key: cord-0740067-u64vfv1c
authors: Hall, Celeste A.; Jacobs, Jeffrey P.; Stammers, Alfred H.; St. Louis, James D.; Hayanga, J. W. Awori; Firstenberg, Michael S.; Mongero, Linda B.; Tesdahl, Eric A.; Rajagopal, Keshava; Cheema, Faisal H.; Patel, Kirti; Coley, Tom; Sestokas, Anthony K.; Slepian, Marvin J.; Badhwar, Vinay
title: Multi-institutional Analysis of 505 COVID-19 Patients Supported with ECMO: Predictors of Survival
date: 2022-02-18
journal: Ann Thorac Surg
DOI: 10.1016/j.athoracsur.2022.01.043
sha: e0ec48e042acc8434e455aff18cf717ab99c08a0
doc_id: 740067
cord_uid: u64vfv1c

Background We reviewed our experience with 505 patients with confirmed COVID-19 supported with ECMO at 45 hospitals and estimated risk factors for mortality. Methods A multi-institutional database was created and utilized to assess all patients with COVID-19 who were supported with ECMO. A Bayesian mixed-effects logistic regression model was estimated to assess the effect on survival of days between COVID-19 diagnosis and intubation, as well as age at ECMO cannulation. Results Median time on ECMO was 18 days (interquartile range=10-29). All 505 patients have separated from ECMO: 194 patients (38.4%) survived and 311 patients (61.6%) died. Survival with veno-venous ECMO was 184 of 466 patients (39.5%), while survival with veno-arterial ECMO was 8 of 30 patients (26.7%). Survivors had lower median age (44 versus 51 years, p<0.001) and shorter median time interval from diagnosis to intubation (7 days versus 11 days, p=0.001). Adjusting for several confounding factors, we estimated that an ECMO patient intubated on day 14 post COVID-19 diagnosis vs day 4 had a relative odds of survival of 0.65 (95% Credible Interval [CrI]:0.44-0.96, posterior probability of negative effect: 98.5%). Age was also negatively associated with survival: relative to a 38-year-old we estimated that a 57-year-old patient had a relative odds of survival of 0.43 (95% CrI:0.30-0.61, posterior probability of negative effect: >99.99%). Conclusions ECMO facilitates salvage and survival of select critically ill patients with COVID-19. Survivors tend to be younger and have shorter time from diagnosis to intubation. Survival of patients supported with only veno-venous ECMO was 39.5%.

As of September 29, 2021, 232,909,046 patients around the world have been diagnosed with Coronavirus Disease 2019 (COVID- 19) , with 4,768,139 associated deaths (2.05% mortality worldwide) [1] . Meanwhile, in the United States of America, as of September 29, 2021, 43,235,477 patients have been diagnosed with confirmed COVID-19, with 693,076 associated deaths to date (1.60% mortality in the USA) [1] . Most deaths in patients with COVID-19 are due to severe respiratory failure, with a small group succumbing to combined pulmonary and cardiac failure [2, 3] .

Extracorporeal membrane oxygenation (ECMO) emerged as a vital therapeutic strategy for severely ill COVID-19 patients, with inadequate oxygenation via conventional and ventilatory means [4] [5] [6] [7] . As such, the role of ECMO in the management of severely ill patients with COVID-19 continues to be defined. We previously published analyses of our initial 32, 100, and then 200 COVID-19 patients with severe pulmonary compromise supported with ECMO [8] [9] [10] . These prior analyses documented the evolution of the use ECMO to support patients with COVID-19 and supported the concept that "ECMO facilitates survival of select critically ill patients with COVID-19." [8] [9] [10] Although substantial variation exists in drug treatment of COVID-19, ECMO offers a reasonable rescue strategy [9, 10] .

Several previously published analyses describe cohorts of COVID-19 patients supported with ECMO [8] [9] [10] [11] [12] [13] [14] [15] [16] . Early data from Wuhan, China reported an alarmingly high rate of mortality of 83% (5 out of 6) in COVID-19 patients supported with ECMO [11, 12] . More recent data, however, reveal improved survival of COVID-19 ECMO patients [8] [9] [10] [13] [14] [15] [16] . Both individual institutional reports [13] , as well as reports from multi-institutional registries [14] , present detailed analyses with promising results. Our previous reports from our multi-institutional database [8] [9] [10] corroborate these findings from individual institutions [13] and multi-institutional registries [14] , but additionally provide more granular, detailed information than largescale registries [14] and more generalizable information than can be garnered from analysis from a single institution [13] . Increased knowledge about risk factors for mortality in COVID- 19 This analysis includes 505 patients with confirmed COVID-19 who were supported with and separated from ECMO between March 17, 2020, when our first COVID-19 patient was placed on ECMO, and October 11, 2021, when our last patient in this series was decannulated. Thirty-four patients who were cannulated but transferred to other hospitals on ECMO were not included in this analysis ( Figure 1 ). Data analyzed included patient characteristics, pre-COVID-19 risk factors and comorbidities, confirmation of COVID-19 diagnosis, features of ECMO support, specific medications utilized in an attempt to treat COVID-19, and short-term outcomes through hospital discharge.

Criteria for placement on ECMO were determined by the individual patient care team(s) at each of the contributing 45 hospitals; all patients who were placed on ECMO had the diagnosis of COVID-19 with severe respiratory failure deemed to be refractory to conventional management. The decision to initiate ECMO, the mode of therapy (i.e., veno-venous, veno-arterial, etc.), and the cannulation strategy were each determined by the individual ECMO teams, in keeping with their respective individual institutional protocols and guidelines. This analysis includes all patients with COVID-19 placed on ECMO at the 45 hospitals participating in this study during the period of this analysis. None of these 505 patients were placed on ECMO during cardiopulmonary resuscitation (CPR). Extracorporeal CPR (ECPR) was not utilized for COVID-19 patients at these 45 hospitals.

Descriptive summaries of the data were tabulated according to survival group using median and interquartile range with Kruskal-Wallis rank sum test for continuous variables or count and percent with chi-square test for categorical variables. The primary outcome of interest was mortality during the index hospitalization.

Missing data on covariates of interest was addressed by means of multiple imputation with chained equations as implemented by Harrell and colleagues [17] . Less than 10% of cases had missing data for most covariates of interest. A total of 25 imputed data sets were created, modeled, and combined into a single set of regression results to properly account for uncertainty due to missing data.

In order to assess the effects of patient and care variables upon survival, we estimated a Bayesian mixed-effects logistic regression model which included terms for age, sex, the presence of one or more key comorbidities (among asthma, cancer, chronic renal failure, diabetes, heart disease, hypertension and obesity), days between COVID-19 diagnosis and intubation, days between intubation and the initiation of ECMO, whether or not a patient was placed in the prone position pre-ECMO, with a random effect term controlling for the hospital at which care was given. Effects for age, days between diagnosis and intubation, and days between intubation and initiation of ECMO were all modeled using restricted cubic splines with three knots placed at the 10 th , 50 th , and 90 th percentiles of each variable's distribution, respectively. Prior distributions for each covariate were relatively uninformative: normal with a mean of zero and standard deviation of 100, allowing for a wide range of possible effects to be identified within the regression analysis. Individual model effects were summarized using the posterior predictive mean and 95% credible interval (CrI), with marginal effect contrasts between observations at the 25 th and 75 th percentile of each variable's distribution, unless otherwise noted. Predictive ability of the overall model was assessed using the c-index and Somers' Dxy. All analyses were conducted within the R statistical computing environment version 4.0.3 [18] with the use of the 'Hmisc' [17] and 'rmsb' [19] packages.

Institutional Review Board approval and waiver of the need for consent were obtained. The human subjects research protocol for this study was reviewed and approved by an independent Institutional Review Board. Institutional ethics review board approval was obtained for the use of data from the SCOPE TM Registry (Protocol #012017, ADVARRA Center for IRB Intelligence, Columbia, Maryland). This study involved a retrospective review of data contained within the SCOPE TM Registry; the reviewed data documented the individualized ECMO care provided at the direction of each patient's medical team. Consent for ECMO treatment was managed according to local hospital protocols. ECMO care was not altered for purposes of this study. ECMO records were archived in the SCOPE TM Registry for quality review purposes. A full waiver of the need for patient consent for retrospective research through the SCOPE TM Registry was approved by the ADVARRA Institutional Review Board (Protocol #012017). Figure 3 depicts the predicted probability of survival by age and shows improved survival with younger age. Figure 4 depicts the predicted probability of survival by days between COVID-19 diagnosis and intubation age and shows improved survival with a shorter time interval between COVID-19 diagnosis and intubation. Figure 5 depicts the predicted probability of survival by days between intubation and ECMO initiation shows a somewhat improved survival with a shorter time interval between intubation and ECMO initiation, but with a less consistent relationship that appears to be important only during the first five days after intubation. Substantial variation exists in the use of adjunctive drugs and therapies in the treatment of COVID-19, but these findings support the selective use of veno-venous ECMO as a reasonable rescue strategy.

It is not surprising that we found that the time interval from COVID Diagnosis to ECMO Cannulation is inversely related to survival after ECMO for COVID-19. Indeed, as documented in Table 1 , median "Days from COVID Diagnosis to ECMO Cannulation" was 10 days in survivors versus 15 days in non-survivors (p<0.001). A plausible biological explanation for this finding is based J o u r n a l P r e -p r o o f on the concept that lungs have substantial intrinsic regenerative capacity [20] [21] [22] [23] . Fulminant acute respiratory failure, occurring over a short time interval following infection and diagnosis, is a manifestation of extensive and rapid lung damage. When utilizing ECMO as a bridge to recovery of the lungs, ECMO provides pulmonary support, during which time lung recovery is dependent upon regenerative functions and may occur in a setting of minimally traumatic mechanical ventilation [24] , or even in the absence of mechanical ventilation [25] . Patients who have a longer time interval from COVID-19 diagnosis to respiratory failure severe enough to warrant support with ECMO, definitionally have (at least by physiological criteria) slowly progressive lung damage. Slowly progressive lung damage necessarily means that the pathogenic process exceeds lung generative function over a prolonged time period. If this is true, then it is entirely predictable that survival after ECMO cannulation is inversely related to the time interval from disease diagnosis to cannulation.

Although it is not surprising to some that "Days from COVID Diagnosis to ECMO Cannulation" is inversely related to survival after ECMO for COVID-19, it may be surprising to some that "Days from COVID Diagnosis to Intubation" is a more important predictor of outcome than "Days from Intubation to ECMO Cannulation". As documented in Table 1 , median "Days from COVID Diagnosis to Intubation" was 7 days in survivors versus 11 days in non-survivors (p=0.001), while median "Days from Intubation to ECMO Cannulation" was 3.5 days in survivors versus 4 days in non-survivors (p=0.001). In other words, as documented in Table 2 , when comparing the 25 th and 75 th percentile of "Days from COVID Diagnosis to Cannulation" (4 days versus 14 days), we find that patients in the 75 th percentile have a relative odds of survival of 0.65 (95% CrI: 0.44-0.96, posterior probability of negative effect: 98.5%); however, when comparing the 25 th and 75 th percentile of "Days from Intubation to ECMO Cannulation" (1 day versus 6 days), we find that patients in the 75 th percentile have a relative odds of survival of 0.82 (95% CrI: 0.49-1.38, posterior probability of negative effect: 77.7%). This observation might be explained by the theory that the "clock" on lung damage starts before the diagnosis of the disease, and it is the development of symptoms that typically trigger the test for the disease. From diagnosis of disease (mild lung damage typically, unless the patient presents with fulminant respiratory failure) to intubation is a period during which lung damage is ongoing, without net regeneration (because if there were sufficient regeneration, the patient's pulmonary function would not deteriorate). However, from intubation to ECMO cannulation is going from bad lung function to extremely poor lung function, which may not be that different quantitatively.

One might consider an analogy to perioperative kidney injury. The change from normal urine output and normal creatinine immediately after surgery to oliguria with a creatinine of 2.0 represents a much larger loss of renal function than the change oliguria with a creatinine of 2.0 to anuria and dialysis J o u r n a l P r e -p r o o f dependence. (That stated, dialysis dependence is more predictive of mortality than is oliguria alone, but our current study does not include patients who required mechanical ventilation but not ECMO).

Another important observation is that the length of time supported with ECMO for COVID-19 is inversely related to survival after ECMO for COVID-19. As documented in Table 1 , median "Days on ECMO" was 15 days in survivors versus 20 days in non-survivors (p=0.009). Patients with COVID-19 extremely long periods of time on ECMO without evidence of lung recovery have often developed de facto end-stage lung disease, and should truly be considered for ongoing ECMO support only if they might be reasonably hypothesized to be or become candidates for lung transplantation [26, 27] .

Our study adds to the body of knowledge and the literature by providing more granular multiinstitutional data about our cohort of 505 patients with COVID-19 supported with ECMO at 45 hospitals. As previously described, several published analyses have studied the outcomes of ECMO in patients with COVID-19, and these outcomes have been quite heterogenous [8] [9] [10] [11] [12] [13] [14] [15] [16] . Our analysis of the SCOPE TM Registry adds another dataset of multi-institutional data to the growing body of literature about the use of ECMO in patients with COVID-19 and demonstrates that support with ECMO facilitates salvage and survival of select critically ill patients with COVID-19. Survivors had lower median age (44 versus 51 years, p<0.001) and shorter median time interval from diagnosis to intubation (7 days versus 11 days, p=0.001).

Much remains to be learned about the role of ECMO in these patients. From our analysis, no specific demographic, clinical, or laboratory data, to date, is predictive of outcome with ECMO in patients with COVID-19, with the exception of younger age and shorter time from diagnosis to intubation. Survivors tend to be younger and have a shorter duration from diagnosis to intubation.

Meanwhile, the role of multiple medications in the treatment of COVID-19 remains unclear: none of the adjunct therapies appeared to be associated with survival. More information is needed to better determine which patients with COVID-19 will benefit from ECMO and which patients with COVID-19 will benefit from lung transplantation. Lessons learned from the use of ECMO to support patients with COVID-19 will inform the management of other patients with different forms of severe respiratory failure.

This analysis is based on the available data in our database. Potential limitations include patient selection bias, institutional bias, confounding bias, and potentially under-powering of the analysis. Additional follow-up is required on all surviving patients. Further patient accrual will enhance continued analysis of outcomes. We plan to continue gathering data to provide additional insight as to guideposts for patient selection and predictors of outcomes. It is our hope that by sharing our experience, other centers and patients may benefit.

Our experience and analysis of 505 consecutive patients at 45 hospitals reveal that ECMO facilitates salvage and survival of select critically ill patients with COVID-19. Survivors tend to be younger. Survival of patients supported with only veno-venous ECMO is 39.5% in our cohort.

Survivors had a shorter median time interval from the diagnosis of COVID-19 to ECMO Cannulation, driven mostly by the observation that survivors also had a shorter median time interval from the diagnosis of COVID-19 to intubation for mechanical ventilation. Substantial variation exists in drug treatment of COVID-19, but ECMO offers a reasonable rescue strategy. Additional gathering and analysis of data will inform appropriate selection of patients and provide guidance as to best use of ECMO in terms of timing, implementation, duration of support, and best criteria for discontinuation.

Expansion of studies such as the current analysis presented here will provide a means to further define the role of ECMO in the management of severely compromised patients with COVID-19 and will serve to refine the optimal use of ECMO in these patients, with the goal of continuing to enhance survival. J o u r n a l P r e -p r o o f 

Coronavirus COVID-19 Global Cases by the Center for Systems Science and Engineering

COVID-19 and Cardiovascular Disease

Potential Effects of Coronaviruses on the Cardiovascular System: A Review

Advanced Pulmonary and Cardiac Support of COVID-19 Patients: Emerging Recommendations From ASAIO-A "Living Working Document

Advanced Pulmonary and Cardiac Support of COVID-19 Patients: Emerging Recommendations From ASAIO-a Living Working Document. Circ Heart Fail

Initial ELSO Guidance Document: ECMO for COVID-19 Patients with Severe Cardiopulmonary Failure

Extracorporeal Membrane Oxygenation for COVID-19: Updated 2021 Guidelines from the Extracorporeal Life Support Organization

Extracorporeal Membrane Oxygenation in the Treatment of Severe Pulmonary and Cardiac Compromise in Coronavirus Disease

Multi-institutional Analysis of 100 Consecutive Patients with COVID-19 and Severe Pulmonary Compromise Treated with Extracorporeal Membrane Oxygenation: Outcomes and Trends Over Time

Patients treated with ECMO: Outcomes and Trends

Clinical course and outcomes of critically ill patients with SARS-CoV-2 pneumonia in Wuhan, China: a single-centered, retrospective, observational study

COVID-19, ECMO, and lymphopenia: a word of caution

Extracorporeal Membrane Oxygenation Support in Severe COVID-19

Extracorporeal membrane oxygenation support in COVID-19: an international cohort study of the Extracorporeal Life Support Organization registry

Venovenous extracorporeal membrane oxygenation for patients with refractory coronavirus disease 2019 (COVID-19): Multicenter experience of referral hospitals in a large health care system

Outcomes of Extracorporeal Membrane Oxygenation in Patients With Severe Acute Respiratory Distress Syndrome Caused by COVID-19 Versus Influenza

with contributions from Charles Dupont and many others

R: A language and environment for statistical computing. R Foundation for Statistical Computing

rmsb': Bayesian Regression Modeling Strategies

Distal airway stem cells yield alveoli in vitro and during lung regeneration following H1N1 influenza infection

Lung regeneration: mechanisms, applications and emerging stem cell populations

Lung regeneration by multipotent stem cells residing at the bronchioalveolar-duct junction

The Cellular and Physiological Basis for Lung Repair and Regeneration: Past, Present, and Future

Pumpless extracorporeal interventional lung assist in patients with acute respiratory distress syndrome: a prospective pilot study

Early Usage of Extracorporeal Membrane Oxygenation in the Absence of Invasive Mechanical Ventilation to Treat COVID-19-related Hypoxemic Respiratory Failure

Commentary: Extracorporeal membrane oxygenation for patients with refractory Coronavirus Disease 2019: What do we know and what do we need to learn?