Physiological Basis of Extracorporeal Membrane Oxygenation and Extracorporeal Carbon Dioxide Removal in Respiratory Failure

Extracorporeal life support (ECLS) for severe respiratory failure has seen an exponential growth in recent years. Extracorporeal membrane oxygenation (ECMO) and extracorporeal CO2 removal (ECCO2R) represent two modalities that can provide full or partial support of the native lung function, when mechanical ventilation is either unable to achieve sufficient gas exchange to meet metabolic demands, or when its intensity is considered injurious. While the use of ECMO has defined indications in clinical practice, ECCO2R remains a promising technique, whose safety and efficacy are still being investigated. Understanding the physiological principles of gas exchange during respiratory ECLS and the interactions with native gas exchange and haemodynamics are essential for the safe applications of these techniques in clinical practice. In this review, we will present the physiological basis of gas exchange in ECMO and ECCO2R, and the implications of their interaction with native lung function. We will also discuss the rationale for their use in clinical practice, their current advances, and future directions.


Introduction
Respiratory failure is a condition in which the respiratory system is unable to maintain adequate gas exchange to satisfy the body's metabolic demands. Of patients with respiratory failure, over 33% receive mechanical ventilation for more than 12 h and a significant proportion (10%) develop Acute Respiratory Distress Syndrome (ARDS) with a mortality of 35-50% [1]. Furthermore, in patients with very large shunt fraction and dead space, hypoxaemia or hypercapnia may be refractory to changes in inspiratory oxygen and high "intensity" mechanical ventilation (i.e., high airway pressures, respiratory rate, mechanical power) [2,3]. In addition, those who develop more severe or life-threatening respiratory failure cannot always be managed safely with conventional mechanical ventilation. In these patients, the use of extracorporeal support via a "membrane lung"-extracorporeal membrane oxygenation (ECMO)-permits to achieve sufficient oxygen delivery and carbon dioxide removal. At the same time, it allows a reduction in the intensity of ventilation, thus diminishing the risk of ventilator-induced lung injury (VILI) and allowing native lung recovery.
The use of ECMO, the number of ECMO centres and the use of mobile ECMO has increased exponentially since 2009 [4] following the H 1 N 1 influenza A pandemic [5,6] reduction in the intensity of ventilation, thus diminishing the risk of ventilator-induced lung injury (VILI) and allowing native lung recovery.
The use of ECMO, the number of ECMO centres and the use of mobile ECMO has increased exponentially since 2009 [4] following the H1N1 influenza A pandemic [5,6] and the publication of the CESAR trial [7]. Peek at al showed greater survival without disability (63% vs. 47%; relative risk 0.69) and greater quality-adjusted life-years (QALYs) at 6 month when ECMO was offered. Several non-randomised, but matched cohort studies showed safety and efficacy of ECMO [5,6,8]. More recently, the largest randomized trial on ECMO (EOLIA Trial) [9] included 249 patients and showed a clinically large-although not statistically significant-reduction in mortality with ECMO (35% vs. 46%; RR, 0.76). A subsequent post-hoc Bayesian analysis of the EOLIA trial showed that early VV-ECMO provides mortality benefit in patients with very severe ARDS as per EOLIA inclusion criteria [10]. This benefit has been further confirmed by a recent meta-analysis which reported a significantly lower 60-day mortality in patients who received ECMO (RR, 0.73 [95% CI, 0.58-0.92]; p = 0.008) [11].
ECMO is a supportive treatment and can represent a "bridge" either to healing of natural organs (bridge-to-recovery), to long-term devices (bridge-to-destination) or to organ replacement (bridge-to-transplantation). Occasionally, when the aetiology is unclear, it becomes a bridge to make a decision by allowing initial stabilisation of the patient and subsequent re-evaluation. Eventually, for some patients, ECMO becomes a bridge to palliation when no further therapeutic options are available.

Definitions
Extracorporeal membrane oxygenation (ECMO) and extracorporeal carbon dioxide removal (ECCO2R) are two modalities of extracorporeal life support (ECLS) which can partially (ECCO2R) or fully (ECMO) substitute the gas exchange function of the native lung ( Figure 1) [12]. (1) In veno-venous extracorporeal membrane oxygenation (VV-ECMO), deoxygenated blood is drained from a central vein (e.g., inferior or superior vena cava), pumped through a "membrane lung" or oxygenator, and then reinfused-fully oxygenated and decarboxylated-into a central vein. The gas exchange is driven by the diffusion gradient across the membrane. Fresh gas (sweep gas flow, SGF) flows, (1) In veno-venous extracorporeal membrane oxygenation (VV-ECMO), deoxygenated blood is drained from a central vein (e.g., inferior or superior vena cava), pumped through a "membrane lung" or oxygenator, and then reinfused-fully oxygenated and decarboxylated-into a central vein. The gas exchange is driven by the diffusion gradient across the membrane. Fresh gas (sweep gas flow, SGF) flows, countercurrent to the blood, in the lumen of the membrane's hollow fibers, therefore maintaining a favourable gradient for oxygenation and decarboxylation. To achieve this, the blood flows used in ECMO range between 3 and 7 L/min.
(2) Extracorporeal carbon dioxide removal (ECCO 2 R) operates in a similar way to VV-ECMO but at much lower blood flow (usually < 1-1.5 L/min). The main difference between VV-ECMO and ECCO 2 R consists in the fact that CO 2 removal has a linear kinetic and is more efficient than oxygenation: for this reason, ECCO 2 R can be delivered at much lower flows and with smaller cannulae. Although ECCO 2 R does not provide significant oxygenation, it is able to remove 25-50% of the metabolically produced carbon dioxide. (3) Veno-arterial extracorporeal membrane oxygenation (VA-ECMO): VA-ECMO provides support for the heart (e.g., biventricular failure). Hybrid configurations for peripheral ECMO (V-VA-ECMO), where there is drainage of the venous system and return into a central artery and a central vein to assist both heart and lung (e.g., severe ARDS and septic cardiomyopathy with cardiogenic shock).
From this point forward when we mention ECMO we intend the veno-venous (VV) configuration.

Oxygen Delivery and Demand
Oxygen demand in critically ill patients can vary between 200 and 450 mL/min depending on the underlying disease and metabolic state. Therefore, to maintain an oxygen delivery to oxygen demand ratio (DO 2 /VO 2 ratio) of~3, the amount of oxygen delivered to the tissues must be 600-1350 mL/min. The total DO 2 is equal to: In patients with ARDS not on VV-ECMO, the CaO 2 depends on the shunt fraction of the native lung, and on the oxygen content of the mixed venous blood: where Qt−Qs is the portion of the cardiac output going through the ventilated lung parenchyma, and Qs is the portion of the cardiac output shunting through non ventilated lung areas. Oxygen consumption (VO 2 ) can be estimated as: In patients fully dependent on ECMO (i.e., with no residual native lung function) the CaO 2 -in its simplest form, without accounting for recirculation (see below)-is: This formula is analogous to the formula (2) where ECMO blood flow is noted as ECBF and CO is the cardiac output of the patient, CvO 2 is the content of oxygen in the venous blood and C post-oxy O 2 content of the blood exiting the oxygenator.
The VO 2 (mL/min) of the membrane lung (VO 2ML ) is proportional to the ECMO blood flow (ECBF) and to the difference between the oxygen content of the blood entering the oxygenator (C pre-oxy O 2 ) and the content of the blood exiting the oxygenator (C post-oxy O 2 ) ( Figure 2):  By rearranging Equation (4): Using this formula, the oxygen content is expressed in terms of the ratio between the ECBF and the cardiac output-in a similar way as the shunt equation of the native lung. ECBF represents the equivalent of Qt-Qs and CO-ECBF is the equivalent of Qs, as blood is effectively shunted from the ECMO.
It becomes clear that oxygen content depends not only on the ECMO blood flow and the content of oxygen in the venous blood (i.e., the relationship between oxygen delivery and tissue oxygen extraction), but also on the ratio between ECBF and the patient's cardiac output. To understand this concept, we have to consider how the venous return, equal to the patient cardiac output, is "split" into two components: (1) one part-equal to the ECBF-will pass through the oxygenator and therefore will return to the right atrium fully saturated with oxygen (Spost-oxy = 100%; Ppost-oxyO2 ~ 60-70 kPa or 450-525 mmHg); (2) the second part of the venous return-which is equal to the amount of flow that exceeds the ECBF (i.e, − ) will have the saturation of the venous blood. Therefore, the mixed venous blood of the patient (the oxygenation of the blood in the pulmonary artery) will be a mixed "weighed average" of the two in a proportion that will depend on: the ratio between ECBF and CO; the venous oxygenation and the functioning of the membrane (i.e., the ability to fully oxygenate the venous blood).

Shunt of the Membrane Lung
Membrane function is determined by its characteristics but, over time, is also affected by "aging": clots accumulation and protein binding will occlude some of the hollow fibres, preventing the flow of fresh gas through them. As a result, part of the membrane lung (similarly to the native lung) will be perfused, but not "ventilated"-i.e., not in contact with the sweep gas flow. This phenomenon will generate a "shunt of the membrane lung", that can be calculated utilising Riley's model of shunt of the natural lung [13][14][15]: Ccapillary-oxyO2 is the oxygen content of the "capillary" blood (in the oxygenator) that is in equilibrium with the "alveolar" gas (in the hollow fibers). It can be calculated from the "alveolar" oxygen partial pressure: By rearranging Equation (4): Using this formula, the oxygen content is expressed in terms of the ratio between the ECBF and the cardiac output-in a similar way as the shunt equation of the native lung. ECBF represents the equivalent of Qt-Qs and CO-ECBF is the equivalent of Qs, as blood is effectively shunted from the ECMO.
It becomes clear that oxygen content depends not only on the ECMO blood flow and the content of oxygen in the venous blood (i.e., the relationship between oxygen delivery and tissue oxygen extraction), but also on the ratio between ECBF and the patient's cardiac output. To understand this concept, we have to consider how the venous return, equal to the patient cardiac output, is "split" into two components: (1) one part-equal to the ECBF-will pass through the oxygenator and therefore will return to the right atrium fully saturated with oxygen (S post-oxy = 100%; P post-oxy O 2~6 0-70 kPa or 450-525 mm Hg); (2) the second part of the venous return-which is equal to the amount of flow that exceeds the ECBF (i.e, CO − ECBF) will have the saturation of the venous blood. Therefore, the mixed venous blood of the patient (the oxygenation of the blood in the pulmonary artery) will be a mixed "weighed average" of the two in a proportion that will depend on: the ratio between ECBF and CO; the venous oxygenation and the functioning of the membrane (i.e., the ability to fully oxygenate the venous blood).

Shunt of the Membrane Lung
Membrane function is determined by its characteristics but, over time, is also affected by "aging": clots accumulation and protein binding will occlude some of the hollow fibres, preventing the flow of fresh gas through them. As a result, part of the membrane lung (similarly to the native lung) will be perfused, but not "ventilated"-i.e., not in contact with the sweep gas flow. This phenomenon will generate a "shunt of the membrane lung", that can be calculated utilising Riley's model of shunt of the natural lung [13][14][15]: Membranes 2021, 11, 225 5 of 17 C capillary-oxy O 2 is the oxygen content of the "capillary" blood (in the oxygenator) that is in equilibrium with the "alveolar" gas (in the hollow fibers). It can be calculated from the "alveolar" oxygen partial pressure: Changes in the membrane lung performance can be detected by monitoring P post-oxy O 2 , VO 2ML and the Qs/ECBF over time.

Recirculation
If we go one step further: the amount of oxygen entering the membrane lung (ECBF × C pre−oxy O 2 ) derives from two sources. The first one is represented by the systemic venous blood from the tissues (ECBF − Qr) × CvO 2 . The second one is constituted by recirculating blood flow ( Q r × (C post−oxy O 2 )): a variable portion of the fully oxygenated blood ejected by the return cannula is immediately re-aspirated by the drainage cannula; this is due to physical proximity of the two cannulae. Recirculation is defined as the fraction of the ECBF that derives from the blood ejected from the return cannula-which is already oxygenated and decarboxylated by the first passage through the membrane lung. Recirculation raises the oxygen content of the blood entering the membrane lung and therefore reduces the efficiency of the gas transfer; firstly, because the oxygen gradient across the membrane will be reduced, and secondly because the effective ECBF (i.e., the ECBF that reaches the right atrium) is no longer the entirety of the ECBF but ECBF − Qr. These two effects will jointly reduce the VO 2ML (Equation (5)).
Clinically significant recirculation should be suspected when, despite sweep gas flow F d O 2 = 1: • Low SaO 2 and poor increase in SaO 2 at higher ECBF • Increase in S pre-oxy O 2 and reduction in SaO 2 over time • P pre-oxy O 2 > 10% of P post-oxy O 2 • S pre-oxy O 2 > 75% with SaO 2 < 85%

Gas Exchange in Partially Dependent Patients
As already discussed, the oxygen reaching the right atrium is the result of the admixture of blood coming from the membrane lung and blood coming from the tissues. If we also account for recirculation, we will have: In patients partially dependent on ECMO, the blood flow distributes once again through the pulmonary arteries. A portion perfuses the healthy lung, capable of gas exchange, and it equilibrates with the PO 2 of the open ventilated alveoli (CcO 2 × (Qt − Qs)). However, part of the blood flow will perfuse the diseased non-ventilated (shunted) lung areas (CvmixO 2 × Qs). Once more, the arterial oxygen content (CaO 2 ) will result from the sum of the two contents (10): The content of the non-shunted fraction of the cardiac output (left term of the addition) and the content of the shunted fraction (right term of the addition) Based on the above equations, we can conclude that the native lung shunt fraction is the major determinant of the arterial oxygen content in partially dependent patients. Changes in cardiac output, oxygen consumption and haemoglobin concentration will affect the oxygen content in venous blood. A lower systemic venous oxygen content reduces the oxygen content in the pulmonary artery after mixing with the blood coming from the membrane lung, but increases the oxygen transfer in the ML, given that a lower venous saturation facilitates oxygen uptake ( Figure 3). The relationship between ECBF and cardiac output is important as with maximal ECBF, decreasing the cardiac output (e.g., with the use of beta-blockers, lower temperature, sedation) can improve oxygenation in conditions when the cardiac output is high, and the oxygen extraction is impaired (e.g., fever or sepsis and relatively preserved high central venous oxygenation). However, clinicians need to be aware that a reduction in cardiac output can decrease total oxygen delivery and can therefore reduce venous oxygen content. The combination of reduced cardiac output (which should lead to increased PaO 2 ) and decreased saturation of the venous blood (which should lead to decreased PaO 2 ) can lead to a variable effect on the arterial oxygen content and partial pressure.
Membranes 2021, 11, x 6 of 16 fever or sepsis and relatively preserved high central venous oxygenation). However, clinicians need to be aware that a reduction in cardiac output can decrease total oxygen delivery and can therefore reduce venous oxygen content. The combination of reduced cardiac output (which should lead to increased PaO2) and decreased saturation of the venous blood (which should lead to decreased PaO2) can lead to a variable effect on the arterial oxygen content and partial pressure. . This is influenced by the degree of shunted deoxygenated blood from native lung, the ratio of ECBF to intrinsic cardiac output and amount of recirculation. Essentially, venous blood is divided into blood that is pumped through the ECMO membrane (ECBF) and an amount (Cardiac output minus ECBF) that bypasses the ECMO and mixes in the right atrium. Part of the ECBF (Qr) 'recirculates' back into the ECMO, decreasing the effective ECBF. Therefore, the effective ECBF (ECBFeff) is the difference between the ECBF and Qr. The oxygen content in the different compartments is illustrated.

Determinants of CO2 Removal
Carbon dioxide transfer across the membrane lung is dependent on: (1) blood flow; (2) sweep gas flow; (3) membrane lung surface; (4) PCO2 gradient between the blood entering the membrane and the sweep gas flow [16,17]. We will discuss each individual element in the ECCO2R section, however it is worth pointing out here that, at the ECBF rates generally used in ECMO, CO2 removal-and therefore PaCO2-are dependent on the SGF rate. The SGF is effectively analogous to the minute ventilation in the native lung and its rate is in fact the only component used in clinical practice to adjust ECMO CO2 clearance.
ECMO can remove all of the metabolically produced CO2 and therefore maintain PaCO2 within normal range even in the absence of native lung ventilation (apnoea). The CO2 content of the venous blood is approximately 500 mL/L (400-450 mL/L in arterial blood), whilst the O2 content is 150 mL/L (200 mL/L in arterial blood). Therefore, in a perfectly efficient system the total CO2 production (VCO2) of approximately 250 mL/min could be removed by a blood flow of 0.5 L/min [16,[18][19][20][21][22]. However, the O2 uptake of the same blood volume would only be 25 mL/min. This makes the respiratory quotient (VCO2/VO2) of the membrane lung highly variable and dependent upon the extracorporeal blood flow [23][24][25][26].
The total amount of CO2 removed will also depend on the function of the native lung. Therefore, the total amount of CO2 removed (VCO2 TOT) is: . Schematic overview of the determinants of arterial PaO 2 . This is influenced by the degree of shunted deoxygenated blood from native lung, the ratio of ECBF to intrinsic cardiac output and amount of recirculation. Essentially, venous blood is divided into blood that is pumped through the ECMO membrane (ECBF) and an amount (Cardiac output minus ECBF) that bypasses the ECMO and mixes in the right atrium. Part of the ECBF (Qr) 'recirculates' back into the ECMO, decreasing the effective ECBF. Therefore, the effective ECBF (ECBFeff) is the difference between the ECBF and Qr. The oxygen content in the different compartments is illustrated.

Determinants of CO 2 Removal
Carbon dioxide transfer across the membrane lung is dependent on: (1) blood flow; (2) sweep gas flow; (3) membrane lung surface; (4) PCO 2 gradient between the blood entering the membrane and the sweep gas flow [16,17]. We will discuss each individual element in the ECCO 2 R section, however it is worth pointing out here that, at the ECBF rates generally used in ECMO, CO 2 removal-and therefore PaCO 2 -are dependent on the SGF rate. The SGF is effectively analogous to the minute ventilation in the native lung and its rate is in fact the only component used in clinical practice to adjust ECMO CO 2 clearance.
ECMO can remove all of the metabolically produced CO 2 and therefore maintain PaCO 2 within normal range even in the absence of native lung ventilation (apnoea). The CO 2 content of the venous blood is approximately 500 mL/L (400-450 mL/L in arterial blood), whilst the O 2 content is 150 mL/L (200 mL/L in arterial blood). Therefore, in a perfectly efficient system the total CO 2 production (VCO 2 ) of approximately 250 mL/min could be removed by a blood flow of 0.5 L/min [16,[18][19][20][21][22]. However, the O 2 uptake of the same blood volume would only be 25 mL/min. This makes the respiratory quotient (VCO 2 /VO 2 ) of the membrane lung highly variable and dependent upon the extracorporeal blood flow [23][24][25][26]. The total amount of CO 2 removed will also depend on the function of the native lung. Therefore, the total amount of CO 2 removed (VCO 2 TOT ) is: At equilibrium and for any given PaCO 2 , the total CO 2 removed is the sum of VCO 2 ML + VCO 2 NL and is equal to the metabolic CO 2 . According to Equation (12), an increase in VCO 2ML will allow a reduction of VCO 2 NL and alveolar ventilation with consequent reduction in the mechanical ventilation load.

Carbon Dioxide Transfer across the Membrane
Two methods can be used to calculate the amount of CO 2 removed by the membrane lung (VCO 2ML ): (1) The first method is based on the difference in CO 2 content of the blood at each side of the membrane: Equation (13): VCO 2ML (mL/min) as calculated from the trans-membrane CO 2 content difference. C post-oxy CO 2 is the CO 2 content in the blood that exits the membrane lung (mmol/L), C pre-oxy CO 2 is the CO 2 content in blood entering the membrane (mmol/L). Blood flow is measured in L/min and the correction factor (25) is in mL/mmol.
(2) Alternatively, we can measure the concentration of CO 2 from the sweep gas outlet of the membrane: Equation (14): VCO 2ML (mL/min) is calculated from the partial pressure of CO 2 in the effluent gas, PCO 2ML(exp) (kPa). The correction factors are 7.5 mm Hg/kPa and 713 mm Hg (barometric pressure at the sea level minus vapour pressure); SGF is measured in L/min.
In order to describe the VCO 2 in standardised conditions, corrected VCO 2 can be expressed as: Equation (15): VCO 2ML (corr) is corrected for an inlet PCO 2 of 6 kPa using the Douglas equation. VCO 2 (corr) is measured in mL/min. PCO 2ML (pre) is the partial pressure of CO 2 content in blood before the membrane (kPa).

Membrane Dead Space
Membrane lungs are known to be affected by inequalities in the distribution of sweep gas and blood flows leading to local ventilation-perfusion mismatch [27]. Dead space in the membrane lung (i.e., proportion of the membrane receiving sweep gas flow but not in contact with blood flow) can be caused by water saturation or clotting of the gas capillaries. Analogously to the dead space of the NL, the dead space for the ML can be calculated as follows: DS ML = P post−oxy CO 2 − P exp CO 2 P post−oxy CO 2 (15)

Monitoring of Gas Exchange
The following measurements are useful tools to track the progress of an individual patient and to assess the relative contribution of the membrane and native lung: • A pre-oxygenator blood gas • A post-oxygenator blood gas These should be assessed daily-from these values, it is then possible to calculate all the variables described in the ECMO physiology section.

ECMO Initiation: Physiological Considerations
Once a patient is placed on ECMO, the native lung function may deteriorate as a consequence of two main physiological phenomena: (1) Abolition of hypoxic vasoconstriction due to the increased oxygen in the mixed venous blood as a consequence of ECMO; (2) Changes in the respiratory quotient of the natural lung (that is consequent to the decarboxylated blood that arrives to the pulmonary artery); The abolition of the hypoxic vasocontriction increases the physiological shunt of the native lung (Figure 4), while the reduction in the respiratory quotient (RQ) of the natural lung causes of a progressive fall in the alveolar PO 2 based on the modified alveolar gas equation: Membranes 2021, 11, x 8 of 16 • Native lung minute ventilation • VCO2 native lung (via volumetric capnography) These should be assessed daily-from these values, it is then possible to calculate all the variables described in the ECMO physiology section.

ECMO Initiation: Physiological Considerations
Once a patient is placed on ECMO, the native lung function may deteriorate as a consequence of two main physiological phenomena: (1) Abolition of hypoxic vasoconstriction due to the increased oxygen in the mixed venous blood as a consequence of ECMO; (2) Changes in the respiratory quotient of the natural lung (that is consequent to the decarboxylated blood that arrives to the pulmonary artery); The abolition of the hypoxic vasocontriction increases the physiological shunt of the native lung (Figure 4), while the reduction in the respiratory quotient (RQ) of the natural lung causes of a progressive fall in the alveolar PO2 based on the modified alveolar gas equation: Figure 4. ECMO initiation. The delivery of VV-ECMO optimises oxygenation and allows a reduction of mechanical power delivered to the lungs. ECMO produces venous hyperoxia which in turn reduces pulmonary vascular resistance and hypoxic pulmonary vasoconstriction. This improves right ventricular function and cardiac output. A consequence of this physiological change is a larger shunt fraction in the native lung and a lower ECBF/CO ratio, both resulting in a lower arterial oxygen tension. This is further compounded by "lung rest" strategies leading to progressive de-recruitment and alveolar hypoventilation. In order to maintain adequate oxygenation ECBF may need to be increased to re-balance this ratio.

Weaning, "Trial Off" and Decannulation
As the condition causing respiratory failure improves, the ECMO flow rate is reduced to a minimum of 2.5-3 L/min (to minimise the risk of thrombus formation at lower ECBF), while the SGF can be gradually reduced until it is turned off. In this phase, Figure 4. ECMO initiation. The delivery of VV-ECMO optimises oxygenation and allows a reduction of mechanical power delivered to the lungs. ECMO produces venous hyperoxia which in turn reduces pulmonary vascular resistance and hypoxic pulmonary vasoconstriction. This improves right ventricular function and cardiac output. A consequence of this physiological change is a larger shunt fraction in the native lung and a lower ECBF/CO ratio, both resulting in a lower arterial oxygen tension. This is further compounded by "lung rest" strategies leading to progressive de-recruitment and alveolar hypoventilation. In order to maintain adequate oxygenation ECBF may need to be increased to re-balance this ratio.

Weaning, "Trial off" and Decannulation
As the condition causing respiratory failure improves, the ECMO flow rate is reduced to a minimum of 2.5-3 L/min (to minimise the risk of thrombus formation at lower ECBF), while the SGF can be gradually reduced until it is turned off. In this phase, the patient is essentially "off ECMO" and can be decannulated if the gas exchange is stable and the ventilation is lung protective after 12-24 h. It is important to have a standardised weaning plan in place [28]. The pre-requisites for weaning assessment are haemodynamic stability and spontaneous or assisted ventilatory mode (e.g., cPAP/PS). Other additional parameters are: PaO 2 > 30 kPa (225 mm Hg) after a 100% test (i.e., the systemic arterial PaO 2 taken after 15 min on FiO 2 = 1.0); P0.1 < 5 cmH 2 O.
Step 1: Decrease F d O 2 FdO 2 is decreased from 100% to 60%, 30% and 21% in 5-10 min steps. A peripheral oxygen saturation (SpO 2 ) > 88% and P0.1 < 5 cm H 2 O will need to be maintained throughout the test. If this step is successful, FdO 2 can be kept at 21%, and the test can proceed to step 2.
Step 2: Reduction in Sweep Gas Flow The SGF is reduced by 30% every 5-10 min, whilst monitoring the patient's respiratory efforts. Weaning failure and test interruption are indicated by SpO 2 < 88%, RR > 35 bpm, P0.1 > 10 cm H 2 O, oesophageal pressure swings ≥15 cm H 2 O (when available), or if any signs of distress/instability are evident. If the patient's response remains within set limits at 0 SGF, the weaning test is successful, and the clinical team will consider whether to remain off ECMO or reintroduce a variable degree of extracorporeal support pending decannulation.
3.6. Physiology of ECCO 2 R ECCO 2 R, unlike ECMO, provides decarboxylation without oxygenation. CO 2 removal is a more efficient process compared to oxygenation and can be achieved with lower blood flow rates (0.5-1.5 L/min) than the ones required for ECMO (3-7 L/min). For this reason, ECCO 2 R is sometimes referred to as "low-flow ECMO" [18,29]. The possibility of employing low flows allows the use of smaller size cannulae. However, at the same time, there is an increased thrombotic and haemolytic risk. ECCO 2 R has yet to find a definite place in standard care, with its use mainly limited to case series or within clinical trials.
The artificial membrane lung allows for dissociation between VCO 2 and VO 2 . In the native lung, VO 2 and VCO 2 are physiologically strictly associated, with the metabolic quotient (R = VCO 2 /VO 2 ) ranging between 0.7 and 1. In the presence of extracorporeal support, CO 2 removal and O 2 supply become independent given that R for the membrane lung may range between 0 and infinity, depending on the characteristics and composition of blood and gas flows. R = 0 is the equivalent of "apnoeic oxygenation of the membrane lung", where there is no removal of CO 2 in the presence of positive oxygen supply. Conversely, R = infinity is a consequence of extracorporeal CO 2 removal in the absence of oxygenation via the membrane lung. This is achieved when the PO 2 of the SGF and the PO 2 of the venous blood perfusing it (inflow PO 2 ) are equal. In these conditions, ECCO 2 R provides an R close to infinity, as it primarily removes CO 2 while meaningful oxygenation is provided only by the native lung.
We will now focus on the physiological basis of carbon dioxide removal, exploring the ways in which it differs from the oxygenation process and the clinical implications that follow.

Principles of CO 2 Diffusion and Transport
The total CO 2 stores in the body are about 1.7-1.8 L/kg (e.g., 122-123 L of CO 2 in a 70 kg individual), distributed within the lungs, blood, and other tissues, most of it present in the bones. The CO 2 transported in the blood-in the form of bicarbonate ions, carbamino compounds, carbonic acid and a small amount of physical CO 2 -only accounts for 2.5-2.7 L. The vast majority of total body CO 2 is contained in body tissues as bicarbonate, carbamino-CO 2 and carbonate. This CO 2 reserve is in continuous equilibrium with the blood component, which is the fraction of CO 2 directly mobilised in the form of volatile gas by the natural lung and, when present, the membrane lung.
The use of low blood flow for ECCO 2 R is justified by physicochemical principles that make CO 2 removal a much more efficient process than oxygenation. Carbon dioxide is more soluble and diffusible than oxygen. CO 2 enters blood from the tissues and combines with water, forming carbonic acid (H 2 CO 3 ), which in turn dissociates into hydrogen (H + ) and bicarbonate ions (HCO 3 − ). This reaction is catalyzed by carbonic anhydrase (CA), an enzyme present inside the red blood cell. As a result, most of the CO 2 is transported in blood as bicarbonate: The reverse reaction, that generates CO 2 from HCO 3 − , follows linear kinetics and, unlike the bonding of oxygen to haemoglobin, does not become saturated [29]. Therefore, CO 2 diffuses more efficiently than O 2 and is not affected by haemoglobin concentration [30].
Haemoglobin plays an important role in the transport of CO 2 as deoxyhaemoglobin increases the amount of carbon dioxide that is carried in venous blood by working as proton recipient.
The binding of CO 2 to Hb has relevant applications in increasing the efficiency of extracorporeal CO 2 removal in low-flow systems.

Determinants of CO 2 Transfer
The main determinants of CO 2 transfer across the artificial membrane (VCO 2ML ) are: Typically, during high-flow veno-venous extracorporeal support, VCO 2ML increases linearly with gas flow and logarithmically with blood flow for a given inflow pCO 2 [31]. In lower ECBF systems in the range of 0.1 to 0.5 L/min, an increase in ECBF will determine a linear increase in the CO 2 removal (see also Equation (13)) [32]. As mentioned earlier, given that the CO 2 content in 1L of venous blood is approximately 500 mL and the CO 2 production is around 250 mL/min, an ECBF of 0.5 L/min would be theoretically sufficient to match the metabolic CO 2 production. In reality, no system is perfectly efficient, and the currently available devices can achieve the removal of about 50-60% of the total CO 2 production at best.
Despite the theoretical linearity of their relationship, the impact of blood flow on CO 2 clearance is limited by the membrane size: if the surface area is not sufficient, increments in blood flow will not result in greater CO 2 removal. The reverse is also true: low blood flow would not allow clinically significant CO 2 removal regardless of the membrane size [33].

Sweep Gas Flow (SGF)
SGF is one of the key determinants of CO 2 transfer across the membrane lung: increments in SGF provide greater CO 2 clearance. However, it is worth noting that SGF and VCO 2 are not linearly related and this becomes especially apparent in low blood flow conditions: unlike ECMO, where SGF is the main parameter used to control CO 2 removal, increasing the SGF in ECCO 2 R devices above a certain threshold (usually 4-5 L/min) does not result in further CO 2 removal [34]. This ceiling effect is mainly determined by ECBF and membrane size.

Membrane Characteristics
Membrane characteristics are another main determinant of CO 2 transfer. The surface area has the greatest impact [33], but also membrane's geometry and material affect efficiency of the process. Larger membranes are preferable in terms of CO 2 clearance, given that a bigger surface is available for gas exchange. However, the larger the membrane, the greater its thrombotic risk. On the contrary, if the membrane is too small, there is higher blood velocity and increased potential for haemolysis.
In the experimental setting-where the measurement of VCO 2 is more common than in clinical practice-the obtained VCO 2ML ranged between 35 and 75 mL/min for devices of 0.32 m 2 [35] to 0.67 m 2 [36], while it was 86 [37] to 170 mL/min when the lung surface was 1.8 m 2 [32].
Membrane design can vary, too, with the sweep gas flow being either countercurrent or cross-current to the blood. Immobilized carbonic anhydrase on hollow fibers within membrane lungs can also be used to accelerate the conversion of carbonic acid into CO 2 and water [38,39].

Transmembrane pCO 2 Gradient
The difference in pCO 2 between blood and sweep gas is what drives the diffusion process, according to Fick's diffusion law. Since CO 2 concentration in the SGF is either very low or completely absent, the partial pressure of CO 2 in the venous blood (PvCO 2 ) captured by the inlet cannula of the ECCO 2 R (inflow PCO 2 ) is what ultimately determines the diffusion gradient across the membrane lung. As CO 2 diffuses and achieves equilibrium almost instantaneously, SGF rate is important to keep CO 2 low on the gas side of the membrane.
For a given ECBF rate and membrane size, we can say that the VCO 2 depends on the degree of hypercapnia-or transmembrane CO 2 gradient. Indeed, a recent animal study on the use of an ECCO 2 R with membrane area of 1.8 m 2 showed that the VCO 2 was independent from SGF, while it increased linearly with ECBF. Inflow pCO 2 (which drives the transmembrane PCO 2 gradient) and ECBF were the primary determinants of VCO 2 , which was not affected by increasing gas flow above 4 L/min. This is most likely explained by the high efficiency of CO 2 removal in the range of ECBF (100-400 mL/min) and inflow pCO 2 (30-80 mm Hg) used in that study, as demonstrated by outflow pCO 2 being close to its asymptote in all given conditions (approximately 1.3 kPa) [32].

CO 2 Removal Rate
To maximize CO 2 removal rate, research efforts are under way to artificially promote CO 2 transfer by either driving the reaction (Equation (18)) towards increased volatile CO 2 for techniques involving gaseous removal, or towards HCO 3 − for approaches based on decarboxylation via a liquid medium.

Gaseous CO 2 Removal
The partial pressure of CO 2 is only 5% of total CO 2 content, corresponding to 12.5 mL of dissolved CO 2 in 500 mL of blood. The conversion of bicarbonate and carbonic acid into CO 2 is a slow reaction which is not sufficient to promote further CO 2 removal [40]. An artificial increase in CO 2 removal can be promoted by increasing FdO 2 from 21% to 100% in the sweep gas, promoting the displacement of CO 2 from HCO 3 − through the Haldane effect [41].
Blood acidification represents another approach. Acidification of haemoglobin by oxygen binding has two main effects. Firstly, it facilitates CO 2 -Hb dissociation by reducing the binding affinity of haemoglobin for carbon dioxide, hence increasing the amount of CO 2 available for gas exchange in the membrane. Secondly, it promotes the formation of carbon dioxide and water (Equation (18)), increasing again the amount of physical CO 2 (PCO 2 ) available.
The dilution of sulfur dioxide (SO 2 ) in oxygen sweep gas acidifies blood, displacing the bicarbonate to CO 2, and allows 17% more CO 2 removal while maintaining blood pH within physiological range [42].
Respiratory electrodialysis is a regional acidification technique which is able to achieve removal of 50% of total CO 2 metabolism production with a blood flow of 250 mL/min, without systemic acidification. However, this strategy remains experimental and is not yet available on the market [43].

Liquid CO 2 Removal
Another strategy to remove CO 2 is to drive the same reaction (Equation (18)) backwards by removing HCO 3 − using a liquid alkaline dialysate with zero bicarbonate. Federspiel and colleagues were able to remove up to 117 mL/min of CO 2 (inlet pCO 2 100 mm Hg and blood flow 421 mL/min) without affecting the systemic pH [44]. Later, using an innovative membrane with an integrated impeller in the oxygenator, they were able to remove 74 mL/min of CO 2 with a blood flow of 250 mL/min and minimum haemolysis [41].

Potential Clinical Applications
Given its strong physiological rationale, ECCO 2 R undoubtedly represents an appealing technique, but one that still lacks sufficient evidence to support widespread clinical use [45]. Two main areas have been proposed for its clinical application: (1) acute respiratory distress syndrome (ARDS), where ECCO 2 R allows protective lung ventilation strategies; (2) acute type 2 respiratory failure, where hypercapnia represents the primary problem.
(1) Gattinoni and Kolobow first described the use of ECCO 2 R in ARDS in the late 1970s [23,24,26], following the affirmation of the "baby lung" concept and the development of lung protective ventilation strategies [46,47]. ECCO 2 R was suggested as an adjunctive tool that enabled ventilation with low tidal volumes (V T 6-8 mL/Kg PBW) to prevent further lung injury, whilst addressing the hypercapnia resulting from a reduced minute ventilation. However, protective lung ventilation remains feasible in many situations without the use of ECCO 2 R, particularly when tolerating a moderate degree of hypercapnia ("permissive hypercapnia"). More recently, the role of ECCO 2 R in ARDS has gained renewed interest [48][49][50], often in parallel with the predicament of ultra-protective lung ventilation (V T 3-4 mL/Kg PBW) [51][52][53]. Such approach does not provide adequate CO 2 clearance and an alternative strategy for CO 2 removal, e.g., extracorporeal support, becomes necessary. (2) Another potential application is in acute type 2 respiratory failure, such as acute exacerbations of obstructive airway disease (COPD and severe acute asthma). The postulated role for ECCO 2 R is to correct the acute element of respiratory acidosis, allowing the avoidance of intubation in patients with COPD exacerbation, or facilitating extubation in those patients already on invasive ventilation [54][55][56][57]. A theoretical benefit has also been suggested in severe acute asthma [58].

Effects of ECCO 2 R on Intensity of Mechanical Ventilation
The rationale for ECCO 2 R use in ARDS is to allow ultra-protective lung ventilation, with lower tidal volumes and lower driving pressures. These are both associated with improved survival. Goligher and co-authors describe how to predict changes in minute ventilation and driving pressure after applying ECCO 2 R [59]. The following equation is based on a theoretical analysis of the equation that describes the relationship between alveolar ventilation and CO 2 clearance: Equation (20): DP is the predicted difference in driving pressure following initiation of ECCO 2 R. C RS , respiratory compliance. V D alv/ V T , alveolar dead space fraction. PaCO 2 , CO 2 partial pressure in arterial blood. VCO 2ML , amount of CO 2 cleared by membrane lung. k, correction factor (0.863).
According to this equation, the predicted change in driving pressure-at a given amount of extracorporeal CO 2 removal-will be greater in those patients that have lower respiratory compliance and higher alveolar dead space fraction. To validate their hypothesis, they tested the prediction model utilising data from the SUPERNOVA trial [60], a pilot trial that aimed to assess efficacy and safety of ECCO 2 R in achieving ultra-protective lung ventilation in ARDS. The accuracy of the model was found to be only moderate, however the study confirmed that ECCO 2 R reduces the requirements for tidal volume and driving pressure in proportion to respiratory system compliance and alveolar dead space fraction [59]. The authors suggest important implications for future clinical trials: patients could be enrolled on the basis of their likelihood of obtaining benefit from the intervention-what is known as predictive enrichment. Namely, the enrollment criteria for ECCO 2 R in ARDS would include low respiratory compliance and high alveolar dead space fraction, as demonstrated above. This should increase the statistical power of trials and allow a smaller sample size. Ongoing research in this direction might help to define a specific clinical role for this extracorporeal modality.

Conclusions
The use of extracorporeal life support in respiratory failure is growing rapidly, both in numbers and indications. Understanding the physiology of native and membrane lungthe individual role they play as well as their interaction at different stages of disease and recovery-is key to successful use of ECLS in clinical practice.
Funding: This research received no external funding.

Conflicts of Interest:
The authors declare no conflict of interest. Pre-oxygenator O 2 partial pressure P post-oxy O 2 Post-oxygenator O 2 partial pressure PAO 2 Alveolar O 2 partial pressure PaCO 2 Arterial CO 2 partial pressure C pre-oxy CO 2 Pre-oxygenator CO 2 content C post-oxy CO 2 Post-oxygenator CO 2 content P post-oxy CO 2 Post-oxygenator CO 2 partial pressure PCO 2ML (exp) CO 2 partial pressure in the membrane effluent gas PCO 2ML (pre) Pre-oxygenator (or pre-membrane) CO 2 partial pressure P exp CO 2 CO 2 partial pressure in the membrane effluent gas FiO 2 Fraction of inspired O 2