Effects of extracorporeal CO₂ removal on gas exchange and ventilator settings: a systematic review and meta-analysis

Extracorporeal carbon dioxide removal (ECCO₂R) implies the removal of carbon dioxide (CO₂) from the blood across a gas exchange membrane without influencing oxygenation to a clinically relevant extent. ECCO₂R can be provided by heterogenous techniques, thus the method has to be rather regarded as a therapeutic intention than as a specific technical procedure [1]. The term ECCO₂R has been proposed first during the late seventies by Kolobow and Gattinoni in a study using an arterio-venous pumpless circuit in an animal model [2]. In 1986, Gattinoni`s group reported on patients suffering from severe acute respiratory distress syndrome (ARDS) undergoing ECCO₂R by low-flow venovenous (VV) extracorporeal gas exchange to enable more protective ventilator settings [3]. The revival of ECCO₂R was spurred by awareness of mechanical ventilation risks, aiming for ultraprotective ventilation with tidal volumes well below 6 mL/kg/predicted body weight. This led to the development of devices for CO₂ removal, like the arterio-venous Interventional Lung Assist (ILA®, Novalung, Heilbronn, Germany), increasing its use in ARDS patients [4]. Reducing ventilation invasiveness in ARDS patients by ECCO₂R has been the main therapeutic target under investigation to date [5, 6]. ECCO₂R is used in hypercapnic lung failure, such as in acute exacerbated chronic obstructive pulmonary disease (AECOPD), targeting intubation avoidance or weaning, and in terminal fibrosis for lung transplantation (LTX) bridging, promoting spontaneous breathing and ambulation [1, 7‐9]. As effective extracorporeal elimination of CO₂ can be achieved at much lower blood flow than necessary for oxygenation, specially designed low-flow set-ups have been developed for the purpose of ECCO₂R using smaller cannulas and membranes based either on continuous renal replacement therapy (CRRT) or on extracorporeal membrane oxygenation (ECMO) technology [10, 11].

The efficacy of CO₂ removal depends on the partial pressure gradient at the membrane, the diffusion coefficient, the cross-sectional area of membrane lung, as well as blood flow and sweep gas flow [10, 12‐14]. CO₂ extraction tends to be less efficient in CRRT-based systems due to their lower blood flow and membrane surface, unlike ECMO-based systems with centrifugal pumps and larger membranes. A post hoc analysis of the SUPERNOVA trial [6] investigating the effects of ECCO₂R in patients with moderate ARDS compared the subgroups treated with “lower extraction” and “higher extraction” systems [15]. Although the goal of reduced tidal volumes could be achieved in both groups, this was more frequently the case in the “higher extraction” group.

The therapeutic goal of ECCO₂R depends on indication: In ARDS, it is to enable less invasive ventilation; in chronic obstructive pulmonary disease (COPD) or LTX bridging, it aims to reduce ventilatory strain, promoting spontaneous breathing or avoiding mechanical ventilation.

No systematic analysis of ECCO₂R`s clinical effects, varying by indication and technology, exists yet. We conducted a systematic review to assess its effects on gas exchange and respiratory settings dependent on the different indications and its extraction capacity (“higher” versus “lower”).

The review protocol was registered on PROSPERO (Registration No: CRD 42020154110) on 24th January 2021. The reporting adhered to the PRISMA guidelines [16]. The PRISMA checklist is provided as Additional file 1: File A.

Our objective was to examine the effect of contemporary ECCO₂R systems on gas exchange and respiratory settings in critically ill adults, and in subgroups defined by technology and indications.

After searching the databases, 5405 articles met our inclusion criteria for further screening (Fig. 1 and Additional file 1: File B and C). After removing duplicates and ineligible records with electronic tools, 2,079 papers were excluded by the screening team as irrelevant based on title and abstract. After a full text review of 256 studies, 207 were excluded, resulting in 49 studies for inclusion. These comprised three prospective open-label randomized studies, 18 prospective observational, and 28 retrospective observational studies, totalling 1,661 ECCO₂R patients. The median number of patients per study was 20 (range 7–186). Additional file 1: Table S1 summarizes the main characteristics of the studies included. Notably, two studies reported separate cohorts for ARDS and COPD patients undergoing ECCO₂R with distinct therapeutic goals [18, 19]. Since the results were reported separately without pooling, each cohort was treated as an individual study, leading to 51 datasets being analysed independently. Additional file 1: Table S1 marks two such trials as (a) and (b). In Augy et al.’s study [19], 70 ECCO₂R patients with various indications were included, but parameters on gas exchange and ventilation for only ARDS and COPD patients (cohort a: n = 24, cohort b: n = 30) were analysed. Device characteristics used in ECCO₂R are detailed in Additional file 1: Table S2.

Of the six ventilation parameters of interest (plateau pressure, PEEP, tidal volume, respiratory rate, PaCO₂, PaO₂/FiO₂ ratio, and pH), only PaCO₂ could be extracted from all studies, except for one [20]. Only 16 studies reported on all six parameters [4, 6, 18, 21‐33]. Additional file 1: Table S3 shows the parameters available for each included study. Risk of bias assessment (Robins-I tool) revealed 4 studies with a low risk [31, 34‐36], 14 with moderate risk [6, 8, 19, 30, 33, 37‐44], and 30 with serious risk [4, 18, 21‐29, 32, 45‐60]. Three studies were categorized as critical risk [61‐63] (Additional file 1: Figure S1).

ECCO₂R, using specifically designed devices has been in use for about two decades, with a variety of devices introduced for different clinical indications and therapeutic goals. However, no systematic analysis of the effects of ECCO₂R has yet pooled data from studies across all devices and indications.

Our main finding is that the primary goal of ECCO₂R, i.e., elimination of carbon dioxide, reduction of PaCO₂, and acidosis can be achieved irrespective of device and indication. Devices designed for higher extraction and those for lower extraction both produce similar effects, though higher extraction devices do so more markedly. In ARDS patients, higher extraction devices more efficiently reduce PaCO₂, tidal volume, and respiratory rate, while in COPD/LTX patients, they more effectively lower plateau pressure.

A retrospective subgroup analysis of data from a prospective cohort study on ARDS patients indicated that “ultraprotective” ventilation was more commonly achieved using devices with higher CO₂ extraction capacity [15]. “Higher extraction” is not a well-defined term, but instead refers to ECCO₂R systems with larger gas exchange membrane surface operating at blood flows over 600 mL/min. Additional file 1: Table S2 indicates that “lower extraction” devices operate at blood flows below 500 mL/min and are mainly based on CRRT technology. The Hemolung RAS system falls in between, operating around 500 mL/min, with a relatively small membrane surface of 0,59 m². In agreement with Combes et al. [15] we thus categorized the Hemolung RAS system as “lower extractor”. Higher extraction systems typically operate above 800 mL/min of blood flow and utilize larger membranes. Our results suggests that in scenarios such as spontaneously breathing patients (e.g., in AECOPD or during bridging to LTX to avoid mechanical ventilation) or in individuals with a very high carbon dioxide burden, lower extraction devices may not be adequate to achieve therapeutic goals. Lower extraction devices are promoted as easier to use and less invasive, yet there is no proven reduction in side effects such as bleeding. Notably, bleeding rates were high with the Hemolung RAS system [15, 31], potentially due to hemolysis induced by the centrifugal pump, which appears to increase at low blood flow rates [64].

Despite the proven beneficial effects on gas exchange and mechanical ventilation, evidence remains debatable. The concept of “ultraprotective” ventilation in ARDS patients enabled by ECCO₂R, although shown to reduce biotrauma [65], has failed to improve clinical outcomes [20, 31]. In the light of more favourable evidence with respect to VV ECMO in severe ARDS [66, 67], the role of ECCO₂R in ventilating below standard protective settings remains questionable. When pooling the so far published three randomized prospective trials, we could not detect any effect on mortality (Table 2). The trials, however, are heterogeneous: The Xtravent Study on ARDS patients was stopped prematurely and thus underpowered but showed at least a positive effect on duration of ventilation in the subgroup with more severe ARDS [20]. The large and well-conducted REST Trial demonstrated no mortality benefits from ECCO₂R and revealed a negative impact on ventilation duration [31]. Adverse events related to ECCO₂R were frequently recorded. The third RCT involved a small cohort of eighteen spontaneously breathing patients with AECOPD undergoing NIV with a high risk of failure. ECCO₂R improved physiological parameters and reduced the duration of NIV. However, no mortality benefits were observed. Despite varied ECCO₂R indications and study designs, all studies reported on a longer hospital stay in ECCO₂R patients, a finding that became significant upon data pooling (Table 2). This increased hospitalization may stem from different clinical management of ECCO₂R patients in open-label studies [36], and a higher incidence of adverse events such as bleeding [31].

Overall, the reported adverse events rate was as high as 19%. This number has to be regarded with caution as adverse events were defined heterogeneously and often assessed from retrospective studies. Severity of adverse events was not categorized according to standard criteria in most studies. The reported adverse events rate ranged between zero to 77%, pointing towards a considerable heterogeneity between studies concerning definition and documentation of adverse events. Our data however underline that ECCO₂R can lead to severe adverse events, many of them coagulation associated like major bleeding or thrombotic events. While for bleeding, we detected no major difference between higher and lower extraction systems, rate of thrombotic or ischemic events occurred more often in patients treated by higher extractions systems. It has to be taken into account however, that a part of the ischemic events were specifically due to arterial cannulation using a pumpless system. The only large, prospective, randomized trial assessing outcome and complications in 412 patients with acute hypoxemic respiratory failure [31] found no mortality benefit, and a high rate of serious adverse events associated with ECCO₂R. These results indicate that ECCO₂R is not appropriate for broad clinical adoption in ARDS and should be used with extreme caution, most likely in the setting of rigorously designed research protocols.

It seems that hypercapnic lung failure represents a more rewarding indication for ECCO₂R, which has been shown to be a useful tool to prevent mechanical ventilation in patients suffering from AECOPD and failure of NIV [34, 35, 41], as well as a therapeutic option to bridge patients with terminal hypercapnic lung failure to LTX [8, 9]. ECCO₂R has also been reported as successful therapy in refractory status asthmaticus. Only case reports on the use of low-flow ECCO₂R systems have been published so far. There are however retrospective studies on the use of ECMO [68‐70]. VV ECMO with the primary goal of treating hypercapnia has been shown to be a very successful option for refractory asthma [70]. As high-flow extracorporeal gas exchange systems were not within the scope of our review and ECMO settings were not reported in most of these studies, we chose to exclude this indication from our analysis. Our findings however suggest that in hypercapnic patients suffering from AECOPD or during bridging to LTX, higher extraction devices may be superior regarding their effects on plateau pressure and especially the more pronounced effect on respiratory rate, which could contribute to a reduction of overinflation in patients with obstructive lung diseases. Again, high-quality evidence supporting the effectiveness of ECCO₂R in treating hypercapnic lung failure remains lacking, thus classifying it as experimental therapy.

Our findings are subject to several limitations. First, the therapeutic goals of ECCO₂R throughout the trials included were heterogeneous: Some studies included hypercapnic patients with the goal to reduce acidosis, while others set out to reduce tidal volume and ventilation pressures enabled by ECCO₂R. Studies on AECOPD and/or bridging to LTX aimed for avoiding mechanical ventilation or assisting weaning. As one might expect, the effect on CO₂ levels was more pronounced in studies including hypercapnic patients. Interestingly though, when pooling all available data, the effects were quite homogenously directed in the same direction. Moreover, not all studies reported on all data (Additional file 1: Table S3). In spontaneously breathing patients, ventilatory settings were not reported and changes in respiratory rate were either dependent on the patients themselves (if breathing spontaneously) or on the ventilation protocol applied. Second, we found a considerable risk of bias in most of the studies included in our work. Only four studies were categorized as yielding a low risk of bias. When analysing the data according to risk of bias category, however, results were quite uniform in each category.

In summary, we found a robust effect of ECCO₂R on CO₂ removal and related parameters. Data from three RCTs, however, did not indicate a significant mortality benefit. Additionally, ECCO₂R was associated with a high rate of serious adverse events. Based on existing evidence, ECCO₂R cannot be recommended for ARDS outside of clinical trials. While it may show greater effectiveness in hypercapnic lung failure, it remains experimental.