Introduction

Neurally adjusted ventilatory assist (NAVA) is a mode of mechanical ventilation which applies assist in proportion to the electrical activity of the diaphragm (Edi). NAVA and non-invasive NAVA (NIV-NAVA) have been shown to improve patient-ventilator interaction, both in terms of timing and responding to increased demand, regardless of leaks [1,2,3,4]. NAVA can be considered to act as an external respiratory muscle, sharing the work in synchrony with the patient’s own inspiratory muscles, in order to maintain adequate minute ventilation. An important feature of NAVA is that despite its ability to reduce respiratory drive with increasing assist, full respiratory muscle unloading does not eliminate the Edi completely. In fact, about 30–50% of the initial Edi remains during the highest levels of assist [4,5,6,7].

Veno-venous extracorporeal CO2 removal (vv-ECCO2R) is part of a spectrum of techniques [8] that can rapidly remove CO2 and reverse respiratory acidosis, thereby reducing the central respiratory drive at lower than normal minute ventilation [9,10,11,12,13,14,15,16]. In patients with COPD, vv-ECCO2R in combination with NIV has the potential to prevent the complications associated with the use of invasive mechanical ventilation. One major drawback of invasive mechanical ventilation in COPD compared to non-invasive ventilation is the higher rate of nosocomial infections (i.e., ventilator associated pneumonia, central venous line, and urinary tract infections [17]) and prolonged weaning with high morbidity and mortality [18, 19]. Two recent case-controlled studies by Braune et al. [20] and Del Sorbo et al. [21] demonstrated the feasibility of avoiding invasive mechanical ventilation with vv-ECCO2R, although substantial device-related side effects and the subsequent need for intubation due to worsening oxygenation [20] were observed. However, we lack rigorous data on the efficacy of ECCO2R in patients with acute exacerbation of COPD beyond these proof-of-concept studies [11,12,13,14, 22,23,24,25].

The combination of NAVA and vv-ECCO2R offers a novel approach to treating severely hypercapnic patients, as it allows for simultaneous monitoring of respiratory drive via the Edi and offers both the support of the respiratory muscles with reversal of hypercapnia through the extracorporeal system. In fact, the combination of invasive NAVA and ECMO with even higher CO2 removal capacity is synergistic when used together in patients with severe ARDS [26]. During support with a constant NAVA level, reductions in sweep gas flow from a level that maintained normal PaCO2, increased both Edi and peak inspiratory airway pressure. Furthermore, Pisani et al. [27] and Diehl et al. [28] demonstrated a reduction of the work of breathing by ECCO2R in COPD patients in the weaning process from invasive mechanical ventilation.

The application of vv-ECCO2R in combination with awake NIV-NAVA has not been explored in patients with severe COPD exacerbation and combined hypercapnic/hypoxemic respiratory failure. Knowing that there is a direct relationship between Edi and CO2 load in healthy subjects and outpatients [29, 30], we hypothesized that Edi would respond to ECCO2R and could be an important predictor of weaning from extracorporeal systems in awake patients. The aim of the present study, therefore, was to evaluate the association between Edi and tolerance to weaning from vv-ECCO2R in patients with severe acute exacerbation of COPD supported with NIV-NAVA.

Methods

The Ethics Committee of the University of Witten/Herdecke provided approval for this study. Written informed consent was obtained from all enrolled patients. Patients were recruited over a period of more than 2 1/2 years. The study was registered retrospectively (German Clinical Trials Register/DRKS 00012737).

Patient inclusion criteria

As a part of our routine clinical practice, intubated COPD patients failing a spontaneous breathing trial (SBT) and remaining acidotic (pH 7.25–7.35) despite invasive mechanical ventilation or if they required re-intubation within 48 h after extubation were placed on vv-ECCO2R prior to study inclusion. Patients were approached for inclusion after being extubated to NIV-NAVA and were clinically stable (i.e., pH ≥ 7.40, minimal secretions, low dose or no catecholamines, adequate level of consciousness). Further details on patient characteristics and vv-ECCO2R are given in the online data supplement.

vv-ECCO2R

Standard configuration of the system consisted of a 19Fr/38 cm femoral-draining cannula and a 17 Fr/15 cm inlet-flow cannula (Maquet, Rastatt, Germany). All vv-ECCO2R runs were performed with the Cardiohelp system (HLS set advanced 5.0). The use of a vv-ECCO2R system with a surface area of 1.3 m2 allowed for an increase in blood flow of up to 2–3 L/min if necessary, to remove more CO2 and lower the respiratory drive, or even improve oxygenation. Standard anticoagulation with heparin was used to target an aPTT of 1.8–2.0-fold of the normal range.

Main intervention (weaning from vv-ECCO2R) and measurements

Readiness to wean from vv-ECCO2R was assessed clinically according to the criteria noted in Fig. 1. On the day weaning readiness was reached, the intervention phase of the study was carried out by the following protocol: NIV-NAVA level was set to 1.0 cm H2O/μV, the peak pressure limit to 27–30 cmH2O, and positive end-expiratory pressure (PEEP) to 6–7 cmH2O. FiO2 was titrated to maintain a PaO2 of 60–80 mmHg. Once achieved, FiO2 was kept stable throughout the weaning process, even if PaO2 increased. Subsequently, the following five 30-min conditions were studied: (i) an initial “first” baseline period, with vv-ECCO2R settings of 1000 mL/min blood flow and 10 L/min sweep gas flow (O2); (ii) reduced vv-ECCO2R blood flow to 750 mL/min; (iii) reduced vv-ECCO2R blood flow to 500 mL/min; (iv) a pause of sweep gas flow; and (v) return to a second baseline period. If the patients did not tolerate reduced blood flow, a pause of sweep gas flow, or were clearly stating to stop the weaning protocol, the settings were returned to baseline.

Fig. 1
figure 1

Flow chart of inclusion criteria, study protocol, and outcome. See the main text for details

Full size image

After the second baseline, patients were decannulated if they reported no dyspnea while having the sweep gas off and if pH values were ≥ 7.35. Dyspnea was classified as either “acceptable dyspnea” (which lead to removal of the system) or “very severe dyspnea,” (no removal of the system). Arterial blood gases were obtained at the end of each condition. Extracorporeal CO2 elimination was measured in the exhaust of the oxygenator as previously described [31].

Data analysis

Data were collected and analyzed off-line as previously described [26]. Mean values for peak Edi, peak airway inspiratory pressure (PIP), and expiratory tidal volume (Vt), breathing frequency, and minute ventilation were calculated for the whole period of each condition. Neuro-ventilatory efficiency (NVE) was calculated by dividing Vt by Edi as described before [32]. Rapid shallow breathing index (RSBI) was calculated by dividing breathing frequency by tidal volume.

Continuous variables are reported as mean (± standard deviation). One-way ANOVA followed by Dunnett’s multiple comparisons test, unpaired t test, and receiver-operating characteristic (ROC) analysis was performed as appropriate. Further details on data analysis are given in the online data supplement.

Results

Twenty patients were studied with a maximum follow-up period of 2 years (Table 1). Prior to hospitalization, 40% of the patients had been on home NIV, and 60% had been on long-term oxygen treatment. All patients had severe emphysema on computed tomography (CT) scans with concomitant bronchiolitis and/or infiltration. A diagnosis of severe COPD was confirmed by history, CT scan, and flow pattern on mechanical ventilation by two experienced pulmonologists from our lung center. Before study inclusion, all patients received invasive mechanical ventilation following initial NIV failure, with mean PaCO2 96.2 ± 20.5 mmHg and mean pH 7.19 ± 0.06. Clinically, vv-ECCO2R was initiated with a mean blood flow of 2.1 ± 0.8 L/min and a mean sweep gas flow of 4.5 ± 1.9 L/min. The relatively higher blood flow rate applied was mainly necessary due to worsening of oxygenation in the initial phase. Blood flow was gradually reduced over time, depending on the patient’s gas exchange and clinical course with daily evaluation for the possibility of weaning from the extracorporeal system.

Table 1 Patients baseline and mortality
Full size table

Prior to study inclusion, vv-ECCO2R was associated with major bleeding events in five patients (two patients with spontaneous retroperitoneal hematoma, one patient with hemothorax without prior intervention, two patients with pulmonary bleeding). These events prolonged intensive care unit (ICU) length of stay but did not affect mortality. Mortality was 15% at 90 days and 25% at 180 days (Table 1 and Additional file 1: Figure S4). Additional file 1: Figure S4 shows the Kaplan-Meier curves for survival (N = 20) after 1 and 2 years post-study.

Two experienced intensivists judged whether patients who were awake and cooperative were ready to wean, according to clinical criteria (detailed description in the online data supplement). During the study, 12 patients could be successfully weaned from the extracorporeal system during the first attempt, while eight patients remained on the system due to severe patient-reported dyspnea while undergoing the weaning protocol. No patient had to be re-intubated. From eight patients failing the primary weaning attempt, seven could be successfully weaned from vv-ECCO2R on their daily assessment at a later date, while one patient died in hospital on vv-ECCO2R (Fig. 1).

During the first baseline (blood flow of 1000 mL/min with a sweep gas flow of 10 L/min), mean extracorporeal CO2 removal was 129 ± 21 mL/min in patients successfully weaned and 142 ± 46 mL/min in patients with unsuccessful weaning (NS) (Additional file 2: Figure S1 A-B).

While breathing on a NIV-NAVA level of 1 cmH2O/μV, most patients maintained their PaCO2 and pH values within normal range regardless of changes in blood flow. However, turning off sweep gas flow resulted in a slight decrease in pH (Fig. 2a–d). Of note, oxygenation increased significantly in patients with successful weaning when sweep gas flow was paused (Fig. 3a, b).

Fig. 2
figure 2

pH value (a, b) and PaCO2 (c, d) in groups of successful and unsuccessful vv-ECCO2R weaning readiness test. From left to right, values obtained during first baseline (blood flow = 1000 mL/min and sweep gas flow = 10 L/min) and at blood flow to 750 mL/min and 500 mL/min with maintained sweep gas flow, followed by turning off sweep gas flow with 1000 mL/min blood flow and a second repeated baseline. For detailed description, see the main text. *Difference compared to baseline (*P < 0.05, **P < 0. 01, ***P < 0.001)

Full size image
Fig. 3
figure 3

PaO2 (a, b) in groups of successful and unsuccessful vv-ECCO2R weaning readiness test. From left to right, values obtained during first baseline (blood flow = 1000 mL/min and sweep gas flow = 10 L/min) and at blood flow to 750 mL/min and 500 mL/min with maintained sweep gas flow, followed by turning off sweep gas flow with 1000 mL/min blood flow and a second repeated baseline. For detailed description, see the main text. *Difference compared to baseline (*P < 0.05, **P < 0. 01, ***P < 0.001)

Full size image

The Edi at first baseline was significantly lower in patients with successful as compared to unsuccessful weaning from vv-ECCO2R (13.4 ± 8.1 μV vs. 26.7 ± 15.7 μV, P < 0.001) (Fig. 4a, b). In both groups, reducing blood flow increased Edi; however, this increase was more pronounced in the unsuccessful group. Compared to first baseline, turning off the sweep gas flow increased Edi by a significantly smaller amount in the successful as compared to unsuccessful weaning group (7.0 μV vs. 29.3 μV, P < 0.001). The Edi without sweep gas flow was significantly lower in the successful group (19 ± 10 μV vs. 56 ± 20 μV, P < 0.001) (Fig. 4a, b).

Fig. 4
figure 4

Peak diaphragm electrical activity (Edi) (a and b) and tidal volume (c and d) in groups of successful and unsuccessful vv-ECCO2R weaning readiness test. From left to right, values obtained during first baseline (blood flow = 1000 mL/min and sweep gas flow = 10 L/min) and at blood flow to 750 mL/min and 500 mL/min with maintained sweep gas flow, followed by turning off sweep gas flow with 1000 mL/min blood flow and a second repeated baseline. For detailed description, see the main text. *Difference compared to baseline (*P < 0.05, **P < 0.01, ***P < 0.001)

Full size image

Tidal volume (Vt) at first baseline was significantly lower in the successful group (360 ± 171 mL [5.7 mL/kg predicted body weight (PBW)] vs. 487 ± 172 mL [8.1 mL/kg PBW], P < 0.001) and was not affected by changes in blood flow (Fig. 4c, d). Tidal volume increased by a significantly lower amount in the successful group after sweep gas flow was turned off (555 ± 183 mL [8.8 mL/kg PBW] vs. 601 ± 196 mL [9.8 ml/kg PBW], P < 0.05).

The mean neuro-ventilatory efficiency (NVE) at first baseline was significantly higher in patients with successful vv-ECCO2R weaning (40.7 ± 28.2 vs. 22.7 ± 12.5 ml/μV, P < 0.001) (Fig. 5a, b). In the unsuccessful group, NVE was significantly reduced at 500 mL blood flow and further decreased to 13.6 ± 8.9 ml/μV when sweep gas flow was turned off. Further data on PIP, breathing frequency, and minute ventilation are given in the online data supplement and Fig. 5c, d/Additional file 3: Figure S3.

Fig. 5
figure 5

a–d Neuro-ventilatory efficiency (NVE) and minute ventilation in groups of successful and unsuccessful vv-ECCO2R weaning readiness test. From left to right, values obtained during first baseline (blood flow = 1000 mL/min and sweep gas flow = 10 L/min) and at blood flow to 750 mL/min and 500 mL/min with maintained sweep gas flow, followed by turning off sweep gas flow with 1000 mL/min blood flow and a second repeated baseline. For detailed description, see the main text. *Difference compared to baseline (*P < 0.05, **P < 0.01, ***P < 0.001)

Full size image

The area under the receiver-operating characteristic curve (AUROC) analysis was 0.95 (95% CI 0.93–0.97) for Edi, 0.89 (95% CI 0.86–0.92) for NVE, 0.77 (95% CI 0.72–0.81) for PIP, 0.57 (95% CI 0.51–0.63) for minute ventilation and Vt, 0.54 (95% CI 0.49–0.59) for rapid shallow breathing index (RSBI) under NIV-NAVA, and 0.53 (95% CI 0.48–0.59) for breathing frequency (Additional file 4: Figure S2).

Discussion

Our study supports the use of vv-ECCO2R in combination with NIV-NAVA to control respiratory drive in patients with acute exacerbations of COPD. Our results show that the Edi is not only a gauge to monitor the effect of CO2 removal but also a predictor of tolerance to weaning from vv-ECCO2R. Patients who fail to wean from vv-ECCO2R have significantly greater respiratory drive (Edi) and dyspnea than patients who can be successfully weaned from extracorporeal support. Taken together, our study demonstrates proof of concept of implementing vv-ECCO2R in combination with NIV-NAVA for the treatment of severe acute exacerbation of COPD.

Predictors of weaning from vv-ECCO2R

Our study included non-sedated patients with severe COPD initially reporting no dyspnea when supported on NIV-NAVA and vv-ECCO2R and whose success or failure of system weaning was based on perceived dyspnea and pH values. Although dyspnea could be considered a subjective measurement, it could discriminate between patients who tolerated removal of sweep gas flow and was very closely predicted by Edi (AUROC 0.95). The reason behind the high degree of discrimination was that halting the sweep gas flow resulted in extremely high Edi values in the unsuccessful group, almost four times that of the successful group. Also, in previous studies, Edi and NVE during spontaneous breathing tests have been described as good predictors for extubation readiness [32,33,34,35]. We did not measure intrinsic PEEP at baseline, and patients that eventually failed weaning from ECCO2R may have been the ones with higher intrinsic PEEP at baseline. However, our results of the high predictive value of the Edi are consistent with a recent study demonstrating that CO2 washout by nasal cannula oxygen therapy also decreases post-extubation neuro-ventilatory drive and work of breathing in patients with COPD [36].

Ability to maintain PaCO2 and pH

The Edi represents the central respiratory drive [29, 37, 38] and is affected by a multitude of neural receptors [39]. The Edi increases in response to CO2 loading [29] and to increased respiratory muscle loading [40]. The Edi also increases with respiratory muscle weakness [41] induced by hyperinflation, typically observed in COPD patients [38, 42, 43]. NAVA supports the inspiratory muscles in response to the Edi and complements the patient’s effort to adjust ventilation. As NAVA is a proportional assist mode, its relative sharing of force generation with the patient remains constant throughout inspiration, regardless if they have a lower or higher Edi. Since NAVA is a proportional mode and all the patients received the same NAVA level (1 cmH2O), the degree of support was variable, depending on patient’s Edi. Thus, NAVA enhances the patient’s “force-generating” capacity to unload the respiratory muscles, which can help to reduce the patient’s sensation of respiratory effort. However, in some patients, the central respiratory drive is affected by metabolic or irritative reasons, leading sometimes to uncontrolled breathing independent of the level of support.

Similar to the present study, a previous small study in severe ARDS patients on NAVA showed that termination of sweep gas flow led to an increased Edi and required adjustments to ventilation to maintain PaCO2 and pH [22]. Of note, the stepwise reduction in blood flow rate had only a minor impact on CO2 removal capacity due to the large membrane lung surface area of our ECCO2R system. However, the present study shows clearly that the combination vv-ECCO2R and NIV-NAVA makes it possible to detect if the patient’s respiratory drive is “uncontrollable” with just mechanical ventilation. The essential need to preserve PaCO2 and pH was supported by the fact that 8 out of 20 patients who failed removal of sweep gas flow responded with an extreme respiratory drive (Edi > 50 μV). When respiratory drive was uncontrollable, it was possible to remove the CO2 load and “need-to-breathe-more” sensation using extracorporeal CO2 elimination.

Respiratory parameter response

Our results showed that the successful group had lower baseline values for Edi, PIP, Vt, and that these parameters were only affected by removal of sweep gas flow, which despite a low increase in Edi, increased PIP, Vt, and minute ventilation to values similar to those of the unsuccessful group. The tolerance of dyspnea in the successful group shows that they had enough reserve to allow removal of the extracorporeal system. Moreover, the maintenance of NVE suggests the lack of an impact on respiratory muscle function and/or mechanics. In the unsuccessful group, both reducing blood flow and removal of sweep gas flow led to a significant increase in respiratory drive as evidenced by the increased Edi. As the Edi was extremely high in the unsuccessful group, their PIP reached the maximum pressure limit, and we can only speculate that the results could have been different if the upper pressure limit has been increased. We hypothesize that despite high Edi and low NVE, the unsuccessful group would not benefit from more non-invasive assist to unload the diaphragm and thus would never have been able to remain on NIV-NAVA unless also supported by the extracorporeal system. Of note, Edi in the unsuccessful group remained increased during the second baseline, which could be attributed to increased CO2 production by the respiratory muscles [28].

Clinical considerations

The CO2 removal capacity during treatment was comparable to the CO2 removal capacity observed in animal experiments using newer oxygenators [31, 44] reflecting the effectiveness of the system used in the present study, even after several days of treatment in patients.

In the present study, five major bleeding events occurred during vv-ECCO2R, which prolonged ICU length of stay but had no impact on mortality. An interesting observation was the increase in PaO2 during vv-ECCO2R weaning. Global vasodilation of the pulmonary vessels following CO2 removal improving right heart function [45] has been suggested to increase the shunt perfusion [45]. Moreover, ECCO2R affects mainly PCO2 in the VCO2/VO2 ratio, which could lead to worsening of the oxygenation [46].

Conclusions

Our study demonstrates the feasibility of implementing vv-ECCO2R in combination with NIV-NAVA for the treatment of severe acute exacerbation of COPD. Patients who fail to wean from vv-ECCO2R have significantly greater respiratory drive (Edi) and dyspnea than patients who can be successfully weaned from extracorporeal support. A high Edi signal of > 50 μV when pausing the extracorporeal system with a stable pH value of 7.4 could be predictive for unsuccessful weaning of vv-ECCO2R. Randomized controlled trials of vv-ECCO2R and NIV-NAVA vs. conventional invasive mechanical ventilation in patients with severe acute exacerbations of COPD are needed to confirm these findings and the efficacy of this approach.