The Role of VV-ECMO in Severe COVID-19 ARDS

Abstract

Although an established practice in potentially reversible severe respiratory failure, extracorporeal membrane oxygenation (ECMO) support remains controversial. Over the last 50 years, only 4 large scale randomised controlled trials relating to ECMO have been conducted in patients with ARDS. A meta-analysis of only 2 studies has demonstrated survival benefit in those supported with ECMO compared to optimal conventional therapy. With the advent of the COVID pandemic, ECMO utilisation increased, the guidelines evolved, and an unprecedented number of patients were referred for and managed with ECMO support. Approximately 15,000 patients have been supported to date, predominantly using veno-venous ECMO, with an overall in-hospital 90-day mortality of 47%. Although published data reported an increase in ECMO mortality to nearly 60% as the pandemic progressed, this was likely multifactorial, as subsequent data has demonstrated more promising mortality results. This highlights the unique challenges pertaining to patient selection and implementation of this finite support amid an evolving pandemic with many unknowns. Judicious and ethical patient selection is essential to ensure use for the greatest benefit. In this chapter we will outline the unique pathophysiology and clinical features of COVID-ARDS, indications for ECMO referral and patient selection, and implementation during the COVID-19 pandemic.

1. Introduction

With the outbreak of COVID, extracorporeal membrane oxygen support (ECMO) utilisation exponentially increased, and the guidelines on ECMO referral, selection, and patient management rapidly evolved to cope with the unprecedented scale of the pandemic [1]. To date, an unparalleled number of patients have been referred for and managed with ECMO. According to the Extracorporeal Life Support Organisation (ELSO) COVID-19 ECMO registry, approximately 15,000 patients have been supported so far, predominantly using veno-venous ECMO (VV ECMO), with an overall in-hospital 90-day mortality of 47% [2]. Although the pandemic led to an upscale in ECMO use, ECMO mortality actually increased to nearly 60% as the pandemic progressed over 2020 [3]. However, this increase is likely a function of multiple interconnected factors as more recent mortality data has been more optimistic with an estimated survival probability of 87% on Day 7 ECMO and 78% at 90-days, compared with 83% and 64% respectively in the conventional management group [4]. The type of variant, patient demographic and comorbidity, a more severe COVID-ARDS phenotype and lack of reversibility, increased and/ or more resistant co-infections (some of which are possibly associated with steroids and novel COVID-19 therapies), and treatment in a low vs. high volume ECMO centre may have contributed to this mortality variance. The inconsistency in patient outcomes highlights the unique challenges pertaining to judicious and ethical patient selection, and appropriate application of this expensive and finite mode of support amid an evolving pandemic with many unknowns in order to ensure its use for the greatest benefit.

2. Background

2.1 An overview of COVID-19 pathophysiology

Since its isolation in December 2019 in Wuhan China, the novel severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), and its resultant syndrome COVID-19, have affected nearly 567 million people across the globe, with a death toll of over 6 million according to the latest data from the World Health Organisation [5]. Its clinical presentation has been variable, ranging from asymptomatic “happy hypoxaemia”, to one of refractory severe acute respiratory failure in the intensive care unit requiring potentially lifesaving extracorporeal support. Approximately 20% of patients with COVID-19 develop severe COVID pneumonitis, which is similar to conventional acute respiratory distress syndrome (ARDS) as defined by the Berlin criteria [6, 7].

Severe COVID-19 is the consequence of a virally triggered cytokine storm, resulting in initial endothelial inflammation and hypercoagulopathy, rapidly followed by pulmonary oedema, progressive lung parenchymal consolidation, diffuse alveolar damage, and pulmonary fibrosis. The interaction between the SARS-CoV-2 spike proteins and the angiotensin converting enzyme-2 (ACE2) receptor present on type 2 pneumocytes via receptor-mediated endocytosis, has been postulated as the fundamental mechanism driving this cytokine response [8]. Although patients generally present with isolated respiratory failure, progression to multiorgan failure may be rapid. COVID-related multiorgan failure and secondary infection related multiorgan failure are the leading causes of mortality, accounting for 37% and 26% of deaths respectively [9]. A hypercoagulable state is particularly common, and is reflected by increased fibrinogen and D-dimer levels in almost all patients, with a concomitant increased incidence of both venous and arterial thrombosis, pulmonary thromboembolism, and associated increased mortality [10]. Post-mortem studies have demonstrated that the histological hallmark of COVID ARDS is diffuse alveolar infiltration with varying degrees of pulmonary vascular thrombosis [11, 12, 13].

2.2 Hypoxaemia and respiratory system compliance

Although there is progressive hypoxaemia and dyspnoea, the hypoxaemia in COVID-19 is often more severe than that expected from the anatomical shunt alone. In fact, despite a PaO2/FiO2 ratio < 200 mmHg (26.7 kPa) in moderate-to-severe disease, approximately 40–50% of patients have preserved respiratory system compliance (CRS), with peripherally distributed ground glass opacification and minimal parenchymal consolidation, which is in stark contrast to the majority of non-COVID ARDS [14]. There is a small subgroup of classic ARDS that may have high compliance [15]. The underlying mechanisms for the disproportionate hypoxaemia are multifactorial, determined by a temporal and spatial heterogeneous mismatch of pulmonary ventilation and perfusion, a loss of hypoxic pulmonary vasoconstriction, and dysregulated pulmonary blood flow, mainly associated with immunothrombosis, endothelial inflammation and neovascularisation [16, 17, 18].

COVID-19 patients tend to have preserved CRS despite significant pulmonary fibrosis, pulmonary infiltration, pulmonary vascular micro and/ or macro thrombosis and resulting hypoxaemic respiratory failure. A range of clinical phenotypes may exist. Maiolo et al. proposed 3 distinct groups. Firstly, a high elastance-poor compliance group (“H-type”) with a CRS < 40 ml/cm H₂O, increased right-to-left shunt, increased lung weight, and potentially more recruitable lung. This group accounts for 20–30% of COVID ARDS patients in critical care [19]. The second group is “Intermediate CRS”, defined as a compliance of 40–50 ml/cm H₂O, and finally a reduced elastance group (“L-type”) characterised by preserved/ high compliance (CRS >50 ml/cm H₂O), low ventilation-perfusion ratio, low lung weight, and minimal recruitable lung. Camporota et al. have proposed that the earlier phase of COVID ARDS is characterised by a high compliance phenotype, with a transition to a poorer compliance state as the disease progresses [20]. However, further data has questioned the phenotype concept. Several single centre observational studies, including the recent COVADIS study, have demonstrated that the mean CRS is actually rather poor, c. 30–40 mL/cm H₂O. The COVADIS results demonstrated a unimodal distribution of CRS around a mean value of 37 ml/ cm H₂O, similar to that observed in non-COVID-19 ARDS. CRS decreased from day 1 to day 14, and interestingly patients with higher CRS did not demonstrate faster weaning of mechanical ventilation or increased survival in multivariate analyses [21]. Ferrando et al. demonstrated a similar mean CRS distribution of 35 ml/cm H₂O (IQR: 27–45). However, their findings were likely limited by a high proportion of incomplete data [22]. Factors that may partly explain some of the variability in compliance data may be the time since disease onset and time from disease onset to intubation. Early in the pandemic, intubation tended to occur based on hypoxaemia alone. However, as the pandemic progressed, intubation was often deferred until more clinical disease progression occurred at which point compliance would have been poor, and of a more typical ARDS nature. Mortality in COVID-ARDS does however correlate with poorer compliance (CRS < 48 ml/cm H₂O) and increased driving pressure, independent of tidal volume per kilogramme based on ideal body weight, and even with tidal volumes above the accepted 6–8 ml/kg threshold [23]. Notably, patients with COVID ARDS who have a reduced CRS together with increased D-dimer concentrations have a worse survival prognosis [14].

2.3 Pulmonary hypertension

Pulmonary hypertension leading to acute right ventricular (RV) dysfunction +/− failure may occur in COVID ARDS and is associated with a significantly increased mortality (48.5% versus 24.7% in patients with and without RV dysfunction respectively; 56.3% versus 30.6% in patients with or without RV dilatation). Mortality is high even in patients with pulmonary hypertension (PH) without RV strain (52.9% versus 14.8%) [24]. The underlying mechanism is primarily an increase in RV afterload due to increased pulmonary vascular resistance (PVR). Multiple factors in COVID ARDS contribute to PH, elevated PVR, increased RV afterload and, eventually, RV failure. These include hypoxaemia, hypercapnia, acidosis, hypoxic pulmonary vasoconstriction, endothelial inflammation, pulmonary vascular thrombosis, and vascular remodelling. RV dilatation increases the RV distending pressure, thereby increasing the pressure gradient for subendocardial myocardial perfusion, resulting in impaired RV contractility. Pressure volume overload consequently impairs left ventricular function and cardiac output. Although possibly more severe in COVID-19 ARDS, RV dysfunction can be alleviated by improving gas exchange with VV ECMO [25].

2.4 Management principles in severe COVID-19

The management of severe COVID-ARDS is multimodal, and the intensity of support required depends on the phenotype and severity of the disease at presentation. Potentially reversible severe COVID pneumonitis that is refractory to protective lung ventilation, ventilatory adjuncts i.e., prone positioning, inhaled pulmonary vasodilators, neuromuscular blockade etc., and also to targeted pharmacological therapy i.e., steroids, immunotherapy, and antimicrobial agents may require extracorporeal membrane support oxygenation as a bridge to recovery or in exceptional cases, lung transplantation [26]. Venovenous (VV) ECMO support is the modality of choice in 95.9% of cases, as demonstrated in a systematic review and meta-analysis of 18,211 COVID-19 patients by Ling et al. [27] ECMO may also represent an efficient support in cases of severe cardiogenic/septic shock refractory to maximal therapy in these patients. However, venoarterial (VA) ECMO, conversion to VA ECMO from VV ECMO, or use of hybrid ECMO circuits are rare in COVID-19, accounting for <5% of all cases [27]. It is also important to consider that patients with other potential indications for ECMO support, such as massive pulmonary embolism, myocarditis or acute myocardial infarction, may also be COVID-19 positive [28]. Although seen as an established therapy in potentially reversible severe respiratory failure, ECMO remains controversial. Over the last 50 years, only 4 large scale randomised controlled trials have been conducted in patients with non-COVID-ARDS [29, 30, 31, 32]. Overall, these studies have not demonstrated superiority of ECMO over maximal conventional support i.e., protective lung ventilation, prone positioning etc. However, a meta-analysis of the CESAR and EOLIA studies demonstrated a 90-day mortality benefit in the ECMO group (36% vs. 48%; relative risk, 0.75, 95% confidence interval 0.6–0.94; p = 0.013), in addition to more ventilator free days and days out of ICU [33].

3. Referral, patient selection, and ethics

ECMO is a complex and resource intensive intervention. Its use is mostly restricted to specialist centres globally. Disease severity, reversibility and patient reserve are important aspects when considering suitability for ECMO. Selection of patients who will ultimately benefit most is crucial to avoid suffering and prolonged futile ICU admission, in addition to appropriate allocation of an expensive, finite, and labour-intensive resource.

ECMO played a crucial role in previous respiratory viral outbreaks, such as the Middle East Respiratory Syndrome coronavirus (MERS-CoV2) in 2012, and the influenza A virus subtype hemagglutinin 1 neuraminidase 1 in 2009 (H1N1), with acceptable survival rates ranging from 65% to 77% [34]. However, it was clear from the start of the COVID-19, that this global outbreak would place an even greater strain on healthcare systems worldwide in comparison with its predecessors. Early in the pandemic, definitive data to guide clinical decision making for patient selection in severe COVID-19 was lacking, and established protocols for the initiation of VV ECMO in COVID-19 were therefore largely based on 2 randomised controlled trials of ECMO for non-COVID ARDS [29, 30]. As pandemic phases evolved, ELSO adapted its guideline recommendations in this regard [1]. The 2020 ELSO guidelines, the 2021 ELSO update, and guidelines from other international bodies recommended that VV ECMO should be considered in all patients with COVID-19 and severe refractory hypoxaemia despite optimal conventional therapy [1, 35, 36, 37, 38].

Providing complex, finite, and resource intensive therapies such as ECMO during a pandemic has unique challenges [39]. During COVID, referral and selection criteria for VV ECMO had to be redefined during essential resource planning and allocation in order to ethically deploy finite resources. A better understanding of the disease process developed as the pandemic progressed, allowing dynamic modification of these criteria. As a result, regional ECMO services developed individualised approaches to patient selection with a unified aim to ensure that ECMO is offered to those patients who are more likely to reap the most benefit [34, 40, 41]. There is however, some heterogeneity in published selection criteria based on regional variations in demographics, pandemic phase, and resource availability [34, 41].

3.1 Referral criteria

Referral for ECMO (Table 1) should be considered in any COVID-19 patient with potentially reversible acute hypoxaemic respiratory failure, defined as a PaO2/FiO2 ratio < 80 mm Hg, refractory to maximal conventional therapy as per the ELSO recommended algorithm i.e. treatment of underlying cause, protective lung ventilation, diuresis; followed by prone positioning, increased PEEP, use of neuromuscular blockade, and inhaled pulmonary vasodilators +/− recruitment manoeuvres [1]. Severity of hypoxemia in COVID-19 respiratory failure is characterised by the partial pressure of arterial oxygen to fraction of inspired oxygen (PaO2/FiO2) ratio using thresholds recommended in the EOLIA trial [30]. Early referral is particularly important for patients deteriorating in non-ECMO centres. Referral should be made as early as possible in advance of further deterioration to facilitate timely retrieval and transfer to the designated ECMO centre. ECMO initiation should not be delayed due to resource constraints; delays in initiation are associated with increased mortality in both COVID and non-COVID ARDS [1, 42, 43]. There is no consensus on absolute contraindications for referral in COVID ARDS, except in cases of end-stage respiratory failure unsuitable for lung transplantation, and also when critical care system capacity is at crisis point. The optimal timing of ECMO initiation in relation to intubation remains debatable. In non-COVID ARDS, a duration of mechanical ventilation of 7 days or longer is considered to increase the likelihood of irreversibility [29, 33]. Contrary to this, as discussed in detail below, there is emerging evidence that this has minimal impact on mortality in COVID-19 patients supported with ECMO. However, it is still reasonable to assume based on the available evidence, that earlier initiation is associated with improved survival [34, 40, 41, 44]. Of note, high intensity and prolonged non-invasive ventilation (NIV) in an attempt to avoid intubation during the second wave resulted in delayed initiation of evidenced based lung protective ventilation with an increased incidence of barotrauma. Referral criteria were redefined by some ECMO experts to consider days of high intensity continuous positive airway pressure (CPAP) or non-invasive ventilation (NIV) as days of mechanical ventilation in an attempt to encourage early referral [41].

1. Refractory severe hypoxaemia; PaO2/FiO2 ratio < 80 mmHg for >6 hours; OR; PaO2/FiO2 ratio < 50 mmHg for >3 hours

+/− pH < 7.25

+/− PaCO2 > 60 mmHg for >6 hours

2. Uncompensated refractory severe hypercapnia, with or without severe hypoxaemia (PaO2/FiO2 < 150 mmHg)

*Respiratory rate increased to 35; *Plateau pressure < 32cmH20

3. Lung Injury Score ≥ 3

PLUS

4. Failure of optimal protective lung ventilation and adjunctive therapy

5. Absence of contraindications to extracorporeal support

6. Thorough discussion with a national ECMO centre

Table 1.

Referral criteria for VV ECMO in severe refractory COVID-19.

Adapted from Badulak et al. [1], Dalia et al. [34], Camporota et al. [41].

3.2 Patient selection

Given the required judicious approach to patient selection during a pandemic with finite resources, how do we determine who will benefit most? Risk factors for poor survival have been identified and are further discussed in the section on mortality and morbidity. Overall, increasing age, ECMO centre experience and pre-existing concomitant disease are substantial factors congruent worldwide. These factors, in addition to premorbid functional capacity, must be considered during the process of referral and when deciding to initiate ECMO, in order to determine a realistic survival and rehabilitation potential, and also expected quality of life after ECMO. For patients with challenging considerations or potential relative contraindications (Table 2), it would be reasonable to suggest that at least two ECMO centres should agree that it is appropriate to proceed to ECMO in these cases.

	Recommended	Consider in individual cases^**	Not indicated
Age (years)	<65	65–69	No definitive consensus; But >70 is associated with much higher mortality
Comorbidity	No significant comorbidity	Immunosuppression (within 6 months prior)	BMI > 35
		Recent neurosurgical procedure	Refusal of blood products or anticoagulation
		Renal failure	Documented end-stage chronic organ failure (e.g., COPD, cirrhosis); not for device therapy or transplant
			Severe acute neurological injury with poor prognosis for recovery
			Active malignancy
			Cardiac arrest >15 min without tissue perfusion
Frailty	Low	Moderate	High
Organ failure	Isolated respiratory failure	“Mild” additional organ failure	Severe acute multiorgan failure with anticipated death despite ECMO support
		Secondary infections with multidrug resistant organisms
Mechanical ventilation	<7 days	>7 days	No agreed absolute contraindication
	High driving pressure > 15 cm H2O	Unsuccessful trial of prone ventilation ≥6 h
Scoring system		Indices of low potential to recover (e.g., a respiratory ECMO Survival Prediction [RESP] score of ≤ 3)
Ethics	A declared or presumed patient’s will in favour	Patient’s will unclear, next of kin undecided	A declared or presumed patient’s will against ECMO
System capacity	Normal/expanded capacity; Normal/restricted criteria	Near saturation capacity; Highly selective ECMO use only	Crisis capacity; No further cannulations possible

Table 2.

Selection criteria for VV ECMO in severe refractory COVID-19—A graduated approach.

Adapted from Badulak et al. [1], Dalia et al. [34], Camporota et al. [41], Karagiannidis et al. [40], and Herrmann et al. [44].

3.3 Ethics and ECMO in a pandemic

Although there has been intense debate regarding ethical allocation of critical care resources during the COVID-19 pandemic, there has been a paucity of professional guidance specifically relating to ethics and ECMO allocation, with the notable exception of a general ethical guidance document published by ELSO in May 2020 [45]. An international survey of ECMO practitioners (primarily from the ECMOCard group) during the early stages of the pandemic has shed some light on the current ethical climate. Probability of survival if treated, pre-existing disability, functional status, and patient age were the most cited discriminating factors in decision making around maximising patient survival benefit when considering suitability for ECMO support. Most participants stated that their criteria for starting ECMO had changed during the COVID-19 pandemic, with lower age limits, decreased ECMO machine availability, and stricter inclusion criteria. Interestingly, the majority of those surveyed agreed that it would be ethical to give extra priority to a healthcare worker who had contracted COVID-19 due to occupational exposure. Decision making pertaining to futility and cessation of therapy were predominantly guided by lack of benefit to the patient being supported, however, as the pandemic progressed, there was an increased move to a more utilitarian approach i.e., they would consider discontinuing ECMO in a patient with poor prognostic outcome, in order to provide ECMO to a patient more likely to survive [46]. The current ELSO guidance suggests that treatment may be regarded as futile and discontinued after 21 days. However, it is clear from the published data that COVID patients frequently require more prolonged ECMO support compared with other cohorts of up to 5 weeks duration [3, 47].

The process of patient selection is complex and multifactorial and for this reason optimal indications for ECMO support in patients with severe COVID-19 ARDS remain unknown. It is clear however that the indications for ECMO are moving away from rescue therapy and more into an extended form of standard conventional therapy.

Key points

• Patients with severe COVID-19 ARDS should be referred for ECMO early, preferably within the first week of mechanical ventilation.

• Age, ECMO centre experience, and concomitant disease are major risk factors for poor survival.

• Patient selection should be decided on a case-by-case basis guided by multiclinician shared decision making and careful assessment of several immunologic, biographic, medical, and prognostic parameters.

• ECMO in COVID-19 is time and resource-intensive, and not all patients will benefit from this invasive support. We have an ethical responsibility to select the right patient in order to avoid harm.

4. Outcomes in patients supported with VV ECMO during the COVID-19 pandemic

4.1 Predicting outcomes in ECMO and COVID ARDS

Prognostication in COVID ARDS patients being considered for ECMO is very challenging, particularly due to the inconsistent mortality data reported over the course of the pandemic. Various scoring systems have been suggested as an aid to improve risk stratification, prognostication, and allocation of resources, particularly when the healthcare system is constrained. These include the Respiratory ECMO Survival Prediction score (RESP) and the PRedicting dEath for SEvere ARDS on VV-ECMO (PRESERVE) score, both of which were developed exclusively for outcome prediction in patients requiring VV ECMO. Other scoring systems studied in ECMO patients include the Roch score, and general critical care scoring systems such as Sequential Organ Failure Assessment (SOFA) and SAPS II [48, 49, 50]. The RESP score was used by some centres during the pandemic to encourage shared decisions making amongst experts when faced with particularly high-risk challenging cases [41]. A prediction model development study by Moyon et al. demonstrated acceptable discrimination and a good calibration with the RESP score in comparison with both the PRESERVE score and more traditional critical care scores e.g., SOFA [51]. However, other studies have demonstrated poor predictive ability of all the above scoring systems, including the RESP score, with prognostic accuracy ranging between approximately 0.5 and 0.6 (based on the area under the receiver operating characteristic curve—AUROC) [52, 53]. The general consensus is that existing scoring tools appear to perform poorly in COVID-19 patients, both in terms of those being considered and those being supported with VV ECMO. They should not be used in isolation to guide patient selection or refusal, but should be applied judiciously, in conjunction with clinical judgement, expertise, and guideline recommendations.

4.2 Mortality

Initial case studies and case reports of VV ECMO in COVID-19 were discouraging. They suggested a high mortality and raised significant concerns regarding its potential use in this patient population [54, 55, 56]. However, as the pandemic evolved, subsequent ECMO COVID-19 outcome data published from the ELSO Registry reported an estimated cumulative incidence of 90-day in-hospital mortality of 37.4% (95% CI 34.4–40.4), comparable to outcomes after ECMO in non-COVID ARDS [30, 47]. This was supported by several other multicentre observational cohort studies [42, 57, 58]. One of these, a systematic review and meta-analysis by Ramanathan et al., examined the use of ECMO in adult patients with COVID-19 during the first year of the pandemic (Dec 2019 to Jan 2021). This included 22 observational studies, with 1896 patients included in the meta-analysis, and VV ECMO was used in 98.6% of cases. In-hospital mortality in patients receiving ECMO support for COVID-19 was 37.1% during the first year of the pandemic, similar to those with non-COVID-19-related ARDS [42].

As the pandemic progressed, mortality with VV ECMO utilisation in COVID-19 began to increase from the summer of 2020. Barbaro et al. demonstrated that prior to May 1st 2020, the COVID-19 ECMO mortality rate was 36.9% (95% CI 34.1–39.7) compared with 51.9% (50.0–53.8) for patients who started ECMO after this date. Mortality was even higher at 58.9% (55.4–62.3) for patients treated at centres that only offered ECMO after this date. This large multicentre retrospective study of 4812 patients broadly categorised patients as to whether they were managed at early adopting (groups A1 and A2) vs. late adopting centres (group B). Not only did they demonstrate a 15% increase in mortality over the course of the pandemic, but also an increase in the median duration of ECMO support by 6 days. Compared with to patients in group A2, group A1 patients had a lower adjusted relative risk of in-hospital mortality 90 days after ECMO (hazard ratio 0.82 [0.70–0.96]), whereas group B patients had a higher adjusted relative risk (1.42 [1.17–1.73]) [3]. The large multicentre French cohort study of the ECMOSARS registry data has one of the highest in-hospital mortality rates to date of 51%, although this may be due to several factors including an older population compared with the ELSO registry and STOP-COVID studies, a more severe COVID-ARDS phenotype (99% of patients met the Berlin definition criteria compared with 79% in the ELSO study, and a longer duration of mechanical ventilation before ECMO cannulation (median 6 days vs. 4 days in the ELSO study population) [47, 57, 59]. A meta-analysis of 6 studies by Bertini et al. found that the COVID-19 ECMO cohort had a 1.34 increased relative risk of mortality when compared to patients with influenza (44% vs. 38%; RR 1.34; 95% CI 1.05–1.71; p = 0.03) [60]. A robust systematic review and meta-analysis by Ling et al., which included a cohort of >18,000 COVID-19 patients receiving ECMO between Dec 2019 and Jan 2022, found a pooled mortality rate of 48.8%, higher than that reported in the review by Bertini et al. [27, 60].

However, other studies have shown more promising results. Shaefi and colleagues conducted an emulated target trial using observational data to assess the efficacy of ECMO compared with conventional mechanical ventilation in COVID-19. They included patients with severe hypoxemia and observed a reduction in mortality with ECMO (hazard ratio, 0.55; 95% confidence interval, 0.41–0.74) [57]. Urner et al. also performed an emulated target trial similar to Shaefi et al. They conducted a multicentre retrospective observational study of 7345 patients with severe COVID-ARDS admitted between January 2020 and August 2021, 844 of whom received VV ECMO support. They demonstrated a significant reduction in hospital mortality by 7.1% compared with conventional mechanical ventilation without ECMO. Secondary analyses revealed several factors that were significantly associated with reduced ECMO efficacy which are discussed below [61]. Most recently, Hajage et al. have added to the above emulated trial data with their multicentre observational cohort study of 2858 patients, 269 of whom were supported with ECMO. Overall survival at day 7 of ECMO support was high, and comparable between ECMO and non-ECMO survivors (87% vs. 83% respectively). Mortality increased as time progressed, with a reported survival rate of 63% at 90-days which was not significantly different to that of the conventional management group. However, they did demonstrate a significantly improved 90-day survival rate in high volume centres where ECMO was initiated early (within the first 4 days of intubation) in severe COVID ARDS; survival was 78% on ECMO vs. 64% in the conventional arm [4]. Finally, Whebell and colleagues also provided a more optimistic outlook in their multi-centre matched retrospective study of COVID-19 patients from 111 hospitals, referred to two specialist ECMO centres in the United Kingdom between March 2020 and February 2021 [62]. Of 1363 patients referred, 243 were retrieved on mobile ECMO to the quaternary centre. They demonstrated a marginal odds ratio (OR) for mortality of 0.44 (95% CI 0.29–0.68, p < 0.001) and absolute mortality reduction of 18.2% (44% vs. 25.8%, p < 0.001) for treatment with ECMO in a specialist centre, compared with patients managed conventionally in the referring hospital. The findings from Whebell et al. differ compared with other similar cohort studies. In their study, mortality did not increase significantly in the ECMO group during the second wave (22.9% vs. 26.1%, p = 0.672), however it increased significantly in those managed with conventional support (51.9% vs. 62.4%, p = 0.001). This is likely a factor of increased early adjunctive therapy e.g., immunomodulatory agents, and a greater implementation of more discerning ECMO selection criteria. Selected patients were also younger, with lower SOFA and higher RESP scores, and had less duration of organ support prior to ECMO. Notably, a higher proportion of patients with documented ‘perceived futility’ and a lower proportion of ECMO treated patients were seen in the second wave [62].

The variability in mortality seen in studies of ECMO support in COVID-19 patients to date is likely multifactorial (Table 3). Early studies were limited by the inclusion of unselected populations and the lack of adequate controls. In addition, there were substantial changes in the management of COVID-19 as the pandemic evolved, in line with rapidly emerging evidence from large multicentre platform studies, which may have affected the category of patient progressing to ECMO support [63, 64, 65, 66]. Patients were frequently supported with high-flow oxygen, non-invasive ventilation, awake prone positioning, and immunomodulatory therapy as part of standard care which may have mitigated the need to advance to more invasive support therapy [67, 68, 69, 70, 71, 72, 73]. These developments occurred in parallel with a significant increase in the number of centres providing ECMO support to patients with COVID-19 [74]. In general, older age, increasing burden of comorbidity, increased vasopressor requirement and need for renal replacement therapy (RRT), and increased bleeding complications are more common in COVID-19 patients supported with ECMO who die compared with those who survive [75].

Patient factors	Treatment factors	Organisational factors
Age ≥ 65	Pre-ECMO ventilatory therapy; *duration, intensity; *PaO2/ FiO2 ratio < 80 mmHg; * DP >15cmH20	High volume centre experience
ARDS Phenotype		Time interval to ECMO; i.e. symptoms to cannulation
		Cannula Fr size
≥ 2 Comorbidities; * Hypertension; * Obesity; * Ischaemic heart disease; * Diabetes	Refractory to ventilatory adjuncts; *Proning; *High PEEP; *NMB; * Inhaled pulmonary vasodilators
		Duration of ECMO support
		Number of ECMO runs
Pulmonary hypertension +/− right ventricular dysfunction	Novel COVID-19 therapies; *steroids; *IL-6 inhibitors
Male

Table 3.

Summary of factors impacting mortality in COVID-19 patients supported with ECMO.

It is still unclear whether the use of VV ECMO definitively confers improved survival in patients with COVID-19 ARDS. Mortality rates and the duration of support required have so far been inconsistent. Different studies have shown variable outcomes for COVID-19 patients supported with VV ECMO depending on the phase of the pandemic [3]. Although Barbaro et al. reported a 90-day in-hospital mortality of 37% for COVID-19 patients supported with ECMO, we still do not know the long-term outcomes of COVID-19 ECMO patients who have survived [3, 47]. The most recent 60-day and 90-day ECMO survival data from the more recent emulated target trials is however very reassuring [4, 57, 61]. However, the findings from these studies must be taken in the context of certain limitations such as lack of random treatment allocation and unmeasured confounders which may have biased the study results in either direction [76].

4.3 Morbidity

Complications in critically ill patients receiving ECMO are well described, with higher incidence associated with longer duration of ECMO support, increased duration of mechanical ventilation, and coagulopathy associated with both pharmacotherapy and the prothrombotic environment within the ECMO circuit [47, 77, 78, 79]. The overall incidence of complications, in COVID-19 patients supported with ECMO (when defined as complication rates per 1000 hours of ECMO support), excluding renal replacement therapy, is 50–60%. Renal complications in general account for 10–30% overall, depending on the pandemic phase, definition/ parameters used, and other patient and treatment factors [3, 42].

Bleeding and thrombosis are common and are associated with increased mortality in patients supported with ECMO [80]. A retrospective ELSO registry study analysing bleeding and thrombotic events (BTEs) in 7579 VV-ECMO patients between 2010 and 2017, the largest multicentre study of its kind to date, reported a 40.2% incidence of ≥1 BTEs in patients supported with VV-ECMO. The in-hospital mortality rate associated with bleeding and/ or thrombosis was 34.9% in this cohort overall. Thrombotic events were more common than bleeding and comprised 54.9% of all BTEs. This contrasts with VA-ECMO where bleeding events tend to predominate [81]. The most common thrombotic events were circuit clotting (31.8%) and oxygenator/pump failure (12.7%). Bleeding is common and complicates the course of approximately 16–60% of patients managed with ECMO [82, 83]. In the aforementioned ELSO registry study, cannulation sites (15.5%) and surgical bleeding (9.6%) were the most common sources, with medical bleeding accounting for 18.7% of events [80]. Intracranial haemorrhage was more common than ischemic stroke (4.5% vs. 1.9%, respectively). This is consistent with the findings from previous studies in this area, which have also reported incidences of other significant bleeding events such as gastrointestinal (5.5%) and pulmonary (6.1%) haemorrhage [82, 84]. Major bleeding requiring transfusion occurs in 39%, and the mortality associated with bleeding may be as high as 48.5% [85, 86]. The overall incidence of neurological complications in ECMO is approximately 7–9%, with intracranial haemorrhage accounting for 38% of these cases [87]. Of note, during the H1N1 influenza pandemic, intracerebral haemorrhage was reported as the commonest cause of death in patients supported with VV ECMO [88, 89].

Immune-mediated thrombosis has been postulated as a key mechanism in the pathophysiology of COVID-19, and its associated increased thrombotic risk profile. Much research has been dedicated to this area, and in finding the optimal anticoagulation strategy for these patients. However, there is little hard evidence to inform us of the specific bleeding and thrombosis risk in COVID-19 patients supported with ECMO. To date most of the evidence has come from case series reports. In the multicentre ECMOSARS cohort study of approximately 500 COVID-19 patients, haemorrhagic complications occurred in 40% patients, thrombosis occurred in 37%, and neurological complications in 11%. 80% of neurological complications were due to haemorrhagic stroke [59]. Interestingly, the incidence of haemorrhagic complications was higher compared with the Study of the Treatment and Outcomes in Critically Ill Patients with COVID-19 (STOP-COVID), 28% vs. 40%, and also that of the ELSO registry study [47, 57]. A follow-on ECMOSARS registry study by Mansour et al. analysed all patients in this registry over the course of the first and second pandemic waves from February 2020 to the end of March 2022 [90]. In this review, 65.5% of patients experienced either bleeding or thrombosis. Interestingly, thrombosis rates remained stable over the course of the pandemic (approximately 35%), while bleeding increased. Bleeding events (49% of patients) were associated with a significantly higher in-hospital 90-day mortality of 71.8%, unlike thrombosis which was not associated with a significantly increased mortality (adjOR = 1.02 [0.68–1.53]. The commonest bleeding and thrombosis sites were similiar to previous reported studies. Intracranial haemorrhage was independently associated with an increased mortality risk (adjOR = 13.5 [4.4–41.5]. Massive transfusion was required in 10% of bleeding events, and successive bleeding events increased mortality fourfold. Independent risk factors for bleeding included the duration of ECMO support and ventilation duration ≥7 days prior to ECMO cannulation, whilst a fibrinogen >6 g/ L at cannulation was predictive of thrombosis. Barbaro et al. reported similar rates of intracranial, pulmonary and gastrointestinal bleeding in their retrospective multicentre ELSO registry study, however the prevalence of cannula site bleeding events was lower (approximately 6% across the study subgroups) compared with the aforementioned ECMOSARS studies. In addition, they found that haemolytic, haemorrhagic, ischaemic, neurological, and mechanical complications were broadly similar in both early-adopting vs. late-adopting centres over the course of the pandemic [3]. The increased bleeding and lower thrombosis incidence reported by Mansour et al. compared with the ECMOSARS and Nunez et al. ELSO studies, particularly in relation to device thrombosis and membrane failure, is possibly related to the generalised augmentation of anticoagulation therapy over the course of the pandemic.

Notably, randomised controlled trials have not demonstrated a clear benefit for therapeutic heparin in critically ill patients with COVID-19. However, observational study data does suggest evidence of benefit for prophylactic dose low molecular weight heparin (LMWH) in non-critically ill COVID-19 patients in terms of organ support free days, even in those without a documented thrombotic event, albeit with an increased risk of bleeding, and this has formed the basis for widespread prophylactic anticoagulation in these patients [91, 92]. Despite increased knowledge of the risks of thrombosis and bleeding in immobilised COVID-19 patients supported with ECMO, the optimal anticoagulation regimen remains to be fully elucidated. There is no consensus on the optimal choice of anticoagulant, dosing, and duration of treatment; and there is significant regional and institutional variability in clinical practice. It becomes an even more complex scenario with the addition of ECMO, where a minimum threshold of systemic heparinisation and possibly antiplatelet cover are required to prevent circuit thrombosis. However, it is a double-edged sword, as ECMO also depletes host antithrombin (AT) levels through haemodilution, coagulation factor activation, and consumption by unfractionated heparin, thereby reducing heparin efficacy and potentially increasing the risk of thrombosis. Bleeding risk is also increased in AT depletion due to clotting factor consumption by the circuit, and also a relative increased inflammatory coagulopathic response due to a lack of AT anti-inflammatory activity [93]. To date, ELSO have not made any specific recommendations in this arena beyond usual recommended anticoagulation practice for patients receiving ECMO support [1].

Rates of infectious complications in COVID-19 patients on ECMO have been variable. The ECMOSARS Investigators demonstrated a much higher incidence of ventilator associated pneumonia (VAP) and bacteraemia of 51% and 41% respectively, compared with STOP-COVID (Study of the Treatment and Outcomes in Critically Ill Patients with COVID-19) which reported a 35% incidence of VAP and 18% incidence of other infections [57, 59]. This vulnerability to increased infection is likely multifactorial, related to the increased duration of mechanical ventilation, increased used of immunomodulatory agents e.g. steroids, IL-6 inhibitors etc., increased multidrug resistant organisms, and difficulties around maintaining sterility in a high stress and resource constrained environment. In the ECMOVIBER (The use of ECMO during the coVid-19 pandemic in the IBERian peninsula) study, co-infection at ECMO initiation was recorded in 29.8% of cases, although this was not significantly associated with increased mortality [43]. Unsurprisingly, the incidence of complications overall is significantly higher in non-survivors compared with survivors [59, 75].

4.4 Risk factors associated with morbidity and mortality

Age is the major factor predictive of increased mortality in COVID-ARDS patients supported with extracorporeal therapy, with age over 65 increasing mortality four-fold. The relationship between older age, defined in the ELSO guidelines as ≥65 years, and poorer survival is a constant finding in the COVID-19 literature, with OR doubling to 2.7 at 60–70 years and doubling again between 70 and 80 years [35, 94]. However, increasing age alone should not automatically exclude suitability for ECMO but should be reviewed in combination with pre-existing comorbidity and concomitant disease, including organ failure. A male preponderance, ≥2 comorbidities (particularly hypertension, diabetes, ischaemic heart disease, obesity, immunosuppression), the variant subtype, and a more severe ARDS phenotype with lack of reversibility are also associated with worse outcomes [3, 27, 58 , 59, 61, 75, 90]. Of note, there is some conflicting evidence to suggest that obesity is not associated with poorer outcomes in COVID ECMO patients, including increased 90-day mortality [95, 96].

As the pandemic evolved, the nature of pre-ECMO therapies also changed. The intensity and duration of these treatments may have impacted on morbidity and mortality in patients who went on to be supported with ECMO. For example, in the second wave compared with the first, there was a significant difference in the uptake of adjunctive therapies i.e., steroids 99.3% vs. 12%; IL-6 inhibitor 13.1% vs. 1%; NIV 78.2% vs. 49.6%; prone positioning 82.5% vs. 68.7%; and nitric oxide 16% vs. 6.1% [62]. In a recent systematic review and meta-analysis, most patients received neuromuscular blockade (96.2%) and were positioned prone (84.5%) prior to initiation of ECMO [42]. Many of these therapies have an established mortality benefit in COVID-19, steroids being the main example, and some patients may have a more responsive phenotype [97]. Therefore it is possible that the high mortality of approximately 40% in COVID ARDS patients supported with ECMO may stem from a selection bias for a more treatment resistant phenotype, given that patients who responded well to protective lung ventilation and adjunctive therapies may not have progressed to require ECMO. Immunomodulatory therapy may be associated with increased secondary infection, which may also have contributed to a worse ECMO survival rate [98].

The duration of mechanical ventilation pre cannulation has also been a topic of debate. Mechanical ventilation for longer than 7–10 days prior has traditionally been considered a contraindication to initiation of ECMO support as recommended in the 2017 and 2020 ELSO guidelines [35]. However, emerging evidence suggests that the duration of mechanical ventilation pre ECMO has no significant impact on mortality. It is actually the time interval from symptom onset to ECMO cannulation, and the driving pressure that are associated with a higher in-hospital mortality in this group [42, 43]. The comparative effectiveness of a PaO2/FiO2 ratio-guided vs. driving pressure guided ECMO initiation trigger has also been studied, with higher driving pressures at cannulation associated with poorer survival [43, 61]. The findings from a large registry study of approximately 7000 patients suggest that ECMO is possibly most effective if consistently provided to patients with more severe hypoxaemia i.e. PaO2/FiO2 < 80 mmHg, or driving pressure > 15 cm H₂O [61]. The relationship of mechanical ventilation therapy to ECMO survival may be influenced by the timing of intubation and IPPV (which tended to be early during the first wave compared with subsequent waves), and also as various non-invasive ventilation modalities e.g. high flow nasal cannula oxygen, CPAP, NIV etc., began to dominate the initial management phase of COVID ARDS.

COVID-19 patients require a prolonged duration of ECMO support compared with non-COVID ARDS to achieve successful weaning and ICU survival [30]. Approximately 25% of patients required at least 5 weeks or more of ECMO support [3, 47]. Interestingly, a longer duration of ECMO is not associated with increased mortality [27, 42]. This may be partly due to survival bias i.e., patients must survive a certain duration of time while supported with ECMO to fulfil the criteria for weaning, compared with other patients who may have had ECMO stopped earlier for futility or died [27, 42, 99]. Bridging to lung transplantation may also have skewed this data, as this would have removed some of the more critically unwell cohort who would probably not have survived without transplantation. However, studies reporting the use of lung transplantation in COVID-19 have so far been limited primarily to case reports and case series [100, 101, 102].

Organisational factors are also important to consider when examining mortality in COVID-19 patients supported with ECMO, especially given the heterogeneity across the major studies in relation to geographical, resource allocation, and temporal differences therein. A massive surge in capacity combined with a rapid upskilling of non-intensive care staff was required to deliver prompt and effective care to COVID-19 patients both in the main critical care ward and in non-critical care environments. Critical care surge capacity increased up to 155% in total, and 40% of COVID patients overall were managed in surge capacity beds. The patient to nurse ratio increased by about 25%, with most units requiring non-ICU clinicians and non-ICU nurses to aid with the increased workload (58% and 85% respectively). However, ECMO was generally employed only in the standard critical care bed setting [94]. Criteria for patient selection also varied over the course of the pandemic as knowledge of the disease evolved and as availability of resources changed [35]. Even now, ethical patient selection and timing of optimal ECMO initiation remains challenging.

High-volume ECMO centre experience, as measured by the number of ECMO runs greater than 30 per year, has a significant benefit on ICU mortality in COVID-ARDS [3, 44]. This was clearly demonstrated in the emulated target trial study by Hajage and colleagues where in high-volume experience centres, survival was 78% on ECMO vs. 64% for conventional management [4]. In other studies, survival in low vs. higher volume centres has been reported as 20% vs. 38% respectively [44]. The volume-survival relationship extends not only to those patients that receive ECMO support, but also those who are retrieved or transferred to the specialist high volume ECMO centre and do not receive ECMO. This has been demonstrated in non-COVID ARDS also [29]. Paradoxically, some healthcare systems demonstrated a higher in-hospital mortality across all phases of the pandemic despite being well resourced e.g. in Germany, there was an in-hospital mortality of about 70% over the entire pandemic, however this may be due to differences in patient selection criteria. The clinical and organisational factors associated with mortality in COVID-ARDS patients supported with ECMO are summarised in Table 3.

5. Conclusion

The COVID-19 pandemic has presented the most unparalleled global healthcare challenge since the influenza pandemic of 1918. Our knowledge of the disease and potentially efficacious therapy is constantly evolving, and yet patients with COVID ARDS remain at significantly increased risk of poor outcomes, including mortality. Although ECMO is a possible lifesaving option in those who are refractory to optimal conventional therapy, it is still unclear whether the use of VV ECMO definitively confers improved survival in patients with COVID-19 ARDS based on the mortality data thus far, although the more recent data is encouraging. We do know however, that improved survival depends on numerous factors, including resource allocation, patient selection criteria, timing of ECMO initiation, and centre volume experience. Judicious and ethical patient selection is therefore paramount to ensure the greatest benefit and least harm in the face of a constrained healthcare environment with finite resources.

Conflict of interest

No conflicts of interest to declare.