Neurologic Complications of Patients With COVID-19 Requiring Extracorporeal Membrane Oxygenation: A Systematic Review and Meta-Analysis

OBJECTIVES:

In COVID-19 patients requiring extracorporeal membrane oxygenation (ECMO), our primary objective was to determine the frequency of intracranial hemorrhage (ICH). Secondary objectives were to estimate the frequency of ischemic stroke, to explore association between higher anticoagulation targets and ICH, and to estimate the association between neurologic complications and in-hospital mortality.

DATA SOURCES:

We searched MEDLINE, Embase, PsycINFO, Cochrane, and MedRxiv databases from inception to March 15, 2022.

STUDY SELECTION:

We identified studies that described acute neurological complications in adult patients with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection requiring ECMO.

DATA EXTRACTION:

Two authors independently performed study selection and data extraction. Studies with 95% or more of its patients on venovenous or venoarterial ECMO were pooled for meta-analysis, which was calculated using a random-effects model.

DATA SYNTHESIS:

Fifty-four studies (n = 3,347) were included in the systematic review. Venovenous ECMO was used in 97% of patients. Meta-analysis of ICH and ischemic stroke on venovenous ECMO included 18 and 11 studies, respectively. The frequency of ICH was 11% (95% CI, 8–15%), with intraparenchymal hemorrhage being the most common subtype (73%), while the frequency of ischemic strokes was 2% (95% CI, 1–3%). Higher anticoagulation targets were not associated with increased frequency of ICH (p = 0.06). In-hospital mortality was 37% (95% CI, 34–40%) and neurologic causes ranked as the third most common cause of death. The risk ratio of mortality in COVID-19 patients with neurologic complications on venovenous ECMO compared with patients without neurologic complications was 2.24 (95% CI, 1.46–3.46). There were insufficient studies for meta-analysis of COVID-19 patients on venoarterial ECMO.

CONCLUSIONS:

COVID-19 patients requiring venovenous ECMO have a high frequency of ICH, and the development of neurologic complications more than doubled the risk of death. Healthcare providers should be aware of these increased risks and maintain a high index of suspicion for ICH.

KEY POINTS

Question: What is the frequency of neurologic complications and their impact on mortality in COVID-19 patients requiring extracorporeal membrane oxygenation (ECMO)?

Findings: This systematic review and meta-analysis revealed a high frequency of intracranial hemorrhage in COVID-19 patients requiring ECMO (11%) and a low frequency of ischemic stroke (2%). Neurologic complications ranked as third most common cause of death and their presence more than doubled the risk of death.

Meaning: Initiation of ECMO in COVID-19 patients is associated with a high frequency of intracranial hemorrhage and neurologic complications more than doubled the risk of death. Clinicians should be aware of these increased risks when consenting and managing these patients. Future research should explore strategies to improve neurologic outcomes in this high-risk population.

Since 2019, the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has spread across the world to cause the COVID-19 pandemic. Severe COVID-19 infection is associated with acute respiratory distress syndrome (ARDS), which can be refractory to conventional mechanical ventilation. Consequently, there has been increasing utilization of venovenous and venoarterial extracorporeal membrane oxygenation (ECMO) in these critically ill patients. Management of patients with severe COVID-19 infection is complicated by direct and immune-mediated inflammatory response and endothelial damage that predispose patients to thromboembolic and bleeding events (1, 2). Balancing the risk of thrombosis and bleeding poses an additional layer of complexity to managing COVID-19 patients requiring ECMO support.

While ECMO can be life-saving, mortality and morbidity remain high (3). Previous meta-analysis revealed that 7% of non-COVID-19 patients on venovenous ECMO for respiratory failure developed a neurologic complication. Intracranial hemorrhage (ICH) is the most common neurologic complication in patients with ECMO and can be further subdivided into intraparenchymal hemorrhage, subdural hemorrhage, and subarachnoid hemorrhage. Other commonly reported neurologic complications include ischemic stroke and seizures. Recently, Siow et al (4) observed a relatively high mortality rate of 32% among all COVID-19 patients who developed an acute neurologic complication. However, the frequency of neurologic complications in COVID-19 patients requiring ECMO remains unknown.

In this systematic review, we collated existing published evidence regarding neurologic complications in adult COVID-19 patients requiring ECMO. Our primary aim was to estimate the frequency of ICH. Secondary aims were to estimate the frequency of ischemic stroke, to explore association between higher anticoagulation targets and ICH, and to estimate the association between neurologic complications and in-hospital mortality. By summarizing existing evidence regarding neurologic complications in adult COVID-19 patients requiring ECMO, our systematic review informs critical care and ECMO providers, identifies knowledge gaps, and helps inform future research.

METHODS

Study Selection

We performed a systematic literature search of MEDLINE, Embase, PsycINFO, and Cochrane databases using a predetermined search strategy (Supplementary 1, http://links.lww.com/CCX/B160) to identify full-text publications that described the neurologic complications of COVID-19 patients on ECMO. We also performed systematic literature search of MedRxiv database to identify relevant preprint articles. We included studies that enrolled adult patients (18 yr or older) with a SARS-CoV-2 infection confirmed by real-time reverse-transcription polymerase chain reaction assay, who received ECMO for acute hypoxemic respiratory failure, and reported at least one neurologic complication during ECMO or up to 7 days after decannulation. The systematic review searched for any neurologic complication, including ICHs, seizures, cerebral edema, microbleeds, myopathies, CNS infections, neuropathies, plexopathies, acquired demyelinating events, myelitis, and encephalitis. All neurologic complications must be must objectively confirmed by electroencephalogram, electromyography, CT, or MRI. We included case reports, case series, observational studies, cohort studies, and randomized control trials published in English between December 1, 2019, and March 15, 2022. In studies that included overlapping patient populations, we included the study with the largest sample size; the other overlapping studies were excluded to avoid duplication. We registered the study protocol in the International prospective register of systematic reviews database on October 25, 2021 (Identification number: CRD42021278261).

Two independent reviewers screened all titles and abstracts, and conducted full-text reviews of the relevant articles. Conflicts were resolved by consensus discussion or adjudicated by a third independent reviewer.

Data Extraction

Two reviewers independently extracted data. We extracted the study title, name of first author, year of publication, study type, study setting, patient demographics, medical treatments administered for COVID-19, days from initiation of mechanical ventilation to ECMO cannulation, duration of ECMO, type of ECMO, anticoagulation protocol, reported neurologic complications (incidence, type, and timing), and mortality (time and cause of mortality). To assess bias, we used the Newcastle-Ottawa Scale (NOS) for case-control and cohort studies and the Joanna Briggs Institute (JBI) critical appraisal tool for case reports and case series. Conflicts or missing data were automatically flagged and resolved by consensus discussion or by an independent reviewer.

Data Analysis

We performed a meta-analysis when three or more studies reported the same outcome for COVID-19 patients on ECMO. To reduce between-study heterogeneity in the primary meta-analyses, we only included studies with 95% or more of its patients on venovenous ECMO and explicitly-defined ARDS using the Berlin Criteria (5). We conducted meta-analyses of single proportions using a random-effects, generalized linear mixed model (GLMM) to determine the proportion of COVID-19 patients requiring venovenous ECMO that died in hospital and/or experienced neurologic complications. The GLMM method was chosen to account for studies with small sample sizes and zero events (6). We estimated the between-study variance using the maximum likelihood method. We compared the risk of mortality in patients with and without neurologic complications with a random-effects model using the Mantel-Haenszel method and the Paule-Mandel estimator for between-study variance. We performed a prespecified sensitivity analysis to include studies with 90% or more of its patients on venovenous ECMO. A subgroup analysis was performed where we stratified the studies by their anticoagulation protocol to determine the effect of anticoagulation targets on the incidence of ICH. All included studies in the subgroup analysis used unfractionated heparin infusions. Studies were designated as either “higher” or “similar” to the 2018 “ECMO to Rescue Lung Injury in Severe ARDS (EOLIA)” trial (anti-Xa levels 0.2–0.3 and/or activated partial thromboplastin time [aPTT] 40–55 s) (7). A higher target was defined by the lower range of anti-Xa greater than or equal to 0.3 or aPTT greater than or equal to 55 seconds, whereas a similar anticoagulation target was defined by an overlap with EOLIA targets. A meta-regression of the influence of ECMO duration on the proportion of neurologic outcomes was not performed due to a lack of studies consistently reporting this variable. All meta-analyses were performed in R Version 4.0.5 using the “meta” package (8, 9).

RESULTS

We included 54 studies (n = 3,347 patients) from the 840 titles identified using the search criteria (Fig. 1). All studies were observational by design, including 26 case reports or case series (48%), 24 cohort studies (44%), and four case-control studies (7%). There were no randomized control trials. There were 39 studies (72%) that used the Berlin criteria of ARDS (Supplementary 2, http://links.lww.com/CCX/B160).

Figure 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses diagram. Based on the search criteria, 840 titles and abstracts were identified. After removing duplicates and screening, 202 full-text studies were reviewed by two independent reviewers and 54 studies were included in the systematic review.

Of the 54 studies, 11 studies reported on ischemic strokes and 41 studies reported on ICHs. There were two case reports on mononeuritis multiplex (10, 11), one case report on brachial plexopathy (12), three case reports on critical illness myopathy or neuropathy (12–14), five studies on cerebral microbleeds (12, 15–18), and one case report of posterior reversible encephalopathy syndrome (18). Lastly, there were two studies that reported venous causes for neurovascular complications (19, 20) and two studies that used electroencephalogram to investigate the electrophysiologic changes of patients with COVID-19 on ECMO (21, 22). There were no reports of CNS infections, acquired demyelinating events, myelopathy, myelitis, or encephalitis in COVID-19 patients on ECMO.

There were 33 studies that reported the initial configuration of ECMO with a total of 3,172 patients. Approximately 96.8% of patients were on venovenous ECMO, 2.3% were on venoarterial ECMO, and 0.9% were on other types of ECMO (Supplementary 2, http://links.lww.com/CCX/B160).

Medical Interventions and Anticoagulation Practices

The use of corticosteroids, interleukin-6 inhibitors, and inhaled pulmonary vasodilators was reported in 38%, 5.6%, and 31% of patients, respectively. Approximately 74% and 67% of patients were paralyzed and placed in prone position, respectively, prior to ECMO cannulation (Supplementary 3, http://links.lww.com/CCX/B160).

Among the 32 studies that reported on the use of anticoagulation, 22 studies used only unfractionated heparin infusion, nine studies used a combination of anticoagulants, and one study did not specify choice of anticoagulant. There were 21 studies that included anticoagulation monitoring targets: nine studies used aPTT only, six studies used anti-Xa levels only, one study used aPTT ratio only, and five studies used a combination of the three anticoagulation markers. In most studies, the anti-Xa target level was between 0.3 and 0.7, and the aPTT target was between 50 and 75 seconds; however, the specific range varied significantly between studies. There were five studies that used a higher anticoagulation target (n = 605) and five studies that used a similar anticoagulation target to EOLIA (n = 164). Only one study reported adherence to the anticoagulation targets.

Meta-Analyses of Intracranial Hemorrhages in COVID-19 Patients on ECMO

There were 30 studies that reported the frequency of all types of ICH, and 17 studies were included in the initial meta-analysis. The results revealed that 11% (95% CI, 8–15%) of COVID-19 patients had an ICH on venovenous ECMO (Fig. 2A). There was considerable between-study heterogeneity with I² = 77%.

Figure 2.

Meta-analysis of frequency of intracerebral hemorrhage (ICH) in patients with COVID-19 on venovenous extracorporeal membrane oxygenation (ECMO). Studies with 95% or more of its patients on venovenous ECMO were included in the meta-analysis. A 5% exception was arbitrarily selected to minimize the negligible effects of other types of ECMO on the overall analysis. A, Seventeen studies were included in the initial meta-analysis of ICH frequency with considerable between-study heterogeneity of I² = 77% (p < 0.01). The ICH frequency was estimated to be 11% (95% CI, 8–15%). B, A prespecified sensitivity analysis was performed with studies with 90% or more of its patients on venovenous ECMO. Twenty-two studies were included with considerable between-study heterogeneity of I² = 74% (p < 0.01). The sensitivity analysis reaffirmed that the frequency of ICH was 11% with a slight increase in precision (95% CI, 8–14%).

In studies where neuroimaging was routinely performed, the frequency of ICH ranged from 11% to 33% (19, 22–25). The most common types of ICH were intraparenchymal and subarachnoid hemorrhages, which accounted for 73% (163/273) and 8% (14/273), respectively, of all reported ICHs in cohort and case-control studies.

There were 22 studies that reported the frequency of intraparenchymal hemorrhages, and 14 studies were included in the initial meta-analysis. The frequency of intraparenchymal hemorrhage was 10% (95% CI, 7–15%), and there was considerable between-study heterogeneity with I² = 73% (Supplementary 4A, http://links.lww.com/CCX/B160).

There were 13 studies that described or included the features of the intraparenchymal hemorrhage (Supplementary 5, http://links.lww.com/CCX/B160). There were 14 reports of lobar hemorrhage, compared with five reports of subcortical hemorrhage. There were also nine reports of intraventricular extension, seven reports of multifocal ICH, and seven reports of fluid levels within the region of hemorrhage.

Meta-Analysis of Ischemic Stroke in COVID-19 Patients on ECMO

There were 19 studies that reported the frequency of ischemic stroke, and 10 studies were included in the initial meta-analysis. The estimated frequency of ischemic stroke was 2% (95% CI, 1–3%). The between-study heterogeneity was 40% (Fig. 3A).

Figure 3.

Meta-analysis of frequency of ischemic stroke in patients with COVID-19 on venovenous extracorporeal membrane oxygenation (ECMO). Studies with 95% or more of its patients on venovenous ECMO were included in the meta-analysis. A 5% exception was arbitrarily selected to minimize the negligible effects of other types of ECMO on the overall analysis. A, Eleven studies were included in the initial meta-analysis of ischemic stroke frequency with a heterogeneity of I² = 40% (p = 0.08). The frequency of ischemic stroke was estimated to be 2% (95% CI, 1–3%). B, A prespecified sensitivity analysis was performed with studies with 90% or more of its patients on venovenous ECMO. Fifteen studies were included with a heterogeneity of I² = 18% (p = 0.25). The sensitivity analysis reaffirmed that the freqency of ischemic stroke was 2% (95% CI, 1–3%).

Meta-Analysis on Anticoagulation Targets in COVID-19 Patients on ECMO

There were five studies that used a higher anticoagulation target than EOLIA (n = 605) and five studies that used a similar anticoagulation target to EOLIA (n = 164). In studies with a higher anticoagulation target, the frequency of ICH was 9% (95% CI, 5–13%) with considerable between-study heterogeneity of 68% (Fig. 4). In studies with a similar anticoagulation target, the frequency of ICH was 17% (95% CI, 10–27%) and the between-study heterogeneity was 34% (Fig. 4). There was no statistically significant difference in the frequency of ICH between higher and similar anticoagulation targets (p = 0.06).

Figure 4.

The frequency of intracranial hemorrhage (ICH) with higher and similar anticoagulation targets as the “ECMO to Rescue Lung Injury in Severe ARDS (EOLIA)” trial. The 2018 EOLIA trial targeted anti-Xa levels between 0.2 and 0.3 and activated partial thromboplastin time (aPTT) levels between 40 and 55 s. COVID-19 patients on IV heparin for venovenous extracorporeal membrane oxygenation with specified anticoagulation targets were included. Studies were designated as either “higher” or “similar” to the 2018 EOLIA trial, where a higher anticoagulation target is defined by the lower range of anti-Xa greater than or equal to 0.3 and/or aPTT greater than or equal to 55 s. There were five studies that used a higher target (n = 605) and five studies that used a similar target to EOLIA (n = 164). The frequency of ICH in studies with a higher anticoagulation target was 9% (95% CI, 5–13%) with substantial between-study heterogeneity (I² = 68%). The frequency of ICH in studies with a similar anticoagulation target was 17% (95% CI, 10–27%) with moderate between-study heterogeneity (I² = 34%). There was no difference in the frequency of ICH between the two anticoagulation targets (p = 0.06). df = degrees of freedom.

Meta-Analysis of Mortality in COVID-19 Patients on ECMO

Twenty-six studies (n = 2,769) reported overall mortality in COVID-19 patients on all types of ECMO; the total number of reported deaths was 1,193 out of 2,769 patients. The leading causes of mortality are multiple organ failure (41%), sepsis (20%), and neurologic complications (15%). Neurologic cause of mortality ranks as the third greatest cause of death in COVID-19 patients on ECMO (Supplementary 6, http://links.lww.com/CCX/B160).

Eight studies were included in the meta-analysis of in-hospital mortality in COVID-19 patients on venovenous ECMO. The in-hospital mortality was 37% (95% CI, 34–40%) and the between-study heterogeneity was low with I² = 0% (Fig. 5).

Figure 5.

Meta-analysis of in-hospital mortality in patients with COVID-19 on venovenous extracorporeal membrane oxygenation (ECMO). Studies with 95% or more of its patients on venovenous ECMO were included in the meta-analysis. A 5% exception was arbitrarily selected due to negligible effects of other types of ECMO on the overall analysis. A, Seven studies were included. The in-hospital mortality of COVID-19 patients requiring venovenous ECMO was 37% (95% CI, 34–40%) with low between-study heterogeneity (I² = 0%; p = 0.63). B, A prespecified sensitivity analysis was performed with studies with 90% or more of its patients on venovenous ECMO. Nine studies were included. The sensitivity analysis revealed higher in-hospital mortality and reduced accuracy of results, including increased heterogeneity (I² = 65%; p < 0.01) and a wider 95% CI of 35–50%.

There was a higher in-hospital mortality rate of 84.9% (95% CI, 72–93%) among patients with neurologic complications. Further meta-analysis demonstrated that the risk ratio of mortality in COVID-19 patients on venovenous ECMO with neurologic complications compared with those without neurologic complications was 2.24 (95% CI, 1.46–3.46) (Fig. 6).

Figure 6.

Risk of mortality in COVID-19 patients with or without neurologic complications during venovenous extracorporeal membrane oxygenation (ECMO). Studies with 95% or more of its patients on venovenous ECMO were included in the meta-analysis. A 5% exception was arbitrarily selected to minimize the negligible effects of other types of ECMO on the overall analysis. The risk ratio for mortality was 2.24 (95% CI, 1.46–3.46) in patients with neurologic complications.

Sensitivity Analyses of Neurologic Complications in COVID-19 Patients on ECMO

The proportion of patients with an ICH (Fig. 2B), intraparenchymal hemorrhage (Supplementary 4B, http://links.lww.com/CCX/B160), and ischemic stroke (Fig. 3B) did not differ in sensitivity analyses where the inclusion criteria for meta-analysis included studies with 90% or more of patients on venovenous ECMO. The between-study variances were also similar.

The sensitivity analysis using a reduced threshold of 90% ECMO revealed an increased in-hospital mortality of 42% (95% CI, 36–49%) and the between-study heterogeneity was 67% (Fig. 4B).

Quality of Studies

The quality of the studies were widely heterogeneous. Based on the NOS for cohort and case-control studies, 16 (57%) studies were considered high quality and 12 (43%) were considered low quality (Supplementary 7, http://links.lww.com/CCX/B160).

Based on the JBI critical appraisal tool, five (31%) case reports were deemed high quality, four (25%) were fair quality, and seven (44%) were low quality (Supplementary 8, http://links.lww.com/CCX/B160). Among the 10 case series, the median number of responses marked as “low risk of bias” was 6 (interquartile range, 4–8). Many case series did not describe clear inclusion criteria, report complete or consecutive inclusion of patients, or include details of setting (Supplementary 9, http://links.lww.com/CCX/B160).

DISCUSSION

Our systematic review and meta-analysis summarized the frequency and type of neurologic complication in COVID-19 patients requiring ECMO. We identified 54 studies: the majority of patients were on venovenous ECMO, and most studies reported the frequency of ICH and ischemic strokes. ICH was the most common neurologic complication (11% of patients), followed by ischemic stroke (2% of patients). To help benchmark these findings, To help benchmark these findings, recent literature reported 0.9% of COVID-19 cohorts that did not require ECMO had an ICH, Moreover, previously published literature reported that 8% and 2% of non-COVID-19 cohorts requring ECMO had an ICH and ischemic stroke, respectively (26–29). Less commonly reported neurologic complications included cerebral microbleeds, mononeuritis multiplex, and posterior reversible encephalopathy syndrome. In keeping with the recently identified association between bleeding and mortality in COVID-19 patients requiring ECMO (30), we found that patients with neurologic complications had more than a two-fold increase in the risk of death compared with those without neurologic complications.

Intraparenchymal hemorrhage accounted for more than two-thirds of all ICHs, and there were frequent reports of fluid levels and multifocal ICH (31–35). Fluid levels are suggestive of impaired clot formation, either from pathologic coagulopathy or iatrogenic anticoagulation, and are associated with hematoma expansion (36–38). Similarly, multifocal hemorrhages are associated with coagulopathy, typically in the context of systemic anticoagulation (39). The high frequency of ICH in COVID-19 patients is likely multifactorial, and potential contributors include COVID-19–associated endotheliopathy and higher anticoagulation targets (1). The choice of higher anticoagulation targets was likely dictated by earlier reports of a high thrombotic rate among COVID-19 patients and its increase in mortality (40, 41). This is further supported by the higher prevalence of catheter-associated thrombosis in one included study (19). However, the development of neurologic complications—namely, ICHs—posed an even higher risk of mortality than previously reported with thrombotic events (42), and the balance between these risks should be carefully considered in critically ill COVID-19 patients.

Interestingly, the similar frequency of ICH between higher and standard anticoagulation targets may suggest that higher anticoagulation targets do not significantly increase the risk of ICH. However, interpretation of this subgroup analysis is limited by its wide CIs, which reduces the certainty of this finding. Elevated d-dimer and fibrinogen levels found in many of these patients may represent early signs of disruption to the coagulation cascade, and autopsy findings of cerebral vasculopathy further support the role of COVID-19–associated endotheliopathy (32–34). Recent studies demonstrated benefit of therapeutic anticoagulation in patients with moderate COVID-19 infections, but this benefit was not seen in critical or severe COVID-19 infections (43, 44). In fact, there is increasing evidence that venovenous ECMO without anticoagulation or with low-dose systemic anticoagulation may be associated with fewer hemorrhagic complications and a similar frequency of ECMO circuit thrombosis (45, 46). Future well-designed studies are needed to assess whether difference in anticoagulation targets is associated with the incidence of ICH in COVID-19 patients requiring venovenous ECMO.

An important, but less commonly reported, neurologic complication is the development of cerebral microbleeds in COVID-19 patients requiring ECMO. Characterized as small hypointense foci on gradient echo sequences or susceptibility weighted imaging, cerebral microbleeds associated with ECMO are typically found in the juxtacortical white matter and splenium of the corpus callosum. Three case reports and one case-control study noted cerebral microbleeds on MRI (12, 15–17). There are several proposed mechanisms for ECMO-associated microbleeds, including hypoxia-induced microbleeds, gas emboli from ECMO, or infection-mediated microangiopathy. Lersy et al (15) noted that the severity of COVID-19 ARDS may be associated with presence of microbleeds, which may be an independent risk factor for ICH. Furthermore, multiple microbleeds are associated with cognitive dysfunction and increases the risk of dementia (47). While long-term data on microbleeds is not available in COVID-19 patients requiring ECMO, clinical outcomes are overall less favorable in COVID-19 patients with microbleeds (48).

Limitations

Our study had several limitations. The quality of the available studies was variable, with approximately half of the studies rated as good quality and the other half rated as poor quality. At the time of literature review, there were no randomized controlled trials that investigated the effects of anticoagulation strategies on the frequency and nature of neurologic complications in COVID-19 patients on ECMO. The small number of studies and their small sample sizes resulted in less precision and greater uncertainty in the meta-analysis, as showcased by the wide CIs.

Included studies were heterogeneous in study size and design (case reports, case series, cohort studies, and case-control studies), interventions (venoarterial ECMO and venovenous ECMO), other medical interventions (e.g., anticoagulation targets), and patient outcomes (e.g., ICH is a composite outcome that encompasses subdural, subarachnoid and intraparenchymal hemorrhages), which reduces certainty in the interpretation of our results. To reduce heterogeneity, the meta-analyses only included studies where 1) patients were explicitly diagnosed with ARDS and 2) at least 95% of its patients on venovenous ECMO. While these stipulations reduced heterogeneity in meta-analyses of ischemic stroke and mortality, considerable heterogeneity remained a limiting factor in meta-analyses where ICH was the outcome of interest. Only 10 studies were included in subgroup analysis on anticoagulation targets, rendering the results to be less precise with wide CIs.

Lastly, only studies published in English were included, which may lead to a systematic omission of data published in other languages. While we extracted data on venoarterial ECMO patients, there were not enough data for meta-analysis of venoarterial ECMO outcomes. Previous meta-analysis in non-COVID-19 patients showed that venoarterial ECMO, compared with venovenous ECMO, carried a similar risk of intraparenchymal hemorrhage, but there was a higher rate of ischemic stroke and brain death (28).

Future Directions

Our results suggest that COVID-19 patients requiring venovenous ECMO have a high frequency of ICH. Whether this high risk of bleeding is associated with higher anticoagulation targets, COVID-19 endothelial injury, or coagulopathy warrants further investigation. Future studies should compare whether the choice of anticoagulation targets impacts bleeding and thrombosis rates and whether the degree of endothelial injury and inflammation contribute to the risk of bleeding in both COVID-19 and non-COVID-19 patients requiring ECMO.

CONCLUSIONS

COVID-19 patients requiring venovenous ECMO have a high frequency of ICH. Intraparenchymal hemorrhage was the most common type of ICH, and neurologic complications more than doubled the risk of death in COVID-19 patients requiring venovenous ECMO. Healthcare providers should be aware of these increased risks and maintain a high index of suspicion for ICH. Future studies should confirm whether the choice of anticoagulation targets influences the rate of neurologic complications in this patient population.

ACKNOWLEDGMENTS

We thank Alla Iansavitchene for her assistance in systematic literature search.