Veno-venous extracorporeal membrane oxygenation in coronavirus disease 2019: a case series.

BACKGROUND: The use of veno-venous extracorporeal membrane oxygenation (VV-ECMO) in severe hypoxaemic respiratory failure from coronavirus disease 2019 (COVID-19) has been described, but reported utilisation and outcomes are variable, and detailed information on patient characteristics is lacking. We aim to report clinical characteristics, management and outcomes of COVID-19 patients requiring VV-ECMO, admitted over 2 months to a high-volume centre in the UK.
METHODS: Patient information, including baseline characteristics and clinical parameters, was collected retrospectively from electronic health records for COVID-19 VV-ECMO admissions between 3 March and 2 May 2020. Clinical management is described. Data are reported for survivors and nonsurvivors.
RESULTS: We describe 43 consecutive patients with COVID-19 who received VV-ECMO. Median age was 46 years (interquartile range 35.5-52.5) and 76.7% were male. Median time from symptom onset to VV-ECMO was 14 days (interquartile range 11-17.5). All patients underwent computed tomography imaging, revealing extensive pulmonary consolidation in 95.3%, and pulmonary embolus in 27.9%. Overall, 79.1% received immunomodulation with methylprednisolone for persistent maladaptive hyperinflammatory state. Vasopressors were used in 86%, and 44.2% received renal replacement therapy. Median duration on VV-ECMO was 13 days (interquartile range 8-20). 14 patients died (32.6%) and 29 survived (67.4%) to hospital discharge. Nonsurvivors had significantly higher d-dimer (38.2 versus 9.5 mg·L-1, fibrinogen equivalent units; p=0.035) and creatinine (169 versus 73 μmol·L-1; p=0.022) at commencement of VV-ECMO.
CONCLUSIONS: Our data support the use of VV-ECMO in selected COVID-19 patients. The cohort was characterised by high degree of alveolar consolidation, systemic inflammation and intravascular thrombosis.

Introduction

A significant cohort of patients with coronavirus disease 2019 (COVID-19) go on to develop severe respiratory failure, requiring critical care admission. Reports have described the use of veno-venous extracorporeal membrane oxygenation (VV-ECMO) in a subset of critically ill patients, with utilisation ranging from 11% to 32% [1-3]. VV-ECMO is indicated for patients with potentially reversible, refractory, life-threatening hypoxaemia or hypercapnia or in patients where acceptable oxygenation or decarboxylation can be obtained only with injurious ventilatory settings. While VV-ECMO was associated with improved outcome during the H1N1 influenza pandemic [4, 5], COVID-19 demonstrates features unique from other respiratory infections and early case series have reported high mortality in patients on ECMO [6-8].

Given the lack of detailed information about patient characteristics and their clinical course, balanced with the need for judicious use of resources in the context of a pandemic, it is important to understand the role of VV-ECMO in COVID-19. We aim to describe, in detail, the clinical characteristics, management and outcomes of COVID-19 VV-ECMO patients from a high-volume UK ECMO centre, over a 2-month period of the pandemic.

Methods

Case selection

All COVID-19 patients admitted for VV-ECMO to Guy's and St Thomas' NHS Foundation Trust (GSTFT) in London, over a 2-month period (3 March 2020 to 2 May 2020) covering the peak of the pandemic, are included. Suitability for VV-ECMO was assessed in line with UK national commissioning criteria [9], requiring a lung injury (Murray) score ≥3 [10], or uncompensated hypercapnic acidosis with pH<7.2. National criteria were adapted for the COVID-19 pandemic on 10 April 2020 [11] to include clinical frailty scale ≤3 [12], the use of the Respiratory ECMO Survival Prediction (RESP) score [13] to aid pre-ECMO decision-making (with RESP score ≤3 requiring agreement between at least two centres), and an exclusion of “refractory multi-organ failure”. At the time of this series, detection of SARS-CoV-2 RNA on nose and throat swabs or bronchoalveolar lavage (BAL) using multiplexed-tandem PCR technology for detection of two gene targets, ORF 1a and ORF 8 (AusDiagnostics, Mascot, Australia), remained the gold standard. All patients, at point of referral, had either laboratory-confirmed or clinically suspected COVID-19 pneumonia; four patients without a positive result at time of referral subsequently tested positive from admission samples at GSTFT.

Patient clinical pathway

GSTFT is a national VV-ECMO centre commissioned to provide regional ECMO retrieval and provision [9]. At the start of the pandemic, GSTFT ECMO capacity was doubled through adaptation of each bedspace to accommodate two patients on ECMO. Patients were retrieved from referring hospitals via a previously described standard pathway [14], with no deviation in practice, aside from use of recommended personal protective equipment. Standard GSTFT practice is bifemoral percutaneous cannulation at the referring hospital, and use of Maquet Cardiohelp (Maquet, Rastatt, Germany) consoles. Following retrieval, all patients underwent computed tomography (CT) imaging of head, thorax (including CT pulmonary angiogram), abdomen and pelvis. Lung recruitment CT imaging at ventilator pressures of 5 cmH2O and 45 cmH2O were performed, unless pneumothorax was detected on initial scan or pulmonary air leak was suspected, to assess lung recruitment potential and delineate underlying lung parenchyma [15]. Diagnostic bronchoscopy and BAL for bacterial culture, viral and SARS-CoV-2 PCR was performed on all patients within the first 24 h. Patients without haemorrhagic complications were anticoagulated with unfractionated heparin infusion, targeting anti-Xa levels (0.3–0.7 UI·mL−1).

Patients were ventilated with protective lung parameters. Mechanical ventilation was generally initiated using standardised settings: positive end-expiratory pressure (PEEP) 10–15 cmH2O, tidal volume 2–4 mL·kg−1 of predicted body weight provided that driving pressure (plateau minus PEEP total) could be kept at 10 cmH2O, and plateau pressure <25 cmH2O. Initial respiratory rate was generally maintained at 10 breaths·min−1 to limit overall mechanical power [15]. In patients with high potential for lung recruitment, as demonstrated on low/high pressure CT imaging, higher PEEP (near 15 cmH2O) or time-controlled adaptive ventilation was used, with mean airway pressures of 23–26 cmH2O depending on body mass index and small airway closures (following assessment of a low-flow pressure-volume loop). Inspiratory oxygen fraction (FIO) on the ventilator was kept at 30–40%, and ECMO support titrated to achieve arterial oxygen tension (PaO)>60 mmHg and pH 7.35–7.40.

Patients received a course of broad-spectrum antibiotics on arrival, targeted to known microbiology where possible. A subset of patients with failure to progress and signs of a sustained hyperinflammatory state (fevers, persistently elevated C-reactive protein and/or ferritin, sustained organ dysfunction and hypoxaemia), in the absence of untreated active infection (bacteria or fungal species detected on blood culture and BAL, low procalcitonin and galactomannan), were treated with low-dose methylprednisolone regimens of 1–2 mg·kg−1·day−1 for 5–7 days, with halving in dose every 5–7 days, similar to published protocol [16]. This dosing regimen was chosen for its relative safety profile [17, 18]. Patients with persistent hyperinflammatory disease behaviour despite corticosteroids, or those with an “H score” greater than 169 suggesting secondary haemophagocytic lymphohistiocytosis [19, 20] were considered for treatment with the interleukin (IL)-1 receptor antagonist anakinra [21, 22]. Patients with persistent hypoxaemia and radiological abnormality despite low-dose corticosteroids, or patients who demonstrated early fibrosis on CT, were treated with high dose “pulsed” methylprednisolone at doses of 1 g for 3 days, followed by 1 mg·kg−1 per day, followed by a weaning regimen. Treatments were given in consultation with local lung inflammation specialists.

Patients generally remained paralysed for an initial 24 h, particularly if strong inspiratory efforts persisted despite adequate sedation, or if asynchronies due to deep sedation were noted (e.g. reverse triggering). Daily sedation wean was then undertaken in stable patients to maximise wakefulness. A specialist physiotherapy team assessed patients on a daily basis for both secretion clearance, and early rehabilitation. Ventilation weaning was based on daily assessment of lung mechanics, as well as ability to spontaneously ventilate without injurious tidal volumes, respiratory rate and inspiratory effort (including measurements of P0.1–100 ms airway occlusion pressure), that might contribute to patient self-inflicted lung injury. Criteria for decannulation from VV-ECMO in this cohort included maintained fractional inspired oxygen <0.5 and noninjurious ventilatory effort, with ECMO sweep gas turned off for at least 24 h. The full protocol of weaning from VV-ECMO is described and available [23].

Data collection and analysis

Data were collected retrospectively from electronic records, including the IntelliVue Clinical Information Portfolio (Philips, Eindhoven, the Netherlands). Pre-ECMO data were obtained from ECMO referral systems [24], paper records or direct interview with members of retrieval teams. RESP score was calculated at the time of referral. Sequential Organ Failure Assessment score was calculated from pre-ECMO clinical parameters and day 0 ECMO laboratory results, with Glasgow Coma Scale presumed to be 15 unless deranged consciousness pre-intubation was confirmed. Data were collected until all patients had reached either death (nonsurvivors) or discharge from hospital (survivors). Quantitative variables are presented with median and interquartile range (IQR), and categorical variables are presented as frequencies and percentages. Missing data were not imputed and are recognised in tables by adjusted n-values. Comparison of nonparametric continuous variables used Mann–Whitney U-tests with significance at p<0.05 (Python: package SciPy v.1.4.1). The study had institutional approval and waiver of individual informed consent (reference no.10 796), qualifying as service evaluation defined by the UK National Health Service Health Research Authority (www.hra.nhs.uk).

Results

Demographic and pre-ECMO characteristics

Forty-three patients with COVID-19 were accepted and admitted for VV-ECMO based on the listed criteria, out of 215 patients referred to GSTFT during the study period. Patient characteristics are shown in table 1. Median age was 46 years (IQR 35.5–52.5), ranging from 26 to 66. Most patients were male (33, 76.7%), 28 (65.1%) patients came from a Black, Asian and Minority Ethnic background and (21, 48.8%) were obese. Refractory and life-threatening hypoxia was an indication for VV-ECMO in all: median partial pressure of oxygen to fraction of inspired oxygen ratio (PaO/FIO) was 67.5 mmHg (IQR 58.9–77.8) at referral. 11 (25.6%) patients additionally had uncompensated respiratory acidosis with pH <7.20. Median static compliance was 30 mL·cmH2O−1 (IQR 21.5–33.6) in 24 patients with sufficient data for the calculation. Airway pressure release ventilation was used widely pre-ECMO (24 patients, 55.8%). Forty (93%) patients had undergone prone positioning prior to ECMO, and all patients had received neuromuscular blockade. Twenty-one (48.8%) patients received vasopressors at referral, and 3 (7%) required acute renal replacement therapy. Excluding one patient who acquired COVID-19 nosocomially, the median time from hospital admission to VV-ECMO was 7 (IQR 5–9) days. Median time from reported start of COVID-19 symptoms to VV-ECMO was 14 (IQR 11–18) days, and median from invasive ventilation to VV-ECMO was 5 (IQR 2–6) days.

TABLE 1

Baseline and pre-extracorporeal membrane oxygenation (ECMO) characteristics

	Survivors (n=29)	Nonsurvivors (n=14)	Total (n=43)
Age years	45 (35–49)	52.5 (43.8–53)	46 (35.5–52.5)
Sex
Female	9 (31.0)	1 (7.1)	10 (23.3)
Male	20 (69.0)	13 (92.9)	33 (76.7)
BMI kg·m-2	29 (28–34)	31 (26–34)	29 (27–34)
Ethnicity (ISARIC)
White	9 (31.0)	6 (42.9)	15 (34.9)
Black	7 (24.1)	3 (21.4)	10 (23.3)
South Asian	11 (37.9)	2 (14.3)	13 (30.2)
East Asian	2 (6.9)	2 (14.3)	4 (9.3)
Mixed	0 (0)	1 (7.1)	1 (2.3)
Comorbidity
Obesity#	14 (48.3)	7 (50.0)	21 (48.8)
Hypertension	5 (17.2)	5 (35.7)	10 (23.3)
Diabetes	2** (6.9)	6** (42.9)	8 (18.6)
Asthma	2 (6.9)	3 (21.4)	5 (11.6)
Cardiac arrest	0 (0)	1 (7.1)	1 (2.3)
RESP score	4 (3–5)	3 (3–5)	4 (3–5)
PaO2/FIO2¶	67.5 (61.6–77.6)	61.5 (53.7–79.5) (n=13)	67.5 (58.9–77.8) (n=42)
pH	7.34 (7.23–7.39) (n=28)	7.25 (7.17–7.30) (n=13)	7.30 (7.19–7.36) (n=41)
PaCO2 mmHg	66.5 (52.5.0–70.5)	69.8 (60.3–78.4)	67.5 (53.1–75.8)
Ventilation days before ECMO	5 (2–6)	4 (1–6)	5 (2–6)
SOFA score	6 (4–8)	9 (6–12)	7 (4–10)

Data are presented as n (%) or median (interquartile range). BMI: body mass index; ISARIC: International Severe Acute Respiratory and Emerging Infection Consortium; RESP: Respiratory ECMO Survival Prediction; PaO: arterial oxygen tension; FIO: inspiratory oxygen fraction; SOFA: Sequential Organ Failure Assessment. #: Obesity defined as BMI ≥30 kg·m-2; ¶: PaO/FIO in mmHg at point of referral. **: p <0.01.

Clinical diagnostics and features

The majority of patients (41, 95.3%) had extensive pulmonary consolidation on CT in either all lobes (figure 1a), or in a dependent distribution (figure 1b) with minimal sparing. Nineteen patients underwent lung recruitment imaging, 17 (89.5%) showed at least moderate recruitability as assessed visually by a radiologist. Twelve (27.9%) patients had pulmonary embolism (PE), five involving at least one main pulmonary artery. Five (11.6%) additional patients demonstrated areas of pulmonary infarction without PE. Left ventricular impairment on admission echocardiography was rare, in one patient (2.3%) with mild-to-moderate impairment only. In six (14%) patients, right ventricular dysfunction was found concurrently with PE. Six (14%) patients had small foci of subarachnoid haemorrhage on CT head, two of which demonstrated a local mass effect. Three (7%) patients had ischaemic infarction. Patients on day 0 of VV-ECMO had severe lymphopenia (median 0.6×109·L−1 (IQR 0.5–1.1)), elevated neutrophil to lymphocyte ratio (median 12.8 (IQR 9.2–22.5)), and high C-reactive protein (median 326 mg·L−1 (IQR 245–368)), ferritin (median 1907 μg·L−1 (IQR 1153–4083)), and d-dimer (median 11.7 mg·L−1 fibrinogen equivalent units (IQR 6.4–41.7). Acute kidney injury (creatinine ≥105 μmol·L−1) was a common feature (21 patients, 48.8%). Data are shown in table 2.

FIGURE 1

a) Almost completely consolidated lungs, with minor sparing at the right middle lobe periphery where there is scattered ground-glass opacification. Bilateral small pneumothoraces and moderate pneumomediastinum. b) Dense consolidation involving most of the lower lobes, further areas of consolidation seen in all other lobes in a patchy distribution, with ground-glass opacification.

TABLE 2

Extracorporeal membrane oxygenation (ECMO) admission investigation findings

	Survivors (n=29)	Nonsurvivors (n=14)	Total (n=43)
CT thorax
Alveolar consolidation	29 (100)	14 (100)	43 (100)
Ground-glass opacities	29 (100)	14 (100)	43 (100)
Pneumothorax	3 (10.3)	2 (14.3)	5 (11.6)
Pneumomediastinum	4 (13.8)	4 (28.6)	8 (18.6)
Pulmonary embolism	7 (24.1)	5 (35.7)	12 (27.9)
Echocardiogram
Left ventricular impairment	1 (3.4)	0 (0)	1 (2.3)
Right ventricular impairment	3* (10.3)	5* (35.7)	8 (18.6)
CT head
Subarachnoid haemorrhage	3 (10.3)	3 (21.4)	6 (14.0)
With sulcal effacement	0* (0)	2* (14.3)	2 (4.7)
Subdural haemorrhage	1 (3.4)	0 (0)	1 (2.3)
Ischaemic infarction	0** (0)	3** (21.4)	3 (7.0)
Laboratory values#
Haemoglobin	99 (92–102)	95 (86–107)	97 (89–106)
Lymphocyte count	0.8 (0.5–1.3)	0.6 (0.4–0.8)	0.6 (0.5–1.1)
Neutrophil:lymphocyte ratio	11.2 (9.2–19.3)	17.8 (12.1–24.2)	12.8 (9.2–22.5)
Platelet count	247 (203–344)	214 (121–329)	245 (163–339)
Creatinine	73* (53–192)	169* (98–347)	96 (66–237)
Bilirubin	16 (9–22)	15 (12–20)	15 (11–22)
Ferritin	1903 (904–3800)	2251 (1308–5435)	1907 (1153–4083)
C-reactive protein	291 (217–388)	344 (328–361)	326 (245–368)
Procalcitonin	1.24* (0.76–6.43) (n=28)	6.3* (2.99–16.27)	3.96 (1.15–9.01) (n=42)
D-dimer	9.5* (5.2–31.7)	38.2* (10.9–60.5)	11.7 (6.4–41.7)
Troponin	26 (16–40) (n=24)	98 (25–132) (n=7)	28 (15–70) (n=31)

Data are presented as n (%) or median (interquartile range). CT: computed tomography #: Units for laboratory values: haemoglobin in g·dL−1, lymphocyte count in 109·L−1, platelet count per µL, creatinine/bilirubin µmol·L−1, ferritin in µg·L−1, C-reactive protein in mg·L−1, procalcitonin in ng·mL−1, d-dimer in mg·L−1 fibrinogen equivalent units and troponin in ng·L−1. *: p=0.01–0.05; **: p<0.01.

ECMO course and outcomes

Twenty-nine patients (67.4%) were successfully decannulated from VV-ECMO and survived until hospital discharge. Twelve (27.9%) patients died on ECMO, and two (4.7%) died following decannulation. Causes of death and complications are summarised in table 3. Median days on ECMO was 14 (IQR 8–21) for survivors. Nonsurvivors had significantly greater procalcitonin (6.30 versus 1.24 ng·mL−1; p=0.028), d-dimer (38.2 versus 9.5 mg·L−1 fibrinogen equivalent units; p=0.035) and creatinine (169 versus 73 μmol·L−1; p=0.022) at admission. Common complications were haemorrhage requiring ≥1 packed red blood cell transfusion (8, 18.6%) including two bleeds after tracheostomy and two spontaneous retroperitoneal haemorrhages (one leading to death), and tension pneumothorax (5, 11.6%) that contributed to mortality in three patients. One patient developed myocarditis, with cardiac tamponade resulting in death. Nineteen (44.2%) patients required renal replacement therapy and the majority (86%) required noradrenaline at moderate doses during ECMO (peak dose median 0.19 µg·kg−1·min−1 (IQR 0.11–0.35)). Seventeen (39.5%) patients required at least one circuit change during their ECMO for membrane thrombosis; mean ECMO days per each circuit change (numerator as sum of all days on ECMO of entire cohort) was 29.6. In 31 patients decannulated from ECMO, 17 (54.8%) demonstrated cannula-related peripheral venous thrombosis on Doppler imaging. The majority of patients (79.1%) received immunomodulation, typically methylprednisolone at 1–2 mg·kg−1·day−1 although 9 (20.9%) received higher “pulsed” doses and 10 (23.3%) received anakinra (table 4).

TABLE 3

Time course, causes of death and complications on extracorporeal membrane oxygenation (ECMO)

	Survivors (n=29)	Nonsurvivors (n=14)	Total (n=43)
Days on ECMO	14 (8–21)	12.5 (5.75–18)	13 (8–20)
Causes of death
On ECMO		6 (42.9)
Full active treatment#		2 (33.3)
Limits of care		4 (66.7)
Withdrawal of ECMO		6 (42.9)
Progression of ICH with brainstem herniation		1 (16.7)
Extensive cerebral infarction and multi-organ failure		2 (33.3)
Poor respiratory prognosis and multi-organ failure		3 (50.0)
Post-ECMO		2 (14.3)
Progression of extensive cerebral infarction and multi-organ failure		1 (50.0)
Refractory respiratory failure with severe fibrotic lung disease		1 (50.0)
Complications on ECMO
Bleeding leading to transfusion	4 (13.8)	1 (7.1)	5 (11.6)
Bleeding leading to transfusion and intervention	3 (10.3)	0 (0)	3 (7.0)
Bacteraemia	3 (10.3)	1 (7.1)	4 (9.3)
Tension pneumothorax	2 (6.9)	3 (21.4)	5 (11.6)
Myocarditis and cardiac tamponade	0 (0)	1 (7.1)	1 (2.3)

Data are presented as n (%) or median (interquartile range). ICH: intracerebral haemorrhage. #: Causes of deaths on ECMO: 1) spontaneous retroperitoneal haemorrhage 2) myocarditis with cardiac tamponade.

TABLE 4

Organ support, therapeutics, other interventions

	Survivors (n=29)	Nonsurvivors (n=14)	Total (n=43)
Organ support
Noradrenaline	23 (79.3)	14 (100)	37 (86.0)
Peak dose	0.15 (0.09–0.32)	0.23 (0.18–0.41)	0.19 (0.11–0.35)
Milrinone	4 (13.8)	5 (35.7)	9 (20.9)
Renal replacement therapy	11 (37.9)	8 (57.1)	19 (44.2)
Persisting requirement at step-down	6 (20.7)
Immunomodulation	25 (86.2)	9 (64.3)	34 (79.1)
Methylprednisolone 1–2 mg·kg−1	24 (82.8)	8 (57.1)	32 (74.4)
Anakinra	7 (24.1)	3 (21.4)	10 (23.3)
Methylprednisolone 1 g “pulsed”	5 (17.2)	4 (28.6)	9 (20.9)
Hydrocortisone	1 (3.4)	0 (0)	0 (0)
Pharmacological intervention
Remdesivir	4 (13.8)	0 (0)	0 (0)
Hydroxychloroquine	2 (6.9)	0 (0)	0 (0)
ECMO circuit change	11 (37.9)	6 (42.9)	17 (39.5)
ECMO days per circuit change (mean)	33	29	30
Tracheostomy	20* (69.0)	5* (35.7)	25 (58.1)
Post-decannulation lower limb DVT	16 (55.2)	1 (50.0) (n=2)	17 (54.8) (n=31)
Occlusive thrombus	1 (6.3)	0 (0)	1 (3.2)
Nonocclusive thrombus	12 (75.0)	1 (100)	13 (41.9)
Mural thrombus only	3 (18.8)	0 (0)	3 (9.7)

Data are presented as n (%) or median (interquartile range), unless otherwise stated. DVT: deep vein thrombosis. *: p=0.01–0.05.

Microbiology

During their period of ECMO support, 15 (34.9%) patients developed bacterial respiratory infection with Klebsiella spp. (K. aerogenes (6), K. oxytoca (1), and K. pneumoniae (8)). In 2 patients this grew from BAL on day 0 of ECMO, whereas later infection was detected in 13 patients. The most frequent fungal isolate was Candida spp. (15 patients, 34.9%), all light or scanty growth from BAL. There were four significant bloodstream infections, two with K. oxytoca, one with Proteus mirabilis and a persistent vancomycin-resistant Enterococcus secondary to a deep-seated focus (retroperitoneal haematoma). Aspergillus fumigatus grew from BAL in two patients, and two others developed cytomegalovirus viraemia with IgG positivity.

Discussion

The mortality described in this VV-ECMO series (14 of 43, 32.6%) is lower than in early descriptions. Patients exhibited particular characteristics including poor lung compliance, persistent hyperinflammation and high incidence of thrombosis. Survival in this series is comparable to more recent data from the USA [25] and France [26]. Since the study period, a further 13 patients have completed VV-ECMO at GSTFT, with overall survival to intensive care unit discharge at 71.4%.

The pattern of disease seen in our cohort has been previously described. Exudative lung disease with poor compliance (as described by Gattinoni et al. [27]), persistent hyperinflammation [28, 29], and increased thrombosis incidence may demonstrate a particular phenotype that defines a later stage of the disease process. Median ferritin and d-dimer seen at ECMO initiation were comparable to values seen after 2 weeks in a cohort of nonsurviving hospitalised patients [30]. A majority of our patients were given immunomodulation [19] after risks of immunosuppressing critically ill patients [31] were weighed against lack of clinical progress and ongoing inflammatory lung insult. Recent trial data showing benefits of dexamethasone in ventilated COVID-19 patients may support the wider use of steroids, although their role in patients on ECMO is unclear [32].

The incidence of PE (27.9%) was substantially higher than in pre-COVID-19 (9.6%) in the same centre [33], carrying substantial morbidity in our cohort with RV dysfunction in 50%. Cannula-related thrombosis rates (54.8%) were greater than baseline prevalence [34], and ECMO membrane complication rate was similarly high. Thrombosis risk is a known entity in severe COVID-19 [35], but adjusted anticoagulation targets must be balanced against higher risk of haemorrhagic complication in ECMO [36], the cause of multiple complications and one death in our cohort.

At time of writing, no published literature specifically addresses secondary or coinfection in COVID-19 ECMO. These individuals may represent a distinct cohort microbiologically. The unusual predominance of Klebsiella spp. has been seen elsewhere, as has Candida spp. [37, 38], but remains a focus of further analysis in GSTFT regarding infection control consequences of doubling bedspace usage. Admission procalcitonin was elevated in all patients, but significantly greater procalcitonin in the nonsurvivor group may have limited earlier use of steroids. Re-activation or flares of chronic viral infections including cytomegalovirus must also be considered, especially in those receiving immunomodulation.

Following new commissioning criteria in the UK, the threshold for acceptance of patients onto VV-ECMO has been reinforced by the inclusion of the RESP score. This predictive score is validated in patients already on ECMO [13, 39], but not as a pre-ECMO decision tool. The RESP score was one component of a multi-tool assessment process when deciding which patients should be offered VV-ECMO, and cases with low RESP scores were discussed with a second centre if ECMO was felt to be indicated. Validation of this tool in the UK ECMO population may help to guide future usage.

This series has the inherent limitations of a single-centre study, conducted in a well-resourced and experienced centre, during the early stages of our understanding in a new disease. It is likely that aspects of management will differ over time and between centres, as our understanding of how to treat particular phenotypes improves in any future pandemic waves.

This case series suggests that VV-ECMO, when offered to patients with COVID-19 respiratory failure refractory to conventional ventilatory management, can be associated with a favourable outcome. In COVID-19 patients with severe respiratory failure, early consultation with an ECMO centre and joint decision-making on suitable support modality is a key strategy.