Functional recovery in a cohort of ECMO and non-ECMO acute respiratory distress syndrome survivors

Veno-venous extracorporeal membrane oxygenation (ECMO) is a salvage therapy used in part for ARDS patients with severe hypoxemia. To date, there have been nearly 17,000 patients with COVID-19 ARDS placed on ECMO [1]. While the use of ECMO increased as a result of the COVID-19 pandemic, the mortality benefit is widely debated [2‐5]. The use of ECMO as a salvage intervention in the most severe ARDS may rescue patients from fatal hypoxemia and mitigate the potential harms of mechanical ventilation [6]. The 2023 European Society of Intensive Care Medicine (ESICM) guidelines on ARDS make a strong recommendation in favor of ECMO in severe COVID-19 and non-COVID-19 ARDS when performed in an ECMO center [7]. However, ECMO remains a limited, resource intensive, and costly resource [2, 4, 8].

Studies of survivors of ARDS have demonstrated that while spirometry and lung volume have normalized by one year in most patients [9], there are persistent impairments in the diffusion capacity of the lung for carbon monoxide (DLCO), six-minute walk test (6MWT), and psychological outcomes for up to five years [10]. Prior studies have explored the functional and pulmonary recovery of ECMO survivors and found them to be comparable to patients with ARDS who were managed without ECMO [11‐14]. These studies predate current practice patterns for ARDS including the routine use of corticosteroids for COVID-19 ARDS, efforts to prioritize reductions in sedation, and the overall avoidance of continuous neuromuscular blockade. Additionally, these studies precede the COVID-19 pandemic which has led to overall longer lengths of hospital and ICU stay as well as more prolonged ECMO durations [15].

Given the scarcity, cost, and potential morbidity of ECMO as a resource, it is important to study its outcomes beyond in-hospital mortality to better inform decisions about ECMO allocation and utilization. We report the long-term pulmonary, physical, and neurocognitive recovery of present-day survivors of severe ARDS who received ECMO in comparison with a contemporaneous cohort of patients with ARDS who were not managed on ECMO.

This study was performed from January 2020 through January 2023. This study was determined to be exempt by the University of Maryland and R Adams Cowley Shock Trauma Center Internal Review Board (IRB).

From January 2020 through December 2022, there were 3005 patient encounters with a P/F ratio of ≤ 300 mmHg at the University of Maryland Medical Center, R Adams Cowley Shock Trauma Center, and University of Maryland Baltimore Washington Medical Center in the adult Emergency Department or Intensive Care Unit. During the same time period, 211 patients who met the Berlin Criteria for ARDS with refractory hypoxemia were cannulated for VV-ECMO at the University of Maryland Medical Center/R Adams Cowley Shock Trauma Center. Of the 3005 patients with a P/F ratio of ≤ 300 mmHg, 1997 were noted to be alive at the time of discharge, including 141 patients who had been cannulated for ECMO. A total of 110 patients were discharged alive and presented for post-ICU follow-up after requiring VV-ECMO or admission to the ICU for ARDS; these patients represent the focus of this analysis (Fig. 1).

Table 1 describes the population characteristics of these 110 patients on initial presentation to the ICU, the ICU based interventions, and the hospital outcomes. When comparing the ECMO (n = 34) to the non-ECMO (n = 76) cohort, the ECMO patients were younger with a median age of 35 years old (28, 50) compared to 51 years old (44, 61) in non-ECMO patients (p < 0.01). There were no significant differences in sex, race, ethnicity, or BMI between the two groups of survivors. A smaller proportion of patients in the ECMO group had COVID-19 (56%) compared to the non-ECMO group (96%) (p < 0.01). The etiology of ARDS in the ECMO cohort is summarized in Additional file 1: Table S1. Patients on ECMO had higher SOFA scores on admission (7 [5, 9] vs. 4 [3, 6], p < 0.01) and lower admission PaO₂/FiO₂ ratios (104 [81, 158] vs. 150 [103, 210] mmHg, p = 0.02). Patients on ECMO were more likely to be on salvage modes of ventilation and had significantly higher peak inspiratory pressures, mean airway pressures, dynamic driving pressures, FiO₂ requirements, and PaCO₂. ECMO patients also had a significantly lower dynamic compliance and higher oxygenation index. Significantly more patients in the ECMO group were intubated, received neuromuscular blockade, received inhaled vasodilators, and received vasoactive medications compared to the non-ECMO group. The ECMO group had a longer median duration of hospitalization (46 [27, 62] vs. 16 [12, 31] days, p < 0.01), duration of ICU stay (29 [19, 43] vs. 10 [5, 17] days, p < 0.01), and days of mechanical ventilation (24 [14, 42] vs. 10 [7, 17] days, p < 0.01) (Table 1).

To the best of our knowledge, we report the largest contemporaneous comparison of functional outcomes comparing survivors of ARDS with and without ECMO. We also report a novel direct comparison of patients who were managed using ECMO to patients who were ECMO eligible, based on retrospective assessment, but did not get cannulated for ECMO. Both ECMO and non-ECMO patients had a mild restrictive pattern observed by spirometry at an average of 92 days after hospital discharge. The median FVC% predicted of the non-ECMO cohort was 70% compared to 72% in the ECMO cohort. This finding aligns with prior work by Herridge et al. in survivors of ARDS which demonstrated a FVC% predicted of 72% at three-months [9]. There were no statistically significant differences in any markers of recovery when comparing ECMO to non-ECMO survivors, even when adjusting for age, SOFA, COVID-19 status, and length of hospital stay.

Large, randomized control trials and emulated trials have not demonstrated a mortality benefit from the use of ECMO at 60 or 90 days [2, 15]. However, these analyses are not without limitations, including low sample sizes, high cross-over rates, concern with study designs, and lack of long-term follow-up [4, 6, 9]. There is general agreement that lives have undoubtedly been saved in circumstances when even aggressive modes of conventional ventilation, prone positioning, and neuromuscular blockade could not adequately oxygenate and ventilate patients. However, the longer-term impact of ECMO on more functional outcomes like pulmonary function, anxiety, depression, and PTSD has been less clear. A novel aspect of the presented analysis is the direct comparison of patients on ECMO to those that met eligibility criteria for but were not cannulated for ECMO (“ECMO eligible” cohort). While the sample size of the study precludes the use of propensity score matching methods, the creation of a clearly defined “ECMO eligible” subset in combination with regression modeling adjusting for baseline differences between the two cohorts attempts to minimize the influence of confounding. Interestingly, even with this comparator group, the ECMO-treated patients were significantly sicker based on higher average SOFA scores, the need for salvage modes of mechanical ventilation using higher pressures, higher FiO₂, impaired CO₂ clearance, lower lung compliance, longer ICU stay, hospital stay, and an increased utilization of inhaled vasodilators and vasopressors. Each of these factors has the potential to impair recovery. It might be expected, then, that this sicker ECMO population would have worse functional recovery. Our findings, however, suggest that the functional recovery of patients who required ECMO was similar to that of the non-ECMO patients, despite the ECMO patients being more severely ill.

To the best of our knowledge, this is the first study to compare pulmonary, psychiatric, and neurocognitive function of ECMO patients to a contemporaneous cohort of non-ECMO survivors of ARDS since the work published by Grasselli et al. which enrolled patients from 2013 to 2015 [12]. This study is also the first to our knowledge to include both COVID-19 and non-COVID-19 patients. There have been important changes to the management of ARDS in the past decade, including the tendency for lighter sedation, less frequent use of neuromuscular blockade, and the standard use of corticosteroids (in COVID-19 patients). The use of a contemporaneous non-ECMO cohort ensured that both groups were exposed to the same ARDS clinical practice patterns as well as the same COVID-19 related lockdowns, masking policies, vaccine access, and resource limitations. Our study included a diverse population of patients, comprised of nearly 36% Black and 59% White patients, 43% female, and 15% Hispanic or Latino population. This diversity increases the generalizability of our findings to populations of patients known to be more severely afflicted by COVID-19 and to have worse mortality from ARDS and frequently underrepresented in the literature [21]. The University of Maryland is an urban quaternary care center. As a safety net hospital, the University cares for patients who are marginalized–frequently uninsured, lacking primary care, low-income, and highly vulnerable with innumerable barriers to medical care. These social determinants of health may increase the risk for poor outcomes. Yet, the findings of this study demonstrate that most survivors of ARDS (both ECMO and non-ECMO) exhibited only a mild to moderate impairment in pulmonary function and no significant neurocognitive or psychiatric impairment at hospital follow-up despite the fact that these patients often have a paucity of resources and experience barriers to longitudinal care after hospital discharge.

FVC% predicted was used as the primary outcome as a surrogate of restrictive lung physiology. There is no difference in the FVC% predicted between the ECMO group when compared to either the non-ECMO or the “ECMO eligible” group, suggesting that this group does not have more severe restrictive lung disease. However, it is important to consider that the ECMO cohort is consistently significantly younger than the non-ECMO cohort. Thus, if this degree of mild restriction persists, ECMO patients will carry the burden of this morbidity for more of their working, reproductive, and functional lifetime when compared with the older non-ECMO or “ECMO eligible” cohort [21, 22]. Work by Herridge et al. demonstrated normalization of FVC% predicted by six months in survivors of ARDS [20]. Thus, longitudinal follow-up is needed to assess whether this trajectory is seen in ECMO patients.

This study has some limitations, the most important of which is its observational nature. The initial allocation of ECMO to younger patients with fewer comorbidities and contraindications to cannulation who are often earlier in the disease course reveals an inherent selection bias which we can only partially adjust for in our regression analysis. There was additional selection bias in that this study only included individuals who were still alive and who had the ability, resources and desire to complete outpatient long-term follow-up and functional testing, potentially missing both the sickest and most disabled patients and the most recovered patients. Without knowing the demographics of all the patients who died on ECMO or with ARDS, our study may be subject to differential survivorship bias. This study was conducted in an academic medical center which does a high volume of ECMO and cares for a large number of ARDS patients. Improved post-ECMO outcomes have been reported in centers with higher volumes [23], so these data may overestimate the functional recovery of a broader ECMO population. Spirometry is a unidimensional outcome which can be insensitive in detecting restriction, may be impacted by ICU acquired weakness, and which does not quantify global functional impairment or necessarily translate to quality of life. We did not obtain ventilator parameters of our patients after ECMO cannulation. A presumed benefit of ECMO is the use of “ultra low-volume ventilation” which is thought to spare additional ventilator induced lung injury (VILI) in patients with poor lung compliance [24, 25]. Whether this strategy was utilized and how this contributes to pulmonary function at follow-up is not known. The majority of the non-ECMO patients (96%) had COVID-19. No data were available on the variant of SARS CoV-2 that the COVID-19 patients were infected with. Different variants resulted in varying degrees of ARDS severity [26]. The precise impact of COVID-19 on pulmonary recovery after ARDS is unknown, but recent work from Hodgson et al. showed no significant difference in new six-month disability, quality of life, neurocognitive, or psychiatric function when comparing survivors of COVID-19 to non-COVID-19 ARDS [27]. Finally, this study was limited by its relatively small sample size as well as missing data, particularly with regard to some details pertaining to outside hospitalizations and the neurocognitive and psychiatric data. Given the high mortality for ARDS and accounting for tremendous loss to follow-up ICU recovery clinics, the population of ARDS survivors is limited.

In this sample of 110 contemporary patients with ARDS who were able to come to ICU follow-up clinic, we did not identify any differences in functional recovery when comparing ARDS survivors who were managed on ECMO to patients who did not receive ECMO. ECMO patients were younger but also had significantly more ventilator requirements and pulmonary physiology derangements prior to cannulation, had significantly longer duration of mechanical ventilation, ICU and hospital length of stay, use of neuromuscular blockade, inhaled vasodilators, and vasopressors. These findings are reassuring regarding the impact of this resource intensive therapy. Future studies should focus on enrolling more patients to ensure adequate power and on longitudinal multidimensional assessments of morbidity. Furthermore, future ECMO randomized control trials should look beyond in-hospital mortality as the outcome of interest and consider evaluating patients’ overall recovery, return to work, and quality of life to better determine which patients would receive the greatest overall benefit from ECMO.