Death by Neurologic Criteria in Children Undergoing Extracorporeal Cardiopulmonary Resuscitation: Retrospective Extracorporeal Life Support Organization Registry Study, 2017–2021

Abstract

OBJECTIVES:

To determine factors associated with brain death in children treated with extracorporeal cardiopulmonary resuscitation (E-cardiopulmonary resuscitation).

DESIGN:

Retrospective database study.

SETTINGS:

Data reported to the Extracorporeal Life Support Organization (ELSO), 2017–2021.

PATIENTS:

Children supported with venoarterial extracorporeal membrane oxygenation (ECMO) for E-cardiopulmonary resuscitation.

INTERVENTION:

None.

MEASUREMENTS AND MAIN RESULTS:

Data from the ELSO Registry included patient characteristics, blood gas values, support therapies, and complications. The primary outcome was brain death (i.e., death by neurologic criteria [DNC]). There were 2,209 children (≥ 29 d to < 18 yr of age) included. The reason for ECMO discontinuation was DNC in 138 patients (6%), and other criteria for death occurred in 886 patients (40%). Recovery occurred in 1,109 patients (50%), and the remaining 76 patients (4%) underwent transplantation. Fine and Gray proportional subdistribution hazards’ regression analyses were used to examine the association between variables of interest and DNC. Age greater than 1 year (p < 0.001), arterial blood carbon dioxide tension (Paco₂) greater than 82 mm Hg (p = 0.022), baseline lactate greater than 15 mmol/L (p = 0.034), and lactate 24 hours after cannulation greater than 3.8 mmol/L (p < 0.001) were independently associated with greater hazard of subsequent DNC. In contrast, the presence of cardiac disease was associated with a lower hazard of subsequent DNC (subdistribution hazard ratio 0.57 [95% CI, 0.39–0.83] p = 0.004).

CONCLUSIONS:

In children undergoing E-cardiopulmonary resuscitation, older age, pre-event hypercarbia, higher before and during ECMO lactate levels are associated with DNC. Given the association of DNC with hypercarbia following cardiac arrest, the role of Paco₂ management in E-cardiopulmonary resuscitation warrants further studies.

RESEARCH IN CONTEXT.

• Extracorporeal cardiopulmonary resuscitation (E-CPR) has significantly improved survival in children with refractory cardiac arrest, despite being associated with a high burden of neurologic complications.

• Death by neurologic criteria (DNC) is the result of severe neurologic complications following cardiac arrest requiring CPR and the diagnosis can be challenging.

• This 2017–2021 Extracorporeal Life Support Organization Registry study aims to determine the before and during ECMO factors associated with DNC in children undergoing E-CPR.

AT THE BEDSIDE.

• In the 2017–2021 ELSO Registry dataset of pediatric patients who underwent E-CPR (from 1 mo to 18 yr old), we found that survival was 54%.

• DNC occurred in 6% of all patients who underwent E-CPR and accounted for 13% of all mortality in E-CPR.

• The physiologic factors associated with greater hazard of subsequent DNC in patients who received E-CPR, included: higher pre-ECMO lactate levels and Paco₂ above the 75th-percentile (> 82 mm Hg); as well as elevated lactate levels after 24 hours of ECMO. The question of whether these are causal requires further study.

Extracorporeal cardiopulmonary resuscitation (E-CPR) is the use of venoarterial extracorporeal membrane oxygenation (VA-ECMO) during cardiac arrest in patients who fail to achieve return of spontaneous circulation (ROSC) with conventional cardiopulmonary resuscitation (CPR) alone. Among pediatric patients, E-CPR is primarily used in the setting of in-hospital cardiac arrest, and contemporary survival rates are as follows. In an Extracorporeal Life Support Organization (ELSO) Registry study (cohort 2017–2019), E-CPR survival in pediatric patients without congenital heart disease (CHD) is 41% (1). In a Virtual Pediatric System database (VPS, LLC) study, (cohort 2010–2018), E-CPR survival in the pediatric cardiac population is 51% (2). However, this potential survival benefit of E-CPR in children with refractory cardiac arrest does not preclude the risk of neurologic complications, as shown in a systematic review and meta-analysis of literature published 1990–2020 (3). In another ELSO study (cohort, 2011–2019), the prevalence of neurologic complications in children undergoing E-CPR was 14.1%, and mortality on ECMO was 35.7% (4). In this context, progression to brain death or death by neurologic criteria (DNC) (5) after E-CPR is largely unknown in the contemporary pediatric population. However, we do know that in the ELSO Registry (cohort, 2010–2019) out of non-E-CPR pediatric patients (from 1 mo to 18 yr old) undergoing ECMO, DNC occurred in 2.5% of the patients, accounting for approximately 7% of all deaths (6). Furthermore, the determination of DNC by standard clinical criteria (i.e., apnea testing) is challenging in patients supported by ECMO, because no standardized procedure exists. In addition, sedation, hypothermia, and severe metabolic disturbances can impede the ability to perform an accurate DNC examination to avoid false diagnosis of DNC (7). In these circumstances, an ancillary study for diagnosis of DNC may be required if these prerequisite conditions cannot be corrected (5).

Therefore, the aim of this new ELSO study (cohort 2017–2021) was to characterize the before and during ECMO factors associated with DNC in children undergoing E-CPR. The identification of these factors may indeed help lead to earlier recognition of catastrophic outcomes in the E-CPR population and thus enable better communications with families.

MATERIALS AND METHODS

Study Design

This study was a retrospective analysis of data obtained from the ELSO Registry, 2017–2021. The ELSO Registry collects information about patients and devices during the use of ECMO to support cardiorespiratory function. Approximately 580 international member centers report data to the Registry. An agreement between member centers and ELSO allows the release of limited deidentified datasets for purposes of research without the need for further approval from individual centers.

Patient Selection

We performed a retrospective review of the ELSO Registry using cases entered from January 1, 2017, to December 31, 2021. All children (29 d old or older and 18 yr younger) undergoing E-CPR were included. We excluded newborns and neonates because of challenges in determining DNC in this population and the limited accuracy of making the diagnosis in a retrospective database (5, 8). Furthermore, neonatal neurologic injury may be the result of different pathophysiologic processes. For patients with more than one ECMO run, only the last run was included.

Study Variables

As per the ELSO Guidelines, E-CPR is defined as use of VA-ECMO during CPR (9, 10). The Registry collects pre-ECMO variables including patient age, comorbidities as defined by International Classification of Diseases (ICD) 9 and 10 Clinical Modification (11) (Table S1, http://links.lww.com/PCC/C449), hemodynamic and arterial blood gas (ABG) values. Pre-ECMO data are collected within the 6 hours preceding ECMO. Pre-ECMO partial pressure of arterial oxygen (Pao₂) may be low in patients with CHD regardless of the presence of respiratory insufficiency and was excluded from the analysis. Pre-ECMO support, including mechanical circulatory support devices, respiratory support, vasoactive infusions, and IV medications. On-ECMO variables, such as cannulation sites, laboratory information collected close to 24 hours after ECMO onset, complications, ECMO duration, and reason for ECMO discontinuation (DNC, other criteria for death or withdrawal, recovery, or transplant) were also collected. The primary outcome was in-hospital mortality. As the exact timing of determining DNC is not available in the ELSO Registry, the time to ECMO discontinuation was modeled for DNC patients.

Data Categorization and Outcome

The ICD-10 Clinical Modification was used to create the “cardiac” diagnostic category including “myocarditis and cardiomyopathies,” “CHD,” “pulmonary hypertension,” and “other cardiac.” Other “noncardiac” diagnostic categories, included: “respiratory,” “nonpulmonary infection,” and “other,” regrouping patients that could not be assigned to one of these groups. Patients were classified in the “cervical cannulation” group if either the carotid artery or the jugular vein was one of the cannulation sites reported. Reason for ECMO discontinuation was categorized as “death by other causes,” meaning other criteria for death, if ECMO was discontinued because of complication, resource limitation, or poor prognosis (excluding DNC); and as “transplant” if heart transplant, lung transplant or ventricular assist device placement were performed. ABG variables were categorized according to quartiles of distribution. Reference values were determined according to the closest physiologic ranges. Age was categorized into five groups (1–3 mo, > 3–6 mo, > 6–12 mo, > 1–5 yr, > 5–18 yr). Variables included in the E-CPR addendum such as initial rhythm and cardiac arrest location were considered for analysis. Mechanical cardiac support encompasses temporary/permanent pacemaker, intra-aortic balloon, ventricular assist devices, and Berlin Heart. The other forms of acute brain injury (ABI) such as seizures, intracerebral hemorrhage (ICH), and ischemic stroke (CNS diffuse ischemia and/or CNS infarction), were analyzed.

Statistical Analysis

The study was performed and analyzed following the Strengthening the Reporting of Observational Studies in Epidemiology statement (12). Descriptive statistics were used to describe unadjusted patient characteristics, including: count (%) and median (interquartile range [IQR]). Quantitative data were compared between groups using a simple linear regression model and qualitative data were compared between groups using chi-square test or Fisher exact test, as appropriate. Fine and Gray proportional subdistribution hazards regression was used to examine the univariate and multivariable associations between variables of interest and the outcome of DNC. Competing risk analysis was used to estimate marginal probability of DNC in the presence of competing events such as recovery, transplant, and other criteria for (i.e., not DNC) death. Two separate multivariable models were used to analyze pre-ECMO versus on-ECMO associations with DNC. Only patients with at least 24 hours of support were included in the on-ECMO model. The multivariable models were adjusted for a priori defined clinically relevant variables, which were age (yr), weight (kg), diagnosis category (cardiac vs. noncardiac), initial cardiac rhythm, pre- or on-ECMO ROSC, cardiac arrest location, and pre-ECPR support (cardiac vs. other). As age and weight are collinear in the pediatric population, only age was modeled. A no-selection variables procedure was used to avoid overfitting. Covariates with greater than or equal to 40% missing data were removed from the model. Missingness precluded analysis of delta values (changes over 24 hr between pre- and on-ECMO) for Pao₂ and Paco₂. For the covariates with less than 40% of missing data, we performed multiple imputations by substantive model compatible fully conditional specification using the R package smcfcs (R Foundation for Statistical Computing, Vienna, Austria; https://www.R-project.org). Imputations were drawn from a logistic regression model for dichotomous variables and from normal linear regression for continuous ones. Variables used for imputation were the covariates of the multivariable model as well as the dependent variable. We generated 10 datasets corresponding to a compromise between the different proportions of missing values for the covariates of interest. The resulting proportional subdistribution hazards for each imputed dataset were combined using Rubin’s rules (13). Proportional hazard assumption was assessed with Schoenfeld-type residuals against failure time plots. Complete case analysis, that is, analysis that only includes participants without missing data, was also performed (see Supplemental Digital Content, http://links.lww.com/PCC/C449). Competing risks regression for clustered data to capture center effects was also performed. Statistical significance was set at a two-sided p value of less than or equal to 0.05 value for all analyses. All analyses were performed using R statistical software, version R4.0.2 (R Foundation for Statistical Computing).

RESULTS

During 2017–2021, there was a total of 2,209 pediatric patients who received ECMO patients for E-CPR in the ELSO Registry. Of these, 691 experienced ABI (31%), the most common types of ABI included the following list, which were not mutually exclusive: seizures in 253 (11%), ICH in 247, (11%), ischemic stroke in 151 (7%), and DNC in 138 (6%). One hundred seventy-one patients (8%) underwent more than one E-CPR run. Pre- and on-ECMO patients’ characteristics, ABG values, comorbidities, and support are shown in Table S2 (http://links.lww.com/PCC/C449). Children who suffered DNC had lower pH, higher Paco₂, lower bicarbonate, and higher lactate levels before cannulation compared with those who did not. DNC was also associated with higher on-ECMO lactate levels; however, we failed to identify an association with on-ECMO Paco₂ value. Furthermore, a higher percentage of patients with subsequent DNC, compared with those not, had had an out-of-hospital cardiac arrest. The median time spent on ECMO was 3.4 days (IQR 1.6–6.2). Overall, ECMO was discontinued within 15 days in 2,091 patients (95%). The reason for ECMO discontinuation was DNC in 138 patients (6%), other criteria for death in 886 patients (40%), recovery in 1,109 (50%), and transplantation in 76 patients (4%). Figure 1 shows the cumulative prevalence of DNC and the other reasons for ECMO discontinuation. Most patients progressing to determination of DNC (115, 83%) were diagnosed after 24 hours of ECMO. The prevalence of DNC increased over time with a plateau effect of new cases after 5 days of ECMO. We compared patients who died of a diagnosis of DNC and those who met other criteria for death (Table 1). We failed to identify an association between criteria for death and the prevalence of ABI.

Figure 1.

Cumulative prevalence of death by neurologic criteria versus death by other criteria. ECMO = extracorporeal membrane oxygenation.

TABLE 1.

Comparison of Pre- and On-Extracorporeal Membrane Oxygenation (After 24 hr of Support) Characteristics Between Patients Who Met Death by Neurologic Criteria and Those Who Met Other Criteria for Death

Variables	Death by Neurologic Criteria (n = 138)	Other Criteria for Death (n = 886)	p
Characteristics
Age (yr), median (IQR)	3.7 (1–9.6)	1.6 (0.4–9.7)	0.184
Weight (kg), median (IQR)	15.6 (9.1–36.2)	11.9 (5.5–37)	0.693
Female sex, n (%)	70 (51)	394 (45)	0.208
Diagnosis of cardiac origin, n (%)	56 (43)	445 (55)	0.017
Comorbidities, n (%)
Cardiovascular	41 (30)	318 (36)	0.187
Respiratory	27 (20)	213 (24)	0.295
Renal	1 (1)	21 (2)	0.344
Gastrointestinal	8 (6)	79 (9)	0.290
Metabolic	11 (8)	85 (10)	0.652
Neuromuscular	31 (22)	115 (13)	0.005
Hematological	9 (7)	83 (9)	0.354
Immunodeficiency	0 (0)	13 (1)	0.236
Nonrespiratory infections	2 (1)	44 (5)	0.102
Oncological	4 (3)	20 (2)	0.533
Other congenital or genetic defects	18 (13)	127 (14)	0.785
Out-of-hospital cardiac arrest, n (%)	11 (16)	25 (7)	0.018
Pre-ECMO duration of mechanical ventilation (hr), median (IQR)	2 (1–7)	2 (1–21)	0.014
Mean arterial pressure (mm Hg), median (IQR)	15 (11.8–20.5)	13 (10–18)	0.054
Pre-ECMO blood gas values, median (IQR)
pH	6.9 (6.8–7.2)	7.1 (6.9–7.2)	< 0.001
Paco2 (mm Hg)	76.5 (44–104)	60 (42–87)	< 0.001
Bicarbonate (mmol/L)	16 (10–22)	18 (12–24)	0.361
Lactate (mmol/L)	14.9 (8.8–18.6)	11.7 (5.3–16.4)	0.002
Pre-ECMO supportive therapies, n (%)
Mechanical cardiac support	16 (12)	99 (11)	1
Other supporta	25 (18)	183 (21)	0.565
Vasoactive drug infusion	104 (75)	613 (69)	0.170
IV bicarbonate	59 (43)	345 (39)	0.448
Initial cardiac rhythm, n (%)
Sinus bradycardia or complete heart block	20 (29)	112 (29)	1
Other	50 (71)	270 (71)
Neurologic complications, n (%)
Seizure	12 (9)	105 (12)	0.347
Hypoxic-ischemic brain injury	14 (10)	103 (12)	0.715
Intracerebral hemorrhage	13 (9)	129 (15)	0.136

ECMO = extracorporeal membrane oxygenation, IQR = interquartile range.

^aOther support includes renal replacement therapy, conventional mechanical ventilation, and high-frequency oscillatory ventilation.Boldface values indicate statistical significance.

Risk factor analysis of early DNC diagnosed within the 24-hour window (23, 17%) can only be descriptive/univariable due to the limited number of cases (Table S3, http://links.lww.com/PCC/C449).

Results of regression analyses are shown in Tables 2 and 3. Age greater than 1 year (SHR 3.43), Paco₂ value greater than 82 mm Hg (75th percentile) (SHR 1.78), and lactate greater than 15 mmol/L before ECMO (SHR 1.80) were independently associated with subsequent DNC. In contrast, the presence of cardiac disease was associated with a lower hazard of DNC (SHR 0.57). Among patients with more than 24 hours of ECMO support, only a value of lactate greater than 3.8 mmol/L (75th percentile) (SHR 3.18) was independently associated with subsequent DNC.

TABLE 2.

Pre-Extracorporeal Membrane Oxygenation Factors Association With Death by Neurologic Criteria by Multivariable Analysis

Variables	Unadjusted SHR (95% CI)	Adjusted SHR (95% CI)	p
Age: 3–6 mo	1.62 (0.68–3.90)	1.42 (0.59–3.45)	0.434
Age: 6–12 mo	2.24 (0.98–5.12)	2.25 (0.97–5.21)	0.060
Age: 1–5 yr	4.89 (2.39–10.0)	3.92 (1.90–8.08)	< 0.001
Age: 5–18 yr	4.30 (2.14–8.67)	3.43 (1.67–7.01)	< 0.001
Diagnosis of cardiac origin	0.41 (0.29–0.58)	0.57 (0.38–0.84)	0.004
pH < 6.9a	3.04 (1.97–4.69)	1.53 (0.86–2.72)	0.138
Paco2 (mm Hg) > 82b	2.51 (1.72–3.65)	1.78 (1.08–2.94)	0.022
Bicarbonate (mmol/L) < 13	1.69 (1.03–2.76)	1.11 (0.63–1.96)	0.712
Lactate (mmol/L) > 15a	2.65 (1.76–3.99)	1.80 (1.03–3.13)	0.034
Mechanical cardiac support	0.83 (0.50–1.39)	1.42 (0.83–2.44)	0.200
Other supportc	0.76 (0.49–1.17)	0.88 (0.55–1.41)	0.606
Vasoactive drug infusion	1.19 (0.81–1.75)	1.51 (0.99–2.31)	0.058
IV bicarbonate	1.22 (0.87–1.71)	1.08 (0.75–1.56)	0.686

SHR = subdistribution hazard ratio.

^a25th percentile.

^b75th percentile.

^cOther support includes renal replacement therapy, conventional mechanical ventilation, and high-frequency oscillatory ventilation.

Boldface values indicate statistical significance.

TABLE 3.

On-Extracorporeal Membrane Oxygenation Factors Association With Death by Neurologic Criteria by Multivariable Analysis

Variables	Unadjusted SHR (95% CI)	Adjusted SHR (95% CI)	p
Cervical cannulation	0.80 (0.55–1.15)	0.80 (0.55–1.16)	0.232
pH < 7.35a	1.49 (1.00–2.22)	1.29 (0.85–1.95)	0.230
Pao2 (mm Hg) > 186b	1.40 (0.93–2.11)	1.23 (0.79–1.92)	0.358
Paco2 (mm Hg) < 37.5a	1.45 (0.95–2.22)	1.23 (0.78–1.94)	0.378
Lactate (mmol/L) > 3.8b	3.49 (2.30–5.28)	3.18 (1.96–5.14)	< 0.001

SHR = subdistribution hazard ratio.

^a25th percentile.

^b75th percentile.

Boldface values indicate statistical significance.

Complete case analyses and clustered analyses (14) for center effects yielded similar results and are summarized in Supplemental Tables S4-S6 (http://links.lww.com/PCC/C449).

DISCUSSION

In this ELSO Registry study of pediatric who were treated with E-CPR (from 1 mo to 18 yr old), we have examined a contemporary cohort (2017–2021) and found that overall survival was 54%. DNC occurred in 6% of all E-CPR patients and accounted for 13% of all mortality. Age older than 1 year was one of the patient demographic characteristics associated with a greater hazard of subsequent DNC. Of the physiologic factors associated with greater hazard of subsequent DNC pre-ECMO included: higher lactate levels and Paco₂ above the 75th percentile (> 82 mm Hg) and after 24 hours of ECMO, elevated lactate levels. Last, the presence of cardiac disease (acquired or congenital) in patients who underwent E-CPR was associated with almost halving the hazard of subsequent DNC.

The overall survival rate reported in our study is consistent with the pediatric E-CPR data from the ELSO Registry and the VPS dataset (1, 2). We also discovered that within our study group, the prevalence of DNC was nearly double, and the associated mortality rate was more than two times compared with the reported deaths in pediatric ECMO cases not involving E-CPR indications (6). This expected difference may be explained by the higher risk of neurologic complications during E-CPR (3, 4).

Lactate peak before cannulation and after 24 hours on ECMO were both independently associated with DNC in our study. Elevated baseline lactate concentration has been associated with poor performance of CPR and with higher risk of ABI in adults undergoing ECPR (15). Furthermore, in children supported by ECMO for various indications, baseline lactate has been shown to correlate with ABI severity (16). Lactate clearance is commonly employed as an indicator of tissue perfusion, reflecting the adequacy of ECMO support and cerebral perfusion (17). A postmortem investigation of infants who received VA-ECMO support revealed a connection between the inability to effectively clear lactate and the occurrence of hypoxic-ischemic brain injury (18). In adults undergoing E-CPR, lactate clearance is a strong predictor of unfavorable neurologic outcomes (19). The high lactate levels in our cohort may potentially reflect disease severity, low CPR quality, high systemic vascular resistance related to epinephrine dosing, and inadequate ECMO flows leading to low cerebral perfusion (17).

Higher pre-cannulation Paco₂ was associated with higher hazard of DNC in our population. The role played by Paco₂ in the development of neurologic injuries on ECMO remains unclear. A U-shaped relationship between Paco₂ before cannulation and mortality has already been described in adults supported by ECMO (20). During CPR, end-tidal Pco₂ has been shown to provide an estimation of cardiac output and organ perfusion (21). Elevated Paco₂ before ECMO support may indeed be a surrogate of relatively inefficient resuscitation and poor brain perfusion before ECMO resulting in higher rate of cerebral complications (22). Our results are consistent with a recent analysis of the ELSO Registry (cohort 2010–2019) by Shah et al (6), in which initial hypercapnia (> 70 mm Hg) was associated with greater odds of ABI in children supported by VA-ECMO. The same authors also found that a relative decrease in Paco₂ greater than 30% in the first 24 hours of ECMO support, especially when combined with an increase in mean arterial pressure (MAP), is associated with a higher risk of neurologic complications including DNC (6). Hypocapnia and increases in blood pressure both increase cerebral vascular tone and resistance, reducing cerebral blood flow (23). Furthermore, the variation rate of blood pressure may also influence autoregulation resulting in disruption of the blood–brain barrier and edema in case of abrupt elevation of the blood pressure (6, 23). In our cohort, a significant number of values were missing for ABG and MAP. As a result, we were unable to analyze association of changes in Paco₂ and MAP and the subsequent hazard of DNC. Our results thus confirm the necessity for future prospective studies relying on accurate ABG data (24).

We also found that age older than 1 year was associated with greater hazard of subsequent DNC in E-CPR cases. In contrast, known cardiac diseases were independently associated with almost a halving of subsequent hazards of DNC. These features are not modifiable, but they agree with previously published VPS data (2). The lower prevalence in smaller children may reflect the exclusion of neonates from our study population. Better neurologic outcomes after ECPR were also reported in children with cardiac diseases (2, 25). Cardiac patients have previously been reported to have more often shockable rhythms which in turn is associated with better survival and lower rates of neurologic complications (25). Furthermore, the awareness of an underlying cardiac disease might have led to an earlier initiation of CPR and a more rapid deployment of ECPR thus reducing the likelihood of ABI (26).

Several limitations should be considered when interpreting our analyses. Our retrospective study used an ELSO Registry dataset, which is based on voluntary reporting and therefore subject to reporting bias. We did not have access to important data such as duration and quality of CPR, etiology of cardiac arrest, and anticoagulation regimen on ECMO, which may have had an important influence on the subsequent progression to DNC. Additionally, the ELSO Registry database does not provide information on ABI severity and how and when ABI was diagnosed. In particular, the use of ancillary studies and the compliance with currently accepted U.S. National and American Academy of Pediatrics guidance on determining DNC are not reported (5, 8). We found that 6% of patients in our cohort were diagnosed with DNC. The limitation or withdrawal of circulatory support in children with catastrophic brain injury before such a diagnosis and the fact that some centers may choose to only determine DNC in cases of organ donation may all have led to an underestimation of the primary outcome in our study. Also, many variables in the E-CPR Addendum such as initial rhythm, arrest location, and ROSC were not used as predictors in the multivariable models because of high percentage of missing values.

CONCLUSIONS

In our cohort of children undergoing E-CPR, we found that the prevalence of DNC was nearly twice as high compared with non-ECPR ECMO cases. Older age, pre-ECMO hypercarbia (> 82 mm Hg), and higher before and during ECMO lactate levels are associated with greater hazard of subsequent DNC. Given the risk of severe neurologic consequences including DNC following cardiac arrest, the role of Paco₂ management in E-CPR warrants further studies.

ACKNOWLEDGMENTS

We thank the Extracorporeal Life Support Organization for technical support. We also thank Antoine Poncet and Christophe Combescure (Clinical Research Center and Division of Clinical Epidemiology, Geneva University Hospitals, Geneva, Switzerland) for their help and support with statistical analyses.