The (broken) promise of extracorporeal carbon dioxide removal—failure or future?

Extracorporeal carbon dioxide removal (ECCO2R) was conceived to decrease the burden of mechanical ventilation on injured lungs and to avert the need for invasive mechanical ventilation (IMV) in acute exacerbations of obstructive lung disease.1 The concept was that by removing carbon dioxide (CO2) directly from the blood, less CO2 needed to be removed by the lungs, and hence minute ventilation could be reduced. In this way, ECCO2R could help facilitate “ultra-protective” mechanical ventilation, minimizing ventilator-induced lung injury, while avoiding or ameliorating respiratory acidosis. In obstructive lung disease, the reduced need for minute ventilation could unburden failing respiratory muscles, thereby decreasing CO2 production, auto-PEEP (positive end-expiratory pressure), and work of breathing. Given the known harms of mechanical ventilation in acute respiratory distress syndrome (ARDS), along with the knowledge that lung-protective ventilation (4-8 mL/kg tidal ventilation with plateau airway pressures 25-30 cm H2O) decreases mortality,2 enthusiasm that ECCO2R might further improve outcomes grew in recent decades.3 Similarly, in the setting of acute exacerbations of chronic obstructive pulmonary disease (COPD) or asthma, there was excitement that direct removal of CO2 from blood could help avoid or shorten the duration of IMV.4,5 Yet decades later, ECCO2R has not lived up to this promise. The REST trial in patients with acute hypoxemic respiratory failure (AHRF), 61% of whom had ARDS, was halted for futility and revealed more serious adverse events in the ECCO2R group than the conventional ventilation group.6 A 2025 cost–utility analysis added further weight, showing that ECCO2R increased costs by tens of thousands of pounds per patient without improving quality-adjusted life-years.7 Taken together, these results have been viewed by some as the final nail in the coffin for ECCO2R, fueling calls to abandon the technology as both costly and hazardous. But critical care history reminds us that therapies such as prone positioning and lung-protective ventilation were once dismissed, before they transformed practice. We argue here that ECCO2R deserves the same caution: Its promise has certainly not yet been realized, but neither is the technology ready to be laid to rest. ECCO2R dates back to the late 1970s, when the concept of low-frequency positive pressure ventilation combined with ECCO2R was first introduced. It used much lower blood flow rates than what was typically used for cardiopulmonary bypass or extracorporeal membrane oxygenation (ECMO),8 creating a form of “CO2 dialysis.” In 1986, Gattinoni, Pesenti, and colleagues showed that low-flow venovenous ECCO2R could support “lung rest” ventilation (tidal volumes [VT] 3-4 mL/kg) with acceptable gas exchange.9 The underlying principle is based on the difference in oxygen and carbon dioxide dynamics in blood. Oxygen delivery is limited by the sigmoidal shape of the oxyhemoglobin-dissociation curve; once hemoglobin is fully saturated, increasing PaO2 provides minimal increases in O2 content. Thus, extracorporeal oxygenation support requires blood flow rates approaching 50%-60% of the cardiac output to provide meaningful increases in arterial oxygen content, since dissolved oxygen contributes minimally to total content.10 By contrast, carbon dioxide is carried in blood in multiple compartments (dissolved gas, bicarbonate, and carbaminohemoglobin) with total CO2 content increasing almost linearly with PCO2, making it relatively easy to unload large volume of CO2 at low flow rates (<1 L/min). As a result, low-flow circuits (eg, 0.3-1.5 L/min) can eliminate a substantial proportion of a patient’s metabolic CO2 production when sweep gas flow is sufficient. With ECCO2R, the intention is to focus on meaningful removal of CO2, which may be achieved at blood flow rates as low as 0.3-1.5 L/min.1 This physiological difference in gas transport forms the conceptual basis for the use of ECCO2R: It is not intended to provide full respiratory support, but rather to remove CO2 and, in the case of AHRF, facilitate lung-protective ventilation. In addition, because ECCO2R functions at much lower blood flow rates than ECMO, the implicit assumption was that complications of ECCO2R would be substantially lower than with ECMO, which typically requires higher blood flow rates. Unfortunately, this assumption has been shown to be incorrect, at least with existing technology. Initial clinical use of ECCO2R focused on patients with ARDS. Small series in the 1980s-1990s confirmed feasibility, but enthusiasm waned after a clinical trial demonstrated high complication rates and no survival benefit.11 A resurgence came in the 2000s-2010s, with improved membrane lungs, pumps, and compact devices. The Novalung ILA, a pumpless arteriovenous circuit, and venovenous systems like the Hemolung device entered clinical testing.12,13 Using smaller cannulas and moderate blood flow rates (∼0.5-1 L/min), these and other similar devices revived interest in ECCO2R. Armed with better technology and insight into ARDS, investigators tested ECCO2R in ARDS and obstructive lung diseases, hoping to translate physiological promise into improved outcomes. Early case reports in the late 2000s demonstrated that adequate gas exchange was possible with ECCO2R using VT approximately 3 mL/kg, and suggesting biological benefit (eg, reduced IL-6).12 The Xtravent pilot randomized controlled trial (2013) randomized 79 ARDS patients to ultra-protective ventilation plus ECCO2R versus conventional lung-protective ventilation.14 While overall ventilator-free days did not differ, >80% of the patients in the ECCO2R arm were normocapnic with VT approximately 3 mL/kg. A post hoc analysis hinted at benefit in more severe forms of ARDS (PaO2/FiO2 ratio ≤150). Around the same time, Terragni et al. confirmed that ECCO2R could lower VT and driving pressures safely in ARDS patients, although clinical outcomes remained unchanged.15 These studies established feasibility and relative safety, although there was an increased need for transfusion of red blood cells.14 The multicenter SUPERNOVA study enrolled 95 moderately severe ARDS patients across 33 intensive care units (ICUs).16 Within 24 hours, 82% achieved VT <4 mL/kg and plateau pressures ≤25 cm H2O, confirming the feasibility of ECCO2R to achieve ultra-protective ventilation. However, complications were common: 39% had ECCO2R-related adverse events, including a fatal intracranial hemorrhage and a tension pneumothorax.16 Though uncontrolled, and not designed to assess outcomes such as mortality, SUPERNOVA highlighted both the physiologic potential and the substantial risks of ECCO2R, setting the stage for larger definitive trials. The REST trial was a multicenter randomized study in patients with AHRF testing whether ultra-low VT ventilation (∼3 mL/kg predicted body weight [PBW]) enabled by ECCO2R would improve outcomes versus standard lung-protective ventilation (∼6 mL/kg PBW).6 It planned to enroll >1100 patients, but was stopped for futility after 412 patients, with limited physiologic separation since controls already had relatively “safe” plateau pressures. ECCO2R did not reduce mortality or other major clinical outcomes. In addition, the ECCO2R arm had more serious adverse events, especially bleeding, raising safety concerns about low-flow ECCO2R in this setting. In acute exacerbations of COPD or asthma with hypercapnic ­respiratory failure, ECCO2R has been proposed as a way to avoid ­intubation when noninvasive ventilation (NIV) is failing and to facilitate earlier endotracheal extubation in intubated patients. Early uncontrolled studies were encouraging: Del Sorbo et al. found higher NIV success and a trend to lower mortality with ECCO2R versus historical controls,5 while Abrams et al. demonstrated the feasibility of early extubation in intubated patients with acute exacerbations of COPD.4 The ECLAIR study (2016) showed improved gas exchange and reduced intubation rates, though bleeding complications were frequent.17 These reports fueled optimism that ECCO2R could help COPD patients avoid invasive ventilation by reducing gas trapping and the work of breathing, which in turn lowers respiratory muscle CO2 production18 and relieves much of the patient’s dyspnea.4 There was a paucity of high-quality evidence, however, until the VENT-AVOID trial in 2024.19 In 113 patients with acute exacerbations of COPD who were deteriorating on NIV, or who were already intubated but failing to wean, ECCO2R failed to improve ventilator-free days. Importantly, among those in the failing NIV arm, mortality was substantially higher with ECCO2R (22% vs 0%).19 No benefit was seen in the intubated group of patients, although there was a trend toward more ventilator-free days. Thus, despite physiological appeal, the best data to date suggest real risk in using ECCO2R in COPD patients with acute respiratory failure. By the late 2010s, it became clear that in both ARDS and COPD patients that ECCO2R was able to achieve its physiologic goals—that is, lower PaCO2, correct or partially correct pH, and reduce VT or minute ventilation.4,5,16,17 But this physiological success did not translate into improved clinical outcomes. Complications were frequent: bleeding (including catastrophic intracranial hemorrhage), thrombosis and circuit clotting, vascular injury, and hemolysis. Even in expert centers, a substantial number of patients experienced some adverse event, with serious complications several times more common than with standard care. If the physiological rationale for ECCO2R was so compelling, why have trials failed to show benefit? There are several possibilities, summarized in Figure 1. Possible reasons why ECCO2R may have failed in clinical trials. Abbreviations: ARDS, acute respiratory distress syndrome; ECCO2R, extracorporeal carbon dioxide removal. First, patient selection may have been suboptimal. ARDS studies often enrolled moderately severe cases where conventional lung-protective ventilation was reasonably effective, leaving little room for ECCO2R to add value. Furthermore, severity was often measured using the PaO2/FiO2 ratio, missing the opportunity to enrich the population for those most likely to benefit, based on additional physiologic criteria, such as dead space or respiratory system compliance.20,21 And, in a subset of severe cases, patients with more severe hypercapnia who might potentially benefit most were underrepresented. Similarly, with respect to COPD, many patients in VENT-AVOID were not severely acidotic and might have recovered without intubation, diluting any signal of benefit. Second, timing may be important. Starting ECCO2R after lung injury is already well established may attenuate the potential benefit. Using it earlier, at ARDS onset, might have been more ­impactful, but this was not tested. In COPD patients, NIV is such an effective intervention, and the exact timing of failure of NIV is so poorly understood, that the use of ECCO2R was often confined to those who had already failed NIV and were intubated.19 Third, standard care in control arms was already excellent. Modern ARDS management with low VT and airway pressures, reasonable application of positive end-expiratory pressure, and permissive hypercapnia may have left little room for additional benefit. In REST, plateau pressures were already within what are currently considered safe ranges, so the physiological difference between groups was modest. The intervention arm would likely need greater separation from controls than seen in REST. Fourth, trial designs may have limited conclusions. Many studies were small, underpowered, or focused on physiological endpoints. REST planned for >1100 patients, yet enrolled only 412 after being stopped early for futility. The study was also criticized because it used a pragmatic approach, including nonexpert centers, with heterogeneous populations, although those centers underwent substantial training in preparation for the trial.6 Fifth, technical limits constrained efficacy. Low-flow ECCO2R systems typically clear only 30%-50% of metabolic CO2, adequate for moderate hypercapnia but insufficient in patients with high CO2 production or severe acidosis. Even in REST, many patients receiving VT in the range of 4 mL/kg remained mildly hypercapnic. In patients with COPD, PaCO2 often fell only modestly, sometimes not enough to reverse acidosis or obtundation. Finally, and perhaps most importantly, complications were greater than anticipated. Across trials, 20%-40% of ECCO2R patients experienced adverse events, with serious events several-fold higher than controls. In REST, the excess of fatal bleeds underscored the imbalance between risk and unproven benefit.6 These complications likely relate to a key limitation of current ECCO2R systems: the centrifugal pumps. These have often been repurposed from ECMO systems and were designed to work at higher blood flow rates. When run at the much lower flows used during ECCO2R, these pumps do not operate in their optimal flow range, creating excess shear stress, hemolysis, and recirculation within the pump.22 Using an in silico model, Schops and colleagues23 found that recirculation increased by a factor of 3.7 when flow was decreased from 4 to 1 L/min, and using an in vitro model they found hemolysis was 6.2 greater at the 1 L/min flow rate compared to 4 L/min. The resulting blood trauma is likely to increase the need for anticoagulation, which in turn drives bleeding and clotting complications—a key reason ECCO2R has not proven safer than ECMO. Thus, paradoxically, the very feature meant to make ECCO2R less invasive—low blood flow rates—likely increased complications with much of the current technology. Despite its setbacks, ECCO2R may yet have a future. Critical care history is replete with therapies that appeared to fail at first, only to demonstrate benefit later when applied with better technology and/or smarter use. Low VT ventilation,2,24 prone position in moderate ARDS,25,26 and ECMO for the most severe cases11,27,28 all followed this trajectory, from early failure to later success. Given the strong physiologic rationale discussed above, we believe that ECCO2R has a reasonable chance of success in future for a number of reasons (Figure 2). Reasons that ECCO2R may still have a viable future. Abbreviations: ARDS, acute respiratory distress syndrome; COPD, chronic obstructive pulmonary disease; ECCO2R, extracorporeal carbon dioxide removal; ECMO, extracorporeal membrane oxygenation. As discussed, earlier devices used pumps designed for full-flow ECMO or specifically for dialysis, creating conditions conducive to shear stress, hemolysis, clotting, and the need for substantial systemic anticoagulation. Newer pumps aiming to minimize blood trauma are incorporating low-shear designs. For example, miniaturized centrifugal pumps with magnetic levitation or hydrodynamic bearings that eliminate mechanical contact are being developed that reduce friction, heat, and zones of stasis inside the pump.29 In addition, advances such as coated circuits and regional anticoagulation, along with other potential innovations, may reduce or even eliminate the need for systemic anticoagulation.30 Membrane efficiency has also improved, with modern cartridges clearing more CO2 at lower blood flow rates,31 and novel study designs are actively being investigated.32 These advances may markedly mitigate many of the major complications observed in trials to date. Large-bore cannulas once limited ECCO2R use. Newer dual-lumen catheters (13-15 Fr) or dialysis-like circuits reduce bleeding risk and may allow easier and safer mobilization, and lend themselves to portable or wearable systems. Making ECCO2R less invasive could extend use to less critically ill patients, perhaps preventing ICU transfer in COPD or providing longer-term support. Lessons from prior trials suggest that ECCO2R should target patients most likely to benefit.20 In COPD, the right candidates are either those with severe hypercapnic acidosis (pH <7.20) despite optimal NIV or those already receiving IMV, not milder cases. In patients with ARDS, a physiological analysis suggested that patients with very low compliance or high dead space are more likely to have meaningful drops in driving pressure when CO2 is removed via extracorporeal circuits,21 and a reanalysis of the REST study demonstrated that patients with ventilatory ratios (a rough surrogate for dead space) ≥3 had a posterior probability for benefit (from ECCO2R) of 82% (Relative Risk, 0.83 [90% credible interval, 0.58-1.16]).33 This suggests that even with a device associated with significant complications, there was a subgroup of patients who might have benefited from ECCO2R. Of course, such a study might be logistically difficult in practice to carry out because only 14% of the patients in REST had a ventilatory ratio >3. Nonetheless, it is a good example of how enrichment strategies could be used to increase the likelihood of success. Such enrichment strategies, along with novel safer/more efficacious devices, could lead to success in the future. Predictive enrichment, potentially aided by machine learning on ventilator and laboratory data, could guide future trials. The REST trial’s failure in all-comers from an AHRF population6 does not rule out benefit in carefully selected subgroups. ECMO’s trajectory is instructive. Once written off after negative trials in the 1970s-1980s, ECMO later showed survival benefit with better devices and stricter selection, and became central during the COVID-19 pandemic.34 ECCO2R may simply be on the same slow path of refinement, with the need to address very similar problems (device imperfections and patient selection). Indeed, reanalysis of the REST trial demonstrated heterogeneity of treatment effect, suggesting subgroups worthy of further study.35 In summary, ECCO2R remains physiologically sound: Removing CO2 to permit lung rest or to decrease dyspnea and offload the respiratory muscles makes great physiologic sense. The first wave of trials showed that its risks outweighed its benefits in broad populations, but with better technology and refined targeting, that equation could change. Declaring ECCO2R dead now risks repeating the mistake once made with ECMO and abandoning a therapy before it has matured. In ARDS, this was once true of prone positioning and low volume/low pressure ventilation, both of which failed early before proving transformative once refined. ECCO2R remains physiologically sound but clinically underwhelming. Trials in ARDS and COPD have shown no clear outcome benefits and sometimes harm, raising doubts about its value. Yet history cautions against premature abandonment. Similar to ECMO, better devices, safer anticoagulation, and more targeted patient selection may yet reveal benefit. With improved tools and smarter application, ECCO2R could still evolve from “broken promise” to valuable therapy. However, for now, most applications of ECCO2R should be limited to clinical trials. A.S. and D.B. worked on the outline together. AS wrote the first draft, and then both authors worked on subsequent versions of the paper. Supplementary material is available at American Journal of Respiratory and Critical Care Medicine online. Please see the ICMJE disclosure forms, which have been provided as supplementary material. A.S.S. has been or is a for and is of the and is on the of the ECMO D.B. for has been on the for and is the of the Extracorporeal and of the of and for We used to the