Patients with severe coronavirus disease 2019 (COVID-19) infection frequently present acute respiratory distress syndrome (ARDS).1 In addition to the widely used “protective pulmonary ventilation” strategies, the use of extracorporeal membrane oxygenation (ECMO) support has been recommended for those with refractory hypoxemia and hypercapnia.2–4 Although early reports indicated a mortality rate above 90% in patients with severe COVID-19 infection undergoing ECMO,5 a subsequent study with data from more than 1,000 patients from the extracorporeal life support organization (ELSO) registry showed an estimated 90-day mortality of 38%,6 which is in line with previous data for ECMO support in patients with ARDS from other causes.2,7 In a recent meta-analysis that included nearly 2,000 patients with COVID-19 undergoing ECMO support, mortality was 37%, and patient age and time on ECMO were directly associated with worse outcomes.8

In Latina America, however, there is a paucity of data on the outcomes of patients supported with ECMO. Additionally, clinical and functional status, including the magnitude of the post-acute COVID-19 syndrome (PACS), quality of life, anxiety, depression, post-traumatic stress disorder, and working status are largely unexplored in COVID-19 patients’ cohorts.

Therefore, we aim to describe in-hospital clinical outcomes in patients hospitalized with the severe form of COVID-19 who received support with ECMO in two private hospitals in Brazil. We also evaluated the burden of symptoms, functional and mental status, and the return to labor activities 30- and 90 days after hospital discharge. We hypothesized that patients with severe ARDS related to COVID-19 infection who needed ECMO support would have significant impairment in clinical and functional status in the posthospital discharge period.

Methods Study Design and Oversight

This retrospective cohort study was reviewed by an Institutional Review Board (IRB), approved with a waiver of informed consent at each participating site, and deemed exempt from IRB oversight (Instituto de Ensino e Pesquisa, IEP, Hospital Sírio-Libanês, Sao Paulo, Brazil). The data were analyzed in a secure, anonymized database physically separate from the main production server. All data collected was reviewed by the study team to assure data quality.

Patient Population

We retrospectively studied consecutive hospitalized adult patients at Hospital Sírio-Libanês in Sao Paulo and Brasilia, from April 2020 to August 2021. All patients had a confirmed diagnosis of COVID-19 by the presence of related symptoms and a positive result of a SARS-CoV-2 polymerase chain reaction assay for nasal and pharyngeal swab specimens. All patients with severe ARDS related to COVID-19 infection admitted to the intensive care unit (ICU) that presented refractory hypoxemia despite all clinical and mechanical ventilation strategies and who required ECMO support were included. Figure 1, Supplemental Digital Content 1, https://links.lww.com/ASAIO/A978 depicts the flowchart comprising the overall number of confirmed COVID-19 patients admitted to the hospital during the study period.

Variables and Clinical Outcomes Collected at the Hospital

All participants had their data collected until hospital discharge or in-hospital death. We collected data on demographic characteristics, past medical history, Simplified Acute Physiology Score III (SAPS-3) at the ICU admission day, laboratory examinations, cardiopulmonary arrest or previous infection, therapies used before ECMO cannulation (use of neuromuscular blocking agents [NMBA], inhaled nitric oxide [iNO], prone ventilation, vasopressors, corticosteroids, tocilizumab and convalescent plasma, anticoagulation regimen [therapeutic, prophylactic or intermediate-dose]), data on ventilatory settings and respiratory parameters reported for the ELSO database following the rules of Pre-ECLS Assessment (fraction of inspired oxygen [FiO2], positive end-expiratory pressure, lung compliance, driving pressure, arterial partial pressure of oxygen [PaO2], arterial partial pressure of carbon dioxide [PaCO2], PaO2/FiO2 ratio), modality of support (veno-venous or veno-arterial), cannulation site (other hospitals for transfer or at the local hospital), relevant time intervals (symptoms onset to hospital admission, ICU admission, invasive mechanical ventilation [MV] start and ECMO cannulation; hospital and ICU admission to ECMO cannulation; MV start to ECMO cannulation; total ECMO duration; total MV duration; and ICU and hospital length of stay [LOS]), major clinical complications (acute kidney injury; need for dialysis; major bleeding; stroke; ECMO-related complications such as thrombosis, hemolysis, gas embolism, limb ischemia and membrane substitution); and institution where the patient was hospitalized. The definitions of anticoagulation regimen, total MV duration, and ECMO-related complications are described in the supplemental material, Supplemental Digital Content 1, https://links.lww.com/ASAIO/A978.

Variables Collected After Hospital Discharge

The clinical outcomes team prospectively applied a standardized questionnaire by phone calls at 30- and 90 days after hospital discharge for survived patients. Members of this team had no access to in-hospital variables. They evaluated the presence of PACS as defined by the WHO,9 including the following symptoms: dyspnea, tiredness, cough, thoracic discomfort, anosmia, dysgeusia, headache, arthralgia, myalgia, and diarrhea at both timeframes. In addition, the following questionnaires were applied: European Quality of Life Five Dimension (EQ-5D),10 Generalized Anxiety Disorder 2-item (GAD-2),9 Patient Health Questionnaire-2 (PHQ-2).11,12 Post-traumatic stress symptoms and the patient’s return to work were also registered. The complete methodological description of these questionnaires can be accessed in the supplementary material, Supplemental Digital Content 1, https://links.lww.com/ASAIO/A978.

Statistical Analysis

Categorical variables were reported as absolute numbers and percentages and continuous variables as mean and standard deviation or median and interquartile range (IQR). The comparisons between in-hospital survivors and nonsurvivors were performed using chi-square or Fischer’s exact tests for categorical variables and Student’s t-test or Mann-Whitney test for continuous variables, according to their distribution.

Multivariable stepwise Cox proportional hazards regression analysis was performed to explore risk factors associated with in-hospital mortality in time-to-event survival analysis and the proportional hazard assumption was tested by using Schoenfeld residuals. We included pre-ECMO variables (age categorized as > or ≤65 years, sex, arterial hypertension, diabetes, history of coronary artery disease, history of cancer, SAPS-3, PaO2/FiO2 ratio, MV duration before ECMO cannulation categorized as > or ≤10 days), clinical complications (dialysis initiation after ECMO cannulation and stroke), and relevant time intervals (total ECMO support duration, time from hospital admission to ECMO cannulation and in-hospital LOS). All analyses were performed with the IBM-SPSS Statistics version 25 (Statistical Package for the Social Sciences, SPSS, Inc, Chicago, IL).

Results In-hospital Data

Eighty-five COVID-19 critically ill VV-ECMO-supported patients were included during the study period. The mean age was 59 ± 13 years, 36.5% were older than 65 years and 15% were women. Hypertension and obesity were the most common comorbidities and more than 85% of patients had at least one chronic condition. Compared to nonsurvivors, patients who survived were younger, and had a lower Charlson’s Comorbidity Index and a lower SAPS-3 (Table 1). Only two patients (2.4%) initially on VV-ECMO were converted to VA-ECMO.

Table 1. - Baseline Demographic Characteristics, Prevalence of Comorbidities and Prognostic Risk Scores at ICU-admission in the Whole Population, Survivors and Non-survivors Total population n = 85 Survivors n = 45 Non-survivors n = 40 p-value Baseline demographics Age, mean ± SD, years 59 ± 13 52 ± 11 66 ± 11 <0.001 >65 years, no. (%) 31 (36.5) 5 (11.1) 26 (65.0) <0.001 >70 years, no. (%) 16 (18.8) 1 (2.2) 15 (37.5) <0.001 Weight, median (IQR), kg 88 (77–99) 90 (77–101) 86 (74–97) 0.535 BMI, median (IQR), kg/m2 29.4 (25–33) 29.9 (25–34) 28.9 (25–32) 0.613 Female sex, no. (%) 13 (15.3) 9 (20.0) 4 (10.0) 0.240 Comorbidities Any chronic condition, no. (%) 72 (85.7) 39 (86.7) 33 (84.6) 1.000 Hypertension, no. (%) 44 (52.4) 19 (42.2) 25 (64.1) 0.052 Diabetes, no. (%) 23 (27.4) 9 (20.0) 14 (35.9) 0.142 Dyslipidemia, no. (%) 21 (25.0) 10 (22.2) 11 (28.2) 0.616 Coronary artery disease, no. (%) 9 (10.7) 2 (4.4) 7 (17.9) 0.075 Obesity, no. (%) 35 (41.7) 19 (42.2) 16 (41.0) 1.000 History of cancer, no. (%) 12 (14.3) 4 (8.9) 8 (20.5) 0.210 Prognostic risk scores SAPS-3, mean, ± SD 54 ± 12 50 ± 10 57 ± 12 0.009 CCI, median (IQR) 2 (1–3) 1 (0–2) 3 (2–4) <0.001

BMI, body mass index; CCI, Charlson Comorbidity Index; SAPS-3, Simplified Acute Physiology Score III.

As depicted in Table 2, the near totality of patients received NMBA, and 74.1% were submitted to prone position ventilation. More than 40% received rescue therapy with iNO and nearly 90% of patients were on vasopressors before ECMO cannulation. There were no differences in ICU therapies used between survivors and nonsurvivors. (Table 2) As for ventilatory settings and respiratory parameters at the day of ECMO cannulation, compared to nonsurvivors, survivors needed lower FiO2 and had a higher PaO2/FiO2 ratio. The duration of invasive MV before ECMO was the same for survivors and nonsurvivors and the proportion of patients with >10 days and >14 days of invasive MV before ECMO was equal in both groups.

Table 2. - Therapies Used Before ECMO Cannulation and Respiratory Parameters at the Cannulation Day Total population n = 85 Survivors n = 45 Non-survivors n = 40 p-value Therapies before ECMO cannulation No. pts who received therapy/no. pts with data (%) Neuromuscular blocking agents 80/82 (97.6) 44/44 (100) 36/38 (94.7) 0.212 Prone position 63/82 (74.1) 36/44 (81.8) 27/38 (71.1) 0.299 Inhaled nitric oxide 38/82 (44.7) 20/45 (44.4) 18/37 (48.6) 0.824 Vasopressors 76/85 (89.4) 39/45 (86.7) 37/40 (92.5) 0.491 Corticosteroids 85/85 (100) 45/45 (100) 40/40 (100) 1.000 Tocilizumab 12/85 (14.1) 7/45 (15.6) 5/40 (12.5) 0.762 Convalescent plasma 9/85 (10.6) 7/45 (15.6) 2/40 (5.0) 0.163 Prophylactic anticoagulation 15/77 (19.5) 8/42 (19.0) 7/35 (20.0) 1.000 Therapeutic anticoagulation 60/77 (77.9) 34/42 (81.0) 26/35 (74.3) 0.588 Intermediate-dose anticoagulation 3/77 (3.5) 1/42 (2.4) 2/35 (5.7) 0.584 Patients transferred on ECMO 22/85 (25.9) 10/45 (22.2) 12/40 (30.0) 0.464 Tracheostomy before ECMO 19/71 (26.8) 7/38 (18.4) 12/33 (36.4) 0.111 Ventilatory settings and respiratory parameters before ECMO cannulation* Total n = 64 Survivors n = 35 Non-survivors n = 29 p-value No. pts with respiratory parameters available FiO2, median (IQR), % 80 (65–100) 70 (60–80) 90 (70–100) 0.008 PEEP, median (IQR), mmHg 10 (8–10) 8 (6–10) 10 (8–10) 0.667 Lung compliance, median (IQR), mL/cmH2O) 22 (18–27) 22 (17–27) 23 (19–27) 0.494 Driving pressure, median (IQR), cmH2O 15 (13–15) 15 (13–16) 14 (13–15) 0.625 PaO2/ FiO2 ratio, median (IQR) 100 (81–120) 107 (92–129) 85 (70–113) 0.003 PaCO2, median (IQR) mmHg 68 (57–80) 68 (56–83) 68 (59–77) 0.622 MV duration before ECMO, median (IQR), days 7 (1–14) 6 (1–13) 7 (3–14) 0.210 MV>10 days before ECMO, no. (%) 31 (37.3) 15 (34.1) 16 (41.0) 0.515 MV>14 days before ECMO, no. (%) 18 (21.7) 9 (20.5) 9 (23.1) 0.772

*data reported for the ELSO database following the rules of Pre-ECLS Assessment.

no. pts, number of patients; FiO2, fraction of inspired oxygen; PEEP, positive end-expiratory pressure; PaO2, arterial partial pressure of oxygen; PaCO2, arterial partial pressure of carbon dioxide; MV, mechanical ventilation; ECMO, extracorporeal membrane oxygenation.

Overall median in-hospital LOS was 58 (39–84) days and compared to nonsurvivors, survivors had a significantly higher in-hospital LOS (62 [51–91] vs. 50 [23–72] days; p = 0.016). Other important time intervals during the clinical course of COVID-19 infection, from symptoms onset until hospital discharge or death are shown in Table 3.

Table 3. - Time Intervals During the Clinical Course of COVID-19 Infection, Since Symptoms Onset Until Hospital Discharge or Death Total population n = 85 Survivors n = 45 Non-survivors n = 40 p-value Symptoms onset - hospital admission, median (IQR), days 8 (5–11) 8 (6–11) 8 (5–11) 0.856 Symptoms onset - ICU admission, median (IQR), days 12 (9–16) 12 (9–15) 12 (9–16) 0.911 Symptoms onset - invasive MV start, median (IQR), days 17 (10–23) 15 (10–23) 17 (11–23) 0.682 Symptoms onset - ECMO, median (IQR), days 25 (21–32) 25 (17–32) 25 (21–34) 0.561 Hospital admission - ECMO, median (IQR), days 17 (10–23) 16 (8–22) 18 (13–32) 0.073 ICU admission - ECMO, median (IQR), days 12 (6–19) 12 (4–17) 13 (7–24) 0.271 ICU admission - ECMO >7 days, no. (%) 51 (65.4) 28 (63.6) 23 (67.3) 0.812 ECMO duration, median (IQR), days 12 (7–18) 10 (8–16) 13 (6–21) 0.724 Total MV duration, median (IQR), days 37 (16–60) 40 (20–59) 37 (14–67) 0.738 ICU LOS, median (IQR), days 39 (23–58) 41 (29–58) 39 (15–61) 0.306 In-hospital LOS, median (IQR), days 58 (39–84) 62 (51–91) 50 (23–72) 0.016

ECMO, extracorporeal membrane oxygenation; ICU, intensive care unit; LOS, length of stay; MV, mechanical ventilation; LOS, length of stay.

In-hospital outcomes are depicted in Table 4, with p-values for univariable analysis. Overall in-hospital mortality was 47.1% and 13 nonsurvivors (32.5%) were successfully decannulated from ECMO and stayed alive for at least 48 h after decannulation. None of the patients diagnosed with a stroke during ECMO support survived and only one patient who initiated dialysis after ECMO cannulation survived, an almost 93% in-hospital mortality rate. Fifty-four patients (63.5%) were supported with ECMO during the first COVID-19 wave in Brazil (April/2020 to March/2021) and the remaining during the second COVID-19 wave (April/2021 to August/2021), without any in-hospital mortality difference (48.1% vs. 45.2%, p = 0.791).

Table 4. - In-hospital Outcomes and Clinical Complications Total population n = 85 Survivors n = 45 Non-survivors n = 40 p-value In-hospital death, no. (%) 40 (47.1) Succcessful decannulation*, no. (%) 59 (69.4) 45 (100) 14 (35.0) <0.001 Dialysis during hospital stay, no. (%) 30 (35.3) 5 (11.1) 25 (62.5) <0.001 Dialysis initiation after ECMO, no. (%) 14 (16.5) 1 (2.2) 13 (32.5) <0.001 Stroke, no. (%) 7 (8.2) 0 (0) 7 (17.5) 0.004 Major bleeding, no. (%) 19 (22.4) 8 (17.8) 11 (27.5) 0.309 Thrombosis, no. (%) 12 (14.3) 7 (15.6) 5 (12.8) 0.765 Hemolysis, no. (%) 6 (7.1) 1 (2.2) 5 (12.5) 0.095 Gas embolism, no. (%) 5 (5.9) 3 (6.7) 2 (5.0) 1.000 Limb ischemia, no. (%) 0 (0) 0 (0) 0 (0) - Membrane substitution, no. (%) 3 (3.5) 0 (0) 3 (7.5) 0.100 ECMO modality change, no. (%) 2 (2.4) 0 (0) 2 (5.0) 0.218

*Patients who were successfully decannulated from ECMO and stayed alive for at least 48h after decannulation.

ECMO, extracorporeal membrane oxygenation.

The association of relevant demographic characteristics, clinical features, respiratory parameters previous to ECMO cannulation, important time intervals and clinical complications during ECMO support with in-hospital mortality evaluated by Cox proportional hazards regression analysis is shown in Table 5. After multivariable stepwise adjustments including all relevant predictors, age >65 years, diabetes, the duration of ECMO support and dialysis initiated after ECMO remained significantly associated with higher in-hospital mortality.

Table 5. - Univariable and Multivariable Stepwise Cox Proportional Hazards Regression Analysis With Predictors Associated With In-Hospital Death Univariable analysis Multivariable analysis HR 95% CI p-value HR 95% CI p-value Age >65 years 3.12 1.63–5.98 0.001 4.8 1.4–16.4 0.012 Diabetes 1.68 0.86–3.27 0.127 6.0 1.8–19.6 0.003 ECMO duration (days) 1.01 1.00–1.02 0.053 1.08 1.05–1.12 0.001 Dialysis started after ECMO cannulation 3.21 1.63–6.32 0.001 3.4 1.1–10.8 0.039 Male sex 1.24 0.43–3.56 0.686 0.93 0.15–5.9 0.934 Hypertension 1.98 1.02–3.82 0.043 1.04 0.29–3.80 0.949 Coronary artery disease 3.38 1.43–7.98 0.005 2.57 0.57–11.6 0.220 History of cancer 2.06 0.94–4.51 0.071 0.31 0.05–1.88 0.200 SAPS3 1.03 1.01–1.06 0.015 1.00 0.93–1.07 0.937 PaO2/FiO2 ratio 0.98 0.97–0.99 0.005 0.99 0.97–1.01 0.186 MV >10 days before ECMO 1.18 0.62–2.23 0.618 0.38 0.71–2.09 0.268 Hospital Admission to ECMO cannulation (days) 1.02 1.00–1.05 0.087 0.98 0.92–1.05 0.551 Stroke during ECMO 1.05 0.46–2.45 0.902 5.73 0.18–181.0 0.322 Hospital LOS (days) 1.00 0.99–1.01 0.821 1.02 1.00–1.04 0.118

ECMO, extracorporeal membrane oxygenation; FiO2, fraction of inspired oxygen; LOS, length of stay; MV, mechanical ventilation; PaO2, arterial partial pressure of oxygen; SAPS-3, Simplified Acute Physiology Score III.

MV duration before ECMO cannulation was not associated with higher in-hospital mortality, not as a continuous variable (unadjusted: p = 0.230), nor when categorized as greater than or ≤10 days (unadjusted p = 0.618) or greater than or ≤14 days (unadjusted p = 0.564, Table 2). Even after multivariable-adjusted analysis, MV duration before ECMO did not show any association with in-hospital mortality (Table 5).

Post-discharge Data

From the 45 survivors, 15 (33%) responded to both 30- and 90-day post-discharge structured questionnaires. Except for a higher proportion of diabetes in the responders, there were no other differences in clinical features between responders and nonresponders (Table 1S, Supplemental Digital Content 1, https://links.lww.com/ASAIO/A978). Around two-thirds and half of the responders reported the persistence of at least one symptom at the 30- and 90-day landmark, respectively. The median EQ-5D score reported was 0.85 (0.70–1.00) and 0.77 (0.66–1.00) at 30 and 90 days, respectively (p-value of 0.463). The median EQ visual analog scale (EQ VAS) was 80 (70–90) at both time points, p = 0.653. Only one patient scored at least 3 points in the PHQ-2 score at 90 days, suggesting a major depressive disorder and one patient scored 6 points in the GAD-2, compatible with generalized anxiety disorder. No post-traumatic stress disorder symptoms were reported.

From the 15 responders, four were not working before hospitalization, five returned to work without restrictions 30 days after hospital discharge, one returned with some restriction and five have not returned. At 90 days post-discharge, all previously working patients, except one, have returned to their labor activities.

Discussion

Our study described demographic aspects and clinical features, exploring their association with in-hospital outcomes in 85 ECMO-supported patients with severe ARDS related to COVID-19 infection, who were treated in two private hospitals in Brazil. Although the ELSO Registry Dashboard reports over 13,000 ECMO-supported COVID-19 adult patients worldwide, less than 900 of these reports come from Latin American ECMO centers.13 In 2021, Diaz et al.14 published the most relevant study with COVID-19 patients supported with ECMO in a Latin American country, a nationwide study with 85 patients treated in Chile. The population’s median age was 48 (41–55) years and the 90-day mortality of 38.8% was comparable to previous reports. In our study, overall in-hospital mortality was 47.1%, similar to that reported by the ELSO global registry.13 However, we should notice that our patients were older and 85.7% of them had at least one previous chronic condition, features strongly associated with poor outcomes in severe forms of COVID-19.15 The median age in our cohort was 59 (47–69) years, more than 10 years higher compared to the ELSO registry and similar to a nationwide German cohort that reported a 73% in-hospital mortality.16

Thirty patients (35.3%) received dialysis during their hospital stay and from the 14 patients (16.5%) who initiated dialysis after ECMO cannulation, only one survived. This strikingly high (nearly 93%) in-hospital mortality rate associated with dialysis initiated after ECMO support initiation should be further evaluated in other databases, but a possible explanation could be that the patients who evolve with kidney failure after ECMO cannulation probably have other organs’ dysfunctions in progress and therefore are at higher mortality risk. And since starting hemodialysis in an ECMO-supported patient is a relatively simple procedure, the ICU teams could have been more permissive on the indication for dialysis therap