Introduction

Extracorporeal membrane oxygenation (ECMO) has been used extensively for coronavirus disease 2019 (COVID-19)-related acute respiratory distress syndrome (ARDS). However, it is highly resource intensive, leading to challenges in provision during the pandemic [1]. A systematic review and meta-analysis examining patients who received ECMO for COVID-19 in 2020 reported a 37% mortality rate [2]. As the pandemic progressed, treatment practices and patterns evolved, and newer variants of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) emerged. Alongside these changes, contemporaneous studies reported increasing mortality rates and longer duration of ECMO runs in patients with COVID-19 ARDS. The mortality rate reported by the Extracorporeal Life Support Organisation (ELSO) registry data for the use of ECMO in COVID-19 increased from 37% in early 2020 to 52% by the end of 2020 [3, 4], demonstrating the dynamic nature of clinical outcomes during the course of the pandemic.

While subsequent single-centre studies have shown similar trends, the mortality rates for patients receiving ECMO for COVID-19 appear variable globally, with reports of rates ranging from 17.5% to 68% in the first 18 months of the pandemic [5]. Several reasons related to patient, disease, and treatment factors have been postulated for this and include increased virulence of SARS-CoV-2 variants [5, 6]; changes in patient selection patterns based, at times, on local resource availability; changes in interventions, including the need of using prolonged noninvasive forms of mechanical ventilation and delays in endotracheal intubation due to the overwhelming number of patients with respiratory failure; and the use of immunomodulators such as corticosteroids and interleukin-6 receptor antagonists [3, 7]. Based on this, we performed an updated systematic review and meta-analysis to summarise outcome data during the first 2 years of the pandemic, including the changes in mortality trends, and identify risk factors for unfavourable outcomes in order to guide clinical decision-making and further research.

Methodology

Search strategy and selection criteria

We registered the protocol with PROSPERO (CRD42021271202) and conducted the review in adherence with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) Statement (Additional file 1: Table S1) [8]. We searched MEDLINE, Embase, Cochrane and Scopus databases from 1 December 2019 to 26 January 2022 using the following keywords and their variations: “extracorporeal membrane oxygenation”, “extracorporeal life support”, “SARS-CoV-2” and “COVID-19” (Additional file 1: Table S2). We also reviewed the reference lists of included studies and review articles on the topic. We included studies or online registries reporting on at least 10 adult patients with COVID-19 requiring ECMO. We excluded any studies primarily reporting on animals or paediatric patients (< 16 years old). In the case of overlap** patient data, we included the largest study and excluded any other overlap** studies.

Data collection and risk of bias assessment

We collected data using a prespecified data extraction form. Authors were contacted for additional data where necessary (Additional file 1: Table S3). We assessed individual study risk of bias using the appropriate Joanna Briggs Institute checklist for case series or cohort studies. We assessed certainty of evidence using the Grading of Recommendations, Assessments, Developments and Evaluations (GRADE) approach [9]. The screening of studies, data collection, and risk of bias assessment were conducted independently and in duplicate by RRL and JJLS, and FA assisted with the risk of bias assessment. Conflicts were resolved by consensus or by KR. Where there was missing data, we contacted the corresponding authors of each study to obtain additional data for analysis.

Data synthesis

The primary outcome was mortality at the longest recorded time of follow-up. Secondary outcomes included ICU and hospital and length of stay, duration of invasive mechanical ventilation, duration of ECMO, and complications during ECMO (which we then classified according to the broad groups described by ELSO). We performed random-effects meta-analyses (DerSimonian and Laird) based on the logit transformation [10,11,12], and computed 95% confidence intervals (CIs) using the Clopper–Pearson method [13]. As inter-study heterogeneity in observational studies tends to be overestimated by I2 statistics, we assessed statistical heterogeneity (inconsistency) as part of the GRADE approach [9], using I-squared but also the Chi-squared test and visual inspection of the forest plots [14]. We assessed for publication bias qualitatively using visual inspection of funnel plots, and quantitatively using Egger’s regression test. We corrected for small-study effects using the random-effects trim-and-fill (R0 estimator) procedure. As some centres which published studies on their patient cohort report that patient data to the ELSO registry, there is a risk of duplicating patient data when including studies reporting on data from the ELSO registry. Hence, we conducted a sensitivity analysis excluding any studies reporting on ELSO registry data. We also conducted a second analysis excluding studies with high risks of bias (defined as JBI score < 7) and analysed the mortality among studies specifically reporting on outcomes of patients receiving venovenous ECMO (VV-ECMO). We present survival outcomes as pooled proportions, while continuous outcomes are presented as pooled means, both with corresponding 95% CIs.

We conducted pre-specified subgroup analysis based on the geographical region (North America, Latin America, Asia–Pacific, Europe, Southwest Asia and Africa), as well as by time period (every six months from 1 January 2020, defined by the date of enrolment of the last patient included in each study). We conducted univariable meta-regression when at least 6 data points were reported, to explore potential sources of heterogeneity, or prognostically relevant prespecified study-level covariates (date of last patient enrolment [per 100 days from 1 January 2020], age [per year], proportion of male patients, and patients receiving corticosteroids and interleukin blockers [percentage], body mass index [per 1 kg/m2], SOFA score [every increase by 1 point], PaO2/FiO2 ratio [increase by 1], duration of ECMO cannulation, time from symptoms to mechanical ventilation and time from mechanical ventilation to ECMO [days]). For continuous variables, we pooled the means from the aggregate data presented in each study as per Wan et al. [15]. A p value of < 0.05 was defined as statistically significant for our analyses. We performed all statistical analyses using R 4.0.2.

Post hoc analysis

We investigated the impact of time of last patient enrolment from Jan 1, 2020 on the duration of ECMO, ICU and hospital lengths of stay using study-level meta-regression. In addition, given the disparity in sample sizes, we conducted an exploratory meta-regression of sample size with mortality rates. As studies might recruit patients over a period of time, we conducted a meta-regression of the mean date of patient enrolment (defined as the midpoint between the date of first and last patient enrolment within each study) and mortality. Finally, we conducted an exploratory subgroup analysis based on the duration of follow-up reported by each study.

Role of the funding source

There was no funding source for this study.

Results

Study selection and characteristics

Of 4522 citations, we reviewed 222 full-texts and included 52 studies totalling 18,211 patients receiving ECMO for COVID-19, in the meta-analysis (Fig. 1, Additional file 1: Table S4) [3, 5, 16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,81, 82]. As such, the findings of this review need to be interpreted in context and clinical practice may evolve further.

Conclusions

In conclusion, our review summarising the updated literature on the use of ECMO for COVID-19 demonstrated an increase in mortality in 2021, likely due to a combination of demographic, disease, and intervention factors. It is evident that a one-size fits all protocolised approach to ECMO, used earlier in the pandemic, may not be as applicable as newer variants emerge, clinical patterns vary and management for severe COVID-19 changes. Despite the increase in mortality over time, ECMO still serves an important role as supportive therapy for select patients. Physicians should carefully weigh the potential benefits and harms of ECMO for each patient in the context of resource availability, the individual’s disease course, and local experience and mortality rates in order to decide on ECMO candidacy [7].