Perioperative oxygenation—what's the stress?

Summary

Oxygen is the most used drug in anaesthesia. Despite such widespread use, optimal perioperative oxygen administration remains highly controversial because of concerns about the competing harms of both hyperoxia and hypoxia.

Notwithstanding a Cochrane review concluding that routinely administering a fractional inspired oxygen concentration (FiO₂) >0.6 intraoperatively might increase postoperative morbidity and mortality, the World Health Organization (WHO) currently recommends all anaesthetised patients receive 0.8 FiO₂ during and immediately after surgery to reduce surgical site infections. Results from the largest trial available at the time of these two reviews (suggesting long-term survival may be worse with high FiO₂, particularly in patients with malignant disease) were considered ‘biologically implausible’ by the WHO's Guideline Development Group. In addition, the integrity of some perioperative oxygen studies has been challenged. Resolving these controversies is of fundamental importance to all perioperative clinicians.

This narrative review is based on the inaugural BJA William Mapleson lecture delivered by the senior author (AC) at the 2023 annual meeting of the Royal College of Anaesthetists in Birmingham. We present the current evidence for perioperative oxygen administration and contrast this with how oxygen therapy is targeted in other specialties (e.g. intensive care medicine). We will explore whether anaesthetists follow the WHO recommendations and consider how oxygen administration affects the stress response to surgery. We reason that novel clinical trial designs in combination with targeted experimental medicine studies will be required to improve our understanding of how best to optimise individualised perioperative oxygenation—a cornerstone of anaesthesia.

Oxygen is the most commonly used drug during the perioperative period. For most surgical procedures, oxygen exposure typically begins before the induction of anaesthesia, and in emergency cases patients may already be receiving additional oxygen well in advance of meeting their anaesthetist. Oxygen supplementation normally continues throughout the intraoperative period and after surgery at least into the recovery room. However, titration of perioperative oxygen is often rudimentary at best, and how anaesthetists should balance the benefits and risks of different oxygen concentrations remains unclear.¹

This narrative review will examine the existing guidelines, evidence, and associated controversies that surround perioperative oxygen therapy and contrast practices in anaesthesia with those of other medical specialties. The current evidence base surrounding the outdated tropes of ‘oxygen is good’ and ‘hypoxia is bad’ will be revisited, highlighting the complexities that are missed by a ‘one size fits all’ approach. This will include a discussion of the mechanistic effects of oxygen at the cellular level, where the potential for causing harm is brought into sharp focus. Finally, we will explore the directions that future work must take if anaesthetists are to attain clarity regarding a subject that is of fundamental importance to all clinicians involved in perioperative patient care.

World Health Organization recommendations and the 30–80% dogma

Oxygen is crucial for optimal cellular and organ function in humans; insufficient oxygen supply can have potentially catastrophic consequences.² Under general anaesthesia, pathophysiological changes occur that may impair the body's ability to oxygenate blood, including altered respiratory muscle function, reduced functional residual capacity, development of atelectasis, disturbance of ventilation–perfusion relationships and the development of hypercapnia.³ The extent to which these occur depends on factors such as the type of anaesthetic, the use of spontaneous or mechanical ventilation, and the duration of the procedure. Traditionally, clinicians' almost exclusive focus has been on the maintenance of oxygen delivery, often through supra-normal oxygenation to achieve a ‘safety-margin’ and guard against hypoxia.

Over time, the discussion has evolved to encompass potential benefits of different oxygen concentrations, but there remains no clear consensus from the research/perioperative community about the best strategy for oxygen supplementation during surgery.⁴ For a drug that is both so widely prescribed and so vital to the care of surgical patients, this is unusual. Part of the reason is the conflicting evidence and controversies that this research area has been embroiled in over the last decade.5, 6, 7 One of the major arguments used to advance the cause for delivering a higher FiO₂ (typically >0.6) intraoperatively has been the potential for this approach to lower the incidence of surgical site infections (SSIs).⁸ The scientific premise to this approach is that neutrophils, one of the body's primary defences against bacterial infection, use oxidative killing to combat pathogens.⁹ Classical teaching would suggest that increasing FiO₂ should not directly increase convective oxygen flux, other than by a very small amount related to oxygen dissolved in plasma. According to the ‘oxygen delivery equation’, oxygen delivery, or more correctly oxygen flux, is the product of cardiac output and arterial oxygen content (CaO₂). CaO₂ is defined as the volume of oxygen per volume of blood and is predominantly determined by haemoglobin concentration and oxygen saturation rather than the partial pressure of oxygen (Pao₂). Moreover, the relationship between FiO₂ and Pao₂ is also dependent on the effectiveness of pulmonary gas exchange (determined by the degree of hypoventilation, diffusion limitation, V/Q mismatching and shunting present).¹⁰ However, studies using both hypoxic and hyperbaric oxygen interventions have demonstrated that NADPH oxidase enzymes driving the ‘respiratory burst’ (the process that generates superoxide anions [O₂^•−] and other reactive oxygen species [ROS] to fulfil the antimicrobial function of neutrophils) depend on Pao₂ and not CaO₂.^8,11,12 Therefore, increasing FiO₂ could potentially increase bactericidal respiratory burst activity and reduce SSIs.

Early trials appeared to support this notion. In 2000, Greif and colleagues¹³ reported a statistically significant reduction in the rate of SSIs in patients undergoing colorectal surgery administered an FiO₂ of 0.8 compared with an FiO₂ of 0.3 intraoperatively and 2 h after surgery {5.2% (95% confidence interval [CI] 2.4–8.0%) vs 11.2% (95% CI 7.3–15.1%), P=0.01}. Further studies soon followed with many also reporting a significant benefit in preventing SSIs when using a higher concentration of oxygen.^14,15 Interestingly, there is no justification for why the FiO₂ values of 0.8 and 0.3 were chosen to represent ‘high’ and ‘low’ concentrations of oxygen anywhere in Greif and colleagues' report,¹³ and the so-called ‘low’ (FiO₂=0.3) group still represents a ‘liberal’ amount of supplemental oxygen when compared with room air (FiO₂=0.21). Yet, these two set FiO₂ values of 0.3 and 0.8 have repeatedly defined the interventional groups used in almost every perioperative oxygen trial conducted since.

In 2016, a World Health Organization (WHO) Guideline Development Group (GDG) (that notably did not include any anaesthetic representation) systematically reviewed the available evidence. This group recommended that adult patients undergoing general anaesthesia with tracheal intubation for surgical procedures should receive an FiO₂ of 0.8 intraoperatively and, if feasible, in the immediate postoperative period for 2–6 h to reduce the risk of SSI.¹⁶

The WHO vs Cochrane controversy and the importance of PROXI

This 2016 WHO recommendation was a stark contradiction to the conclusions published by a Cochrane systematic review just a few months earlier, which concluded that there was ‘insufficient evidence to support the routine use of a high fraction of inspired oxygen during anaesthesia and surgery’.¹⁷ The WHO's recommendation also provoked strong reactions from many in this field who criticised the recommendation for being too general and cited several methodological issues with the review; including the omission of additional relevant studies.^7,18,19 The reliability of some of the included studies was also questioned. Specifically, one group had been forced to retract several publications not focussing on oxygen because of research fraud, raising doubts over the data integrity of their oxygen therapy studies.^6,20

The results of the PROXI trial, still currently the largest randomised trial of perioperative oxygenation to have fully reported, were also questioned by the WHO's 2016 GDG who felt there was no plausible biological mechanism to explain results indicating that long-term survival may be better with normal oxygenation, particularly in patients with malignant disease.⁷ This randomised study included 1400 patients undergoing elective or emergency laparotomy in Danish hospitals to receive an FiO₂ of 0.8 or 0.3 during and for 2 h after surgery and found no difference in SSI rates between the two groups (odds ratio [OR] 0.94, 95% CI 0.72–1.22, P=0.64).²¹ Thirty-day mortality was noted to be higher in the FiO₂ 0.8 group but this was not significant (30 [4.4%] vs 20 [2.9%], OR 1.56, 95% CI 0.88–2.77, P=0.15). However, later analysis showed that long-term mortality (median 2.3, range 1.3–3.4 yr) was significantly higher overall in the FiO₂ 0.8 group (hazard ratio [HR] 1.30, 95% CI 1.03–1.64, P=0.03), predominantly because of an effect seen in patients undergoing cancer surgery (HR 1.45, 95% CI 1.10–1.90, P=0.009).²² Further post hoc analyses also reported significantly higher incidences of postoperative myocardial infarction (HR 2.86, 95% CI 1.10–7.44, P=0.03) and shorter durations of cancer-free survival (HR 1.19, 95% CI 1.01–1.42, P=0.04) in the FiO₂ 0.8 group.^23,24 Other registry studies have also since associated high FiO₂ with a dose-dependent increase in major postoperative respiratory complications (OR 1.99, 95% CI 1.72–2.31, P<0.001) and 30-day mortality (OR 1.97, 95% CI 1.3–2.99, P<0.001).²⁵

In 2018, the WHO GDG re-reviewed the evidence and reported separately on both the safety and effectiveness of administering an FiO₂ of 0.8 to reduce SSI in adult patients undergoing general anaesthesia in two simultaneously published systematic reviews.^26,27

These concluded that no definite signals of harm or benefit could be detected using an FiO₂ of 0.8. There was some evidence a high FiO₂ may be beneficial in patients undergoing general anaesthesia specifically with tracheal intubation, however the certainty of this benefit had become weaker since the 2016 recommendation. Consequently, the wording of the 2018 iteration of the WHO guideline remained unchanged aside from downgrading the strength of the recommendation to a ‘conditional recommendation, moderate quality of evidence’²⁸ (see Fig. 1).

Fig 1.

The 2018 recommendation for perioperative oxygenation from the World Health Organization's global guidelines for the prevention of surgical site infections. SSI, surgical site infection.

Alongside this development, the Centers for Disease Control and Prevention (CDC) issued a guideline supporting the use of high FiO₂ in intubated surgical patients,²⁹ whilst other groups, including the British Thoracic Society and the World Federation of Societies of Anaesthesiologists, have continued to advocate a more conservative approach (with the latter also considering the unnecessary added acquisition costs of both oxygen and delivery systems).^30,31 Consequently, guidelines for perioperative oxygen supplementation remain conflicting with no certainty on the approach anaesthetists are choosing to follow in their routine practice (see Fig. 2).

Fig 2.

Timeline summarising some of the key events in more than two decades of research investigating the effect intraoperative oxygen administration has on surgical site infections (SSIs). After the first study exploring if ‘high’ (0.8) FiO₂ might reduce SSI rates compared with ‘low’ (0.3) FiO₂ reported in 2000, long-term follow-up of patients in the Danish PROXI trial suggested higher rates of adverse outcomes (including higher mortality) with 0.8 FiO₂. The World Health Organization (WHO) recommendations were first published in 2016 and reviewed again in 2018, shortly before several studies reported that anaesthetists often do not follow these recommendations internationally. More recent trials (e.g. the Australasian HOT-ROX) are now including a ‘usual care’ arm (i.e. FiO₂ around 0.5) and ‘high’ and ‘low’ groups.

Routine perioperative oxygenation—current anaesthetic practice

In 2017, a large group of anaesthetic trainees conducted a multicentre retrospective observational study in the UK examining the anaesthetic records of 387 patients having surgery requiring arterial cannulation over a 5-day period across 29 participating centres.³² There was significant variation in the FiO₂ delivered to patients, with the recorded intraoperative FiO₂ ranging from 0.25 to 1.0. Consequently, there was also substantial variation between patients in the Pao₂, the majority of which were above resting physiological values. The median FiO₂ was around 0.5, and this remained consistent across sequential blood gases, with the inter-quartile range of all recorded arterial blood gas values being 0.4–0.55. The authors concluded that an FiO₂ of ∼0.5 represented the standard intraoperative practice for this cohort of patients, alongside a median Pao₂ of 24.7 kPa. These UK results are consistent with wider international practice, as shown in Table 1. A survey of members of the European Society of Anaesthesiology and Intensive Care reported that only 24% of the 798 respondents knew of and followed the WHO's recommendations on perioperative oxygenation.³⁷

Table 1.

Summary of selected studies that describe international typical intra-operative oxygen administration and resulting oxygenation levels in adults during general anaesthesia. IQR, inter-quartile range.

Authors	Year(s) analysed	Location	Method	Study population	Size (n)	Administered FiO2
Staehr-Rye and colleagues25	2007–2014	USA (Massachusetts)	Registry study of three hospitals	Non-cardiothoracic surgery with tracheal intubation	73 922	Median FiO2 0.52 (1st–5th quintile medians 0.31–0.79) Median Spo2 100% (1st–5th quintile medians 99–100)
LAS VEGAS investigators33	2013	International (29 countries)	Prospective, observational study of pulmonary complications	Surgeries not requiring one-lung ventilation or cardiopulmonary bypass	9413	Median FiO2 0.52 (IQR 0.45–0.70) Median Spo2 99.0% (IQR 98.0–100.0)
Karalapillai and colleagues34	2014–2017	Australia	Retrospective single-centre cohort study	>40 yr of age, major surgery (exclusions included cardiac, intracranial and one lung) anticipated to take >2 h, with tracheal intubation and arterial cannulation	373	Median lowest FiO2 0.45 (IQR 0.4–0.5) Median highest FiO2 0.7 (IQR 0.5–0.96) Median pre-emergence Pao2 25 kPa (IQR 19–32)
Suzuki and colleagues35	2015	Japan	Cross-sectional, multicentre study	Surgeries >1 h, requiring mechanical ventilation, and no cardiopulmonary bypass or extracorporeal membrane oxygenation in first hour	1498	Median FiO2 0.47 (IQR 0.4–0.6) Median Spo2 100% (IQR 0.4–0.6)
Morkane and colleagues32	2017	UK	Retrospective multicentre observational study	Operations not requiring cardiopulmonary bypass and necessitating an arterial cannula, with blood gas monitoring	378 Cases	Median FiO2 0.50 (IQR 0.41–0.55) Median Pao2 24.7 kPa (IQR 17.9–30.8)
Frei and colleagues36	2020–2021	Australia and New Zealand	Prospective multicentre observational study	Non-cardiothoracic surgery for ASA 3 or 4 patients, anticipated to be ≥120 min and ≥1 postoperative night stay	150 Anaesthetists	Median FiO2 0.47 (IQR 0.40–0.55) Median Spo2 98.3% (97.5–99.2)

Although significant variation in practice exists with regards to intraoperative oxygen therapy, the median FiO₂ delivered is consistently found to be ∼0.5 internationally. This does not align with any guidelines or recommendations and lies midway between the two FiO₂ values that have been extensively evaluated in trials (0.3 and 0.8), and which now appear to represent relative extremes of what anaesthetists have been shown to deliver in practice.

Oxygen—always a good thing?

The issue of targeting oxygen therapy is not unique to anaesthesia; alongside its use in chronic illness, supplemental oxygen is central to the clinical management of many acutely unwell patients. Approximately 34% of patients in ambulances receive some form of supplemental oxygen, as do around 15% of hospital inpatients in the UK.^38,39 Clear guidelines exist for the management of acutely unwell medical patients, but unlike the perioperative practice of delivering a set FiO₂, more commonly these take the form of targeting a level of oxygenation in the patient (e.g. peripheral oxygen saturation, Spo₂).

However, in 2018 the Improving Oxygen Therapy in Acute Illness (IOTA) systematic review and meta-analysis challenged the assumption that giving more oxygen is always the safest thing to do.⁴⁰ This analysis concluded that for acutely ill adults, high-quality evidence shows that liberal oxygen therapy increases mortality without improving other outcomes, and supported the use of conservative oxygen therapy. It is difficult to extrapolate these findings to specific groups as the 25 trials included in the analysis encompassed a heterogeneous population of patients. In the perioperative cohort, elective surgical patients were excluded, and emergency surgery cases only represented 1.7% of the total patients analysed. Furthermore, a more recent systematic review and meta-analysis that used trial sequential analysis to account for risk of bias, found no effect from liberal oxygen targets on mortality or other secondary outcomes.⁴¹

The British Thoracic Society have not issued new guidelines since the publication of the IOTA review although they are due to review their guidelines imminently. Currently though they recommend a target Spo₂ of 94–96% for most acutely ill patients not at risk of hypercapnic respiratory failure.^30,42 The Thoracic Society of Australia and New Zealand (in guidelines issued more recently) are slightly more conservative, recommending an Spo₂ of 92–96% for most acute medical conditions, and 88–92% for patients with chronic obstructive pulmonary disease (COPD) and other conditions associated with chronic respiratory failure, with supplemental oxygen only to be administered when the Spo₂ is below the lower limit of the appropriate range.⁴³

Throughout their training, clinicians are repeatedly taught to reach for the oxygen first in situations of uncertainty or in emergencies, and with good reason; brain tissue may have a hypoxia tolerance of <3 min.⁴⁴ However, this has led to the rhetoric that ‘oxygen is good’ and ‘hypoxia is bad’ becoming so deeply engrained in the mind of healthcare professionals that it can be difficult to challenge these two reductionist statements even though the reality is considerably more nuanced.

Supplemental oxygen may bring benefits beyond preventing hypoxic damage. For example, hyperoxia has received attention as a proposed strategy for reducing postoperative nausea and vomiting (PONV). One proposed mechanism is that hyperoxic conditions may reduce the occurrence of subclinical intestinal ischaemia and the subsequent release of mediators such a serotonin.⁴⁵ Another is that hyperoxia may result in a central antinausea effect through reducing the release of dopamine from the carotid bodies.⁴⁶ Multiple trials testing these hypotheses have been conducted with conflicting results. Two large systematic reviews and meta-analyses have found no evidence for hyperoxia reducing PONV,^47,48 whereas another found a small benefit for late nausea, in addition to a more pronounced effect on PONV for a subgroup of patients undergoing inhalation anaesthesia without a prophylactic antiemetic.⁴⁹ The most recent iteration of the consensus guidelines for the management of PONV do not recommend the use of supplemental oxygen for this purpose.⁵⁰

The 2015 Difficult Airway Society guidelines advocate preoxygenating all patients with oxygen 100% before induction and emergence from anaesthesia to build a reserve in the lungs in case of unanticipated airway difficulty.⁵¹ Similarly, administration of warm humidified high-flow nasal oxygen (e.g. transnasal humidified rapid-insufflation ventilatory exchange, THRIVE) has been shown to extend apnoea times significantly and high-flow nasal oxygen devices are now commonly used in many anaesthetic rooms and intensive care units.⁵²

One patient group with a larger evidence base is the critically ill. Large retrospective observational studies in Dutch intensive care units reported that in-hospital mortality was strongly associated with both low and high Pao₂ during the first 24 h of intensive care admission.⁵³ Furthermore, the probability of death increased both with longer duration of hyperoxic exposure and with higher average Pao₂ values across the length of stay.⁵⁴ Following this, several large pragmatic interventional trials have been conducted. The ICU ROX and HOT ICU studies both showed no difference in mortality with either conservative or higher oxygen targets.^55,56 However, the LOCO₂ trial was stopped prematurely after five patients in the conservative oxygen group (target Spo₂ 88–92% or Pao₂ 55–70 mm Hg) developed mesenteric ischaemia.⁵⁷ A large meta-analysis of these interventional trials comparing high vs low oxygen targets in ICUs was unable to draw clear conclusions about the effect on all-cause mortality.⁵⁸ However, this may have been complicated by the fact that definitions of ‘high’ and ‘low’ oxygen targets have become progressively more conservative.^56,57,59 A meta-analysis accounting for this tendency towards more conservative targets reported that the highest oxygen targets were associated with higher mortality at longest follow-up, supporting the presence of a ‘U-shaped’ effect on mortality as originally suggested by the earlier retrospective reviews.^60,61

Similar concerns have been raised in other areas of medicine. In patients with a myocardial infarction, hyperoxia has been associated with increased risk of a repeat myocardial infarction and ventricular arrythmia.⁶² Hyperoxia also decreases systemic and coronary blood flow and consequently reduces tissue oxygen consumption.63, 64, 65 Current guidelines now advocate treating normoxaemic patients suffering an acute myocardial infarction with air and not supplemental oxygen.⁶⁶ Guidelines for neonatal resuscitation also recommend using room air initially as this is associated with a reduction in mortality without any evidence of other harm.⁶⁷ Hyperoxia has also been associated with adverse outcomes after out-of-hospital cardiac arrest in adults,⁶⁸ and after stroke.⁶⁹

Mechanisms of oxygen toxicity—just oxidative stress?

Hyperoxaemia (i.e. high blood oxygen concentrations) may mediate harm through its systemic vasoconstricting effect (reversed in the pulmonary and placental circulations) and a decrease in an individual's heart rate. These changes are associated with a decrease in cardiac output and a potential reduction in the delivery of oxygenated blood to tissues. However, excessively high FiO₂ may also induce intense inflammation in the lungs, or in cells in other tissue beds, even in the absence of hyperoxaemia (e.g. in patients with acute respiratory distress syndrome [ARDS] who may still have a low Spo₂ despite receiving a high FiO₂).⁷⁰ ROS may have an important role in triggering these changes. Oxygen is the ‘terminal acceptor’ in the mitochondrial electron transport chain (ETC); an essential process for the release of a cell's universal energy currency, adenosine triphosphate (ATP). Here, oxygen is reduced to water, but the efficiency of electron transfer along the ETC is not 100%. Both electron and proton leaks occur physiologically and are responsible for the generation of ROS, specifically the leakage of incompletely reduced oxygen in the form of superoxide anions (O₂^•−).⁷¹ Superoxide then undergoes spontaneous and enzyme-catalysed dismutation to form hydrogen peroxide (H₂O₂) and oxygen.⁷² Although O₂^•− and H₂O₂ are the two most abundant ROS and can exert their effects directly, their interactions with metals and other biomolecules can produce additional species with far greater reactivity, including the hydroxyl radical(^•OH) and singlet oxygen (¹O₂), see Fig. 3.

Fig 3.

Oxygen is fully reduced to water by gaining four electrons (blue arrows) in the mitochondrial respiratory chain (top line). Incomplete oxygen reduction results in the production of the reactive intermediates superoxide (O₂^•−), hydrogen peroxide (H₂O₂), and hydroxyl radicals (^•OH), respectively. These intermediates can go on to react with other cell constituents (red arrows) such as metal ions (e.g. Fe²⁺ in haemoproteins), thiols (e.g. glutathione) or compounds in the cell nucleus (e.g. DNA).

The proposed mechanism for oxygen's vasoconstricting effect is the scavenging reaction of superoxide with nitric oxide (NO) to form peroxynitrite (ONOO⁻), which reduces the bioactivity of NO.⁷³ This reaction not only reduces the bioavailability of an endogenous vasodilator, but also changes the chemistry of the cellular microenvironment as peroxynitrite is a powerful thiol oxidant that also engages in nitrosation and nitration reactions.⁷⁴ In addition, ROS react with protein and non-protein thiols (including glutathione and hydrogen sulfide, H₂S), leading to the formation of various oxidation products. This greater complexity is captured by the ‘reactive species interactome’ conceptual framework and may contribute to the system-wide effects of COVID-19 and other viral infections, and affect many other bodily functions linked to electron exchange (redox) processes, including immune function, metabolism, and nutrient utilisation.^75,76

We are still learning about the physiological and pathophysiological roles of ROS, a family of molecules which were initially regarded as solely harmful. The potential consequences of their overproduction include damage to biological structures such as DNA and RNA, the impairment of repair systems for these molecules, lipid peroxidation, direct interference with protein function, and cell death.⁷⁷ In addition, ROS have also been implicated in inflammatory processes and disease, including the development of atherosclerosis.⁷⁸ However ROS have also been demonstrated to play key messenger roles in intracellular pathways,79, 80, 81, 82 and as described previously, serve an important function in the killing of microbes by phagocytes.⁹ Antioxidant systems play a vital role in regulating the concentration of different redox species and it is the balance of these, along with the level of ROS production, which determines the consequences for individual cells and the organism as a whole.^83,84 When conditions result in antioxidant mechanisms becoming overwhelmed, this balance is tipped in favour of the oxidants, causing oxidative stress.^71,85 One factor that dynamically modulates the level of ROS produced is the local oxygen tension. Although decreased oxygen delivery to mitochondria does lead to less oxygen being available to accept electrons, this does not necessarily translate into less ROS generation. Hypoxic conditions can also simultaneously lead to a reduction in the efficiency of electron transfer along the ETC and thereby paradoxically increase ROS production.⁸⁶

Hypoxia—always a bad thing?

Hypoxia describes an insufficiency of oxygen at the tissue level to meet the metabolic demands of individual cells. Unrecognised and uncorrected perioperative hypoxia poses a threat to life. Of Barcroft's four types of hypoxia,⁸⁷ hypoxaemic hypoxia (a low Pao₂), is the form most commonly encountered intraoperatively. Prior to the increased use of pulse-oximetry in the 1980s and the subsequent development of monitoring standards in the 1990s, hypoxaemia was the leading cause of anaesthesia-related mortality.⁸⁸ Despite these safety advances, hypoxaemia remains a relatively common occurrence in operating theatres. A retrospective study of 95 407 records from two large centres in the United States looked at episodes of hypoxaemia and severe hypoxaemia during anaesthesia, which were defined as ≥2 min of an Spo₂ of <90% and ≤85%, respectively.⁸⁹ Their findings showed that 6.8% of patients experienced an episode of hypoxaemia, and this was severe for 3.5%, with these episodes lasting ≥5 consecutive minutes in 1.6% and 0.8% of patients, respectively. How these transient hypoxaemic episodes correlate to hypoxia at the tissue level remains unclear, as does their clinical impact.

The body's response to hypoxia is multifaceted as illustrated by successful acclimatisation to high altitude.⁹⁰ Chemoreception leads to several responses from many body systems, including the respiratory, sympathetic nervous, cardiovascular, and microvascular systems. Each of these responses is dependent on the magnitude and temporality of the triggering exposure. Hypoxia-inducible factors (HIFs) act as intracellular sensors and are crucial to achieving oxygen homeostasis. These transcription factors promote the expression of several hundred other genes including erythropoietin, angiogenic, and vasoactive mediators, and metabolic enzymes that optimise oxygen demand and energetic balance. Under normoxic conditions, destruction of HIFs is mediated by the Von Hippel–Lindau protein (VHL), ubiquitination, and proteolysis. Hypoxia prevents this degradation and allows HIF-activated transcription of these other proteins to occur.⁹¹ When successful, these responses allow acclimatised individuals to function well with remarkably low Pao₂ values (e.g. mean 3.28 kPa, range 2.55–3.93 kPa at 8400 m on Mt Everest), values well below what would usually be considered compatible with survival in a clinical perioperative or critical care context.⁹²

Tibetan Sherpas, highly adapted and able to perform exceptionally well under hypobaric hypoxic conditions at high altitude, have evolved mechanisms to enhance tissue oxygen utilisation (i.e. metabolic adaptation) as well as enhancing tissue oxygen delivery.⁹³ These include reduced skeletal muscle capacity for fatty acid oxidation, improved muscle energetics, and protection against oxidative stress. Lowlanders appear to develop similar responses over prolonged altitude exposures; reducing muscle mitochondrial volume density and downregulating ETC complexes possibly to mitigate against ROS production,⁹⁴ with the stress response also appearing to be unique to each individual.⁹⁵

High altitude/hypoxia pre-acclimatisation is already commonly used to improve athletic performance, and its therapeutic potential continues to be explored.⁹⁶ In a small randomised double-blind study, 15 sessions of passive intermittent hypoxia (FiO₂ 0.10–014), across 3 weeks, compared with normoxia, was shown to increase peak oxygen consumption by +6.2% vs −3.0% (P<0.001).⁹⁷ Additionally, similar intermittent hypoxic training (IHT) has been shown to reduce blood pressure in hypertensive patients,⁹⁸ assist in weight loss (when alongside simultaneous exercise),⁹⁹ and enhance glucose homeostasis.¹⁰⁰ More specifically, HIT has been associated with lower fasting glucose concentrations, improved oral glucose tolerance test results, and remission from pre-diabetes.¹⁰¹ HIFs are thought to be driving these effects.^102,103

Other potential therapeutic roles for hypoxia include the treatment of mitochondrial disease such as Leigh syndrome, which results in neurodegeneration and death, often before the age of 3 yr. Compared with controls breathing room air (FiO₂ 0.21), mouse models of this condition reared in FiO₂ 0.11 at normal atmospheric pressure (normobaric hypoxia) extended their overall median survival from 58 days to 270 days (P<0.0001), whilst rearing these mice in FiO₂ 0.55 dramatically reduced survival.¹⁰⁴ This suggests that patients with mitochondrial disease may be highly sensitive to oxygen toxicity, with hypoxia either facilitating repair mechanisms by removing the upstream cause of damage or directly promoting repair mechanisms. Hypoxia may trigger adaptive programs that the disease itself does not, and the altered interaction with other reactive species may also be important.

The above examples together illustrate the folly of a reductionist approach that labels hypoxia universally as ‘bad.’ The type, magnitude, and duration of hypoxia, in combination with the context in which it is experienced and several patient-specific factors, all are key determinants of its eventual impact, as is the individual sensitivity to the hypoxic exposure.

Steps towards personalising perioperative oxygen therapy?

A one-size-fits-all approach is rarely the answer to patients' problems in modern medicine, and we have known for many decades that individual responses to oxygen administration vary. After World War II, experiments exploring the tolerance to increased (hyperbaric) oxygen tensions showed that the time it took for fit young male participants to show signs of oxygen toxicity (nausea, vertigo, lip twitching, paraesthesia, vomiting, convulsions) ranged from 6 to 96 min.¹⁰⁵ Another recent trial in patients undergoing coronary artery bypass grafting concluded that the response of biventricular systolic variables to an FiO₂ of 0.8 was dependent on their systolic function when receiving an FiO₂ of 0.3.¹⁰⁶ In patients with a poorer left ventricular (LV) global longitudinal strain (GLS) than the derived cut-off receiving an FiO₂ of 0.3, systolic function improved with higher FiO₂ (P<0.01). However, the opposite response occurred in patients with LV GLS better than the derived cut-off whilst receiving an FiO₂ of 0.3.¹⁰⁶

Identifying and targeting these different responses and sensitivities to oxygen poses a challenge for perioperative physicians. Although a brief 10-min exposure to oxygen 100% (delivered as preoxygenation before a sedative procedure in the emergency department) decreases cardiac output by at least 10% in ∼25% of patients, it was not possible for clinicians to identify this ‘at risk’ subgroup of patients from physical examination or any patient data.¹⁰⁷ Most pulse oximeters cannot measure hyperoxaemia, by definition the Spo₂ scale stops at 100% and any increase in a patient's oxygenation level beyond this value cannot be detected without invasive and intermittent arterial blood sampling. New biomarkers of oxygen tolerance, sensitivity, and toxicity are urgently needed, along with new technology and monitoring devices to measure them.

As discussed above, although observational studies have demonstrated a dose response to oxygen's effects on bodily functions, few interventional studies have been designed in an optimal way to detect a similar effect. Whilst large pragmatic RCTs carry many methodological advantages, only comparing an arbitrary ‘high’ (e.g. an FiO₂ of 0.8) to an arbitrary ‘low’ (typically an FiO₂ of 0.3) allows no assessment of effective titration and neither of these commonly studied FiO₂ levels represent the amount most commonly delivered by the majority of anaesthetists in routine practice. The HOT-ROX trial, developed in Australia and New Zealand, has recently reported its pilot data confirming the feasibility of recruiting and randomising to three different interventional groups (restricted, usual care, or liberal oxygen therapy).¹⁰⁸ However, even if the HOT-ROX trial does eventually produce clear results, this study alone will not tell us definitively how best to optimise perioperative oxygenation for individual patients.¹ The broad inclusions needed by large pragmatic trials to generate adequate power may also inherently ignore individual phenotypes and response signals amongst background noise. Novel trial designs, such as platform trials, offer an opportunity to test multiple different interventions more efficiently and also can allow for different interventions to be added or removed whilst the trial is in progress.¹⁰⁹ Platform trials have become much more familiar since the COVID-19 pandemic, with the Randomised Evolution of COVID-19 Therapy (RECOVERY) and the Randomized, Embedded, Multifactorial Adaptive Platform trial for Community-Acquired Pneumonia (REMAP CAP) trials both allowing many different potential treatments for COVID-19 to be tested remarkably rapidly.^110,111 However, relatively few perioperative studies have used a platform trial design to date. A platform perioperative oxygen trial could offer many advantages over more classical trial designs. For example, it could allow for the FiO₂ values being tested as interventions to be adjusted and refined as the trial progressed. This design could also make it easier for trialists to target different interventional FiO₂ values to specific patient phenotypes for the first time, including patients from under-represented minority groups, such as pregnant women or children, where research in this area is especially lacking.

Conclusions

Supplemental oxygen will continue to be a common and essential treatment in perioperative and intensive care medicine. However, oxygen will remain a limited and finite resource (as was sadly witnessed in many hospitals internationally during the COVID-19 pandemic) that requires large amounts of energy to produce, making decisions about its appropriate use critical to maintaining adequate supplies and ensuring future sustainability. Research into perioperative oxygen needs not only to resolve the uncertainty created by the WHO guideline controversy but also to equip perioperative physicians with simple and effective ways of personalising the treatment they give to individual patients.

Author’s contributions

Project conception and design: ME, DSM, MF, MPWG, AFC

Drafting manuscript: JL, AFC

Reviewing and editing manuscript: all authors

Declarations of interest

MPWG has received grant support from: National Institute for Health Research, Bill and Melinda Gates Foundation, National Lottery Fund (UK), NHS England (UK), Edwards Lifesciences, Macmillan Cancer Support (UK). MG has received consulting fees (medical advisory board and trial monitoring) from Edwards Life Sciences, Sphere Medical Ltd, South-West Sensors Ltd. AFC has received oxygen monitoring equipment on loan from Masimo (Irvine, CA, USA) to use in investigator-led trials. All other authors have no conflicts of interest to declare.

Funding

MG is in part funded by the NIHR Senior Investigator Scheme and in part by the NIHR Southampton Biomedical Research Centre. AFC is funded through an NIHR Clinical Lecturer award and has previously received funding through the NIHR Southampton Biomedical Research Centre.

Handling editor: Phil Hopkins