Low Spontaneous Breathing Effort during Extracorporeal Membrane Oxygenation in a Porcine Model of Severe Acute Respiratory Distress Syndrome

The current experimental study was approved by the Animal Ethics Committee of Pontificia Universidad Católica de Chile (Santiago, Chile; No. 150811040) and conducted according to the guidelines from the National Institute of Health (Bethesda, Maryland) guidelines. Twelve female pigs (Sus scrofa domestica, 30 ± 5 kg) were included.¹⁷  The study design is summarized in figure 1. A more detailed description of methods is provided in the supplementary material (Supplemental Digital Content methods, https://links.lww.com/ALN/C475).

After the preparation phase,¹⁸  a two-hit model of lung injury was induced under deep anesthesia. First, repeated lung lavages with warm saline (30 ml/kg, 39°C) were performed until Pao₂/fractional inspired oxygen tension (Fio₂) felt less than 100 for at least 15 min. Subsequently, 2 h of injurious ventilation was applied in pressure-controlled ventilation, with positive end-expiratory pressure (PEEP) of 0 cm H₂O, inspiratory pressure of 40 cm H₂O, respiratory rate of 10/min, inspiratory:expiratory ratio of 1:1, and Fio₂ of 1.0. After completing the 2 h, ventilator settings were switched back to those used at baseline, and after 10 min, just before starting extracorporeal membrane oxygenation, a full assessment of all variables was registered (time 0).

Extracorporeal membrane oxygenation cannulation was performed through the right external jugular vein with a 23-F double-lumen cannula (AVALON ELITE, Maquet, USA). The cannula was connected to the circuit after time 0 measurements and extracorporeal circulation started progressively. The pump was adjusted to target a blood flow of 60 to 70 ml · kg^-1 · min^-1. The heat exchanger was set at 38°C. The initial sweep gas flow (Fio₂, 1.0) was adjusted to a partial pressure of carbon dioxide of 30 to 50 mmHg.

During the first 3 h, all animals were ventilated with near-apneic ventilation (pressure-controlled ventilation; PEEP, 10 cm H₂O; driving pressure, 10 cm H₂O; respiratory rate, 5/min; inspiratory:expiratory ratio, 1:2)¹⁸  and keeping muscle paralysis. After time 3, animals were allocated into two groups by simple randomization:

1. Near-apneic ventilation, which continued with the same settings previously described (time 0 to time 3), adjusting sweep gas flow (Fio₂ 1.0) to achieve a partial pressure of carbon dioxide between 30 and 50 mmHg.

2. Spontaneous breathing: In this group, neuromuscular blockade was stopped and sweep gas flow decreased until regaining respiratory efforts, without modifying sedation. Thereafter, ventilation was switched to pressure support ventilation (pressure support, 10 cm H₂O; PEEP, 10 cm H₂O) and sweep gas flow adjusted to keep Paco₂ at 30 to 50 mmHg. The main goals in this group were to keep the animals with continuous spontaneous breathing efforts and without agitation. Although no specific targets were defined for esophageal pressure swings or tidal volumes, in pilot experiments we observed that animals developed frequent agitation when sedation or sweep gas flow was decreased while trying to achieve stronger breathing efforts. In contrast, if sedation and sweep gas flow were kept high, animals were able to maintain regular low spontaneous breathing efforts without agitation at respiratory rates of 30 to 50 breaths/min.

Heart rate, pulse oximetry, core temperature, arterial blood pressure, pulmonary artery pressure, respiratory rate, respiratory mechanics, ventilator and extracorporeal membrane oxygenation settings, anesthetic drugs, and maintenance fluid, as well as infusion drugs for hemodynamic support, were registered at baseline, after completing lung injury (time 0), and at 3, 6, 12, and 24 h of the study period (time 3, time 6, time 12, and time 24).

Electric impedance tomography images were recorded at each time point during the protocol (EIT-Pioneer Set, Swisstom, Switzerland). Lung images were divided into four symmetrical nonoverlapping regions of interest: (1) ventral, (2) central–ventral, (3) central–dorsal, and (4) dorsal regions. Values represent percent changes in local impedance as compared with a peak-expiratory reference image taken at the beginning of each acquisition. Changes in impedance (ΔZ) during the study were defined as the difference between end-inspiratory and end-expiratory lung impedance. End-expiratory lung impedance variations along time relative to baseline values were also assessed and expressed as percent of end-expiratory lung impedance.

At the end of the protocol, animals were euthanized, and the lungs were extracted for histologic analysis. Representative lung tissue samples from six areas (upper, nondependent middle, central middle, dependent middle, nondependent lower, and dependent lower) were dissected.

To assess lung injury, tissue slices were cut from paraffin blocks, stained with hematoxylin and eosin, and observed with light microscopy. A validated score¹⁹  was used to evaluate three parameters of lung injury: (1) intraalveolar neutrophil exudate, (2) alveolar disruption, and (3) intraalveolar hemorrhage; each of these variables received a score ranging from 0 (no pathologic alteration) to 3 (severe pathologic alteration; Supplemental Digital Content fig. 1, https://links.lww.com/ALN/C475). Histologic assessment was performed by a board-certified pathologist who was blinded to time and group assignment.

Interleukin-1b and interleukin-8 levels were evaluated by enzyme-linked immunosorbent assay in lung tissue homogenates from the middle region of the left lung (dependent and nondependent).

As we had no reliable data about the potential impact of a strategy of low spontaneous breathing effort on lung injury compared to a near-apneic ventilation, particularly in the context of ARDS assisted with extracorporeal membrane oxygenation, sample size calculation was based on effect estimates obtained from a previous study in which we compared a near apneic ventilation strategy with a less protective ventilation using the same animal model.¹⁸  In that study, which included six animals per group, the near apneic ventilation group had a mean histologic lung injury score of 0.7 ± 0.3, versus 1.3 ± 0.2 observed in the group ventilated with tidal volume (V_T) of 6 ml/kg. Accordingly, we calculated that six animals per group were required to find a 50% difference in histologic lung injury scores between groups, using t test family, two-sided test, and equal number of animals per group, with 80% power and an alpha level of 5%.

The normality of the variables was tested by the Shapiro–Wilk test. Data measured along time were analyzed, when appropriate, using repeated-measures two-way ANOVA, followed by Tukey’s multiple comparisons test; or Friedman Test of variance for repeated measures, followed by pairwise comparisons using the Dunn post hoc test with Bonferroni correction; both for differences between groups and along time (time 0 compared to baseline and all other time points compared to time 0). Data derived from lung tissue analysis were compared with the Mann–Whitney test. Outliers were detected checking standardized residuals, but no action was taken. Statistical analysis was performed with GraphPad Prism 7 (GraphPad Software, USA) for all analyses, with a two-tailed P value < 0.05 and 95% CI.

Twelve (n = 12) animals completed the study protocol and were included in the analysis; however, due to technical issues in the esophageal catheter during time 12, we missed the data of one animal in the near-apneic group. Lung injury led to severe hypoxemia and decreased compliance, without differences between groups at time 0 (table 1). After starting extracorporeal membrane oxygenation and during the first 3 h of the study period, oxygenation increased and exhaled carbon dioxide production decreased to a minimum, without differences between groups. After randomization at time 3, animals from the spontaneous breathing group started to trigger the ventilator and were switched to pressure support within the first hour after stopping neuromuscular blockade. Thereafter, this group exhibited a breathing pattern characterized by low V_T, comparable to the near-apneic group (2.3 ± 0.8 ml/kg and 2.8 ± 0.4 ml/kg; mean difference, 0.6 ml/kg; 95% CI, –0.4 to 1.4; P = 0.983), but a much higher respiratory rate (52 ± 17 breaths/min and 5 ± 0 breaths/min; mean difference, 47 breaths/min; 95% CI, 34 to 59; P < 0.001; table 1). In the spontaneous breathing group, negative swings in esophageal pressure ranged from –1 to –4 cm H₂O (fig. 2; table 1). Despite the larger minute ventilation in the spontaneous breathing group, no increase was observed in exhaled carbon dioxide levels. Furthermore, sweep gas flow requirements to keep stable Paco₂ remained unchanged throughout the experiment and without differences between groups (8.3 ± 0.83 l/min and 6.9 ± 0.61 l/min; mean difference, –1.35 l/min; 95% CI, –2.0 to 4.6; P = 0.631; table 2). No differences were observed in sedation requirements.

Concerning hemodynamics, tachycardia and hypotension were noted in both groups after connection to extracorporeal membrane oxygenation, requiring the infusion of noradrenaline throughout the study (table 3). After randomization, noradrenaline requirements continued increasing during the study period in the near-apneic group, but not in the spontaneous breathing group.

Induction of lung injury was associated with a decrease in end-expiratory lung impedance in both groups (fig. 3A), with a redistribution of regional ventilation from dorsal toward central-ventral areas (Supplemental Digital Content table 1, https://links.lww.com/ALN/C475). After randomization, the spontaneous breathing group exhibited an increase in dorsal regional ventilation (fig. 3B and Supplemental Digital Content table 1, https://links.lww.com/ALN/C475). In addition, we observed simultaneous inflation in nondependent and dependent areas in both groups without evidence of pendelluft.

Lung histologic analysis showed a mild lung injury without differences between groups (fig. 4 and Supplemental Digital Content table 2, https://links.lww.com/ALN/C475). There were also no differences in interleukin-1b and interleukin-8 lung tissue concentrations between groups. However, interleukin-1b levels were statistically higher in the lung-dependent areas (fig. 5). In terms of regional lung water content, wet–dry weight ratios were similar in both groups, ranging from 5.3 to 7.5 in the near apneic group, and from 5.6 to 9.2 in the spontaneous breathing group (fig. 6).

In the current study, in an experimental model of severe ARDS supported with extracorporeal membrane oxygenation, animals assigned to the spontaneous breathing group developed a rapid shallow breathing pattern with high respiratory rates, but small swings in esophageal pressure resulting in low tidal volumes. When compared to the group treated with controlled near-apneic ventilation, we observed no differences in lung injury or lung function, despite the fact that the spontaneous breathing group had markedly higher respiratory rates and minute ventilation.

One of the first experiences with spontaneous ventilation during extracorporeal membrane oxygenation in ARDS patients was reported two decades ago by Lindén et al., who described the use of pressure-supported ventilation and minimal sedation in sixteen ARDS patients connected to extracorporeal membrane oxygenation, with 76% survival.¹³  A recent international survey described that 27% of extracorporeal membrane oxygenation centers reported using spontaneous breathing modes in ARDS patients placed on venovenous extracorporeal membrane oxygenation, without specifying the time at which this strategy was started.²⁰  One of the theoretical bases for using spontaneous ventilation in severe ARDS connected to extracorporeal membrane oxygenation is that ventilatory drive may be modulated by adjusting extracorporeal carbon dioxide removal. However, there is controversy regarding the feasibility of this approach. Langer et al. showed in an experimental model that spontaneous breathing activity of both healthy and ARDS animals can be controlled via extracorporeal gas exchange according to the amount of carbon dioxide removed, up to complete apnea. Nevertheless, in injured animals, the response was more heterogeneous, with some animals presenting vigorous ventilatory efforts despite high amounts of carbon dioxide unloading.²¹  Similar results were described in clinical studies of “awake” extracorporeal membrane oxygenation, where carbon dioxide removal could relieve the work of breathing and even allow extubation of patients during the bridge to lung transplant and in chronic obstructive pulmonary disease patients, but ARDS patients’ response was more variable. Half of them maintained unexpectedly high respiratory rates and esophageal pressure swings even with the maximum sweep gas flow, corresponding to a high carbon dioxide removal.²²  The variable response observed in ARDS has been attributed to factors such as sedation-agitation, metabolic status, and lung receptor activity, which may influence the ventilatory drive independently of gas exchange. In contrast, Mauri et al. performed a physiologic study in ARDS patients connected to extracorporeal membrane oxygenation and reported that the amount of extracorporeal carbon dioxide removal tightly regulates ventilatory drive and that ventilator-induced lung injury determinants such as minute ventilation, V_T, and airway pressure could be modulated in pressure support ventilation.²³  However, this study was performed in patients who had already been on extracorporeal membrane oxygenation for several weeks, a setting in which the high ventilator drive, characteristic of the acute phase, may have already ceased. In the current experimental study in a porcine severe ARDS model assisted with extracorporeal membrane oxygenation, pressure support ventilation resulted in rapid shallow breathing in which extracorporeal carbon dioxide removal and sedation could not be decreased.

Although our original plan before starting the study was to target a moderate breathing effort with specific values of esophageal pressure swings and tidal volume, in pilot experiments we found that pigs became agitated and were not able to increase their alveolar ventilation if sedation or sweep gas flow were decreased in order to reach moderate breathing efforts. We also tried different sedation strategies, including high doses of opioids or propofol, but none of these sedatives allowed us to maintain deeper breathing with lower respiratory rates. Whatever the sedatives chosen, deep sedation was required to prevent agitation, which resulted in rapid shallow breathing with low effort. Therefore, for the current study, we assessed a low spontaneous breathing effort strategy, without trying to modulate breathing effort and adjusting sedation and sweep gas flows to keep the animals calmed and Paco₂ at 30 to 50 mmHg, respectively. We speculate that this pattern of spontaneous ventilation reflected a low ventilator drive due to the following: (1) deep sedation was maintained to assure animal well-being, which may differ from the clinical scenario in which spontaneous ventilation is usually applied in awake patients; (2) sweep gas flow requirements remained high and unchanged after starting spontaneous breathing despite increased minute ventilation, because animals were unable to generate significant alveolar ventilation due to the low tidal volumes, as reflected by the very low exhaled carbon dioxide output; high extracorporeal CO₂ removal may have inhibited ventilator drive; and (3) PEEP was kept at 10 cm H₂O during spontaneous breathing. High PEEP levels have been shown to decrease the intensity of inspiratory efforts and their potential injurious effects.^9,24  Another possible explanation for the shallow breathing pattern observed may be diaphragmatic dysfunction due to muscle fatigue, which can occur early on after controlled mechanical ventilation,^10,25–29  but its role in the current model is uncertain.

Spontaneous breathing during the acute phase of severe ARDS has been reported to result in further lung injury.^2,8,30,31  However, in the current study, we observed no differences between groups in any of the lung injury markers assessed. A possible explanation for the current results is that the low magnitude of respiratory efforts and low V_T observed in this group, as well as the absence of pendelluft in electric impedance tomography assessment, may prevent the adverse consequences of spontaneous breathing reported in other studies. Several preclinical and clinical studies have described the deleterious effects of cyclic alveolar strain during high tidal volume and how lung injury could be decreased or minimized when a low V_T is applied.^32–35  In our study, both groups developed a very low V_T (3.0 ± 0.2 ml/kg and 2.7 ± 0.4 ml/kg, in near-apneic ventilation and spontaneous breathing groups, respectively). It is reasonable to argue that extracorporeal membrane oxygenation assistance allowed this very low V_T while maintaining normocapnia, and consequently contributed to better protect the lungs both during controlled ventilation and during a low spontaneous breathing effort.

Respiratory rate was the only variable markedly different between groups. This observation does not support the idea that high respiratory rates may favor ventilator-induced lung injury. However, the role of respiratory rate as a multiplying factor of the tidal strain exerted on the lungs may depend on whether the tidal strain is potentially injurious itself or not.³⁶  A very low tidal strain may be harmless, independent of whether it is applied 10 or 100 times per minute. At the same time, the higher respiratory rate exhibited by the spontaneous breathing group determined an increased minute ventilation compared to the near-apneic ventilation group. However, this higher minute ventilation did not increase carbon dioxide clearance, and therefore it did not influence sweep gas flow requirements. As tidal volumes generated in the spontaneous breathing group were so low, the increased minute ventilation was at the expense of dead space ventilation, as indicated by the negligible increase in exhaled carbon dioxide output observed after starting spontaneous breathing ventilation.

On the other hand, concerning the potential benefits associated with spontaneous breathing, we observed that ventilation in dependent lung regions, assessed by electric impedance tomography, was slightly improved in the spontaneous breathing group, but this did not result in better oxygenation. Regarding hemodynamics, arterial pressure was better preserved in the spontaneous breathing group.

We acknowledge several limitations of our study. First, although our model reproduces some of the main functional and histologic features of ARDS,^37–39  the specific impact on ventilator drive may differ from human ARDS, and alternative animal models may have provided a different result. Also, in terms of severity, a subgroup of ARDS patients treated with extracorporeal membrane oxygenation exhibits lower compliance than observed in our experimental model, and this factor may influence the tolerance to spontaneous breathing. In addition, we kept animals with deep sedation to prevent agitation and assure well-being. This may contrast with spontaneous ventilation in ARDS patients on extracorporeal membrane oxygenation, which is usually applied in awake patients.²²  Second, we acknowledge that the rather small sample size may have precluded the possibility of finding smaller and less pronounced differences between groups. Third, the low spontaneous breathing effort observed in the current study may not represent the ventilatory pattern observed in some ARDS patients connected to extracorporeal membrane oxygenation who develop strong breathing efforts when allowed to ventilate spontaneously. A different pattern of spontaneous breathing may have provided a different impact in terms of lung injury. However, as stated, based on previous pilot experiments in which different sedation and ventilatory approaches were used to modulate the intensity of breathing efforts, attempts to achieve stronger efforts were unsuccessful and incompatible with keeping the animals well sedated and calmed. Fourth, we included only female pigs in our study, which could limit the generalization of our results when considering sex as a biologic variable. Fifth, we applied a pressure support mode with fixed settings for the spontaneous breathing group, which may not reflect the best approach to obtain an effective and safe spontaneous effort in the clinical setting. We chose this approach to decrease the variability of the model.

In conclusion, in this animal model of severe ARDS supported with extracorporeal membrane oxygenation, spontaneous breathing characterized by low-intensity efforts, high respiratory rates, and very low tidal volumes did not induce an increase in lung injury when compared to controlled near-apneic ventilation. Nonetheless, our findings must be interpreted with caution as the study was powered to detect only large differences between groups. In addition, the study assessed the impact of low spontaneous breathing efforts, but we cannot rule out that stronger spontaneous efforts may potentially enhance lung injury.

The authors thank Macarena Amthauer, R.N., Melinka Torrejón, R.N., Fernanda Shoenfeldt, R.N., Pamela Quilodrán, R.N., Felipe Rodríguez, R.N., Cristina Perez, R.N., and Gabriel Castro, R.N., from the Intensive Care Unit of Clinical Hospital, UC-CHRISTUS Health Network (Hospital Clínico, Red de Salud UC-CHRISTUS; Santiago, Chile), for their assistance in the care of animals and the extracorporeal membrane oxygenation circuits; Diego Romero, M.Sc., from Department of Pathology, Faculty of Medicine, Pontificia Universidad Católica de Chile (Santiago, Chile), for his valuable help in preparing lung tissue for histologic analysis; and Carlos Martinez, D.V.M., from Center for Simulation and Experimental Surgery (Centro de Simulación y Cirugía Experimental), School of Medicine, Pontificia Universidad Católica de Chile (Santiago, Chile), for his support in placing vascular access and the extracorporeal membrane oxygenation cannulas.

Drs. Cruces, Retamal, Cornejo, Bugedo, and Bruhn acknowledge support from CONICYT (Santiago, Chile) through grants FONDECYT 1161556 and FONDECYT 1191709. Drs. Dubo and Damiani acknowledge partial support from CONICYT (Santiago, Chile) PFCHA/Doctorado Nacional/2018-21181376 and PFCHA/Doctorado Nacional/2017-21171551, respectively.

Dr. Rodrigo acknowledges lecture fees for Medtronic (Santiago, Chile). The other authors declare no competing interests.