Abstract
Objective:
Stroke is commonly reported in patients receiving veno-venous extracorporeal membrane oxygenation (V-V ECMO), but risk factors are not well described. We sought to determine pre-ECMO and on-ECMO risk factors for both ischemic and hemorrhagic stroke in patients with V-V ECMO support.
Design:
Retrospective analysis.
Setting:
Data reported to the Extracorporeal Life Support Organization by 366 ECMO centers from 2013 to 2019.
Patients:
Patients older than 18 years old supported with a single run of V-V ECMO.
Interventions:
None.
Measurements and Main Results:
Of 15,872 V-V ECMO patients, 812 (5.1%) had at least one type of acute brain injury (ABI), defined as ischemic stroke, hemorrhagic stroke, or brain death. Overall, 215 (1.4%) experienced ischemic stroke and 484 (3.1%) experienced hemorrhagic stroke. Overall in-hospital mortality was 36%, but rates were higher in those with ischemic or hemorrhagic stroke (68% and 73%, respectively). In multivariable analysis, pre-ECMO pH (adjusted OR [aOR]=0.10; 95% CI, 0.03–0.35; p<0.001), hemolysis (aOR=2.27; 95% CI=1.22–4.24; p=0.010), gastrointestinal hemorrhage (aOR=2.01; 95% CI=1.12–3.59; p=0.019), and disseminated intravascular coagulation (aOR=3.61; 95% CI=1.51–8.66; p=0.004) were independently associated with ischemic stroke. Pre-ECMO pH (aOR=0.28; 95% CI, 0.12–0.65; p=0.003), pre-ECMO PO2 (aOR=0.96; 95% CI=0.93–0.99; p=0.021), gastrointestinal hemorrhage (aOR=1.70; 95% CI=1.15–2.51; p=0.008), and renal replacement therapy (aOR=1.57; 95% CI=1.22–2.02; p<0.001) were independently associated with hemorrhagic stroke.
Conclusions:
Among V-V ECMO patients in the ELSO registry, approximately 5% had ABI. Mortality rates increased 2-fold when ischemic or hemorrhagic strokes occurred. Risk factors such as lower pH and hypoxemia during the peri-cannulation period and markers of coagulation disturbances were associated with ABI. Further research on understanding pre-ECMO and on-ECMO risk factors and the timing of ABI is necessary to develop appropriate prevention and management strategies.
Keywords: Extracorporeal membrane oxygenation, venovenous ECMO, ischemic stroke, hemorrhagic stroke, brain injury, brain death
INTRODUCTION
Veno-venous extracorporeal membrane oxygenation (V-V ECMO) is a life-saving intervention that provides pulmonary support for patients with severe acute respiratory failure refractory to mechanical ventilation. The use of V-V ECMO has increased in recent years1 and has reduced 60-day mortality in patients with acute respiratory distress syndrome (ARDS) when compared to that in patients who receive conventional mechanical ventilation2. Prior analysis of Extracorporeal Life Support Organization (ELSO) registry data showed that the presence of neurologic complications or acute brain injury (ABI, defined as ischemic stroke, intracranial hemorrhage, and seizure) increases the risk of poor functional outcome and mortality from ECMO3. In that study, 7% of adults with V-V ECMO experienced neurologic complications, with an associated mortality of 80% for intracranial hemorrhage and 68% for ischemic stroke3.
Stroke is often perceived to be less common in patients who receive V-V ECMO than in those who receive veno-arterial (V-A) ECMO, as the V-V ECMO circuit does not involve the arterial circulation and V-V ECMO is often run with less or even without anticoagulation, compared to V-A ECMO. However, in a recent publication, ABI appears to occur with equal frequency regardless of ECMO type4. Furthermore, when V-V ECMO patients underwent routine computed tomography (CT) scans within 24 hours of cannulation, 16% had hemorrhagic strokes5. In the population who underwent postmortem neuropathologic evaluation, 68% of ECMO non-survivors exhibited ABI, suggesting that the true prevalence of ABI is likely higher than that previously reported6. Currently, there is a large gap in knowledge regarding the prevalence, risk factors, timing, and management of ECMO-related ABI4,7. Understanding mechanisms and risk factors of ischemic and hemorrhagic stroke in patients receiving V-V ECMO will provide crucial information to guide prevention and inform appropriate management strategies.
The aim of this study was to identify risk factors for ischemic and hemorrhagic stroke in patients receiving V-V ECMO support in an effort to provide an understanding of potential mechanisms for each stroke subtype. We hypothesized that ischemic stroke and hemorrhagic stroke have different pre-ECMO and on-ECMO risk factors.
Materials and Methods
Study Design and Population
The ELSO registry is a voluntary database that collects information on use, indications, complications, and outcomes of ECMO support in adults and children from 366 member centers worldwide8. Records stored in the registry include patient demographics; clinical characteristics; pre-ECMO conditions, including pulmonary diagnoses; hemodynamic and laboratory values collected before and during ECMO; complications that occur during ECMO support, including neurologic complications; and outcome data such as survival to hospital discharge. Diagnosis and medical history are reported according to the International Classification of Diseases (ICD) 9th edition (ICD-9) and 10th (ICD-10) edition codes.
We carried out a retrospective analysis of the ELSO registry database from 2013 through 2019. We included only patients 18 years and older who received one run of V-V ECMO support for respiratory failure. We excluded patients who underwent multiple runs to avoid complexity and bias in the analysis resulting from severe illness, as we considered that the morbidity during the inter-ECMO periods could introduce significant confounders. Patients treated with V-A ECMO were excluded.
Data Collection and Definitions
For all patients in the study, we extracted the following information from the ELSO registry database: pre-ECMO demographic information; pre-ECMO clinical variables; cardiac diagnoses; pulmonary diagnoses; laboratory values; on-ECMO clinical variables, including hemodynamic data and the use of other supportive cardiac devices; and ECMO-associated morbidity and mortality, including neurologic complications. The pre-extracorporeal life support (ECLS) hemodynamics and arterial blood gas (ABG) values were measured no more than 6 hours before initiation of ECLS; if the database contained multiple values, we chose those closest to the time of cannulation. Twenty-four–hour ABG values were drawn between 18 and 30 hours after ECLS start time. The “ECMO circuit mechanical failure” variable included oxygenator failure, cracks in pigtail connectors, clots in the hemofilter, clots in the circuit component, circuit change, cannula problems, air in the circuit, pump failure, and tubing rupture.
Neurologic complications included ischemic stroke, hemorrhagic stroke, brain death, and seizure. In the ELSO database, ischemic stroke is defined as central nervous system (CNS) infarction determined by ultrasound, CT, or magnetic resonance imaging (MRI). Hemorrhagic stroke is defined as intra- or extra-parenchymal CNS hemorrhage determined by ultrasound, CT, or MRI. The ELSO database has two seizure variables: those determined by clinical assessment and those confirmed by electroencephalogram. We combined these two variables as “seizure” for the analysis. Although new neurologic variables, including “CNS diffuse ischemia,” “intraventricular hemorrhage,” and “neurosurgical intervention performed,” recently have been added to the ESLO database addendum, these were not available in the current analysis.Additionally, subtypes of hemorrhagic stroke such as intracerebral hemorrhage, subarachnoid hemorrhage, and subdural hematoma were not available in the registry. This study was reviewed by the Johns Hopkins Institutional Review Board (IRB) and waived the need for approval.
Statistical Analysis
Demographic and clinical variables in subjects with and without each type of neurologic injury (ischemic stroke, hemorrhagic stroke, and brain death) were compared by unpaired t-test, chi-squared test, or Mann-Whitney U test as appropriate. Results are expressed as the mean with standard deviation if normally distributed or as the median with quartiles if non-normally distributed for quantitative variables and as proportions for categorical findings. Ordinal or categorical data are reported as numbers and percentages. A p value <0.05 was considered statistically significant. Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated by logistic regression analysis. We carefully selected clinically relevant variables that had biologically plausible associations or causality for inclusion in multivariable logistic regression analysis to identify risk factors that were associated with each of ischemic and hemorrhagic stroke compared to patients with no known stroke. Akaike information criterion was calculated for selecting the best model. All analyses were carried out in STATA 15 (College Station, TX, USA).
RESULTS
A total of 15,872 patients (median age 49 years, 61% male) who received V-V ECMO support were included in the study. The use of V-V ECMO increased over time 1,179 in 2013 to 3,011 in 2019, representing a 2.5-fold increase in 7 years (Table 1). Of the 15,872 patients, 812 (5.1%) had at least one type of ABI, including ischemic stroke, hemorrhagic stroke, or brain death (Fig. 1). Overall, 215 (1.4%) patients had ischemic stroke and 484 (3.1%) had hemorrhagic stroke. Brain death was reported in 204 (1.3%) patients. Although the rates of V-V ECMO–associated ischemic stroke remained stable over 7 years (p=0.18), the rates of hemorrhagic stroke and brain death decreased over time (p<0.001 and p=0.002, respectively; Table 1 and Supplemental Table 1). Overall, ABI during V-V ECMO decreased from 6.2% in 2013 to 5.1% in 2019 (Fig. 1). The mortality for all ECMO patients was 36% over 7 years, a rate that increased to 68% in the presence of ischemic stroke and 73% in the presence of hemorrhagic stroke. Overall mortality remained stable over time from 38% to 36% as did mortality due to ABI (77% to 74%) during the study period.
Table 1.
Baseline demographics and ECMO variables in V-V ECMO patients.
| Variables | Without any stroke (n=15,206) | Ischemic stroke (n=215) | P-value | Hemorrhagic stroke (n=484) | P-value |
|---|---|---|---|---|---|
|
| |||||
| Age (years), median (IQR)* | 48 (35–60) | 50 (36–61) | 0.19 | 49 (37–59) | 0.40 |
| Weight (kg), median (IQR)* | 85 (70–103) | 85 (72–100) | 0.75 | 80 (66–100) | 0.002 |
| Male* | 9,108 (61) | 125 (59) | 0.46 | 271 (56) | 0.03 |
| Race* | |||||
| White | 8,633 (58) | 132 (63) | 0.009 | 277 (58) | <0.001 |
| Asian | 1,841 (12) | 26 (12) | 90 (19) | ||
| Black | 1,710 (11) | 17 (8) | 29 (6) | ||
| Hispanic | 1,027 (7) | 20 (10) | 41 (9) | ||
| Others | 1,667 (11) | 15 (7) | 38 (14) | ||
| Year ECLS | |||||
| 2013 | 1,117 (7) | 15 (7) | 0.07 | 49 (10) | <0.001 |
| 2014 | 1,537 (10) | 35 (16) | 59 (12) | ||
| 2015 | 1,746 (11) | 26 (12) | 67 (14) | ||
| 2016 | 2,361 (16) | 35 (16) | 93 (19) | ||
| 2017 | 2,594 (17) | 31 (14) | 60 (12) | ||
| 2018 | 2,915 (19) | 42 (20) | 109 (23) | ||
| 2019 | 2,936 (19) | 31 (14) | 47 (10) | ||
| ECMO indication | |||||
| Acute respiratory distress syndrome | 2,534 (17) | 41 (19) | 0.35 | 94 (19) | 0.11 |
| Influenza | 931 (6) | 12 (6) | 0.74 | 60 (12) | <0.001 |
| Viral Pneumonia | 465 (3) | 7 (3) | 0.87 | 25 (5) | 0.009 |
| Bacterial Pneumonia | 678 (4) | 9 (4) | 0.85 | 30 (6) | 0.07 |
| Chronic obstructive pulmonary disease | 479 (3) | 5 (2) | 0.49 | 4 (1) | 0.004 |
| Asthma | 387 (3) | 4 (2) | 0.53 | 3 (1) | 0.007 |
| Pulmonary edema | 20 (0.1) | 0 (0) | 0.60 | 0 | 0.43 |
| Pulmonary embolism | 93 (1) | 5 (2) | 0.002 | 2 (0.4) | 0.58 |
| Acute respiratory failure | 3,256 (21) | 49 (23) | 0.63 | 79 (16) | 0.007 |
| Pneumonitis | 175 (1) | 1 (0.5) | 0.35 | 3 (1) | 0.28 |
| Pre-ECMO ABG | |||||
| pH*° | 7.24 (7.14–7.33) | 7.20 (7.09–7.28) | <0.001 | 7.22 (7.14–7.30) | 0.002 |
| pCO2, mmHg*° | 58 (47–74) | 64 (49–79) | 0.01 | 62 (51–80) | <0.001 |
| pO2, mmHg*° | 63 (50–80) | 57 (44–79) | 0.002 | 60 (48–75) | <0.001 |
| HCO3, mmol/L* | 24.0 (20.0–29.0) | 23.7 (20.0–28.9) | 0.37 | 24.7 (21.0–29.5) | 0.08 |
| Mixed venous oxygen saturation* | 66 (55–77) | 60 (51–76) | 0.46 | 67 (58–78) | 0.90 |
| Cardiac index, (L/min/m2)* | 3.0 (2.3–3.9) | 3.2 (2.8–4.3) | 0.44 | 2.8 (2.2–4.5) | 0.87 |
| Pre-ECMO hemodynamics | |||||
| Systolic blood pressure, mmHg* | 106 (90–125) | 103 (88–125) | 0.43 | 105 (90–122) | 0.40 |
| Diastolic blood pressure, mmHg* | 58 (50–68) | 57 (48–67) | 0.23 | 57 (48–66) | 0.07 |
| Mean blood pressure, mmHg* | 72 (62–84) | 68 (59–82) | 0.03 | 71 (62–81) | 0.37 |
| ABG at 24 hours | |||||
| pH*° | 7.40 (7.35–7.44) | 7.40 (7.35–7.44) | 0.75 | 7.40 (7.35–7.44) | 0.91 |
| pCO2, mmHg*° | 41 (37–46) | 41 (36–46) | 0.41 | 42 (37–46) | 0.22 |
| pO2, mmHg*° | 80 (65–108) | 81 (61–107) | 0.58 | 74 (60–96) | <0.001 |
| HCO3, mmol/L* | 25.0 (22.0–28.2) | 24.2 (22.0–28.0) | 0.14 | 24.9 (22.0–28.9) | 0.72 |
| Days on ECMO support* | 7.8 (4.0–14.1) | 7.6 (3.3–15.7) | 0.66 | 8.6 (4.0–15.8) | 0.037 |
| ECMO complications | |||||
| ECMO circuit mechanical failure | 3,398 (22) | 55 (26) | 0.26 | 107 (22) | 0.90 |
| Renal replacement therapy | 4,171 (27) | 86 (40) | <0.001 | 204 (42) | <0.001 |
| Glucose < 40 mg/L | 142 (1) | 5 (2) | 0.037 | 11 (2) | 0.003 |
| Hemolysis (plasma hemoglobin >50 mg/dL) | 573 (4) | 19 (9) | <0.001 | 31 (6) | 0.003 |
| Cardiac arrhythmia | 1,068 (8) | 25 (12) | 0.009 | 62 (13) | <0.001 |
| Gastrointestinal hemorrhage | 721 (5) | 23 (11) | <0.001 | 46 (10) | <0.001 |
| Disseminated intravascular coagulation | 205 (1) | 10 (5) | <0.001 | 12 (2) | 0.036 |
| Seizure | 146 (1) | 15 (7) | <0.001 | 17 (4) | <0.001 |
All data are presented as n (%) for categorical variables and median (interquartile range) for continuous variables.
ECMO = extracorporeal membrane oxygenation; IQR = interquartile range; V-V = veno-venous; ABG = arterial blood gas.
Bold text indicates statistical significance.
Missing values for some variables resulted in different denominators for the following variables: age (n=15,814), weight (n=15,092), sex (n=15,367), race (n=15,530), pH (n=13,284), pCO2 (n=12,294), pO2 (n=13,017), HCO3 (n=12,444), mixed venous oxygen saturation (n=1,227), cardiac index (n=680), systolic blood pressure (n=11,719), diastolic blood pressure (n=11,661), mean blood pressure (n=10,627), 24 hour pH (n=14,022), 24 hour pCO2 (13,017), 24 hour pO2 (n=13,695), 24 hour HCO3 (n=13,762) days on ECMO support (n=15,742).
The following values were excluded due to biological implausibility: pH>8, pH<6.5, pO2>760 mmHg, pCO2>240 mmHg, pCO2<20 mmHg.
Figure 1.
The incidence of acute brain injury in patients treated with venovenous extracorporeal membrane oxygenation, 2013–2019. Data were collected from the Extracorporeal Life Support Organization registry.
Associations with ABI: Univariate analysis
Among the 215 patients who had ischemic strokes, the median age was 50 and 59% were male. Compared to patients with no stroke, patients who exhibited a higher pre-ECMO PCO2, lower pre-ECMO pH, PO2, and mean arterial pressure (MAP) were more likely to have ischemic stroke. During ECMO support, higher rates of the following ECMO complications were associated with ischemic stroke: renal replacement therapy (RRT), hypoglycemia, hemolysis, cardiac arrhythmia, gastrointestinal hemorrhage, disseminated intravascular coagulation (DIC), and seizure (Table 1).
Compared to patients without stroke, the 484 patients with hemorrhagic stroke (median age 49 years, 56% male) had lower weight, lower pre-ECMO PO2, higher pre-ECMO PCO2, lower pH and PO2 on ECMO, longer duration of ECMO support, and higher rates of ECMO complications including RRT, hypoglycemia, hemolysis, cardiac arrhythmia, gastrointestinal hemorrhage, DIC, and seizure (Table 1).
Compared to patients without brain death, patients with brain death were younger, had lower pre-ECMO pH and HCO3, higher pre-ECMO PCO2, and lower pre-ECMO blood pressure values. They also had a lower pH and HCO3 on ECMO, higher pCO2 on ECMO, shorter duration of ECMO support, and a higher incidence of ECMO complications including RRT, DIC and seizure, than did patients without brain death (Supplemental Table 1).
Pre-ECMO PO2 values were lower among those who had ischemic stroke than among those without stroke (57 vs. 63 mmHg, p=0.002). Similarly, pre-ECMO PO2 values were lower in patients with hemorrhagic stroke than in those without (60 vs. 63 mmHg, p<0.001). Supplemental Figure 1 demonstrates that there are more patients with hypoxemia in the hemorrhagic stroke group vs. without (B). Even at 24 hours of ECMO support, lower PO2 values were observed in patients with hemorrhagic stroke (74 vs. 80 mmHg, p<0.001).
Associations with ABI: Multivariable analysis
In multivariable analysis, pre-ECMO pH (adjusted OR [aOR]=0.10; 95% CI, 0.03–0.35; p<0.001), hemolysis (aOR=2.27; 95% CI=1.22–4.24; p=0.010), gastrointestinal hemorrhage (aOR=2.01; 95% CI=1.12–3.59; p=0.019), and disseminated intravascular coagulation (aOR=3.61; 95% CI=1.51–8.66; p=0.004) were independently associated with ischemic stroke (Table 2). Pre-ECMO pH (aOR=0.28; 95% CI, 0.12–0.65; p=0.003), pre-ECMO PO2 (aOR=0.96; 95% CI=0.93–0.99; p=0.021), gastrointestinal hemorrhage (aOR=1.70; 95% CI=1.15–2.51; p=0.008), and renal replacement therapy (aOR=1.57; 95% CI=1.22–2.02; p<0.001) were independently associated with in hemorrhagic stroke (Table 3).
Table 2.
Multivariable analysis of ischemic stroke in patients with V-V ECMO
| Ischemic stroke | Odds Ratio | 95% C.I. | P-value |
|---|---|---|---|
|
| |||
| Pre-ECMO | |||
| pH | 0.10 | 0.03–0.35 | <0.001 |
| pO2 (per 10 mmHg) | 1.01 | 0.99–1.04 | 0.311 |
| Mean blood pressure, mmHg | 1.00 | 0.99–1.01 | 0.878 |
| ECMO course | |||
| pCO2 at 24 hours (per 10 mmHg) | 1.03 | 0.90–1.17 | 0.664 |
| pO2 at 24 hours (per 10 mmHg) | 1.02 | 0.99–1.04 | 0.214 |
| HCO3 at 24 hours (per 10 mmHg) | 1.00 | 0.96–1.04 | 0.937 |
| Renal replacement therapy | 1.26 | 0.83–1.84 | 0.296 |
| Glucose < 40 mg/L | 0.73 | 0.10–5.53 | 0.764 |
| Hemolysis (plasma hemoglobin >50 mg/dL) | 2.27 | 1.22–4.24 | 0.010 |
| Cardiac arrhythmia | 0.68 | 0.33–1.43 | 0.312 |
| Gastrointestinal hemorrhage | 2.01 | 1.12–3.59 | 0.019 |
| Disseminated intravascular coagulation | 3.61 | 1.51–8.66 | 0.004 |
VV = veno-venous; ECMO = extracorporeal membrane oxygenation; C.I. = Confidence Interval.
Bold text indicates statistical significance.
Table 3.
Multivariable analysis of hemorrhagic stroke in patients with V-V ECMO
| Hemorrhagic stroke | Odds Ratio | 95% C.I. | P-value |
|---|---|---|---|
|
| |||
| Pre-ECMO | |||
| pH | 0.28 | 0.12–0.65 | 0.003 |
| pO2 (per 10 mmHg) | 0.96 | 0.93–0.99 | 0.021 |
| Mean blood pressure, mmHg | 1.00 | 0.99–1.01 | 0.549 |
| ECMO course | |||
| pCO2 at 24 hours (per 10 mmHg) | 1.06 | 0.92–1.22 | 0.438 |
| pO2 at 24 hours (per 10 mmHg) | 0.99 | 0.96–1.01 | 0.280 |
| HCO3 at 24 hours (per 10 mmHg) | 1.02 | 0.99–1.04 | 0.204 |
| Renal replacement therapy | 1.57 | 1.22–2.02 | <0.001 |
| Glucose < 40 mg/L | 2.12 | 0.83–5.41 | 0.116 |
| Hemolysis (plasma hemoglobin >50 mg/dL) | 1.08 | 0.64–1.82 | 0.776 |
| Cardiac arrhythmia | 1.11 | 0.74–1.65 | 0.612 |
| Gastrointestinal hemorrhage | 1.70 | 1.15–2.51 | 0.008 |
V-V = veno-venous; ECMO = extracorporeal membrane oxygenation; C.I. = Confidence Interval.
Bold text indicates statistical significance.
DISCUSSION
Incidence and Mortality
Our analysis of 15,872 V-V ECMO patients in the ELSO registry showed that approximately 5% had at least one type of ABI, defined as ischemic stroke, hemorrhagic stroke, or brain death. During the study period, reported ABI rates during V-V ECMO remained stable at 5–6%. Although mortality was 36% overall in V-V ECMO patients, not surprisingly, the mortality rate was doubled when ischemic stroke or hemorrhagic stroke was present, as reported previously3,4,7, Mortality associated with ischemic or hemorrhagic stroke remained stable over 7 years. Notably, the observed rates of ischemic stroke and hemorrhagic stroke in the ELSO database were lower (at 1.4% and 3.1%, respectively) than those reported in prior literature (4–14% for ischemic stroke; 4–19% for hemorrhagic stroke)4,5,9–13. However, this perceived low rate of stroke determined by our study is likely an underestimation because ELSO data is reported voluntarily without central or local adjudication of neurological diagnoses, and CT scans of the brain were not mandated to make a stroke diagnosis in the registry. Furthermore, CT scans are insensitive for detecting early cerebral ischemic changes and posterior fossa ischemia. More dramatic was the finding in two recent brain autopsy studies that 69–90% of ECMO non-survivors had ABI on postmortem pathological evaluation6,14, suggesting the true prevalence of ABI is likely much higher than that reported in clinical observational studies. This underscores the importance of protocolized neuromonitoring and systematic neurological care in these patients if we are to better understand the neurologic sequelae of ECMO, and improve timely diagnoses so as to provide appropriate therapeutic interventions for these patients7,15,16.
Ischemic Stroke
Risk factors and mechanisms of ischemic stroke for patients on V-V ECMO are ill-defined3,9,10. Critical illness is strongly associated with cardiac arrhythmias such as atrial fibrillation for example, a known cause for embolic strokes, as well as hypercoagulability and the difficulty in managing anticoagulation therapy4,17,18. Our results showed that cardiac arrhythmias were not an independent risk factor for ischemic stroke. Other biologically plausible causes of V-V ECMO-associated ischemic stroke include emboli from the circuit through a patent foramen ovale as well as cerebral venous sinus thrombosis (CVST)4. The use of the internal jugular vein for cannulation is a known risk factor for CVST if thrombosis extends intracranially, a risk that may be increased with the use of dual-lumen right internal jugular veno-venous catheters, currently in common use, but with significantly larger internal diameter than the single lumen cannulae used routinely in the past4. Interestingly, when CT scans were performed within 4 days after V-V ECMO femoro-jugular decannulation, 53% (42 of 79) of patients were found to have cannula-associated jugular deep vein thrombosis19. The ELSO database does not collect information on CVST, and thus, we could not evaluate its association with ABI.
We identified four independent variables that were associated with ischemic stroke, namely, lower pre-ECMO pH and three on-ECMO complications, hemolysis, gastrointestinal hemorrhage, and DIC. These findings may provide insight into the timing and type of ischemic stroke in patients receiving V-V ECMO. Lower pre-ECMO pH in the ischemic stroke group suggests that severe refractory respiratory failure along with hemodynamic instability before cannulation may be a cause of ischemic stroke. The ELSO registry did not differentiate stroke subtypes, limiting our ability to distinguish global brain ischemia, leading to hypoxic-induced brain injury (HIBI) or watershed infarcts from other stroke subtypes as defined by TOAST criteria (Trial of Org 10172 in Acute Stroke Treatment) including large-artery atherosclerosis, small-vessel occlusion, and cardioembolism20. While pre-ECMO hypoxic respiratory failure may predispose to global brain ischemia, hemolysis and DIC, independent “on-ECMO” risk factors, may lead directly to thromboembolic stroke given its associated microangiopathy21. Notably, hemolysis and subclinical pump thrombosis, (which could cause elevations in plasma free hemoglobin that we observed) are known risk factors for ischemic infarct in the LVAD population22. In contrast, gastrointestinal hemorrhage is more likely a surrogate for anticoagulation disruption or may represent severity of critical illness with ischemic stroke causing stress ulceration.
Hemorrhagic Stroke
We found four independent variables that were associated with hemorrhagic stroke including lower pre-ECMO pH and PO2 and two on-ECMO complications, RRT and gastrointestinal hemorrhage (Table 3). Mechanical circulatory support devices such as LVAD and ECMO cause platelet dysfunction and endothelial cell dysfunction with subsequent acquired von Willebrand syndrome (Factor VIII deficiency), a known risk factor for bleeding24–26. Our finding of RRT increasing hemorrhage risk is consistent with prior studies reporting thrombocytopenia27,28 and renal failure4,9,12 as risk factors for hemorrhagic stroke. We also observed that both pre- ECMO pH and PO2 values were lower in the hemorrhagic stroke group (Table 1, Supplemental Figure 1). The absolute difference in median values of pH and PO2 between the groups (0.02 in pH and 3 mmHg in PO2, Table 1) is unlikely to be clinically meaningful. However, one might hypothesize that hemorrhagic stroke occurs in the peri-ECMO cannulation period with refractory hypoxia and hemodynamic instability. A higher number of RRT also supports worse hemodynamic instability in the hemorrhagic stroke group. This conjecture is supported by recent observational studies in which consecutive V-V ECMO patients received early routine CT scans, which reported a 16% hemorrhagic stroke rate5 and 10% rate in those receiving extracorporeal cardiopulmonary resuscitation (eCPR)29. Given these recent data on “early” hemorrhagic stroke in ECMO, we suggest two hypotheses on potential mechanisms of hemorrhagic stroke in V-V ECMO. First, early hemorrhagic stroke may occur secondary to hemorrhagic conversion of “fresh” hypoxic-ischemic brain tissue30 from refractory respiratory failure with the use of bolus dose heparin during ECMO cannulation and then maintenance anticoagulation during the ECMO run. Second, hemorrhagic stroke may occur in patients with severe ARDS independent of ECMO initiation. Lockie et al.5 reported that ~8% of ARDS patients requiring mechanical ventilation had hemorrhagic stroke without ever being exposed to heparin or the ECMO circuit. However, the incidence of hemorrhagic stroke was twice that (16%) among those treated with V-V ECMO5. Similarly, patients with critical illness and severe respiratory failure may suffer from cerebral microhemorrhages31,32; thus, many hemorrhagic strokes may arguably precede the ECMO run. Additional research is needed to understand the association between severe hypoxic respiratory failure and critical illness leading to cerebral micro- and macro-hemorrhage and its association with anticoagulation intensity.
Limitations
The main limitations of our study were that it was a retrospective analysis and the ELSO registry lacks several important explanatory variables for the diagnoses of ischemic and hemorrhagic stroke including anticoagulation data, occurrence of atrial fibrillation, patent foramen ovale, platelet counts, and other stroke risk factors. We were also unable to measure exposure duration to hypoxemia/acidosis as we had only one ABG data point for each pre- and on-ECMO support. The incidence of ABI in the ELSO registry likely is an underestimation as entries are reported voluntarily without central or local adjudication of neurological diagnosis. Also, it is unclear when and for which indications CT brain studies were performed and the criteria followed for brain death. The findings of our multivariable logistic regression models for ABI can only describe associations and do not imply these factors are predictive, owing to the retrospective design and uncertainty regarding the timing of strokes and covariates. The clinical relevance of some small differences in physiologic variables such as PO2 and pH requires further evaluation and may have occurred by chance due to the high number of values analyzed. Strength of our study was its large sample size and high statistical power. The ELSO registry is the largest international repository of ECMO data, offering a distinctive opportunity to evaluate the incidence, mortality, and risk factors of each type of ABI in a large number of patients.
Conclusions
Among V-V ECMO patients in the ELSO registry, approximately 5% experienced ABI. Mortality rates were high (~70%) when ischemic and hemorrhagic strokes were present compared to those with no strokes. As the use of ECMO is increasing exponentially, future research should be directed toward use of neurologic monitoring for early recognition and development of targeted management strategies especially with regard to timing and intensity of anticoagulation.
Supplementary Material
Supplemental Figure 1. Cumulative probability plots show that lower partial pressure of arterial oxygen (PO2) values are more commonly observed in patients with ischemic stroke (A) and hemorrhagic stroke (B). The y-axis is the proportion of patients with a PO2 less than or equal to the x-axis value. The frequencies are accumulated as it moves to the right on the x-axis. ECLS, extracorporeal life support (prior to ECMO cannulation).
Acknowledgments
Dr. Cho receives a grant from the Extracorporeal Life support Organization. Joe Canner, MHS, has no disclosures. Dr. Caturegli has no disclosures. Dr. Choi has no disclosures. Dr. Etchill has no disclosures. Dr. Giuliano has no disclosures. Dr. Chiarini has no disclosures. Kate Calligy, RN, has no disclosures. Dr. Lorusso is a consultant for Medtronic and LivaNova and an advisory board member for EUROSETS. All honoraria are paid to Maastricht University for research funding. Peter Rycus, MPH, has no disclosures. Dr. Kim has no disclosures. Dr. Sussman has no disclosures. Dr. Suarez has no disclosures. Dr. Geocadin is supported in part by NIH Grants UG3 HL145269, R01 HL071568 and unrestricted fund from the Wenzel Family Foundation.Dr. Ziai is supported by the NIH and receives consulting fees from C.R. Bard, Inc. and Portola outside of the area of work commented on here. Dr. Whitman has no disclosures.
Footnotes
DISCLOSURES:
None of the authors report any conflicts of interest related to this manuscript.
Disclosures: The authors have nothing to disclose
References
- 1.McCarthy FH, McDermott KM, Kini V, et al. Trends in U.S. Extracorporeal Membrane Oxygenation Use and Outcomes: 2002–2012. Semin Thorac Cardiovasc Surg. 2015;27(2):81–88. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Munshi L, Walkey A, Goligher E, Pham T, Uleryk EM, Fan E. Venovenous extracorporeal membrane oxygenation for acute respiratory distress syndrome: a systematic review and meta-analysis. Lancet Respir Med. 2019;7(2):163–172. [DOI] [PubMed] [Google Scholar]
- 3.Lorusso R, Gelsomino S, Parise O, et al. Neurologic Injury in Adults Supported With Veno-Venous Extracorporeal Membrane Oxygenation for Respiratory Failure: Findings From the Extracorporeal Life Support Organization Database. Crit Care Med. 2017;45(8):1389–1397. [DOI] [PubMed] [Google Scholar]
- 4.Cho SM, Farrokh S, Whitman G, Bleck TP, Geocadin RG. Neurocritical Care for Extracorporeal Membrane Oxygenation Patients. Crit Care Med. 2019. [DOI] [PubMed] [Google Scholar]
- 5.Lockie CJA, Gillon SA, Barrett NA, et al. Severe Respiratory Failure, Extracorporeal Membrane Oxygenation, and Intracranial Hemorrhage. Crit Care Med. 2017;45(10):1642–1649. [DOI] [PubMed] [Google Scholar]
- 6.Cho SM, Geocadin RG, Caturegli G, et al. Understanding Characteristics of Acute Brain Injury in Adult Extracorporeal Membrane Oxygenation: An Autopsy Study. Crit Care Med. 2020. [DOI] [PubMed] [Google Scholar]
- 7.Cho SM, Ziai W, Mayasi Y, et al. Noninvasive Neurological Monitoring in Extracorporeal Membrane Oxygenation. ASAIO J. 2019. [DOI] [PubMed] [Google Scholar]
- 8.Thiagarajan RR, Barbaro RP, Rycus PT, et al. Extracorporeal Life Support Organization Registry International Report 2016. ASAIO J. 2017;63(1):60–67. [DOI] [PubMed] [Google Scholar]
- 9.Luyt CE, Bréchot N, Demondion P, et al. Brain injury during venovenous extracorporeal membrane oxygenation. Intensive Care Med. 2016;42(5):897–907. [DOI] [PubMed] [Google Scholar]
- 10.Kon ZN, Bittle GJ, Pasrija C, et al. Venovenous Versus Venoarterial Extracorporeal Membrane Oxygenation for Adult Patients With Acute Respiratory Distress Syndrome Requiring Precannulation Hemodynamic Support: A Review of the ELSO Registry. Ann Thorac Surg. 2017;104(2):645–649. [DOI] [PubMed] [Google Scholar]
- 11.Fletcher Sandersjoo A, Bartek J Jr, Thelin EP, et al. Predictors of intracranial hemorrhage in adult patients on extracorporeal membrane oxygenation: an observational cohort study. J Intensive Care. 2017;5:27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Kasirajan V, Smedira NG, McCarthy JF, Casselman F, Boparai N, McCarthy PM. Risk factors for intracranial hemorrhage in adults on extracorporeal membrane oxygenation. Eur J Cardiothorac Surg. 1999;15(4):508–514. [DOI] [PubMed] [Google Scholar]
- 13.Mazzeffi M, Kon Z, Menaker J, et al. Large Dual-Lumen Extracorporeal Membrane Oxygenation Cannulas Are Associated with More Intracranial Hemorrhage. ASAIO J. 2018. [DOI] [PubMed] [Google Scholar]
- 14.Mateen FJ, Muralidharan R, Shinohara RT, Parisi JE, Schears GJ, Wijdicks EF. Neurological injury in adults treated with extracorporeal membrane oxygenation. Arch Neurol. 2011;68(12):1543–1549. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Cho SM, Choi CW, Whitman G, et al. Neurophysiological Findings and Brain Injury Pattern in Patients on ECMO. Clin EEG Neurosci. 2019:1550059419892757. [DOI] [PubMed] [Google Scholar]
- 16.Cho SM, Ziai W, Geocadin R, Choi CW, Whitman G. Arterial-Sided Oxygenator Clot and Transcranial Doppler Ultrasound Emboli in Venoarterial Extracorporeal Membrane Oxygenation. Ann Thorac Surg. 2019;107(1):326–327. [DOI] [PubMed] [Google Scholar]
- 17.Clayton B, Ball S, Read J, Waddy S. Risk of thromboembolism in patients developing critical illness-associated atrial fibrillation. Clin Med (Lond). 2018;18(4):282–287. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ryu JA, Bang OY, Suh GY, et al. Ischemic Stroke in Critically Ill Patients with Malignancy. PLoS One. 2016;11(1):e0146836. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Parzy G, Daviet F, Persico N, et al. Prevalence and Risk Factors for Thrombotic Complications Following Venovenous Extracorporeal Membrane Oxygenation: A CT Scan Study. Crit Care Med. 2020;48(2):192–199. [DOI] [PubMed] [Google Scholar]
- 20.Adams HP, Bendixen BH, Kappelle LJ, et al. Classification of subtype of acute ischemic stroke. Definitions for use in a multicenter clinical trial. TOAST. Trial of Org 10172 in Acute Stroke Treatment. Stroke. 1993;24(1):35–41. [DOI] [PubMed] [Google Scholar]
- 21.Fagniez O, Tertian G, Dreyfus M, Ducreux D, Adams D, Denier C. Hematological disorders related cerebral infarctions are mostly multifocal. J Neurol Sci. 2011;304(1–2):87–92. [DOI] [PubMed] [Google Scholar]
- 22.Cho SM, Hassett C, Rice CJ, Starling R, Katzan I, Uchino K. What Causes LVAD-Associated Ischemic Stroke? Surgery, Pump Thrombosis, Antithrombotics, and Infectionsa. ASAIO J. 2018. [DOI] [PubMed] [Google Scholar]
- 23.Giroud M, Gras P, Fayolle H, André N, Soichot P, Dumas R. Early seizures after acute stroke: a study of 1,640 cases. Epilepsia. 1994;35(5):959–964. [DOI] [PubMed] [Google Scholar]
- 24.Kalbhenn J, Schlagenhauf A, Rosenfelder S, Schmutz A, Zieger B. Acquired von Willebrand syndrome and impaired platelet function during venovenous extracorporeal membrane oxygenation: Rapid onset and fast recovery. J Heart Lung Transplant. 2018;37(8):985–991. [DOI] [PubMed] [Google Scholar]
- 25.Kalbhenn J, Schmidt R, Nakamura L, Schelling J, Rosenfelder S, Zieger B. Early diagnosis of acquired von Willebrand Syndrome (AVWS) is elementary for clinical practice in patients treated with ECMO therapy. J Atheroscler Thromb. 2015;22(3):265–271. [DOI] [PubMed] [Google Scholar]
- 26.Cho SM, Lee T, Starling RC, Thompson NR, Uchino K. The Impact of Infection and Elevated INR in LVAD-Associated Intracranial Hemorrhage: A Case-Crossover Study. ASAIO J. 2018. [DOI] [PubMed] [Google Scholar]
- 27.Guennec LL, Cholet C, Schmidt M, et al. Brain injury during veno-arterial extracorporeal membrane oxygenation. Annals of Intensive Care. 2017;7(1):8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Fletcher Sandersjöö A, Bartek J, Thelin EP, et al. Predictors of intracranial hemorrhage in adult patients on extracorporeal membrane oxygenation: an observational cohort study. J Intensive Care. 2017;5:27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Zotzmann V, Rilinger J, Lang CN, et al. Early full-body computed tomography in patients after extracorporeal cardiopulmonary resuscitation (eCPR). Resuscitation. 2020;146:149–154. [DOI] [PubMed] [Google Scholar]
- 30.Zhang J, Yang Y, Sun H, Xing Y. Hemorrhagic transformation after cerebral infarction: current concepts and challenges. Ann Transl Med. 2014;2(8):81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Riech S, Kallenberg K, Moerer O, et al. The Pattern of Brain Microhemorrhages After Severe Lung Failure Resembles the One Seen in High-Altitude Cerebral Edema. Crit Care Med. 2015;43(9):e386–389. [DOI] [PubMed] [Google Scholar]
- 32.Fanou EM, Coutinho JM, Shannon P, et al. Critical Illness-Associated Cerebral Microbleeds. Stroke. 2017;48(4):1085–1087. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental Figure 1. Cumulative probability plots show that lower partial pressure of arterial oxygen (PO2) values are more commonly observed in patients with ischemic stroke (A) and hemorrhagic stroke (B). The y-axis is the proportion of patients with a PO2 less than or equal to the x-axis value. The frequencies are accumulated as it moves to the right on the x-axis. ECLS, extracorporeal life support (prior to ECMO cannulation).