Partial pressure of oxygen, hyperoxemia and hyperoxia in the intensive care or anesthesia setting

Authors: Sylvain Diop, Roman Mounier

Abstract

In clinical studies, the partial pressure of oxygen (PaO2) and oxygen pulse saturation are the main variables used to assess blood oxygenation and define the threshold of hypoxia/hyperoxia and hypoxemia/hyperoxemia. Determination of the optimal oxygenation target has generated a lot of interest in recent years, mainly because of the potential risk of worse outcomes associated with hyperoxia, whereas the risk associated with hypoxia has been already well known. In this short narrative review, we recall some fundamental elements of physiology regarding the meaning of PaO2, the diffusion of oxygen to cells, the definitions of hyperoxemia and hyperoxia and the mechanisms of oxygen toxicity to provide a better understanding of these concepts, to which intensive care clinicians are frequently confronted. PaO2 provides only limited information about oxygen concentration carried by blood and does not allow to determine whether cells are exposed to hyperoxia. This should be considered for the design of future studies that aim to determine optimal oxygenation target and by clinicians for their daily practice.

Introduction

Determination of the optimal blood oxygenation target in critical care settings has generated a lot of interest since the last decay.12345678910 The arterial partial pressure of oxygen (PaO2) is frequently used as the main variable to assess blood oxygenation. It is also widely used to define the threshold of hypoxemia/hypoxia and hyperoxemia/hyperoxia. The physiological concepts on which trials comparing a conservative versus a liberal oxygen supplementation strategy were mainly built are based on the potential toxicity of oxygen (O2) through the generation of excessive oxidative stress.1234567

Until now, the results have been conflicting. A recent review found no difference in mortality in the intensive care unit (ICU) between patients receiving lower or higher oxygenation targets.11 Other studies in patients hospitalized for out-of-hospital cardiac arrest and patients receiving venoarterial extracorporeal membrane oxygenation support revealed an association between hyperoxemia and mortality.1213 On the contrary, a recent large randomized trial in adults receiving venoarterial extracorporeal membrane oxygenation reported no difference in ICU-free days and other secondary outcomes, including mortality.14 These discrepancies may result from the difference in the definitions used to characterize hyperoxemia/hyperoxia and the population studied.15 This finding could also reflect an oversimplification of the physiological meaning of commonly used variables, such as PaO2 or O2 saturation, which cannot be blindly used as a marker of hyperoxia and O2 toxicity.

Accordingly, we believe that it is important to recall some fundamental elements of physiology regarding the meaning of PaO2, the diffusion of O2 to cells, the definitions of hyperoxemia and hyperoxia and the mechanisms of O2 toxicity to provide a better understanding of the concepts to which intensive care clinicians are frequently confronted.

Search Strategy

In this narrative review, we included English studies focusing on the effect of O2 toxicity under normobaric and hyperbaric conditions. The authors searched the PubMed database to identify relevant publications using the following oxygen toxicity, hyperoxia, hyperoxemia, supplemental oxygen, liberal oxygen, conservative oxygen, partial pressure of O2, oxygen saturation, hyperbaric oxygen therapy, and reactive oxygen species.

Physiological Consideration

Partial pressure of oxygen

PaO2 is widely used among physicians to assess arterial blood oxygenation at the bedside.5679 However, its real meaning needs to be clarified. Pressure is defined as a force perpendicularly exerted against a surface whose international unit is the Pascal, which corresponds to the force of one Newton on an area of 1 m^2^ (N/m^2^ or kg/m/s2). The partial pressure of a gas, here PaO2, is not a measure of a quantity or a concentration of O2. PaO2 measures the force of the collision of O2 molecules against the “container wall” in the hypothetical gas phase in contact with the arterial blood sample at equilibrium for a given condition of temperature and pressure. Accordingly, for the same concentration of O2 molecules, PaO2 changes with temperature and/or pressure.

The dissolution of O2 in blood corresponds to a thermodynamic process in which O2 molecules are transferred from the alveolar gas into the blood to reach an equilibrium (PO~2gas~ = PO~2blood~). If we consider that the gas is ideal, the partial pressure of the gas is lower than the atmospheric pressure (Patm), and the temperature is constant; at equilibrium, Henry’s law is

P = kH × xi      (1)

where P is the partial pressure of the gas (in our case, the alveolar gas), xi is the molar fraction of the dissolved gas in the liquid (here, the blood) and kH is Henry’s solubility constant that in fact is not a constant. kH depends on the temperature, pressure and liquid properties. For example, an increase in temperature leads to greater kinetic energy and an increase in the entropy (the randomness between the dissolved gas molecules), leading to a decrease in the solubility of the gas in the solvent (because of the more rapid translation of the molecules from the liquid to the gas). Henry’s law can be easily rewritten as

xi = (1/kH) × P      (2)

Equation 2 allows the concentration of the dissolved substance to be linked to the partial pressure of the same substance in the gas phase. When the conditions of Henry’s law are satisfied, the quantity of dissolved gas per unit of liquid can be expressed in units of pressure. Importantly, under standard pressure conditions, until a specific threshold is reached, O2 does not behave like an ideal gas because it interacts with several blood and tissue compounds.

It is also fundamental to underline the following under normobaric condition, the metabolic effect of a substance in a solvent is driven by its concentration, which can be expressed in terms of volume or molar concentration. For example, when we calculate the arterial content in O2 with the following

CaO2 = 1.34 × Hb x SaO2 + α × PaO2      (3)

where CaO2 is the arterial content in O2, SaO2 is the arterial saturation of hemoglobin in O2, and α is the solubility coefficient of O2 in blood at 37°C and 1 atmosphere pressure (ATM; 1 ATM = 760 mmHg, representing the average air pressure at sea level), which gives us the volume of O2 contained in 100 mL of blood. Because is very low (α = 0.0031 mL of O2/mmHg/100 mL of blood, at 37°C and 1 ATM), we can easily see that much of the concentration of O2 is carried by hemoglobin and that the dissolved part of O2 is very low in comparison. This means that even when breathing 100% O2, the concentration of dissolved O2 remains below 2.5 mL of O2/100 mL of blood, which is far lower than the concentration of O2 bound to hemoglobin (i.e., 15–20 mL of O2/100 mL of blood). When PaO2 is higher than 150 mmHg, the hemoglobin is fully saturated, and the concentration of dissolved O2 in the blood (but not inside the cells) can approach Henry’s law. The O2 linked to hemoglobin has no chemical activity, and only the free O2 molecules (O2 dissolved into the blood) are able to diffuse to the cells (Figure 1).16

O2-bound carrier system

Several O2-bound carrier systems exist in living organisms to allow adequate delivery of O2 to cells while limiting its toxicity.17 In mammals, the basal concentration of O2 needed to ensure proper metabolic function is too high to be supplied solely by the diffusion of O2 from the environment.17 Development of the lung and heart to pump and distribute O2 to the tissue through the hemoglobin carrier system helps to overcome this issue. In humans, the hemoglobin present in red blood cells is the most well-known O2 carrier system.17 Other molecules, such as myoglobin, are located inside various types of cells (i.e., myocytes and muscular cells).17 Importantly, the binding of O2 molecules to these carriers is a passive phenomenon that depends on the local dissolved O2 concentration (other local factors, such as temperature, pH and 2,3-diphosphoglycerate concentration, also play a role in modifying the affinity between O2 and hemoglobin).16 Thus, the carrier will bind and release O2 according to the local O2 concentration.16 In physiologic books, it is often referred to as P50, the partial pressure to which 50% of the carrier is saturated. The P50 of hemoglobin (in standard conditions of pressure and temperature) is around 27 mmHg, which is why the venous saturation of O2 is usually above 50%.18

O2 diffusion to the tissue

Diffusion describes the net movement of atoms, molecules or anything due to the random thermal motion of the species considered. It is driven by the variation in Gibbs free energy or in the chemical potential of the chemical species of the system (meaning the rate of change in free energy/chemical potential in the thermodynamic system considered). In our area of interest, the diffusion of a molecule may be approximated by Fick’s

where J is the diffusion flux, D is the coefficient of diffusion of the molecule in the medium (m^2^/s), ∂(Cmol) is the gradient of the concentration of the molecule and ∂(x) is the distance. This law is phenomenological, meaning that it describes the diffusion process under specific conditions (including that the density of molecules is not too low or too high and that its variation is not brutal).

The diffusion flow rate depends on the diffusion coefficient, which is specific to the medium and the molecules considered, with respect to the concentration gradient. The concentration of O2 in the cells is very low (2–4 mmHg in pressure equivalent).16 O2 is continuously consumed inside the cells, so its diffusion also depends on its consumption. The more O2 that the cells need, the more the hemoglobin arriving at the capillaries will unload their O2 molecules (because of the decreased local dissolved O2 concentration and local pH, as more O2 consumption means more carbon dioxide production).10 Notably, the consumption rate of O2 is much faster than its release and diffusion across the cell and mitochondrial membranes. This explains why we need such an O2 carrier system (hemoglobin), whereas the amount of O2 consumed is lower than that transported. O2 delivery to cells can be regulated via several metabolic blood flow, the amount of the carrier system, the speed at which desaturated hemoglobin is supplied by O2, and the affinity of hemoglobin for O2. It is hypothesized that there is a sensory system that is able to detect the dynamic rate of O2 consumption to adapt the quantity of O2 delivered to the tissue.16

Differences Between Hyperoxemia and Hyperoxia

Several trials studying the effects of different PaO2 levels on outcomes use the terms hyperoxemia and hyperoxia indistinctly to characterize high levels of PaO2.1234567891019 Other studies defined hyperoxia or hyperoxemia (again, sometimes without distinction between them) as an oxygen pulse saturation (SpO2) or a fraction of inspired O2 (FIO2) above a specific threshold (Table 1).1920212223 Hyperoxemia refers to excess O2 in the blood, which can be seen either as an excess of dissolved O2 or an excess of total O2 (dissolved O2 and O2 bound to hemoglobin). For example, if a patient has a hemoglobin level of 18 g/dL, a PaO2 of 80 mmHg and a SaO2 of 95%, its arterial content in O2 is 23.2 mL of O2/100 mL of blood; conversely, if a patient has a hemoglobin level of 10 g/dL, a PaO2 of 300 mmHg and a SaO2 of 100% of its CaO2 is 14.3 mL of O2/100 mL of blood. In the first patient, one would say that he is normoxemic and the second one is hyperoxemic, whereas the concentration of O2 in blood is much higher in the first patient. The results are quite different according to the method used to assess blood oxygenation. Again, we must keep in mind that O2 molecules linked to hemoglobin have no physico-chemical activities; only free O2 molecules (already dissolved in the plasma and unloaded by hemoglobin) are able to cross the cell membrane and be metabolized. Therefore, neither the PaO2 nor the CaO2 allows the determination of the effective proportion of O2 molecules reaching the cells where they can potentially induce oxidative stress. Oxygen arterial saturation (SaO2 or SpO2) depends on many factors, and for the same PaO2, it may be quite different (according to the local pH, 2,3-diphosphyglycerate concentration and temperature). Hyperoxia implies that excess O2 molecules are available for cells and may induce adverse effects (i.e., oxidative stress).24 The transfer of O2 from the blood to the tissue depends on both diffusion and perfusion. Perfusion may vary greatly according to the cardiac output, the driving pressure through the organ considered, the local condition at the tissue level (i.e., local pH), the transit time of blood into the capillaries, the drugs administered to the patient, the temperature, etc. A high CaO2 does not necessarily mean that a high proportion of O2 will reach the cells. Under hyperbaric conditions, the O2 supply can be satisfied only by means of dissolved O2 (hemoglobin does not release O2 molecules because the dissolved O2 concentration is above the P100 of hemoglobin and is sufficient to supply the consumed O2).2526 Similarly, during deep hypothermic cardiopulmonary bypass, because of the increased solubility of O2 with hypothermia, most O2 needs are satisfied by dissolved O2, preventing the unloading of O2 from hemoglobin.27 The global quantity of O2 transported from the blood to the tissue may be quantified by the venoarterial difference in O2. In 1965, Whalen et al.25 explored the composition of blood gas in healthy subjects breathing room air and 100% O2 at 1 and 3 ATM. In patients breathing 100% O2 at 1 ATM, the PaO2 and the CaO2 rise to 507 mmHg and 21.2 mL of O2/100 mL of blood, respectively (compared with 89 mmHg and 19.1 mL of O2/100 mL of blood in room air). The mean venoarterial difference in O2 (D(A-V)O2) was 4.2 mL of O2/100 mL of blood and 4 mL of O2/100 mL of blood in 100% O2 and in room air, respectively. Reinhart et al.28 also reported a decrease in the mean D(A-V)O2 in patients exposed to 100% O2, which was not significantly different from the baseline value. Interestingly, the quantity of O2 extracted from the tissue was higher in the breathing room air group than in the breathing 100% O2 group. This example illustrates the fundamental difference between hyperoxemia and hyperoxia and that the former does not systematically imply the latter. Whalen and Nair2930 demonstrated that the cellular PO2 is not modified during exposure to hyperoxia because of adaptive microcirculatory mechanisms.

Understanding the Toxicity of Oxygen

O2 is both essential and sometimes harmful for eukaryote cells.31 The main justification of the large trials investigating the optimal oxygenation target in critically ill patients was that hyperoxemia or hyperoxia could lead to excessive oxidative stress and potentially worse outcomes.12345678910 The physiological basis underlying such a statement is oversimplified and often misleading. First, the generation of superoxyde anion O2^–^ (the precursor of other reactive oxygen species (ROS)) increases when the concentration or variation in the concentration of mitochondrial O2 increases. PaO2 or CaO2 does not correlate linearly with the intra mitochondrial concentration in O2; therefore, they cannot be used as surrogates of mitochondrial [O2] (except when there is a gas–liquid interface, as it is in the lung but not in other parts of the body).293032 Second, numerous studies confound hyperoxia and hyperoxemia.12345678919 Third, the scheme of comparison between the conservative and liberal O2 groups varies greatly from trial to trial.12345678910 Some even compare two different levels of PaO2 that are below the “commonly established” threshold of hyperoxemia (i.e., < 100–150 mmHg).569 Another study compared different targets of SpO2, where values above 96–97% represented the hyperoxemia group.23 These values are hardly interpretable alone, as healthy human breathing room air has a PaO2 of 90–100 mmHg and a SpO2 ranging from 96–100%. We agree that no one would say that this human is hyperoxic and recommend that humans breath a hypoxic gas mixture to lower its PaO2 (or SpO2), thus preventing the generation of excessive oxidative stress. Therefore, in standard pressure conditions, the amount of supplemental O2 administrered to obtain the oxygenation target set up by clinicians is what it is truly important. Indeed, the airways and particularly the lungs are directly exposed to O2 molecules, in contrast to other tissues.33

The toxicity of O2 is thought to be mainly mediated via the generation of ROS beginning with their precursor the superoxide anion (O2^-^).3132 O2 molecules have two unpaired electrons in their distal orbital, making them potentially reactive. However, it is slightly more complicated. In fact, because these two electrons are unpaired (meaning that their spins are parallel), this makes them much less reactive than we usually think. This unusual property of O2 represents a barrier to its toxicity. This process is the univalent reduction of O2 in water, which generates intermediates (i.e., O2^-^ and then H2O2) that are responsible for O2 toxicity.3134 A ROS is a natural product of aerobic metabolism and is produced mainly by the mitochondria; by some enzymes, such as xanthine oxidase, nicotinamide adenine dinucleotide phosphate oxidase, nitric oxide oxidase, among others; and by exogenous triggers (radiation, pollution or tobacco).313235 They are characterized by their high level of instability and capacity to react with other organic compounds and to damage DNA, enzymes and lipid-based molecules. ROS may have either positive or negative effects.31 ROS play important roles in many metabolic pathways (i.e., intracellular and growth factor signaling).31 Some of the metabolic effects of ROS are at the cornerstone of the therapeutic mechanisms of hyperbaric oxygenation therapy (HBOT) (anti-inflammatory effects, antimicrobial effects, angiogenesis, and inhibition of hypoxia-inducible factor).36

Different kinds of O2 toxicity must be distinguished. The clinical manifestations of O2 toxicity are mainly neurologic (the risk increases for exposure to PaO2 > 1.4 ATM; a condition encountered during diving or HBOT; the symptoms are multiples ranging from paresthesia, nausea, vertigo to general seizure), pulmonary (for prolonged exposure to a gas partial pressure of O2 above 0.5 ATM corresponding to FIO2 above 50% at sea level atmospheric pressure) and ocular (chronic exposure leading to myopia or/and cataract and retinal damage).3337383940 Other relevant clinical effects are difficult to assess and separate from the underlying disease leading to O2 supplementation.38 The mechanisms of neurologic toxicity are not completely understood. HBOT impaired the transmission of neurotransmitters and several transport systems of neuronal membranes and increased overall neuronal excitability by increasing NO production and decreasing GABAergic synaptic conductance.4041

Pulmonary toxicity depends on both the intensity and the duration of exposure.333840 The intensity depends on the atmospheric pressure and the FIO2. This toxicity is mediated by the direct interaction between O2 molecules and pulmonary cells (pneumocytes, fibroblasts) and by denitrogenation atelectasis.3338 After hours to days of exposure, it causes histological alveolar damage and increased alveolocapillary membrane permeability.3338 In awake patients, symptoms include chest discomfort, cough and a decrease in vital capacity.3338 Additionally, exposure to supraphysiologic O2 concentrations alters the healthy lung microbiota.42 In addition to the airways, there are no other parts of the body that are exposed to such high concentrations of O2 molecules because of the blood carrier system and the very low solubility of O2 in the blood.33 This allows for the relative protection of other organs against high concentrations of O2.33 Readers can find an exhaustive review on pulmonary O2 toxicity in critically ill patients on the top reference.39 In patients hospitalized and ventilated in the context of acute respiratory failure, it is difficult to differentiate the effects linked to the underlying disease and mechanical ventilation and those linked to O2 toxicity, both of which are likely intricate.39

Supraphysiologic exposure to O2 is not the only means leading to excessive ROS production by cells. The production of ROS is increased when cells are exposed to hypoxia, presumably secondary to the slowdown of the electron mitochondrial chain.43 The acute variation in O2 concentration (as observed during ischemia–reperfusion) also drives the formation of ROS.44 In hypoxic cells, there is an accumulation of reducing agents, which are responsible for the increased production of ROS after the normal O2 concentration is restored.3644

HBOT provides a unique opportunity to investigate the toxicity of O2 and the related generation of oxidative stress. Several studies have investigated the effects of HBOT on ROS generation, lipid peroxidation, DNA damage and inflammation.45 Most of these studies were conducted with healthy subjects. The results are not univocal, with heterogeneous effects of HBOT on the aforementioned markers. Most of the studies reported an increase in superoxide anion and hydrogen peroxide. The results for other ROS, such as reactive nitrogen species, and nitric oxide synthase contrasted more strongly.45 To date, no strong data support the occurrence of increased DNA damage and lipid peroxidation following HBOT.45 HBOT seems to have an anti-inflammatory effect, likely mediated by the inhibition of NF-κB transcription factors. Additionally, some studies have demonstrated a decrease in acute-phase protein concentrations following HBOT.45 Notably, during HBOT, patients are exposed to high levels of alveolar pressure of O2 and arterial pressure of O2 during a short period of time (a few hours per day); therefore, the results cannot be blindly extrapolated to patients exposed to lower O2 concentrations but during prolonged periods, as observed in critically ill patients.

Perspectives

Until now, we highlighted the following key 1) PaO2 can be assimilated to the concentration of dissolved O2 in blood only if the conditions for the application of Henry’s law are met. 2) In standard temperature and pressure conditions, dissolved O2 represents a minor part of the total volume of O2 carried by the blood. 3) A high PaO2 does not indicate high CaO2 and, more importantly, does not indicate hyperoxia. 4) In normobaric conditions, in addition to the lungs, which are directly exposed to O2 molecules, other organs are relatively protected from high O2 concentrations because of the blood carrier system and the low solubility of O2 in the blood. 5) Clinical studies failed to determine an optimal target of blood oxygenation when assessed by PaO2 and/or SpO2. 6) Under normobaric conditions, the potential systemic toxicity of high PaO2 or SpO2 is putative; the pulmonary toxicity of long-term exposure to high fractional inspired O2 is established.

For several years, clinicians have struggled to determine the optimal oxygenation target assessed by PaO2 or SpO2 in critically ill patients, without significant results (Table 1).11 This may be because this is not an adequate question. Based on strong physiological data, we can see that under normobaric conditions, the PaO2 concentration is inadequate to assess the concentration of O2 molecules reaching the tissues and the mitochondria where they can exert their potential toxicity. Therefore, instead of seeking a specific target of PaO2 or SpO2 for all patients, we could simply seek to ensure adequate O2 delivery while limiting the exposure of the lung to a high fraction of inspired O2. How can we do at the bedside? This can be achieved in several ways (Figure 2):

1. First, the decrease in hemoglobin concentration during the ICU stay can be minimized as much as possible. Not exhaustively, it can be achieved by limiting blood samples or using a flushing device allowing blood conservation after sampling by limiting excessive hemodilution through careful monitoring of vascular filling adapted to patients, by limiting metabolic needs (clinical evidence of inadequate cardiac output responding to vascular filling).4647 Additionally, the use of patient blood management strategies in specific populations (i.e., patients undergoing cardiac surgery or other hemorrhagic procedures) is also beneficial.47

2. Second, the balance between O2 delivery and consumption and tracking any potential O2 debt is monitored. Clinical signs of decreased organ blood flow (oliguria, neurological impairment) as well as an increase in the venoarterial difference in O2 or carbon dioxide tension or a decrease in mixed venous saturation in O2 (SvO2) suggest an inadequate match between O2 delivery and consumption and potential tissue hypoxia.4849 These variables are easily available in the ICU through central venous and arterial line samples, and static thresholds can be found in any textbook. However, a dynamic approach that involves evaluating their variation over time is more useful to detect a mismatch between delivery and consumption.

3. Third, when possible, by using the effects of pH on the oxyhemoglobin dissociation curve to facilitate O2 delivery to the tissue, a moderate decrease in blood pH (i.e., pH of 7.30) could be beneficial, as it shifts the oxyhemoglobin curve to the right, allowing a better release of O2 to the tissue. It also slows metabolism (through glycolysis inhibition) and thus O2 consumption.

4. Four, by limiting the occurrence of ischemia–reperfusion injury or equivalent, which are at the core of the generation of ROS, as much as possible.44

These several proposals represent common-sense measures that should also drive our thinking at the bedside. Under normobaric conditions, targeting supranormal levels of O2 is often not indicated because the solubility of O2 is very low and the increase in dissolved O2 is negligible when lung exposure is at a maximum. Of course, in some conditions in which HBOT is the main therapeutic, normobaric O2 therapy is mandatory until transfer to a hyperbaric chamber (i.e., gaseous emboli, carbon monoxide intoxication or decompression sickness).

This review is limited by the complexity and density of the questions addressed; one variable, such as PaO2 or SpO2, does not reflect the complexity of the interaction of O2 with other compounds or the occurrence of oxidative stress. Many mechanisms, such as preconditioning effects and adaptations to hypoxia and hyperoxia, are still poorly understood.50 The aforementioned variables that are easily available at the bedside should not obscure the underlying physiological pathway. They are not markers of hyperoxia. PaO2 could help quantify lung exposure to O2, but it also depends on the venous partial pressure in O2, the cardiac output and the pulmonary shunt, which can vary greatly from one patient to another, within the same patient and according to the pathology. Instead, it is much easier to directly consider FIO2.

Conclusion

On a physiological basis, PaO2 is not the holy graal of blood oxygenation assessment; it provides only limited information about the O2 concentration carried by blood and does not allow the determination of whether cells are exposed to hyperoxia. Under normobaric conditions, only the lungs are exposed to hyperoxia, and other organs are relatively protected because of the low solubility of O2 and the hemoglobin carrier. Hypoxia and severe changes in the mitochondrial O2 concentration, as observed during ischemia–reperfusion, are the major sources of ROS and O2 toxicity. This should be considered in the design of future studies that aim to determine the optimal oxygenation target and by clinicians for their daily practice at the bedside.