Bench to bedside review: Extracorporeal carbon dioxide removal, past present and future

Acute respiratory distress syndrome (ARDS) has a substantial mortality rate and annually affects more than 140,000 people in the USA alone. Standard management includes lung protective ventilation but this impairs carbon dioxide clearance and may lead to right heart dysfunction or increased intracranial pressure. Extracorporeal carbon dioxide removal has the potential to optimize lung protective ventilation by uncoupling oxygenation and carbon dioxide clearance. The aim of this article is to review the carbon dioxide removal strategies that are likely to be widely available in the near future. Relevant published literature was identified using PubMed and Medline searches. Queries were performed by using the search terms ECCOR, AVCO2R, VVCO2R, respiratory dialysis, and by combining carbon dioxide removal and ARDS. The only search limitation imposed was English language. Additional articles were identified from reference lists in the studies that were reviewed. Several novel strategies to achieve carbon dioxide removal were identified, some of which are already commercially available whereas others are in advanced stages of development.


Introduction
Th e reported incidence of acute respiratory distress syndrome (ARDS) ranges from 7 to 59 per 100,000 people [1,2], and is associated with a mortality rate of 40 to 45%. Th is rate remains unacceptably high despite the introduction of lung protective ventilation and, although hospital mortality may be slowly decreasing, ICU and 28 day mortality have remained constant [1,3]. Failure to implement lung protective ventilation (LPV) may be one of the reasons ICU mortality rates have remained unchanged [4][5][6]. When surveyed, health care providers reported that hypercapnia or its related eff ects were signifi cant barriers to achieving LPV [7]. Hypercapnia complicated 14% of patients in the large ARDS network on the use of LPV [8]. However, patients with a high risk of death were excluded. In a study of severe ARDS, where tidal volumes were adjusted to target a mean airway pressure less than 28 cmH 2 O, all patients experienced hypercapnia [9]. As evidence emerges that tidal volumes <6 ml.kg -1 might further reduce mortality [9,10], alternative strategies to manage the inevitable hypercapnia must be considered.
Permissive hypercapnia is one approach, but it only improves mortality when patients are ventilated with high tidal volumes [8]. Such volumes should no longer be used since 6 ml.kg -1 is superior to 12 ml.kg -1 and <4 ml.kg -1 might be superior to 6 ml.kg -1 [9][10][11]. Although hypercapnia might have benefi cial eff ects on oxygen delivery and attenuation of infl ammation [12], it also harms injured lung through immunosuppression and impaired pulmonary epithelial repair [13,14]. Furthermore, hypercapnia perpetuates right heart failure [15] and is undesirable in patients with elevated intracranial pressure. An alternative strategy to manage hypercapnia is extracorporeal carbon dioxide removal (ECCOR), a technology pioneered four decades ago [16] but only recently readily accessible through commercialization of several novel devices. ECCOR therefore deserves a fresh look and this review aims to provide an overview of devices currently available and those that may be available in the near future.

ECCOR in principle
ECCOR is designed to remove carbon dioxide (CO 2 ) and, unlike extracorporeal membrane oxygen (ECMO), does not provide signifi cant oxygenation. A discussion of ECMO is beyond the scope of this article but is well reviewed elsewhere [17,18]. In its simplest form, ECCOR consists of a drainage cannula placed in a large central vein, a pump, a membrane lung and a return cannula ( Figure 1). Blood is pumped through the membrane 'lung' and CO 2 is removed by diff usion. Membrane lungs are permeable to gases but not liquids. A fl ow of gas containing little or no CO 2 runs along the other side of the membrane, ensuring the diff usion gradient favors CO 2 removal.
In contrast to ECMO, where the need for oxygenation requires high blood fl ow rates, ECCOR allows much lower blood fl ow rates, a result of major diff erences in CO 2 and oxygen (O 2 ) kinetics. First, almost all the O 2 in blood is carried by hemoglobin, which displays sigmoidal saturation kinetics. Assuming normal hemoglobin and venous O 2 , each liter of venous blood can only carry an extra 40 to 60 ml of O 2 before the hemoglobin is saturated. Blood fl ows of 5 to 7 L.minute -1 are therefore required to supply enough O 2 for an average adult (250 ml. minute -1 ). Conversely, most CO 2 is transported as dissolved bi carbo nate, displaying linear kinetics without saturation. Th us, 1 L of blood is capable of carrying more CO 2 than O 2, and 250 ml of CO 2 can be removed from <1 L of blood. Second, CO 2 diff uses more readily than O 2 across extracorporeal membranes because of greater solubility [17].

The membrane lung
Th e membrane lung made long-term extracorporeal gas exchange feasible. Before membrane lungs, extracor por eal circuits achieved gas exchange by creating a direct airblood interface, either bubbling air through blood or creating a thin fi lm of blood on the surface of a rotating cylinder/disc. However, blood-air interfaces denature proteins, activate clotting and infl ammatory pathways, and damage circulating cells [19]. Consequently, devices relying on blood-air interfaces cannot be used more than a few hours without serious complications.
Th e concept of placing a barrier between blood and air began with the observation that gas exchange occurred across cellophane tubing in hemodialysis machines [20]. Th is led to the development of membrane lungs consisting of gas permeable silicone-rubber mounted on a nylon mesh [21]. Th e nylon mesh provided structural strength and decreased leakage from random pinhole defects, which occur during the manufacture of thin siliconerubber membranes [19]. Th ree major factors determine the amount of gas crossing membranes: the diff usion gradient, the membrane-blood contact time and the membrane diff usion characteristics.
Th e CO 2 diff usion gradient is determined by the CO 2 content of the blood and the air passing through the membrane lung, as well as the speed of the airfl ow. Membrane-blood contact time is determined by membrane geometry. In early devices, Th eodore Kolobow arranged the membrane into a coil [22] and used a fabric with an irregular surface, increasing the surface area [23]. Hollow fi ber membranes have now replaced coiled silicon-rubber membranes. Early fi bers were constructed with microporous polypropylene. Micropores create microscopic blood-gas interfaces allowing effi cient gas exchange, but also cause plasma leak. Recently, nonmicroporous poly-4-methyl-1-pentene (PMP) has been used; it provides superior gas exchange, better bio compatibility and is less susceptible to plasma leak [24][25][26]. Adding covalently bound heparin to membrane surfaces enhances biocompatibility, and gas exchange has been improved by arranging fi bers into a complex mat and running blood on the outside [27] (Figure 2). Th is arrange ment allows perpendicular blood fl ow to the fi bers, improving mass transfer by reducing the diff usion path length compared to parallel fl ow. Modern membrane lungs achieve adequate gas exchange with surface areas of 1 to 3 m 2 ( Table 1).

The pump
Blood fl ow through ECCOR circuits can be achieved in one of two ways. In patients with suffi cient arterial pressure, a pumpless system can be used where blood is driven out of an arterial cannula by high arterial pressures and returned through a venous cannula, often called arterio venous CO 2 removal (AVCO2R). Pumpless systems result in less blood trauma, but require large bore arterial cannulas and an adequate cardiac output. Th e alternative is to use a mechanical pump.
Early devices used roller or peristaltic pumps. Although cheap and reliable, these pumps were prone to blood trau ma -for example, hemolysis -from compression and heating of blood components. Blood trauma is less of a problem at lower blood fl ow rates -for example, those used in dialysis. Th e introduction of rotary pumps has resulted in simpler yet eff ective systems that cause less blood trauma. Two main types of rotary pumps are used in ECCOR devices, centrifugal and diagonal fl ow pumps. Centrifugal pumps use a radial rotating impeller to create a suction vortex that draws blood into the center of the pump and spins it outwards, imparting centrifugal momentum, which is converted into driving pressure. In diagonal fl ow pumps, impellor design is a mix of radial and axial geometry. Centrifugal pumps tend to generate high pressures and low fl ows, whereas diagonal pumps produce both high fl ows and high pressures [28]. Impellors are connected to a drive shaft, requiring bearings to support the rotational movement. Exposure of blood to typical bearings promotes clotting, causing depo sition of coagulation debris that can seize the bearing. Some pumps use seals to protect the bearings, but these can wear out; other designs use biocompatible materials to construct the bearings. In the most advanced centrifugal pumps impellors are completely suspended in an electromagnetic fi eld, eliminating the need for a drive shaft or bearings and reducing heating, minimizing blood trauma and lowering the incidence of mechanical failure.

Access cannula
Early clinical trials placed separate drainage and return cannulas in the saphenous veins [29,30]. Modern cannulas are placed percutaneously in a femoral-femoral or femoraljugular orientation. To maintain fl ow and minimize blood trauma, heparin-coated wire-reinforced cannulas are used. Recently, a high fl ow, wire-reinforced doublelumen catheter has been developed. It is placed via the right internal jugular vein and the drainage port (tip of the cannula) is advanced into the intra-hepatic inferior vena cava using ultrasound guidance [31]. In this orientation the return port aligns with the right atrium, minimizing recirculation. New ECCOR devices with fl ow rates comparable to those in dialysis use double-lumen cannulas similar to dialysis catheters [32,33].

ECCOR in practice
Th e fi rst clinica l trial of extracorporeal respiratory support was published in 1979, and used the Kolobow spiral-coil membrane lung, a roller pump and venoarterial access to provide ECMO [34]. Th is trial found no diff erence between conventional treatment and ECMO. At about the same time Gattinoni and coworkers introduced ECCOR [35], but did not publish the fi rst clinical trial until 1986, where patients with severe ARDS were selected for LPV combined with ECCOR (Kolobow spiral-coil membrane lung, and a roller pump). Observed mortality was 51% using this technique [29]. Subsequent work was initially encouraging [36] but a randomized controlled study in 1994 concluded that ECCOR conferred no survival advantage [30]. Importantly, complication rates were high with ECCOR, being discontinued in 33% of cases owing to bleeding, and 20% experiencing circuit clotting. Recently, new devices with lower complication rates have demonstrated improved survival when combined with ultra-protective ventilation [9]; some are already available whilst others are in advanced development. Th ey can be broadly categorized into i) arteriovenous devices, ii) venovenous devices, iii) gas exchange catheters and iv) respiratory dialysis.

Arteriovenous carbon dioxide removal
AVCO2R is commercially available through Novalung (GmbH, Hechingen, Germany) and marketed as the interventional lung assist (iLA) membrane ventilator ( Figure 3). Th e membrane lung, frequently called the 'Novalung' , utilizes a low resistance design allowing blood fl ow using the patient's own arteriovenous pressure gradient. Cannulas are placed percutaneously in the femoral artery and vein [37,38]. A similar system has been developed in the United States using the Affi nity NT (Medtronic, Minneapolis, MN) [39,40].
Pumpless systems require an arteriovenous pressure gradient ≥60 mmHg, which is unsuitable for hemo dynamically unstable patients. Further, cannulation of a major artery can result in distal ischemia [37], although measuring the artery diameter with ultrasound and selecting a cannula that occupies no more than 70% of the lumen reduces this risk [38]. AVCO2R has been successfully used to facilitate LPV in patients with ARDS [41][42][43], severe asthma [44] and as a bridge to lung trans plantation [45].

Venovenous carbon dioxide removal
Venovenous carbon dioxide removal (VVCO2R) requires a mechanical pump to propel blood through the circuit and can be broadly divided depending on whether the pump and membrane lung are separate components or incorporated into a single console. When separate components are used, the circuit is set up as described in Figure 1. Table 1 shows some of the diff erent components that can be used. Th ese circuits are more complicated to operate, often need fl ow rates >1 L.minute -1 and may need multidisciplinary support. Th e growth of programs in more general settings has provided impetus to simplify ECCOR, resulting in several devices where the pump and membrane lung are combined into one console.

iLA Activve
Th e iLA Activve mounts the Novalung and a diagonal fl ow pump together in one device. At higher blood fl ow rates this device can provide venovenous ECMO. Conceptually, this is the simplest method of providing ECCOR via a console, and although it does not provide any special benefi ts over separate components, the pump is designed to provide reliable fl ows throughout a large range of fl ow rates.

Decap/Decapsmart
Th e Decap system (Hemodec, Salerno, Italy) uses a membrane lung in series with a hemodialysis fi lter and roller pump (Figure 4). Th e hemodialysis fi lter serves two purposes with regard to CO 2 removal. First, it reduces the chance of bubble formation by increasing resistance within the membrane lung. Second, ultrafi ltrate from the fi lter is returned to the blood stream prior to the membrane lung infl ow. Since ultrafi ltrate contains dissolved CO 2 , recirculating in this way allows additional CO 2 removal by creating a greater fl ow rate through the membrane lung than the fl ow from the patient. Consequently, smaller membrane lungs can be used (0.3 to 1.35 m 2 ) with lower fl ow rates (<500 ml.minute -1 ) than conventional ECCOR [33], resulting in similar anti coagulation requirements to continuous venovenous hemodialysis [46]. Th e Decap has been successfully used in adults and children [9,47,48].

Hemolung
Th e Hemolung (Alung Technologies, Pittsburgh, USA) is the latest device to enter the ECCOR arena. In this device the membrane lung and centrifugal pump are combined together, acting as one unit ( Figure 5). Blood is drawn into the unit via a rotating impeller. Th e center contains a rotating core that accelerates blood towards a surrounding stationary fi ber bundle. Th is is called active mixing; the rotating core generates disturbed blood fl ow patterns subjacent to the fi ber membrane, reducing diff usional resistance and increasing gas exchange. As a result, CO 2 removal is more effi cient and achieved with a smaller membrane surface area and fl ows of 400 to 600 ml.minute -1 , which allows use of smaller double-lumen catheters. Th e smaller membrane surface area, siloxane coating for plasma resistance and covalently bound heparin result in lower anticoagulation requirements [32]. Gas fl ow through the membrane lung is supplied under negative pressure, a safety feature preventing air embolism if the membrane is disrupted. Th e Hemolung enabled a 50% reduction in minute ventilation in animal trials and was recently successfully used in a clinical case series of fi ve adults [49].

Gas-exchange catheters
Several gas-exchange catheters have been developed but only one, the intravenocaval oxygenator and carbon dioxide removal device (IVOX), has been used clinically. Th ese devices package hollow fi ber membrane lungs into a catheter that is small enough to be placed in the vena cava, that is, <15 mm in diameter. Intracorporeal catheters are conceptually attractive because they are exposed to 2 to 3 L.minute -1 of blood fl ow and therefore CO 2 removal is not fl ow limited. Th e IVOX was designed for both oxygenation and CO 2 removal. Orienting 'crimped' membrane fi bers in a spiral arrangement maximized gas exchange by increasing surface area and creating disturbed blood fl ow patterns over the membrane [50]. Disturbed blood fl ow provides convection velocity towards the fi ber surfaces, reducing diff usional resistance. Th e membrane surface of the IVOX ranged from 0.2 to 0.5 m 2 [51] and gas fl ow was applied under negative pressure; an important safety feature in intracorporeal devices since there is no other opportunity to prevent air embolism if the membrane is disrupted.
In animal trials the IVOX consistently removed 40 ml.minute -1 of CO 2 , but oxygen delivery was less reliable. Clinical experience was mixed; the IVOX facilitated lower ventilator settings in some studies [52], but made no diff erence in others [53,54]. On the whole, gas exchange was too limited and placement associated with high complication rates from bleeding and throm bosis [52]. Commercial development has subsequently ceased.

Future directions and devices in development
Several of the above devices are undergoing clinical trials, often in combination with LPV (Table 2). Other promising approaches are still in development, in particular more effi cient gas exchange catheters and respiratory dialysis. Novel methods to maximize CO 2 removal, such as blood acidifi cation, are also under investigation [55].

Gas-exchange catheters in development
Following the IVOX, attention has focused on developing a catheter that meets 50% of adult gas exchange requirement. Several ingenious approaches are being studied. Th e fi rst approach is generation of active mixing within the catheter. Th is was initially attempted using an intraaortic balloon pump close to the shaft of the IVOX catheter [56]. However, the membrane fi bers were not fi xed and fi ber movement opposed active mixing. Th e Hattler catheter solved this using a rigid fi ber mat constructed around a central balloon [57] (Figure 6). Rapid pulsation of the balloon directed blood fl ow over the membrane fi bers, causing active mixing. In this design membrane fi bers do not occupy the whole lumen of the vein, causing less fi ber drag on blood fl ow. In animal trials the Hattler catheter exchanged CO 2 at 305 ml.minute -1 .m -2 , almost double the IVOX rate at similar CO 2 concentrations [58,59].  Active mixing can also be achieved by rotating the fi ber bundle; a strategy used in the dynamic intravascular lung assist device (D-ILAD) [60]. Although the D-ILAD was almost twice as effi cient as balloon-pulsating catheters, rotating fi bers could damage vessel walls upon contact. Recently, the Hattler catheter has been modifi ed by replacing the balloon with a series of small impellers. It has been successfully used in animals and has CO 2 exchange rates similar to the D-ILAD [61].
Finally, in addition to active mixing, CO 2 exchange has been improved by covalent immobilization of carbonic anhydrase to the surface of the hollow fi ber membrane [62]. As a result, CO 2 is more rapidly generated from bicarbonate, facilitating removal.

Respiratory dialysis
In the 1980s, several groups reported the results of animal experiments using dialysis to remove CO 2 in the form of bicarbonate. Th is approach is appealing because CO 2 is transported in the form of bicarbonate, which moves freely across dialysis membranes. Conventional hemodialysis uses bicarbonate-containing dialysates to correct the metabolic acidosis accompanying renal failure, but bicarbonate-free dialysates can remove enough CO 2 to replace pulmonary ventilation in dog models [63]. Currently, respiratory dialysis is limited by the inability to maintain electrolyte concentrations and pH whilst removing bicarbonate. Several approaches to replace bicarbonate have been attempted using sodium hydroxide, tromethamine (THAM), and organic anions. However, fl uid gain, hyperchloremic acidosis, hemolysis, cardiac arrhythmias and acid-base derangements have prevented successful long-term use [64,65].
Recently, hemofi ltration has been used to remove bicarbonate. One group used sodium hydroxide in a post-fi lter replacement fl uid and maintained pH and CO 2  within physiological range for 18 hours in hypoventilated sheep. However, hyperchloremic acidosis developed and blood fl ow rates exceeding 500 ml.minute -1 would be needed to remove suffi cient CO 2 in humans [66]. Another group removed bicarbonate by using pre-fi lter replacement fl uid containing THAM. Physiologic CO 2 levels and pH were maintained for 1.5 hours, but it was not determined whether THAM had the same long-term problems seen in the hemodialysis models [67]. Nonetheless, respiratory dialysis holds much promise if the problems of electrolyte and acid-base disturbances can be solved.

Conclusion
Several modalities of providing ECCOR are now either available or in development. As evidence favoring lowvolume, low-pressure ventilation in ARDS accumulates, the argument for applying these ventilation strategies in all critically ill patients will gather momentum. However, successful application is dependent upon a safe, reliable approach for CO 2 removal.
Simpler more effi cient ECCOR devices requiring lower blood fl ow rates and smaller access cannulas promise to improve safety and ease of use. Novel designs, such as the Decap, can serve the dual purpose of renal support and ECCOR. However, other solutions currently in development, gas exchange catheters and respiratory dialysis, promise to be minimally invasive, easy to initiate and well tolerated. Th ey may even eliminate the need for intubation in some forms of respiratory failure, where CO 2 is the primary problem [68]. Familiarity with devices already available can change our approach to ARDS and prime the ICU for the arrival of devices that may revolutionize our approach to respiratory failure.

Competing Interests
MEC and GM have no competing interests to declare. WJF is head of the scientifi c advisory board at ALung Technologies, and has an equity interest in this company. JAK is a paid consultant for ALung Technologies.