Mortality of COVID-19 Patients Requiring Extracorporeal Membrane Oxygenation During the Three Epidemic Waves

Abstract

Clinical presentation and mortality of patients treated with extracorporeal membrane oxygenation (ECMO) for COVID-19 acute respiratory distress syndrome (CARDS) were different during the French epidemic waves. The management of COVID-19 patients evolved through waves as much as knowledge on that new viral disease progressed. We aimed to compare the mortality rate through the first three waves of CARDS patients on ECMO and identify associated risk factors. Fifty-four consecutive ECMO for CARDS hospitalized at Amiens University Hospital during the three waves were included. Patients were divided into three groups according to their hospitalization date. Clinical characteristics and outcomes were compared between groups. Pre-ECMO risk factors predicting 90 day mortality were evaluated using multivariate Cox regression. Among 54 ECMO (median age of 61[48–65] years), 26% were hospitalized during the first wave (n = 14/54), 26% (n = 14/54) during the second wave, and 48% (n = 26/54) during the third wave. Time from first symptoms to ECMO was higher during the second wave than the first wave. (17 [12–23] days vs. 11 [9–15]; p < 0.05). Ninety day mortality was higher during the second wave (85% vs. 43%; p < 0.05) but less during the third wave (38% vs. 85%; P < 0.05). Respiratory ECMO survival prediction score and time from symptoms onset to ECMO (HR 1.12; 95% confidence interval [CI]: 1.05–1.20; p < 0.001) were independent factors of mortality. After adjustment, time from symptoms onset to ECMO was an independent factor of 90 day mortality. Changes in CARDS management from first to second wave-induced a later ECMO cannulation from symptoms onset with higher mortality during that wave. The duration of COVID-19 disease progression could be selection criteria for initiating ECMO.

Like many countries in the World, France has faced several waves of the COVID-19 pandemic since February 2020.¹ In our center, during the first wave, several patients with COVID-19–related respiratory distress syndrome (CARDS) required extracorporeal membrane oxygenation (ECMO) with survival in more than 50% of cases.² After decreasing detected cases in June 2020, France faced a second wave after summer vacations. As the pandemic progressed, clinical and therapeutic guidelines for managing patients with CARDS have evolved.^3,4 But, despite the beneficial effect of the early administration of corticosteroids (ref Recovery trial), the nonsystematic use of mechanical ventilation, curative-systemic anticoagulation, and the experience gained during the first wave, the mortality of patients remained unimproved for patients admitted in an intensive care unit (ICU)⁵ and even an increase in mortality in patients under mechanical ventilation.⁶

Concerning ECMO, the analysis of cases reported, during the second wave, by the EuroElso group suggested an increase in mortality, weaning failure, and time spent on ECMO.⁷ Substantial changes in the management of COVID-19 patients, notably the early administration of corticosteroids, might have affected the profile of patients with CARDS progressing to ECMO support during the second wave. Despite the arrival of vaccines on the French market and the massive vaccination of the most vulnerable as of January 2021, a new epidemic wave was beginning in March 2021. Our center has faced a critical number of requests for ECMO as in the first wave.⁸ During this third wave, to improve patient survival and avoid futile cases of CARDS requiring ECMO according to EOLIA criteria,⁹ we arbitrarily included in the ECMO criteria the duration of disease progression (time between first COVID-19 symptom and ECMO) based on our clinical experience of the first two waves. To date, there are little data in the literature on the clinical, biologic, and outcome differences of CARDS requiring ECMO according to the three epidemic waves.

The study aimed to compare clinical baseline characteristics, the effect of pre-ECMO symptoms duration, and outcomes of patients with CARDS requiring ECMO during the three epidemic waves of COVID-19 infection in France (Supplemental Digital Content; https://links.lww.com/ASAIO/A840).

Methods

Population

Adult patients (>18 years of age) admitted to the cardiac thoracic vascular and respiratory ICU (CTVR-ICU) of Amiens University Hospital for severe CARDS requiring ECMO during the first epidemic wave (between February 28, 2020 and June 1, 2020), the second epidemic wave (between September 1, 2020 and April 15, 2020) and the third epidemic wave (between December 12, 2020 and June 15, 2021) of COVID-19 infection. Exclusion criteria were patients who died during ECMO implantation and with ECMO after a veno-arterial ECMO.

Ethics

This is an ancillary study of a prospective cohort study of patients with COVID-19 infection hospitalized in ICU at Amiens University Hospital (NCT04354558). This study was approved by the Amiens University Hospital IRB (Comite de Protection des Personnes Nord-Ouest II CHU–Place V. Pauchet, 80054 AMIENS Cedex 1, CNIL Number: PI2020_843_0026). Informed consent was waived by French law on clinical research for non-interventional studies. Still, oral and written information was provided whenever possible to the patients and systematically to their families, specifying that they could oppose using their data.¹⁰

Data

Patient characteristics (age, sex, body mass index, and comorbidities), medical reports, and biologic values were collected prospectively. The severity of illness on ICU admission was evaluated by the sepsis-related organ failure assessment score (SOFA score).¹¹ Acute respiratory distress syndrome grade was defined according to Berlin definition.¹² The respiratory ECMO survival prediction (RESP) score was used to predict survival at ECMO initiation.¹³ We collected the dates of the first symptom, ICU admission, intubation, and ECMO initiation.

Before ECMO implantation, we collected computed tomography scan features (ground-glass opacities, pulmonary embolism, crazy paving, condensation, and percentage of parenchymal lung involvement), ventilator settings (positive end-expiratory pressure [PEEP], a fraction of inspired oxygen, respiratory rate, tidal volume, plateau pressure, driving pressure), arterial blood gas parameters, laboratory values (white blood count, creatinine, hemoglobin, platelets, fibrinogen, C reactive protein, procalcitonin), rescues therapies (prone positioning, inhaled nitric oxide) and specific COVID-19 therapies.

The following ICU-related complications were also collected: ventilator-acquired pneumonia, pneumothorax, stroke, thrombotic events, renal replacement therapy, bacteremia, and cardiac arrest. ECMO-major complications included ECMO circuit change, switch to veno-arterial ECMO, and major bleeding. According to the ELSO definition, major bleeding was defined by bleeding that led to surgical exploration or characterized by its location (central nervous system, hemothorax, retroperitoneal bleeding) or requiring immediate transfusion of at least two units of PRBC for either a sudden fall of hemoglobin of 2 g/dl in less than 24 hours or new hemodynamic instability or overt bleeding. The vital status at 30 and 90 days after ECMO implantation were collected.

Outcomes

The primary outcome was in-hospital mortality in a time-to-event analysis assessed 90 days after ECMO initiation.

Extracorporeal Membrane Oxygenation Program in Our Center

Amiens University Hospital is a tertiary care hospital and the CTVR-ICU is the regional reference center for ARDS treatment. Our 20 bed unit has managed ECMO since 2009, performed about 60 ECMO per year, and set up ECMO in the whole Picardy subregion using a mobile ECMO team.⁸ According to the “ECMO to Rescue Lung Injury in Severe ARDS” (EOLIA) trial,⁹ patients eligible for ECMO had to fulfill ARDS criteria and one of the following disease severity criteria, despite ventilator optimization (fraction of inspired oxygen [FiO₂] ≥80%, tidal volume set at 6 ml/kg predicted body weight and PEEP ≥10 cm of water): 1) partial pressure of arterial oxygen (PaO₂) over a FiO₂ ratio of less than 50 mm Hg for more than 3 hours; 2) PaO2_/FiO₂ less than 80 mm Hg for more than 6 hours; or 3) arterial blood pH less than 7.25 with a partial pressure of arterial carbon dioxide (PaCO₂) of 60 mm Hg or more for 6 hours or more; 4) time under mechanical ventilation under 7 days.¹⁴ Extracorporeal membrane oxygenation contraindications were age older than 75 years, mechanical ventilation for more than 10 days, severe comorbidities (e.g., advanced cardiac, respiratory, or liver failure; metastatic cancer; or hematological malignancies), cardiac arrest, irreversible neurologic injury, and refractory multiorgan failure.¹⁴ Implementation of ECMO was performed once the ECMO team approved the indication. The ECMO team was composed of a cardiac thoracic surgeon, a specialized intensivist, and a trained perfusion nurse. The ECMO procedure has been reported previously.^5,8 Ultrasound-guided percutaneous cannulation was performed with Seldinger’s technique. A large blood drainage cannula (21–29 French, Maquet HLS cannula, Getinge, Germany) in venous femoral and a return cannula (15–19 French, Maquet HLS cannula, Getinge, Germany) through the right internal jugular vein were inserted. Optimal cannula positioning was verified by transoesophageal echocardiography (TEE) and chest x-ray.

Extracorporeal Membrane Oxygenation Management

Our ECMO management was in line with the ELSO guidelines. After an initial bolus of 50–100 UI/kg, continuous intravenous unfractionated heparin (UFN) was administrated for an anti-Xa level target of 0.2–0.4 UI/ml. Since the COVID-19 pandemic, a higher anti-Xa level target was aimed to avoid fatal thrombotic events during cannulation.^14,15 All ECMO implanted by our mobile ECMO unit in remote regional hospitals were hospitalized in CTVR-ICU. After ECMO implantation, we targeted a body temperature over 36.5°C, peripheral oxygen saturation between 92% and 96%, PaCO₂ between 35 and 45 mm Hg, and blood pump flow to ensure at least 70% of the patient cardiac output. Ultra-protective ventilation of 4 ml/kg of predicted body weight was applied during ECMO therapy.

Statistical Analysis

Data are expressed as median (interquartile range) or numbers (percentage), as appropriate. The study population was divided into three groups according to the epidemic wave. Study groups were compared using the Mann-Whitney U test, Fisher exact test, Kruskal-Wallis test, and χ² tests. Bonferonni adjustments were used to adjust the p values between groups.

Multivariable Cox regression analysis was used to assess the relationship between the 90 day mortality and covariates considered to have a potential prognostic impact. Data are presented as hazard ratios and 95% confidence intervals (CIs). Variables associated with 90 day mortality in the univariate analysis with a p value <0.1 were included in the Cox model.

A sensibility analysis by a receiver-operating characteristic curve (ROC) was built to assess the time between first symptoms and ECMO support for predicting in-hospital mortality. The Youden index was used to determine the optimal threshold for predicting in-hospital mortality. The Kaplan-Meier method was used to plot the survival curves compared between the three groups with the log-rank test.

All p values are the results of two-tailed tests. Statistical analyses were performed using SPSS software version 24 (IBM Corp, Armonk, NY). A statistical test was significant when the p value was under 0.05.

Results

Among the 217 consecutive patients admitted to ICU in our center for CARDS between February 1, 2020, and June 15, 2021, 61 of 217 (28%) required ECMO for CARDS. The study population was divided into three groups according to the three French epidemic waves of COVID-19. A total of 54 patients were included in the study, with 26% (n = 14/54) in the first wave, 26% (n = 14/54) in the second, and 48% (n = 26/54) during the third wave (Figure 1, flow chart).

Figure 1.: Flow diagram of the study group. CARDS, COVID-19 acute respiratory distress syndrome; ECMO, extracorporeal membrane of oxygenation.

The percentage of CARDS patients requiring ECMO was 14 of 64 (22%) during the first wave, 14 of 36 patients (39%) during the second wave, and 26 of 117 (22%) during the third wave. During the first wave, the maximum number of patients under ECMO in our ICU was 10 for a maximum capacity of 12 (Supplemental Digital Content; https://links.lww.com/ASAIO/A840).

Figure 2.: Survival analysis assessed by Kaplan-Meir analysis according to the three epidemic waves.

Of the 54 ECMO, 75% (n = 43) were men, with a median age of 61 (48–65) years and with a median SOFA score of 11.^8–12 Pre-ECMO demographic, comorbidities, and biologic data are summarized in Table 1. No significant differences were found between the different epidemic waves except for the serum creatinine, which was higher in the patients of the first wave (113 [75–175] µmol/L vs. 62 [49–71] µmol/L, p = 0.03).

Table 1. - The Population’s Clinical and Demography Characteristics According to the COVID-19 Epidemic Wave

Variables	Overall (n = 54)	First wave (n = 14)	Second wave (n = 14)	Third wave (n = 26)	p value
Age (y)	61 [48–65]	63 [60–67]	61 [52–69]	57 [44–64]	0.07
BMI (kg·m−2)	30.2 [28–33]	29.6 [29.1–32.2]	30 [25–31.5]	31 [28–35]	0.27
SOFA score	11 [8–12]	11 [9–14]	12 [9–12]	9 [8–12]	0.28
Male gender (n; %)	40 (74)	12 (86)	8 (57)	20 (77)	0.21
Medical history(n; %)
Hypertension	3 (1)	6 (42)	8 (57)	17 (65)	0.40
Diabetes	1 (4)	5 (35)	2 (14)	7 (26)	0.42
Dyslipidemia	9 (17)	0	3 (21)	6 (23)	0.15
Smoking (former or active)	5 (9)	2 (14)	2 (14)	1 (4)	0.43
Chronic renal disease	4 (7)	2 (14)	0	2(7)	0.36
COPD/asthma	5 (9)	1 (7)	1 (7)	3 (11)	0.86
Coronary disease	4 (7)	1 (7)	0	3 (11)	0.42
Immunocompromised	3 (6)	2 (14)	0	1 (4)	0.23
Time from first symptom to ICU admission	7 [4–10]	7 [4–10]	9 [3–20]	7 [3–9]	0.51
Biologic data before ECMO
White blood cell count, ×106 cells per L	8,700 [5,900–13,000]	9,400 [6,700–12,080]	13,900 [9,600–23,700]	1,100 [8,850–19,000]	0.47
Lymphocytes, ×106 cells per L	600 [400–900]	676 [325–1,770]	1,000 [700–1,700]	600 [200–800]	0.06
Serum creatinine, (µmol/L)	70 [53–96]	113 [75–175]	62 [49–71]*	68 [55–95]†	0.03
C reactive protein (mg/L)	137 [84–203]	207 [150–295]	152 [40–272]	159 [88–272]	0.18
Procalcitonin (µg/L)	1.7 [1.2–2.2]	2.3 [1.7–2.6]	2.4 [1.9–3.1]	2.0 [1.6–3.3]	0.17
Platelet count ×109/L	234 [173–314]	200 [124–245]	272 [135–468]	213 [159–309]	0.42
Fibrinogen (g/L)	5.9 [4.7–7.4]	5.4 [4.2–8.1]	6.6 [4.2–7.8]	5.6 [4.8–7.3]	0.49
Hemoglobin (g/L)	11.0 [9.3–12.5]	11.2 [9.4–13.1]	10.6 [9.6–11.4]	10.0 [9.2–12.1]	0.82
CT scan before ICU admission (n = 49/54)		12 (85)	14 (100)	23 (96)	0.38
Ground-glass opacity	49 (100)	12 (100)	14 (100)	23 (100)	0.60
Consolidation	24 (44)	6 (50)	3 (21)	14 (61)	0.06
Crazy paving	25 (46)	3 (25)	8 (57)	14 (61)	0.09
Parenchymal lung involvement > 50%	24 (49)	11 (92)	13 (93)	20 (87)	0.45

Data are presented as median [interquartile range] and number (percentage).

^*p < 0.05 for the second wave vs. the first wave.

^†p < 0.05 for the third wave vs. the first wave.

BMI, body mass index; CT, computerized tomography; COPD, chronic obstructive pulmonary disease; ECMO, extracorporeal membrane oxygenation; ICU, intensive care unit; RESP, respiratory extracorporeal membrane oxygenation survival prediction; SOFA, sepsis organ failure assessment.

Pre-Extracorporeal Membrane Oxygenation Settings

Pre-ECMO ventilatory settings were different for patients of the second wave (Table 2). Median static compliance was 20.3 (164–27.4) cm H₂O vs. 30.4 (22.6–33.1) cm H₂O p = 0.02, and median driving pressure was 17 (13–21) vs. 14 (12–15). Glucocorticoids were mostly used after the first wave (8 [57%] vs. 13 [93%], p < 0.05). Femoral-jugular cannulas were inserted in 54 out of 57 95% patients with a median of 3 days (1–6) after endotracheal intubation.

Table 2. - Ventilation Baseline Parameters and Rescue Therapy Before ECMO Support

	Overall (n = 54)	First wave (n = 14)	Second wave (n = 14)	Third wave (n = 26)	p value
RESP score	2 [1–4]	1 [0–4]	1.5 [0–4]	4 [1–6]*	0.01
Modified RESP score	4 [3–6]	3.5 [2–4.5]	4 [4–5]	5.5 [4–6]*	0.02
Ventilator settings pre-ECMO
Tidal volume (ml/kg)	6.2 [5.5–6.6]	6.2 [5.9–6.4]	5.2 [3.4–7.5]	6.2 [5.4–6.6]	0.72
Positive end-expiratory pressure, (cm H2O)	12 [10–14]	14 [12–15]	10 [8–12]†	12 [8–14]	0.002
Plateau pressure (cm H2O)	28 [25–30]	28 [26–28]	27 [24–31]	28 [24–30]	0.63
Respiratory rate	28 [20–30]	30 [28–31]	26 [18–30]†	25 [20–28]*	0.001
Static compliance (ml/cm H2O)	27.1 [21.3–32.5]	30.4 [22.6–33.1]	20.3 [16.4–27.4]†	26.4 [22.6–33.3]	0.02
Driving pressure	15 [12–17]	14 [12–15]	17 [13–21]†	16 [12–17]	0.04
Last blood gas values pre-ECMO
pH	7.35 [7.28–7.46]	7.31 [7.23–7.40]	7.36 [7.28–7.46]	7.37 [7.30–7.46]	0.18
PaO2/FiO2 (mm Hg)	68 [57–80]	69 [61–77]	65 [54–91]	68 [55–90]	0.80
PaCO2, mm Hg	45 [35–55]	58 [45–64]	46 [39–57]	44 [34–48]*	0.02
Plasma bicarbonate, mmol/L	25 [23–29]	28 [22–31]	28 [24–30]	24 [22–27]	0.16
Pre-ECMO co-infection
Bacterial pneumoniae	6 (11)	1 (7)	3 (21)	2 (8)	0.13
COVID-19 therapies and Immunomodulators pre-ECMO
Glucocorticoids	44 (81)	8 (57)	13 (93)†	24 (89)*	0.02
Lopinavir-ritonavir	9 (17)	9 (64)	0	0	-
Hydroxychloroquine	4 (7)	4 (7)	0	0	-
Tocilizumab	2 (4)	0	0	2 (7)	-
Retrieval on ECMO by mobile ECMO team	28 (52)	11 (79)	7 (50)	10 (39)	0.05
Time from ICU admission to intubation	2 [1–4]	1 [1–3]	4 [2–12]†	1 [1–3]	0.02
Time from intubation to ECMO	3 [1–6]	3 [1–7]	4 [1–12]	3 [1–5]	0.95
Time from first symptom to ECMO	11 [9–16]	11 [9–15]	17 [12–23]†	11 [8–15]	0.04
Time from hospitalization to ECMO	5 [3–11]	5 [3–8]	8 [3–17]	5 [3–10]	0.42
Rescue therapy pre-ECMO
Neuromuscular blockade	54 (100)	14 (100)	14 (100)	26 (100)	-
Prone positioning	47 (87)	13 (93)	10 (71)	24 (92)	0.13
Number of session	2 [1–3]	2.5 [1–4]	1 [0–3]	2 [1–2]	0.06
Inhaled nitric oxide	44 (81)	11 (79)	12 (85)	21 (81)	0.88
Almitrine infusion	3 (6)	0	0	3 (11)	0.37
Vasoactive support before ECMO	21 (39)	3 (21)	9 (64)	9 (34)	0.06
Type of ECMO support
Femoral-jugular	51 (94)	14 (100)	12 (85)	25 (96)	0.23
Femoral-femoral	3 (6)	0	2 (14)	1 (4)	-
Percutaneous	54 (100)	14 (100)	14 (100)	26 (100)	-
Cannula outflow size (Fr)	25 [23–25]	25 [23–25]	25 [23–25]	25 [25–29]*	0.04
Cannula inflow size (Fr)	19 [17–19]	19 [17–19]	19 [17–19]	18 [17–19]	0.71

^†p < 0.05 for the second wave vs. the first wave.

^*p < 0.05 for the third wave vs. the first wave.

ECMO, extracorporeal membrane oxygenation; ICU, intensive care unit; RESP, respiratory extracorporeal membrane oxygenation survival prediction.

Time from ICU admission to intubation (4 [1–12] vs. 1 [1–3] days, p = 0.02) and time from first symptoms to ECMO support (17 [12–23] vs. 11 [9–15] days, p < 0.05) were higher during the second wave compared to the first wave.

Outcome

There were no differences between the three waves for the occurrence of thrombotic complication, major bleeding, heparin-induced thrombocytopenia, or the use of renal replacement therapy (Table 3). The median duration of ECMO support (24 [21–39] days vs. 10 [8–21] days, p = 0.001) and ICU stay (36 [24–47] days vs. 26 [14–30], p = 0.008) were higher during the second wave. During their stay in ICU, 100% of the second and the third waves patients developed ventilator-associated pneumonia. Duration of mechanical ventilation was higher for patients in the second wave (37 [24–50] days vs. 25 [6–31] days, p = 0.04).

Table 3. - Clinical Evolution and Outcomes According to the Epidemic Wave

	Overall (n = 54)	First wave (n = 14)	Second wave (n = 14)	Third wave (n = 26)	p value
Respiratory evolution
Pneumothorax	16 (30)	4 (29)	8 (57)*	4 (15)	0.02
Ventilator-associated pneumonia	50 (92)	9 (64)	13 (93)*	26 (100)†	0.002
Prone positioning under ECMO	31 (57)	7 (50)	6 (43)	18 (69)	0.23
Tracheostomy	11 (21)	3 (21)	4 (29)	4 (16)	0.66
Time under MV (d)	28 [19–39]	25 19–32	37 [25–48]	27 [17–37]	0.11
Thrombotic and hemorrhagic complication
Pulmonary embolism	14 (26)	3 (21)	6 (42)	5 (19)	0.25
With cardiac arrest	8 (57)	1 (7)	4 (29)	3 (11)	0.23
Venous thromboembolism	14 (26)	4 (29)	4 (29)	6 (23)	0.9
Major bleeding	29 (54)	8 (57)	9 (64)	12 (46)	0.53
Heparin-induced thrombocytopenia	8 (21)	3 (21)	2 (14)	3 (14)	0.71
Neurologic complication	7 (13)	1 (7)	3 (12)	3 (11)	-
Hemorrhagic stroke	2 (4)	0	0	2 (8)	0.62
Ischemic stroke	5 (9)	1 (7)	3 (12)	1 (4)	-0.60
Renal replacement therapy	30	9 (64)	8 (61)	13 (50)	0.64
Outcome
Switch to veno-arterial-venous ECMO	2 (4)	0	2 (14)	0	0.05
ECMO weaning	32 (60)	9 64	6 (43)	17 (65)	0.36
Time under ECMO	17 [10–26]	12 [9–21]	24 [21–34]*	15 [7–25]	0.009
90 d mortality	28 (52)	6 (43)	12 (85)*	10 (38)‡	0.007
ICU discharge (d)	27 [19–38]	26 [18–33]	33 [22–47]	26 [16–37]	0.25

^*p < 0.05 for the second wave vs. the first wave.

^†p < 0.05 for the third wave vs. the first wave.

^‡p < 0.05 for the third wave vs. the second wave.

ECMO, extracorporeal membrane oxygenation; ICU, intensive care unit; MV, mechanical ventilation.

Complete follow-up on 90 days was available for all patients post-ECMO implantation. Thirty-two (32/54 = 60%) patients were weaned from ECMO, and 52% (n = 28/52) patients died. The 90 day mortality rate significantly increased from the first to second wave (respectively, 6 [43%] vs. 12 [86%], p < 0.05) and significantly decreased from the second to third wave (12 [86%] vs. 10 [38%], p < 0.05).

Pre-ECMO Risk Factors

The multivariable Cox model retained SOFA score, RESP score, and time from first symptoms to ECMO as associated with an increased hazard of death (Table 4). Kaplan-Meier survival analysis demonstrated a significant difference between the three waves. Especially, survival was lower for the second wave (log-rank test p = 0.003, Figure 2).

Table 4. - Univariate and Multivariate Cox Analysis of Variables Associated with 90 Day Mortality

Variables	90 day mortality
	Univariate analysis		Multivariate analysis
	HR (95% CI)	p	HR (95% CI)	p
Age > 60 years	3.52 (1.52–8.1)	0.003	3.36 (1.43–9.1)	0.006
Modified RESP score (for each point)	0.72 (0.56–0.91)	0.007	0.75 (0.57–0.98)	0.044
Lung compliance before ECMO	0.96 (0.92–1.01)	0.12	-
Time to first symptoms to ECMO (for each day)	1.12 (1.05–1.19)	0.001	1.10 (1.02–1.17)	0.013
Prone positioning	1.14 (0.39–3.31)	0.811
Epidemic waves	-	-	-	-
	-	-	-	-
First wave	2.9 (1.2–6.9)	0.012	1.29 (0.51–3.27)	0.58
Second wave	1.07 (0.41–2.7)	0.88	-	-
Third wave
SOFA score	1.9 (0.91–4.1)	0.08	-	-
SOFA CV	1.5 (0.71–3.2)	0.29	-	-

CV, cardiovascular; ECMO, extracorporeal membrane oxygenation; HR, hazard ratio; RESP, respiratory extracorporeal membrane oxygenation survival prediction; SOFA, sepsis organ failure assessment.

Figure 3.: ROC curve analysis of the delay from the first symptom to ECMO.

ROC curve analysis showed that the AUC of the delay between the first symptoms to ECMO for predicting in-hospital mortality was 0.81 (95% CI 0.70–0.93, p < 0.001). A cutoff value of 12 days had a sensitivity of 81% and a specificity of 78% to identify patients who died during hospital stay (Figure 3).

Discussion

The results of our study comparing the different clinical characteristics and outcomes of patients undergoing ECMO for CARD during the three epidemic waves of COVID-19 in France can be summarized as follows: 1) mortality at 90 days was 52%, 2) mortality has increased during the second wave, 3) mortality has improved during the third wave, and 4) disease progression of COVID-19 infection before ECMO seems to be a factor associated with 90 day mortality.

ECMO Mortality

In our cohort of 64 ECMO for CARDS, the 90 day mortality was 52%. This is higher than reported from a recent meta-analysis of 1896 ECMO patients implanted during the first pandemic year, with a pooled in-hospital mortality of 37.1% (95% CI: 32.2–42.0).¹⁶ Age is an independent factor of mortality in ECMO.^13,16 A higher age can partially explain our result in our cohort (61 [48–65] years) than in another French cohort (52 years [45–58])¹ or European registry (52.6 years with a range of 16–80).¹⁷ Nevertheless, despite a higher age, the mortality in our study was lower than that of an extensive German registry (68%) composed of 3,397 ECMO implanted during the first three waves of the COVID-19 pandemic.¹⁸

ECMO Mortality During the Second Wave

Ninety day mortality was higher during the second wave (n = 6/14 vs. n = 12/14, p < 0.05). In the second wave, the rate of ECMO weaning was lower (n = 6/14 vs. n = 9/14, p = 0.36), and ECMO duration (24 [21–34] days vs. 12 [9–21] days, p = 0.009) was higher. A different respiratory phenotype pre-ECMO may explain these differences. During the second wave, patients with CARDS presented a more significant alteration of static compliance, a higher driving pressure, and a more substantial progression of the COVID-19 disease before ECMO implantation (Table 2). Our results are in accordance with those of the international EuroELSO survey that described a change in clinical characteristics of CARD requiring ECMO during the second wave. Especially, the authors found a decreased rate of successful weaning (47%, n = 718/1723 vs. 47%, n = 841/1442, p < 0.0001) and survival (44%, n = 677 vs. 53%, n = 770, p < 0.0001).⁷

Several hypotheses can be considered. First, during the second wave, the ventilatory management of patients with COVID-19 was primarily modified. In the first wave, COVID-19 patients hospitalized in ICU were mostly under invasive mechanical ventilation.⁶ The use of non-invasive ventilation or nasal high flow oxygen therapy was controversial due to the potential harm of aerosolization of viable virus particles and airborne contamination.¹⁹ The increased (and prolonged) use of noninvasive ventilation support during the second wave may have contributed to the altered pulmonary profile of CARDS requiring ECMO. Indeed, ECMO was implanted in patients with more advanced COVID-19 lung injury and potentially more affected lung parenchyma related to patient self-inflicted lung injury during noninvasive support.²⁰

Clinical Evolution and Outcomes

Mortality occurred in 28 (52%) patients (Table 4). Mortality increased during the second wave group (n = 12/14 vs. 5/14, p < 0.05) and decreased during the third wave (n = 10/23 vs. 12/14, p < 0.05). On Cox univariable analysis, the occurrence of the primary outcome was associated with the RESP score (p = 0.007), the time from first symptoms to ECMO (p = 0.001), and length of stay during the second wave (p = 0.012). After multivariable adjustment, RESP score (HR = 0.83; 95% CI: 0.70–0.98, p = 0.03) and time from first symptoms to ECMO (HR = 1.12; 95% CI: 1.05–1.20, p = 0.001) remained independently associated with 90 day mortality. Kaplan-Meier survival curves showed a difference in 90 day mortality during the three French epidemic waves (p = 0.047, Figure 2). Especially 90 day mortality was increased during the second wave.

Time from Onset of Symptoms to ECMO Implantation

The most relevant finding of our report is the time from symptoms onset to ECMO cannulation. The time from symptoms onset to ECMO was confirmed as an independent factor in our Cox model (Table 4). The ROC curve analysis showed that the prognosis worsened after 12 days of disease evolution, making ECMO therapy not beneficial for these patients.

The rationale of ECMO therapy in the management of severe ARDS is to buy time until significant resolution of acute alveolar lung injury. Over 12 days, our data suggest that the resolution of alveolar damage could not be expected, leading to impossible weaning from ECMO. We hypothesize that these patients had already evolved to an irreversible fibrosis stage. During the second wave, all patients received dexamethasone before ICU admission in line with the edited guideline at the light of the RECOVERY trial that confirmed the benefice of dexamethasone on mortality.²¹ Hence, we potentially selected non-responder patients to dexamethasone without avoiding refractory hypoxemia.

Our ECMO program usually considers the time from intubation to EMCO implantation based on a solid rationale.⁹ The RESP score that predicts the mortality risk of ECMO already includes the time under mechanical ventilation. Hence, we consider a time under mechanical ventilation below seven days as a good time window for ECMO therapy.¹³ In our study, the median intubation time from mechanical ventilation to ECMO was 3–4 days and constant through the three waves. However, the RESP score did not include the time from symptoms onset to ECMO cannulation.

After the second wave and considering the high mortality rate, we decided to include the time from symptoms onset to ECMO as an additional criterion for ECMO eligibility in addition to the EOLIA criteria. This more drastic patient selection may explain the improved survival observed during the third wave. Nevertheless, this hypothesis requires further investigations with external validation. We may suggest for COVID-19 patients a “modified” RESP score, including the time from symptom onset to ECMO decision.

Limits

We acknowledge several limitations in our study. First, the limited sample size and the monocentric design may have led to underpowered statistical analyses. However, several monocentric studies with smaller numbers found similar results, especially regarding the increased mortality during the second wave.^22,23

Second, the impact of SARS-CoV-2 variants on ECMO mortality during the three waves could not be analyzed. In our center, systematic screening for SARS-CoV-2 mutations was only performed after November 2020. To date, data on ECMO outcome and SARS-CoV-2 variants are very scarce. Further studies are needed to investigate this association.

Finally, the improvement in survival observed during the third wave was probably due to several non-observational biases in selecting patients requiring ECMO. The higher proportion of young patients may be due to the vaccine policy against SARS-CoV-2. Indeed, the French vaccination policy, implemented in January 2021, has probably decreased the risk of severe CARDS in people over 60 years of age, thus reducing ECMO risk in this population.

It should be noted that ECMO equipment shortage was never considered as a criterion for non-selection. Indeed, thanks to regional management, our center has never encountered any problem with ECMO equipment availability.

Conclusions

This study conducted in the Picardy subregion (northern France) showed that survival of patients with severe CARDS requiring ECMO was 48% during the first three waves. When comparing the first with the second wave, we found that survival of ECMO patients with severe CARDS has declined during the second wave. Systematic corticosteroids administration and multiple lung injuries due to prolonged noninvasive oxygenation strategies before intubation or ventilator-associated pneumonia may explain this outcome. The improvement in survival during the third wave was probably due to a stricter selection of CARDS patients, taking into account the different specific COVID-19 treatment and ventilatory strategies performed before ECMO. Further multicentric studies focusing on time from first symptoms as selection criteria to initiate ECMO for severe CARDS patients are required.

Acknowledgments

The authors thank the paramedical team of the ICU and perfusionists of the UMAC team.