Management of Anticoagulation Therapy in ECMO-Associated Ischemic Stroke and Intracranial Hemorrhage

Abstract

Objectives:

Despite the common occurrence of extracorporeal membrane oxygenation (ECMO)-associated acute ischemic stroke (AIS) and intracranial hemorrhage (ICH), there are little data to guide optimal anticoagulation management. We sought to describe antithrombotic therapy management after stroke and outcomes.

Methods:

A retrospective analysis was conducted of venoarterial (VA) and venovenous (VV) ECMO patients treated a tertiary center from June 2016 to February 2021. Patients with image-confirmed diagnosis of AIS or ICH while receiving ECMO were included for study with data collected regarding anticoagulation management and clinical outcomes.

Results:

Overall, 216 patients (153 VA-ECMO; 63 VV-ECMO) were included in this study. Of 153 patients on VA-ECMO, 13 (8.4%) had AIS and 6 (3.9%) had ICH. Of 63 patients on VV-ECMO, none had AIS and 5 (7.9%) had ICH. One patient (9%) received anticoagulation reversal after ICH. Anticoagulation was discontinued and later resumed in all 5 ICH survivors (median cessation time: 30 hours) and 1 (50%) of 2 AIS survivors (median cessation time: 96 hours). While off anticoagulation, two (18%) of 11 patients had thromboembolic events and none had new AIS. Upon resumption, there were no cases of hemorrhagic transformation of AIS or ICH expansion. There was no difference in in-hospital mortality between patients with ICH and those without in both VA- and VV-ECMO cohorts nor between VA-ECMO patients with AIS and those without.

Conclusions:

Early cessation and judicious resumption of anticoagulation appeared feasible in the cohort of patients with ECMO-associated AIS and ICH.

Introduction

Acute ischemic stroke (AIS) and intracranial hemorrhage (ICH) are life-threatening neurological complications associated with extracorporeal membrane oxygenation (ECMO) therapy. Types of ICH during ECMO therapy include intraparenchymal hemorrhage (IPH), subarachnoid hemorrhage (SAH), and subdural hemorrhage (SDH). For patients supported with venoarterial (VA)-ECMO, frequency of AIS is 4–14% and frequency of ICH is 2–27%.^1–6 For patients supported with venovenous (VV) ECMO, frequency of AIS is 2–6% and frequency of ICH is 4–19%.^1,7–9 The wide range in frequency of AIS and ICH is due to heterogeneity in reporting of neurological complications and the lack of standardized neurological monitoring in ECMO patients.

Blood contact with the non-biological surface of the ECMO circuit activates the coagulation cascade and prompts the use of systemic anticoagulation for safe initiation of support and to reduce the risk of catastrophic clot during circuit operation.¹⁰ Unfractionated heparin is typically used to achieve target levels of anticoagulation given its availability, general familiarity, and existence of a reversal agent.¹¹ Despite the common occurrence of ECMO-associated AIS and ICH, there are little data to guide optimal anticoagulation management strategy regarding cessation and timing of resumption after acute stroke as well as anticoagulation reversal for ICH patients. The resumption of anticoagulation should be carefully considered in ECMO, balancing pro- and anti-coagulatory demands as morbidity and mortality associated with hemorrhagic conversion of AIS and hematoma expansion of ICH is often catastrophic.¹²

In an effort to further insight into the clinical management of ECMO-associated stroke, we conducted a retrospective review of patients maintained on ECMO support at a tertiary-care center. Here, we present anticoagulation management for ECMO patients following diagnosis of ECMO-associated stroke. We further evaluate the risk of ischemic and thromboembolic events following cessation of anticoagulant therapy and coagulopathy reversal, hemorrhagic events after resumption of anticoagulation, and in-hospital mortality.

Patients and Methods

Study design

A retrospective study of patients on VA-ECMO and VV-ECMO at a tertiary center from June 2016 to February 2021 was performed. This study derives from a multidisciplinary effort and an initiative between the Cardiovascular Intensive Care Unit, Cardiac Critical Care Unit, and Neuroscience Critical Care Unit to improve overall clinical care and outcomes for patients with ECMO support.¹³ The study duration was selected given the start of the database in June 2016 and extended to the most recent data from February 2021. This study was approved by Johns Hopkins Hospital Institutional Review Board and informed consent was waived as this was a retrospective observational cohort study.

Participants

We included all adult patients (age ≥ 18 years) who received VA-ECMO and VV-ECMO. We excluded patients who had temporary ECMO support solely in the operating room without needing to extend to the intensive care unit.

Data collection

For all patients in the study, we collected pre-cannulation characteristics including demographics, social history, past medical history, comorbidities and laboratory values. All patients underwent neurocritical care consultations and a standardized neuromonitoring per institutional protocol.¹³ On day 1 of cannulation, all patients received a baseline neurological examination, transcranial doppler (TCD), and electroencephalography (EEG) for 48–72 hours if the patient had a cardiac arrest. On days 3–5, all patients received neurological exams off sedation, TCD, EEG if Glasgow Coma Scale (GCS) was less than 8, and somatosensory evoked potential (SSEP) if GCS was less than 4. Sedation cessation was performed on days 3–5 of ECMO support, as deemed safe by the provider, and comatose patients were clinically indicated to receive a head computed tomography (HCT). All patients with a clinical indication for HCT and who were stable enough to be transported received HCT between day 3 and 5 after cannulation.⁶ Patients with acute ICH had anticoagulation discontinued until stable 6-hour and 24-hour follow-up HCT scans. When anticoagulation was resumed after stroke and two APTT values were within therapeutic range (50–65 seconds), repeat HCT scan and frequent neurological exams were performed to assess hemorrhagic conversion or hematoma expansion. Patients with AIS that were managed with anticoagulation cessation also had a follow-up HCT scan following anticoagulation resumption with two therapeutic aPTTs. This pathway is described in Supplemental Figure 1.

For all patients with AIS or ICH detected by head CT (HCT) scan while on ECMO, we collected data including antithrombotic therapy (heparin, bivalirudin, or neither), activated partial thromboplastin time (aPTT) at time of stroke identification, antithrombotic reversal, time off antithrombotic therapy (hours), and in-hospital mortality. We also collected incidence of the following thromboembolic events that occurred during the period of anticoagulation cessation: pulmonary embolism (PE), deep venous thrombosis (DVT), intracardiac thrombus, and ECMO circuit clotting. Small infarct/hematoma volume was defined as < 5 cc. Volumes of AIS, IPH, and SDH were calculated using the ABC/2 method.^14,15,16 There is no standard way of calculating hematoma volume for SAH. All neurological recommendations were provided by the neurointensive care consult team.

Definitions

For this study, neurological complications focused on AIS and ICH. AIS was defined as a cerebral infarction determined by HCT. ICH was defined as bleeding within the skull, including intra- and extra-parenchymal hemorrhage as well as intraventricular hemorrhage (IVH), identified by HCT scan.¹⁷ ICH was subclassified as IPH, SAH, SDH and IVH. IPH was defined as cerebral hemorrhage within the parenchyma; SAH as hemorrhage accumulated between the arachnoid and pia mater; SDH as hemorrhage accumulated between the dura mater and arachnoid; and IVH as hemorrhage into cerebral ventricular system.¹⁸

Outcomes

The primary outcomes were neurological adverse events including the presence of hemorrhagic conversion of ischemic stroke and hematoma expansion (≥6 mL or ≥33%) for ICH after the resumption of anticoagulation.^19,20 The secondary outcomes were the presence of new AIS while off anticoagulation, thromboembolic events off anticoagulation, and in-hospital mortality.

Statistical analysis

Patients were analyzed in the following groups: AIS on VA-ECMO, ICH on VA-ECMO, and ICH on VV-ECMO. All data were presented as a median and range for continuous variables and absolute numbers with percentages for binary and categorical variables. Demographic and clinical characteristics including ECMO variables in patients with and without neurological events (AIS or ICH) were compared using Wilcoxon rank-sum test for continuous variables and Fisher’s exact test for binary or categorical variables. A p value less than 0.05 was considered statistically significant. Kaplan-Meier survival estimates were calculated. For survival analyses, differences for specific subsets of data were compared by using log-rank testing. Survival was compared between patients with AIS, ICH, or neither complication. Time zero was defined as ECMO cannulation date and patients were followed until discharge or death. All statistical analyses were performed using STATA 15.1 (StataCorp, College Station, TX, USA).

Results

Demographics and Baseline Characteristics

The cohort included adult patients (n=216, 60% male, median age=55) supported with ECMO (VA-ECMO [n=153, 71%] or VV-ECMO [n=63, 29%]) between June 2016 and February 2021 (Figure 1). Baseline demographics and characteristics for the VA-ECMO and VV-ECMO cohorts are described in Tables 1 and 2. The most common ECMO indication was post-cardiotomy shock (n=79 of 153, 52%) for VA-ECMO patients and acute respiratory distress syndrome (n=53 of 63, 83%) for VV-ECMO patients. The median duration of ECMO support was 6 (1–62) days for VA-ECMO and 18 (1–114) days for VV-ECMO.

Figure 1.

Flow chart for study inclusion. *One patient had both ICH and AIS identified on initial head CT. **Two patients died before heparin resumption.

Table 1.

Baseline Characteristics of VA-ECMO Patients

Patient Characteristics	All patients (n=153)	AIS (n=13)	Without AIS (n=140)	p-value	ICH (n=6)	Without ICH (n=147)	p-value
Demographics
Age	59 (18–83)	62 (26–81)	59 (18–83)	0.38	54 (27–68)	59 (18–83)	0.32
BMI	29 (15–51)	28 (20–39)	29 (15–51)	0.23	26 (24–31)	29 (15–51)	0.18
Male	95 (62)	8 (62)	87 (62)	p > .99	4 (67)	91 (62)	p > .99
Past Medical History
Diabetes mellitus	51 (33)	4 (31)	47 (34)	p > .99	2 (33)	49 (33)	p > .99
Hypertension	111 (73)	12 (92)	99 (70)	0.12	3 (50)	108 (73)	0.35
Hyperlipidemia	83 (54)	9 (69)	74 (52)	0.38	3 (50)	80 (54)	p > .99
Congestive heart failure	51 (33)	5 (38)	46 (33)	0.76	4 (67)	47 (32)	0.10
Chronic kidney disease	23 (15)	1 (8)	22 (16)	0.69	2 (33)	21 (14)	0.22
Atrial fibrillation	42 (27)	3 (23)	39 (28)	p > .99	2 (33)	40 (27)	0.53
ECMO Indication
Cardiogenic shock	54 (35)	3 (23)	51 (36)	0.55	3 (50)	51 (35)	0.67
Post-cardiotomy shock	79 (52)	10 (76)	69 (49)	0.08	2 (33)	77 (52)	0.43
ECPR	20 (13)	0 (0)	20 (14)	0.22	1 (17)	19 (13)	0.58
ECMO Variables
Central cannulation	71 (46)	9 (69)	62 (44)	0.14	2 (33)	69(47)	0.69
ECMO support days	6 (1–62)	7 (4–26)	6 (1–62)	0.12	11 (3–26)	6 (1–62)	0.13
SOFA score	11 (3–17)	10 (7–17)	11 (3–17)	0.18	10 (9–17)	11 (3–17)	0.72
Mortality and Discharge
In-Hospital Mortality	104 (68)	11 (85)	93 (66)	0.23	4 (67)	100 (68)	p > .99
mRS at discharge	6 (1–7)	6 (1–7)	6 (1–7)	0.70	6.5 (6–7)	6 (1–7)	0.20

Values are expressed as median (range) for continuous variables and absolute frequency (percentages) for categorical variables. Continuous variables compared using Wilcoxon rank-sum and categorical variables compared using Fisher’s exact test.

Abbreviations: AIS= acute ischemic stroke; BMI= body mass index; ECPR= extracorporeal cardiopulmonary resuscitation; ICH= intracranial hemorrhage; mRS= modified Rankin scale for neurologic disability; SOFA= Sequential Organ Failure Assessment; VA-ECMO= venoarterial extracorporeal membrane oxygenation

Table 2.

Baseline Characteristics of VV-ECMO Patients ^*

Patient Characteristics	All patients (n=63)	ICH (n=5)	Without ICH (n=58)	p- value
Demographics
Age	44 (18–70)	48 (30–58)	44 (18–70)	0.76
BMI	31 (16–46)	28 (23–30)	31 (16–46)	0.11
Male	34 (54)	3 (60)	31 (53)	0.58
Past Medical History
Diabetes mellitus	13 (21)	1 (20)	12 (21)	p > .99
Hypertension	24 (38)	1 (20)	23 (40)	0.64
Hyperlipidemia	24 (38)	1 (20)	23 (40)	0.64
Congestive heart failure	2 (3)	0 (0)	2 (3)	p > .99
Chronic kidney disease	3 (5)	0 (0)	3 (5)	p > .99
Atrial fibrillation	2 (3)	0 (0)	2 (3)	p > .99
ECMO Indication
ARDS/ARF	53 (84)	5 (100)	48 (83)	0.58
Bridge to transplant	7 (11)	0 (0)	7 (12)	p > .99
Other**	3 (5)	0 (0)	3 (5)	p > .99
ECMO Variables
Central cannulation	4 (6)	0 (0)	4 (7)	p > .99
ECMO support days	18 (1–114)	28 (18–60)	14 (1–114)	0.03
SOFA score	10 (3–17)	11 (6–12)	11 (3–17)	0.92
Mortality and Discharge
In-Hospital Mortality	26 (42)	2 (40)	24 (41)	p > .99
mRS at discharge	5 (1–7)	6 (2–7)	5 (1–7)	0.77

^*No ischemic stroke in VV-ECMO

^**Intraoperative for tracheal repair (n=1), IVC stenosis (n=1), pulmonary embolism (n=1)

Values are expressed as median (range) for continuous variables and absolute frequency (percentages) for categorical variables. Continuous variables compared using Wilcoxon rank-sum and categorical variables compared using Fisher’s exact test.

Abbreviations: AIS= acute ischemic stroke; ARDS= acute respiratory distress syndrome; ARF= acute respiratory failure; BMI= body mass index; ECPR= extracorporeal cardiopulmonary resuscitation; ICH= intracranial hemorrhage; mRS= modified Rankin scale for neurologic disability; SOFA= Sequential Organ Failure Assessment; VV-ECMO= venovenous extracorporeal membrane oxygenation

VA-ECMO-associated ICH

Of 153 patients with VA-ECMO, ICH was identified in 6 (3.9%) patients: IPH (n=2), SDH (n=3), SAH (n=1), and IVH (n=2) (Table 3). One patient had both IPH and IVH, and another patient had both SAH and IVH. ICH occurred at a median of 5.5 (2–10) days from ECMO cannulation (Supplemental Figure 2). Most IPHs and SDHs were small-volume hemorrhages (n=4 of 5, 80%) with median volume of 82 (1–164) cc and 1.1 (1–1.5) cc, respectively (Table 3).

Table 3:

Patient Anticoagulation Management, Complications, and Mortality

Imaging Findings, Anticoagulation Management, and Outcomes	VA-ECMO AIS (n=13)	VA-ECMO ICH (n=6)	VV-ECMO ICH (n=5)
HCT Characteristics & Findings
Time to AIS or ICH, days	4 (1–11)	6 (2–10)	5 (1–13)
Intraventricular hemorrhage	-–	2 (33)	2 (40)
Stroke volume*	9.8 (1–83.3)	SDH: 1.1 (1–1.5) IPH: 82 (1–164)	SDH: -– IPH: 1.3 (1–2.5)
Stroke volumes for those with heparin discontinuation and subsequent resumption	1	SDH 1.1 (1–1.5) IPH: 1	SDH: -– IPH: 1.3 (1–2.5)
Anticoagulation at Stroke Identification
Antiplatelet agent	5 (38)	2 (33)	1 (20)
Anticoagulation   Heparin   Bivalirudin+   None∇	9 (70) 2 (15) 2 (15)	5 (83) 0 (0) 1 (17)	5 (100) 0 (0) 0 (0)
aPTT, median	39.2 (22.8–58.8)	50.0 (27.4–57.3)	60.9 (34.1–65.6)
Anticoagulation Cessation & Resumption
Protamine sulfate reversal	0 (0)	1 (17)	0 (0)
No heparin cessation	7 (85)	0 (0)	0 (0)
Heparin cessation	2 (15)	5 (83)	5 (100)
Heparin resumption	1 (8)	4 (67)	5 (100)
Duration of heparin discontinuation, hours	96 (96)	45 (30–96)	33 (6–136)
Thromboembolic Events While Heparin Discontinued
Pulmonary embolism	0 (0)	0 (0)	1 (20)
Deep venous thrombosis	0 (0)	0 (0)	1 (20)
Intracardiac thrombus	0 (0)	0 (0)	0 (0)
ECMO circuit clot	0 (0)	0 (0)	1 (20)
Repeat HCT Findings
New AIS off anticoagulation	0 (0)	0 (0)	0 (0)
ICH: hematoma expansion after resumption	-–	0 (0)	0 (0)
AIS: hemorrhagic conversion after resumption	0 (0)	-–	-–
In-Hospital Mortality
In-hospital mortality	11 (85)	4 (67)	2 (40)
In-hospital mortality if  heparin discontinued	1 (50)	3 (60)	2 (40)
In-hospital mortality if  heparin resumed	0 (0)	2 (50)	2 (40)

^*Sizes of intracerebral and subdural hemorrhages calculated using ABC/2 method

⁺Both patients on bivalirudin continued anticoagulation after stroke identification

^∇Reason for no anticoagulation: massive intraoperative bleeding (n=3)

Values are expressed as median (range) for continuous variables and absolute frequency (percentages) for categorical variables.

Abbreviations: AIS= acute ischemic stroke; aPTT= activated partial thromboplastin time; HCT= head computed tomography; ICH= intracranial hemorrhage; VA-ECMO= venoarterial extracorporeal membrane oxygenation; VV-ECMO= venovenous extracorporeal membrane oxygenation

At time of ICH, 5 (83%) of 6 patients were on full dose heparin infusion with a median aPTT of 50.0 (32.4–57.3) (Supplemental Figure 3). One patient was not on heparin infusion due to intraoperative chest cavity hemorrhage with massive transfusion protocol. After ICH diagnosis, heparin was discontinued in all 5 patients (100%). Protamine sulfate reversal was only used in 1 of 6 patients due to extensive SAH with significant hydrocephalus and dilated, non-reactive pupils. This patient died before heparin resumption or repeat stability HCT. The other four patients, who all had small-volume ICHs, were resumed on heparin infusion (median rate: 875 units/hour, range: 500–1200 units/hour) without boluses following stable HCT after a median cessation time of 45 (30–96) hours.

While anticoagulation was held, there were no thromboembolic events and repeat HCTs performed 1 to 3 days after heparin cessation showed no evidence of new AIS (Table 3). Additionally, after resumption of anticoagulation and two therapeutic aPTTs, HCT did not demonstrate evidence of hematoma expansion or new ICHs for all patients.

In-hospital mortality for all VA-ECMO patients with ICH was 67% (n=4 of 6). Meanwhile, in-hospital mortality was 50% (n=2 of 4) among patients managed with heparin cessation and later resumption (Table 3). With careful resumption of anticoagulation and stability HCTs, the Kaplan-Meier survival curve demonstrates no significant difference in in-hospital mortality between VA-ECMO patients with ICH (n=4 of 6, 67%) compared to those without ICH (n=100 of 147, 68%) (Table 1, Figure 2A).

Figure 2.

A. VA-ECMO cohort in-hospital survival from day of cannulation (Log rank test: χ² =2.68, p=0.26). B. VV-ECMO cohort in-hospital survival from day of cannulation (Log rank test: χ² = 0.02; p=0.88). There were no cases of ischemic stroke in the VV-ECMO cohort.

Abbreviations: AIS= acute ischemic stroke; ICH= intracranial hemorrhage; IPH= intraparenchymal hemorrhage; SAH= subarachnoid hemorrhage; SDH= subdural hemorrhage; VA-ECMO= venoarterial extracorporeal membrane oxygenation; VV-ECMO= venovenous extracorporeal membrane oxygenation

VA-ECMO-associated AIS

Out of 153 patients with VA-ECMO, thirteen (8.4%) had AIS during ECMO support at a median of 4 (1–11) days from ECMO cannulation (Supplemental Figure 2). Of these AIS patients, 69% (n=9 of 13) had infarcts ≥5 cc in volume (median: 10 cc, range: 1.0–83.3). Post-cardiotomy shock was the indication for ECMO initiation in 69% (n=9 of 13) of patients with AIS versus 33% (n=46 of 140) of patients without AIS (p=0.01) (Table 1).

At the time of AIS diagnosis, 11 (85%) of 13 patients were on anticoagulation therapy (9 with heparin; 2 with bivalirudin). Two (15%) of 13 patients were not on anticoagulation due to massive intraoperative bleeding. Of the 11 patients on anticoagulation, 9 patients (median infarct size: 9.8 cc) were continued on anticoagulation (2 with bivalirudin; 7 with heparin). The other 2 of 11 patients (median infarct size: 11.7 cc) had heparin infusion discontinued due to potential for hemorrhagic conversion of AIS (infarct volume: 22.4 cc) and concurrent GI bleed, respectively (Table 3). The former died before anticoagulation resumption, and the latter was resumed on heparin infusion (500 units/hour) without boluses after 96 hours.

During the period of heparin cessation, there were no thromboembolic events and repeat HCT 2–3 days after heparin discontinuation showed no evidence of new AIS. After resumption with two therapeutic aPTTs, HCT was performed per protocol and demonstrated no evidence of hemorrhagic conversion of AIS or new ICH.

In-hospital mortality among all patients with AIS was 85% (n=11 of 13). Kaplan-Meier survival curve demonstrates no significant difference in in-hospital mortality between VA-ECMO patients with AIS (n=11 of 13, 85%) compared to those without AIS (n=93 of 140, 66%) (Table 1, Figure 2A). Notably, the only patient managed with anticoagulation cessation and later resumption was 1 of 2 AIS patients that survived to discharge. Meanwhile, only 1 of 9 patients continued on anticoagulation survived to discharge.

VV-ECMO Associated ICH

Of the 63 patients on VV-ECMO, ICH was identified in 5 (3.9%) patients: IPH (n=2), SAH (n=3), IVH (n=2), and SDH (n=0). The median volume of IPH was 1.3 (1–2.5) cc. ICH occurred at a median of 5 (1–13) days from ECMO cannulation (Supplemental Figure 2). Patients with ICH had a longer duration of ECMO support (median: 28 days) than those without ICH (median: 14 days, p=0.03) (Table 2).

All 5 patients were on full dose heparin infusion (median aPTT: 60.9, range: 34.1–65.6) at time of ICH (Supplemental Figure 3). No patients received protamine sulfate for heparin reversal. Heparin was discontinued and later resumed (median rate: 1000 U/hr, range: 500–1350 units/hour) without boluses in all patients following stable repeat HCT images after median discontinuation time of 33 (6–136) hours (Table 3).

While anticoagulation was held, repeat HCTs performed 1 to 4 days after heparin cessation showed no evidence of new AIS. However, two patients had thromboembolic events diagnosed during the period of heparin cessation: ECMO circuit clot (n=1), PE (n=1) and DVT (n=1). After resumption of anticoagulation with two therapeutic aPTTs, HCT was performed per protocol and revealed no evidence of hematoma expansion or new ICH in all patients.

In-hospital mortality was 40% (n=2 of 5) with a median time from ECMO cannulation to death of 46 (31–60) days. With careful resumption of anticoagulation and stability HCTs, the Kaplan-Meier survival curve demonstrated no significant difference in in-hospital mortality between VV-ECMO patients with ICH (n=2 of 5, 40%) compared to those without ICH (n=24 of 58, 41%) (Table 1, Figure 2B).

Discussion

In this clinical series, we found that early cessation and judicious resumption of intravenous anticoagulation appears safe in patients with ECMO-associated AIS and ICH. While off anticoagulation, two (18%) of 11 patients had thromboembolic events and none had new ischemic infarcts. Upon anticoagulation resumption, there were no cases of ICH expansion or hemorrhagic transformation of AIS. This study represents the first report, to our knowledge, describing anticoagulation reversal and resumption strategies and outcomes after ECMO-associated stroke.

Our study found a prevalence of AIS and ICH in VA-ECMO of 8.4% (n=13 of 153) and 3.9% (n=6 of 153), respectively. Le Guennec et al., in a retrospective observational study of 878 VA-ECMO-treated patients at a tertiary referral center, reported an AIS prevalence of 5.3% and an ICH prevalence of 2.8%.² For VV-ECMO, our study reports incidences of AIS and ICH of 0% and 7.9% (n=5 of 63), respectively. Another observational study of 135 patients on VV-ECMO reported similar results, with a prevalence of 2% for AIS and 7.5% for ICH.⁸ Though our cohort consists of fewer patients than those in the referenced studies, the frequencies of AIS and ICH are similar to prior reports.

For AIS in patients on VA-ECMO, we report an in-hospital mortality of 85%, with no difference in mortality among patients with AIS versus those without AIS. Similarly, a retrospective review of 10,342 VA-ECMO patients from the Extracorporeal Life Support Registry by Cho et al. found an in-hospital mortality of 76% in patients with AIS.¹⁷ Additionally, in our cohort, in-hospital mortality was 67% for VA-ECMO associated ICH and 40% for VV-ECMO associated ICH. Notably, we report no difference in in-hospital mortality among patients with ICH compared to those without ICH in the VA-ECMO and VV-ECMO cohorts. In comparison, a systematic review of ECMO-associated ICH by Fletcher-Sandersjoo et al. found that mortality in ICH cohorts varied between 32% and 100% with a relative risk of mortality of 1.27 to 4.43 compared to non-ICH cohorts.²¹ Hematoma volumes were not reported due to lack of data.²¹ While our cohort’s in-hospital mortality falls within the range reported by Fletcher-Sandersjoo et al., our ICH cohorts did not have an increased mortality risk compared to their non-ICH counterparts. This may be a result of early detection from standardized neuromonitoring with routine imaging,^13,22 early intervention such as cessation of anticoagulation, and the small volume (<5 cc) of the majority of ICHs. Although there are limited data on size of ECMO-associated ICH, small volumes of ICHs in our center may be due to early detection of injury via standardized neuromonitoring protocol and neurocritical care team consultation.^13,22

In both VA- and VV-ECMO cohorts, the majority of patients with ICH were managed with discontinuation and later resumption of heparin anticoagulation (median cessation time: 45 hours in VA-ECMO, 33 hours in VV-ECMO). In contrast, only two of 13 patients with AIS had heparin discontinued. Heparin cessation and resumption appeared safe in our cohort, with no new AIS during the period of anticoagulant cessation regardless of the duration. Thromboembolic complications during the period anticoagulation cessation were only seen in two (18%) of the 11 stroke patients managed with heparin discontinuation, both with ICH while on VV-ECMO. Following anticoagulation resumption, there was no hemorrhagic expansion of ICH or hemorrhagic transformation of AIS in any patient. Reversal of anticoagulation was rare with protamine sulfate reversal given to only 1 (9%) of 11 patients with ICH due to large hematoma size. Despite the almost uniform avoidance of protamine reversal, there were no cases of hemorrhagic expansion among patients with acute ICH.

The majority of ICHs in our VA- and VV-ECMO cohorts were small-volume hemorrhages. Routine imaging that is part of our neuromonitoring protocol may have contributed to the identification of small-volume hemorrhages. By potentially increasing early detection of small-volume hemorrhages or infarcts before they may be clinically significant, our protocol may allow for an early intervention to prevent further worsening of brain injury. Our study suggests that standardized neuromonitoring and neurological expertise in management of patients on ECMO support may lessen neurologic morbidities, but studies at other institutions and in larger cohorts are needed to validate these findings.¹³

Our study is limited by the small number of ECMO patients (n=23) with AIS or ICH. Since AIS and ICH were identified by HCT and not all ECMO patients had brain imaging, the number of patients identified with either is likely an underestimate of the true number of neurological complications, especially given high mortality in ECMO patients in general. Given the small numbers of patients with AIS or ICH, we are unable to draw definitive conclusions about the differences in mortality and thromboembolic risk among those managed with a period of anticoagulation cession versus those managed with continuous anticoagulation therapy. Our study is also limited by a selection bias, favoring anticoagulation resumption in smaller strokes, with the largest stroke volume being 2.5 ccs. One VA-ECMO patient with extensive SAH and one VA-ECMO patient with a 22.4 cc AIS died before resumption, excluding these patients from the study. Additionally, there is not enough data to conclude that this anticoagulation approach is safer than standard-of-care. Although the Kaplan-Meier curve shows no survival difference between ECMO patients with and without ICH, this study may not be sufficiently powered to detect a true difference. However, this study highlights feasibility of anticoagulation reversal and resumption strategies in cases of ECMO-associated stroke and identifies associated rates of mortality and thromboembolic complications.

Conclusions

Early cessation and judicious resumption of anticoagulation appeared feasible in the cohort of patients with ECMO-associated AIS and ICH. Further research is needed to guide standardized anticoagulation practices in management of ECMO-associated stroke.

Abbreviations

AIS

acute ischemic stroke

ICH

intracranial hemorrhage

VA-ECMO

venoarterial extracorporeal membrane oxygenation

VV-ECMO

venovenous extracorporeal membrane oxygenation