauma (FORECAST) on 240 patients with ISS ≥ 16 studied the impact of hyperoxaemia during resuscitation (defined as PaO 2 of ≥ 300 mmHg on hospital arrival and/or 3 h after arrival). The results highlighted the importance of the need for mechanical ventilation and/or the presence/absence of circulatory shock: hyperoxaemia was associated with prolonged ICU sty in patients not intubated in the ED, while no effect was present in mechanically ventilated patients [45] . Of note, unadjusted baseline characteristics not only showed a higher proportion of mechanically ventilated patients, but also significantly lower GCS (6 vs. 14) as well as more frequent need craniotomy and transfusion of blood products in the hyperoxaemic group [45] .

Consequently, despite being common practice in daily care, in particular in patients with pronounced blood loss, so far, no specific target for PaO 2 is available.

Two opposing statements ("…The emerging clinical experience demonstrates that hyperoxia is safe and beneficial to the brain…" [46] and "…Hyperoxia may be associated with increased mortality in patients with … traumatic brain injury…" [47] ) highlight the discussion on the impact of supraphysiological PaO 2 in TBI patients during the last decade. This controversy is further mirrored by the results of three clinical studies: In 193 TBI patients (40% with polytrauma) with a GCS of 4 ± 2, Asher et al. reported that maximum values of 250 < PaO 2 < 450 mmHg during the first 72 h were associated with improved outcome [48] . In contrast, Rincon et al. reported in 1,212 mechanically ventilated TBI patients (57% with a GCS < 8) that PaO 2 ≥ 300 mmHg during the first 24 h significantly increased mortality when compared to 60 ≤ PaO 2 < 300 mmHg (33% vs. 23%) [13] . Finally, Raj et al. reporting on 1,116 moderate-tosevere (GCS 3-12) patients with TBI concluded that a PaO 2 > 100 mmHg within the first 24 h of ICU admission had no predictive value for 6-month mortality [49] .

Theoretically, hyperoxaemia may exert beneficial effects during the acute management of traumatic brain injury (TBI) as a result of the above-mentioned vasoconstriction, which would allow for reducing intracranial pressure without compromising tissue oxygenation. This potentially beneficial effect has been demonstrated in severe TBI patients (n = 42, mean GCS = 5.7) undergoing a combined hyperbaric oxygen (HBO) and normobaric (NBO) oxygen treatment. Ultimately, this approach not only significantly reduced mortality, but also improved neurological outcome at 6 months post injury as assessed using the Glasgow Outcome Scale-Extended (GOSE) [50] .

In the context of the above-discussed hyperoxaemia-related increase of systemic O 2 transport capacity, hyperoxaemia may be particularly important when TBI coincides with haemorrhage. Nevertheless, the existing data of the two existing prospective clinical studies again are equivocal: Taher et al. compared the effects of 6 h of F I O 2 = 0.8 and 0.5 early after injury in 68 mechanically ventilated TBI patients with a GCS of 4 ± 2, demonstrating a tendency towards improved neurological outcome at 6 months [51] . However, the generalizability of the findings is limited, since patients with chronic co-morbidity, > 65 years of age, and, in particular, arterial hypotension were excluded.

The BRAINOXY [52] trial investigated the effect of F I O 2 = 0.7 vs. 0.4 in 62 TBI patients (both isolated TBI and TBI in the context of polytrauma) during the time of mechanical ventilation (up to 14 days). As expected, the higher F I O 2 was associated with nearly twice as high PaO 2 levels during the observation period, but there was no inter-group difference for markers of oxidative stress, inflammation, neurological injury and/or pulmonary complications [52] .

Large scale, retrospective studies provided comparably equivocal results: O'Briain et al. reported in > 24,000 mechanically ventilated TBI patients that hyperoxaemia (in PaO 2 category increments up to > 500 mm Hg) during the first 24 h in the ICU did not affect mortality irrespective of the GCS at admission [53] . In contrast, a post hoc analysis of 417 of the 1213 patients of the Citicoline Brain Injury Treatment Trial (COBRIT) showed that 150 < PaO 2 < 250 mmHg (referred to by the authors as "mild" hyperoxaemia) within the first 24 h after injury was associated with significantly lower mortality and, in particular, a better GOSE score and overall long-term functional and cognitive outcomes [9] . This possibly beneficial effect was also suggested by a retrospective study in 115 patients with severe TBI, which assessed the possible relation between PaO 2 and cerebral metabolism as well as pressure-flow autoregulation using cerebral microdialysis and the pressure reactivity index, respectively. The authors concluded that a PaO 2 "…above 90 mmHg and higher may improve oxidative cerebral energy metabolism and pressure autoregulation, particularly in cases of limited energy substrate supply in the early phase of TBI…". However, there was no significant relationship between PaO 2 and clinical outcome as assessed using the GOSE [54] .

In conclusion, it seems to be unequivocal that hyperoxaemia should be avoided after acute brain injury resulting from ischaemia [13] as well as subarachnoid [55] and/ or intracerebral bleeding: Because of the vasoconstrictor properties of O 2 [5] there is a consecutive risk of vasospasm-induced delayed cerebral ischaemia [56, 57] . Given the above-mentioned controversial results reported on TBI, the current knowledge on a possible PaO 2 target in these patients is best characterized by the statements that-with respect to any possible neuroprotective properties of hyperox(aem)ia-there "… is probably a narrow effective dose, and benefit may be limited to at-risk tis-sue…" [58] , and the conclusion that there is the " … need to identify optimal approaches to improve O 2 delivery without exacerbating … oxidative stress or injury…" [59] .

CPR represents whole-body I/R injury, and clearly, two phases of CPR have to be differentiated, i.e. the period of active cardiopulmonary resuscitation, and the postresuscitation phase, which starts after return of spontaneous circulation (ROSC). For obvious reasons, studies on the impact of hyperox(aem)ia during the former are scarce given the common practice of ventilation with 100% O 2 during CPR, but the available data suggest that there is a "dose-dependent" successful incidence of ROSC with incremental PaO 2 values [60, 61] .

The post-resuscitation phase has been thoroughly investigated during the last decade. Two large-scale retrospective American and Australian studies including 6,326 and 12,108 patients, respectively, found equivocal impact of arterial oxygenation on outcome after cardiac arrest. In the American study [10] , hyperoxaemia (PaO 2 ≥ 300 mmHg during the first of 24 ICU hours after ROSC) was associated with a significant, dose-dependent increase of mortality and worsening of neurological outcome for any 100 mmHg rise in PaO 2 [62] ). In contrast, in the Australian study, it did not "have a robust or consistently reproducible association with mortality" [11] . Similar deleterious effects of hyperoxaemia as defined by PaO 2 > 300 mmHg have been reported by others in retrospective analyses of a total of 1,448 patients [14, [63] [64] [65] . Of note, one of these studies concluded that the "…optimal range of PaO 2 for favourable neurological outcome…" could be a PaO 2 interval of 70 < PaO 2 < 240 mmHg [64] . In addition, in the other study, "moderate hyperoxia" (101 < PaO 2 < 299 mmHg) according to the authors' definition was even associated with improved organ function at 24 h [65] . Finally, in 5,258 patients after ROSC, Helmerhorst et al. had observed a similar U-shaped relation between mortality and the maximum PaO 2 during the first 24 h in the ICU [66] as in the above-discussed general ICU population, the nadir mortality coinciding with a PaO 2 ≈ 150-200 mmHg.

The most recent data confirm these conflicting results: In a prospective observational study on 280 patients Roberts et al. [67] concluded that PaO 2 > 300 mmHg during the initial 6 h after ROSC was independently associated with death and poor neurological function. Moreover, any one-hour longer duration of hyperoxaemia was associated with a 3% increase in risk of poor outcome. In contrast, a post hoc sub-study of the Target Temperature Management (TTM) trial, including 939 patients after out-of-hospital-cardiac arrest, did not find any significant relation between hyperoxaemia (defined as PaO 2 > 300 mmHg) within 37 h of ROSC and poor neurological outcome after 6 months [68] . Finally, an additional post hoc analysis of the above-mentioned ICU-ROX study including 161 patients with "suspected hypoxic-ischaemic encephalopathy" after cardiac arrest, a subgroup pre-specified and defined prior to randomization, did not show a statistically significant reduction in death or unfavourable neurological outcomes at day 180 [69] . This finding was in contrast to the analysis of the complete GOSE, which had suggested benefit of the "conservative oxygen" treatment and yielded significantly lower mortality at day 180 in the unadjusted analysis [18] .

Again, as already suggested for ARDS, sepsis and septic shock, there seems to be no ideal target PaO 2 for ROSC after CPR.

Despite the recommendations of the World Health organization from 2016 [70] and subsequently by the US Centers for Disease Control in 2017 [71] , the question whether or not there is evidence for the use of peri-operative hyperoxia to reduce surgical site infections remains a matter of debate. In line with previous ones [72, 73] , the most recent meta-analysis on 17 RCT including 7,817 patients identified a weak signal (Odds ratio 0.80, confidence interval 0.64-0.99) favouring peri-operative hyperoxia in patients undergoing general anaesthesia with endotracheal intubation [74] . No effect was found in the overall analysis. Other authors cautioned this conclusion [75] or even could not confirm this result when analysis "…was restricted to objective-or investigatoridentified low-bias studies, although those analyses were not as well-powered" [76] . Moreover, several concerns have to be raised concerning these recommendations. Most trials investigating the putative impact of perioperative hyperoxia on surgical site infection compared F I O 2 = 80% with F I O 2 = 30%), which would yield a PaO 2 ≈ 350-400 mmHg vs. PaO 2 ≈ 120-150 mmHg, respectively, in patients without major acute or chronic cardiopulmonary disease. However, at least some observations reported standard daily practice frequently using F I O 2 ≈ 40-60% [77, 78] . Finally, even if present, any beneficial effect of peri-operative hyperoxia on the incidence of surgical site infections has to be weighed by potential harm, e.g. pulmonary and/or cardiovascular complications as well as new or recurrent cancer [79] [80] [81] [82] [83] . In fact, the largest available study in this context on almost 74,000 patients, showed a dose-dependent increase in major post-operative respiratory complications and ultimately 30-day mortality of intra-operative F I O 2 increments (median F I O 2 = 0.3, = 0.41, = 0.52, = 0.58, and = 0.79, respectively) [84] .

Hence, albeit a recent review article conjected "…that current evidence is in favour of hyperoxia in noncritically ill intubated adult surgical patients…" [85] , perioperative hyperoxia (and even the possible consecutively enhanced host-defence resulting from increased ROS formation [86] ) remains an open question, because the "good", "bad", and "ugly" [87] of its use are still not fully understood.

As described in detail above, there is no ideal PaO 2 target for critically ill patients so far, no matter the underlying aetiology. Hence, both hyperoxaemia (i.e. supraphysiological PaO 2 ) and, more importantly, due to its short and long-term consequences, hypoxemia (i.e. PaO 2 < 55-60 mmHg) should be avoided. For daily practice, this raises the question whether SpO 2 suffices as a surrogate for PaO 2 . Hyperoxaemia and hypoxemia can be defined as PaO 2 > 100 and < 55-60 mmHg, respectively. However, due to the sigmoid haemoglobin-O 2 dissociation curve and its dependency on pH, PCO 2 , temperature, and erythrocyte 2,3-diphosphoglycerate concentration, the PaO 2 -SaO 2 relation may vary. Moreover, frequently the available SpO 2 devices cannot take into account increased met-or carboxy(CO)-haemoglobin levels or the interference of jaundice, because measurements are based on two wavelengths only. Hence, even normal SpO 2 readings of 94% cannot exclude hypoxemia with PaO 2 < 60 mmHg [88] . This is particularly important in the most severe patients, when vasoactive drugs must be used [89, 90] . To avoid the risk of hypoxaemia, and taking in account both the potential discrepancies between SpO 2 and SaO 2 as well as the above-discussed data, targeting 95 ≤ SpO 2 ≤ 98% appears to be reasonably safe when PaO 2 and/or SaO 2 measurements are not available.

Several authors reported that in ICU patients there is a U-shaped relation between PaO 2 and mortality/morbidity. Interestingly, these mostly retrospective studies found that the lowest mortality was present at PaO 2 ~ 150 mm Hg. Nevertheless, such supraphysiological PaO 2 values cannot be recommended in general, since the absence or presence of circulatory shock and/or the aetiology of tissue dysoxia, i.e. whether it is mainly due to impaired (macro-and/or microcirculatory) O 2 transport and/or disturbed cellular O 2 utilization may determine whether any supraphysiological PaO 2 level is really beneficial and/ or even causes deleterious side effects. 

Understanding the benefits and harms of oxygen therapy

To cope with oxygen: a long and still tumultuous story for life

Mitochondrial formation of reactive oxygen species

The relation of free radical production to hyperoxia

Hyperoxia in intensive care, emergency, and peri-operative medicine: Dr. Jekyll or Mr. Hyde? A 2015 update

Effects of high inspired oxygen fraction during elective caesarean section under spinal anaesthesia on maternal and fetal oxygenation and lipid peroxidation

Brief high oxygen concentration induces oxidative stress in leukocytes and platelets -a randomised cross-over pilot study in healthy male volunteers

Early hyperoxemia is associated with lower adjusted mortality after severe trauma: Results from a French registry

Matching early arterial oxygenation to long-term outcome in severe traumatic brain injury: target values

Association between arterial hyperoxia following resuscitation from cardiac arrest and in-hospital mortality

Arterial hyperoxia and in-hospital mortality after resuscitation from cardiac arrest

Significance of arterial hyperoxia and relationship with case fatality in traumatic brain injury: a multicentre cohort study

Association between hyperoxia and mortality after stroke: a multicenter cohort study

The association between arterial oxygen tension and neurological outcome after cardiac arrest. Ther Hypothermia Temp Manag

Reactive oxygen species: toxic molecules or spark of life?

Mortality and morbidity in acutely ill adults treated with liberal versus conservative oxygen therapy (iota): a systematic review and meta-analysis

Oxygen therapy for acutely ill medical patients: a clinical practice guideline

Conservative Oxygen Therapy during Mechanical Ventilation in the ICU

ICU-ROX Investigators and the Australian and New Zealand Intensive Care Society Clinical Trials Group. Conservative oxygen therapy for mechanically ventilated adults with sepsis: a post hoc analysis of data from the intensive care unit randomized trial comparing two approaches to oxygen therapy (ICU-ROX)

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

The Association between Supraphysiologic Arterial Oxygen Levels and Mortality in Critically Ill Patients. A Multicenter Observational Cohort Study

Time spent in oxygen saturation 95-99% is associated with reduced mortality in critically ill patients with mechanical ventilation

The search for optimal oxygen saturation targets in critically ill patients: observational data from large ICU Databases

Effect of conservative vs conventional oxygen therapy on mortality among patients in an intensive care unit: the oxygen-ICU Randomized Clinical Trial

Incidence of hyperoxia and related in-hospital mortality in critically ill patients: a retrospective data analysis

Metrics of Arterial Hyperoxia and Associated Outcomes in Critical Care

Emergency department hyperoxia is associated with increased mortality in mechanically ventilated patients: a cohort study

The standard of care of patients with ARDS: ventilatory settings and rescue therapies for refractory hypoxemia

Hyperoxemia and excess oxygen use in early acute respiratory distress syndrome: insights from the LUNG SAFE study

Liberal or conservative oxygen therapy for acute respiratory distress syndrome

HOT-ICU Investigators. Lower or higher oxygenation targets for acute hypoxemic respiratory failure

Neuropsychological sequelae and impaired health status in survivors of severe acute respiratory distress syndrome

The adult respiratory distress syndrome cognitive outcomes study: long-term neuropsychological function in survivors of acute lung injury

Oxygen as an antibiotic The effect of inspired oxygen on infection

Hyperoxemia as a risk factor for ventilator-associated pneumonia

Hyperoxia and hypertonic saline in patients with septic shock (HYPERS2S): a two-by-two factorial, multicentre, randomised, clinical trial

Hyperoxia toxicity in septic shock patients according to the Sepsis-3 criteria: a post hoc analysis of the HYPER2S trial

Assessing shock resuscitation strategies by oxygen debt repayment

Bladder tissue oxygen tension monitoring in pigs subjected to a range of cardiorespiratory and pharmacological challenges

Effects of hyperoxia and mild therapeutic hypothermia during resuscitation from porcine hemorrhagic shock

Effects of Hyperoxia During Resuscitation From Hemorrhagic Shock in Swine With Preexisting Coronary Artery Disease

Early exposure to hyperoxia and mortality in critically ill patients with severe traumatic injuries

Supplemental oxygen therapy in trauma patients: an exploratory registry-based study

Japanese Association for Acute Medicine (JAAM) Focused Outcomes Research in Emergency Care in Acute Respiratory Distress Syndrome, Sepsis and Trauma (FORECAST) Study Group. Hyperoxemia during resuscitation of trauma patients and increased intensive care unit length of stay: inverse probability of treatment weighting analysis

Eubaric hyperoxia: controversies in the management of acute traumatic brain injury

Arterial hyperoxia and mortality in critically ill patients: a systematic review and meta-analysis

Survival advantage and PaO 2 threshold in severe traumatic brain injury

Hyperoxemia and long-term outcome after traumatic brain injury

A prospective, randomized Phase II clinical trial to evaluate the effect of combined hyperbaric and normobaric hyperoxia on cerebral metabolism, intracranial pressure, oxygen toxicity, and clinical outcome in severe traumatic brain injury

Effects of normobaric hyperoxia in traumatic brain injury: a randomized controlled clinical trial

A pilot study of hyperoxemia on neurological injury, inflammation and oxidative stress

Early hyperoxia in patients with traumatic brain injury admitted to intensive care in Australia and New Zealand: a retrospective multicenter cohort study

Arterial oxygenation in traumatic brain injury-relation to cerebral energy metabolism, autoregulation, and clinical outcome

Early hyperoxia in the intensive care unit is significantly associated with unfavorable neurological outcomes in patients with mild-to-moderate aneurysmal subarachnoid hemorrhage

Hyperoxia may be related to delayed cerebral ischemia and poor outcome after subarachnoid haemorrhage

Hyperoxemia during the hyperacute phase of aneurysmal subarachnoid hemorrhage is associated with delayed cerebral ischemia and poor outcome: a retrospective observational study

Neuroprotection in acute brain injury: an up-to-date review

Titrating the dose of oxygen after severe traumatic brain injury in the era of precision medicine

Increasing arterial oxygen partial pressure during cardiopulmonary resuscitation is associated with improved rates of hospital admission

Arterial blood gases during and their dynamic changes after cardiopulmonary resuscitation: a prospective clinical study

Relationship between supranormal oxygen tension and outcome after resuscitation from cardiac arrest

Hyperoxia is associated with increased mortality in patients treated with mild therapeutic hypothermia after sudden cardiac arrest

Association between early arterial blood gas tensions and neurological outcome in adult patients following in-hospital cardiac arrest

The association between hyperoxia and patient outcomes after cardiac arrest: analysis of a high-resolution database

Associations of arterial carbon dioxide and arterial oxygen concentrations with hospital mortality after resuscitation from cardiac arrest

Association between early hyperoxia exposure after resuscitation from cardiac arrest and neurological disability: prospective multicenter protocol-directed cohort study

Associations Between Partial Pressure of Oxygen and Neurological Outcome in Out-Of-Hospital Cardiac Arrest Patients: An Explorative Analysis of a Randomized Trial

ICU-ROX Investigators and the Australian and New Zealand Intensive Care Society Clinical Trials Group Conservative oxygen therapy for mechanically ventilated adults with suspected hypoxic ischaemic encephalopathy

New WHO recommendations on intraoperative and postoperative measures for surgical site infection prevention: an evidence-based global perspective

Centers for disease control and prevention guideline for the prevention of surgical site infection

Effect of intraoperative high inspired oxygen fraction on surgical site infection, postoperative nausea and vomiting, and pulmonary function: systematic review and meta-analysis of randomized controlled trials

Effect of intra-operative high inspired oxygen fraction on surgical site infection: a meta-analysis of randomized controlled trials

Effectiveness of 80% vs 30-35% fraction of inspired oxygen in patients undergoing surgery: an updated systematic review and meta-analysis

Supplemental oxygen and the risk of surgical site infection: evidence of compromised data requires correction of previously published meta-analysis

Effect of intraoperative hyperoxia on the incidence of surgical site infections: a meta-analysis

Martin DS

Intraoperative oxygenation in adult patients undergoing surgery (iOPS): a retrospective observational study across 29 UK hospitals

Epidemiology, practice of ventilation and outcome for patients at increased risk of postoperative pulmonary complications: LAS VEGAS -an observational study in 29 countries

Risk of new or recurrent cancer after a high perioperative inspiratory oxygen fraction during abdominal surgery

Perioperative hyperoxia -Long-term impact on cardiovascular complications after abdominal surgery, a post hoc analysis of the PROXI trial

Safety of 80% vs 30-35% fraction of inspired oxygen in patients undergoing surgery: a systematic review and meta-analysis

Perioperative hyperoxia: why guidelines, research and clinical practice collide

Perioperative oxygen therapy: meaningful outcomes and unintended consequences?

High intraoperative inspiratory oxygen fraction and risk of major respiratory complications

Perioperative hyperoxyphobia: justified or not? Benefits and harms of hyperoxia during surgery

Supplemental intraoperative oxygen augments antimicrobial and proinflammatory responses of alveolar macrophages

The three components of hyperoxia

Reliability of pulse oximetry in titrating supplemental oxygen therapy in ventilator-dependent patients

Accuracy of pulse oximetry in the intensive care unit

Do changes in pulse oximeter oxygen saturation predict equivalent changes in arterial oxygen saturation?

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations

Not applicable.

PA and PR drafted the manuscript. 

Not applicable.

Ethics approval and consent to participate Not applicable.

Not applicable.

The authors read and approved the final manuscript.