Recommendations in Extracorporeal Cardiopulmonary Resuscitation Post Resuscitation Care Via An International, Modified Delphi Approach

Abstract

Background

The global adoption of extracorporeal cardiopulmonary resuscitation (ECPR) for cardiac arrest is accelerating, reflecting its potential to improve outcomes in select patient populations. Although research has increasingly focused on patient selection criteria, there remains a critical lack of evidence to guide best practices in post‐ECPR resuscitation management. This study aimed to identify key knowledge gaps and develop consensus‐based recommendations to inform clinical decision‐making in the post‐ECPR setting.

Methods

An international, multidisciplinary steering committee comprising 11 ECPR experts and 1 survivor conducted a targeted literature review to identify key domains and gaps in post‐ECPR care. Experts were recruited based on field impact and snowball sampling. A modified Delphi process was used, comprising 3 survey rounds. Consensus was defined a priori as ≥70% agreement or disagreement on Likert‐scale statements.

Results

A total of 53 experts were recruited, representing diverse health care professions, gender, and levels of clinical experience. Of these, 52 (98%) completed all 3 rounds of the Delphi process. Consensus was reached on 126 individual statements, which were synthesized into 42 summary position statements. These recommendations span key domains of post‐ECPR care, including system‐level considerations, hemodynamic targets and management strategies, ventilation and oxygenation protocols, anticoagulation practices, temperature regulation, complication mitigation, weaning approaches, and neuroprognostication.

Conclusions

This modified Delphi study, informed by a diverse international panel, resulted in 42 consensus‐based position statements addressing key aspects of post‐ECPR care. In the absence of robust empirical evidence, these expert‐derived recommendations offer a valuable framework to support clinical decision‐making and the development of standardized post‐ECPR resuscitation protocols.

REGISTRATION

URL: https://www.clinicaltrials.gov; Unique identifier: NCT06552312.

Nonstandard Abbreviations and Acronyms

ECPR

extracorporeal cardiopulmonary resuscitation

ELSO

Extracorporeal Life Support Organization

OHCA

out‐of‐hospital cardiac arrest

NSE

neuronal specific enolase

PaCO₂

partial pressure of arterial carbon dioxide

PaO₂

partial pressure oxygen

VA ECMO

venoarterial extracorporeal membrane oxygenation

WLST

withdrawal of life support

Clinical Perspective.

What Is New?

• Using a modified Delphi method of extracorporeal cardiopulmonary resuscitation experts, a list of consensus statements in post extracorporeal cardiopulmonary resuscitation resuscitation care has been provided.

What Are the Clinical Implications?

• These findings will help guide for clinicians in the optimal management of extracorporeal cardiopulmonary resuscitation and a basis on which to develop clinical and trial post extracorporeal cardiopulmonary resuscitation protocols.

Extracorporeal cardiopulmonary resuscitation (ECPR) refers to the initiation of venoarterial extracorporeal membrane oxygenation (VA ECMO) during CPR. The adoption of ECPR for refractory in‐hospital and out‐of‐hospital cardiac arrests (OHCA) is increasing globally. ¹ Despite this, there are currently limited data to guide management after initiation of ECPR, ² and widely accepted standards of care have not been established. ³ Moreover, most current guidelines of postcardiac‐arrest care are derived from and provided for conventional CPR patients, and how applicable these are to ECPR patients is uncertain. ⁴ , ⁵ As such, observational studies suggest that a wide variety of management strategies are being used. ³ There is a need to provide structured guidance on the management of the patient who receives ECPR. In the absence of evidence from high‐quality randomized trials, consensus‐based guidelines, using a modified Delphi approach, can provide a framework to support clinical decision‐making and support standardized post‐ECPR resuscitation protocols. The aim of this study was to identify evidence gaps and propose consensus‐based recommendations for post‐ECPR resuscitation management.

Methods

Data Availability

Cumulative Delphi round reports with experts’ feedback are provided in Data S1. Further details can be accessed upon request to the corresponding author.

Consent

Informed consent was gained from all individual participants in the study.

The project was conducted in compliance with the Conducting and Reporting of Delphi Studies standards ⁶ and the Standards for Quality Improvement Reporting Excellence reporting guidelines. ⁷ The study was approved by Sydney Local Health District Human Research Ethics Committee: X24‐0292.

Study Steering Committee

M.D., N.K., and A.B. completed a similar modified Delphi process for training in ECPR scenarios and conceptualized this project. ⁸ After discussion with a broad range of clinicians at local and international meetings, including Prague ECPR School 2023, Minnesota Pinnacle ECMO Symposium 2024, and European Extracorporeal Life Support Organization (ELSO) Congress 2024, we agreed on the need for a modified Delphi consensus process on post‐ECPR treatment. M.D., A.B., and A.S. identified individuals with appropriate expertise, diversity of specialty, leadership in ECPR, and research methodology to be members of the study team. The study team comprised 11 lead authors and included 3 cardiologists (M.D., A.S., J.B.) 5 intensivists/critical care specialists (A.R., A.B., S.L., D.A., T.S.), and 1 each of anesthetics/retrieval (N.K.), cardiothoracic surgery (J.S.S.), and ECPR survivor (M.S.). Steering group members also represented geographical diversity: 3 from Australasia (M.D., N.K., A.B.), 2 from Asia‐Pacific (S.L., J.S.S.), 2 from North America (D.A., J.B.) and 3 from Europe (A.S., A.R., T.S.).

Search Strategy and Literature Review

Steering committee members (T.S.) with collaborators (G.P. and A.M.S.) systematically searched MEDLINE, Embase, and the Cochrane Library for published work. Searches for ongoing clinical trials or unpublished works were performed in the International Clinical Trials Registry Platform, Cochrane Central Register of Controlled Trials and the US clinical trials registry. Reference lists of included studies and relevant reviews were screened for additional articles. The search was completed from inception to May 19, 2025, and the search strategy was for original articles of ECPR postresuscitation care (Data S1, Figure S1).

Survey and Statement Development

Based on a scoping literature review ⁹ (Data S1) and available published reviews, ² existing non‐ECPR cardiac arrest guidelines, ⁴ , ¹⁰ , ¹¹ , ¹² institutional protocols, and their own experience, the steering committee identified a list of topics to be covered in the management of the ECPR patient post cannulation and drafted initial opinion statements for adult patients (≥18 years) who had received ECPR. These topics were presented as clinical statements to the expert panel through Research Electronic Data Capture survey either through a 9‐point Likert‐scale or by a multiple‐choice question format. These statements were discussed, rephrased, and finally approved by all the steering committee members via email correspondence and online meetings before Round 1. After development of the survey, 3 test experts (contributors P.F., M.S., J.S. not included in the expert panel or the study group) reviewed the survey for appropriate coverage length and rigor. Throughout the modified Delphi process survey reports were analyzed by the steering committee after each round in a virtual meeting, changes to the clinical statements were made, and statements were added or dropped based on the comments from the experts, where deemed necessary.

Expert Panel Selection

The steering committee identified potential expert panel members based on clinical experience with ECPR and their impact in the field, including leadership of major ECPR studies or clinical guidelines. To ensure a broad and balanced perspective, particular emphasis was placed on recruiting a diverse panel—representing a range of clinical specialties, experience, geography, and health care settings. To promote transparency and inclusivity, all invited experts were encouraged to nominate additional individuals with relevant expertise. We aimed to recruit a total of 50 experts, in accordance with the recommended panel number of >30 as outlined in the original Delphi method paper. ¹³ All panel members had equal voting rights. To avoid any bias, the steering committee members did not participate in the voting process. The voting process was online and anonymous to avoid any reciprocal influence or dredging effect.

Modified Delphi and Voting Process

The modified Delphi process —(Figure 1) was conducted using a secure, anonymized Research Electronic Data Capture survey platform. In round 1, the survey included both structured, survey‐style questions to gather quantitative data and open‐ended responses to inform the development of subsequent statements. Statements for voting on were presented using a 9‐point Likert scale, allowing experts to indicate their level of agreement or disagreement. The voting process was consistent with the Research and Development/University of California, Los Angeles Appropriateness Method. ¹⁴

Figure 1. Flow chart of the method used to develop expert opinion statements.

Flow chart of the method used by the steering committee to develop expert opinion statements Note: the same expert who partially completed survey round 1, did not complete rounds 2 and 3 of expert’s response.

Consensus was deemed to be achieved when an item reached a high level of agreement for inclusion. High level of agreement required >70% of the survey responses for a particular item meeting a score of strong agreement (ie, a score ≥7 on the 9‐point Likert scale). Statements reaching ≥70% agreement (Likert ≥7) were accepted and removed from subsequent rounds. Statements with ≤30% agreement (Likert ≤3) were excluded. Remaining statements were revised based on expert feedback and reevaluated in subsequent rounds consistent with previously published Delphi processes ⁸ , ¹⁵ , ¹⁶ , ¹⁷ and recommendations. ¹⁸ Statements not meeting high‐level consensus could be removed should the steering committee deem that they were unlikely to meet consensus. The percentage of agreement was quantified as the number of individual scores >7 divided by number of voting experts. Given the limited and low‐quality data available regarding many aspects of the selected topic, the steering committee did not use the Grading of Recommendations Assessment, Development and Evaluation system to prepare the expert statements but referred to the Appraisal of Guidelines for Research and Evaluation statement.

Data Analysis

Results of the voting process were reported as percentage with median and interquartile range. SPSS (version 29.0.1.0, IBM SPSS Statistics for Windows, Version 29.0.1.0. Armonk, NY: IBM Corp; 2023.) was used for the analysis.

Position Statements

Summarized expert statements were drafted by the steering committee from the consensus statements achieving strong consensus, measured by a median of ≥7 (for agreement) to Likert‐scale questions regardless of the round in which they were achieved. Expert position statements and the article were reviewed and approved by the experts before submission for publication.

Results

Of 54 invited experts, 53 (98%) accepted to join and commenced the process, and 52 (96%) completed all 3 rounds—Figure 1. Clinical specialty and geographical region of participating experts displayed a high degree of geographical, professional and experience diversity and are reported in Figure 2 and Tables S1 and S2. The modified Delphi process was conducted through 3 rounds between October 7, 2024 and June 11, 2025—Figure 1 and Data S2. Through round 1, 60 of 102 (59%), round 2, 50 of 84 (60%) and round 3, 16 of 24 (64%) met a high degree of consensus. Eighty‐four (67%) of final consensus statements percentage had ≥80% of experts scoring >7. On completion of the modified Delphi process, 126 (60%) of 211 statements achieved a high degree of consensus (Table S3), from which 42 expert position statements were derived (Table 1 and Tables S3 and S4). Key statements not reaching consensus and identified research priorities are provided in Table 2. Throughout the modified Delphi process, the expert panel emphasized that patient care needs to be individualized depending on the patient and clinical scenario at the time, and that these statements provide only guidance based on expert knowledge and should not be considered definitive management for patients. Making decisions on statements with very limited ECPR‐specific data is very challenging, and it was acknowledged that different approaches were to be expected.

Figure 2. Choropleth depicting the geographical distribution of expert panel participants.

Table 1.

Summarized Expert Consensus Statements

General management considerations

ECPR programs should implement standardized postresuscitation protocols and ensure coordinated care across multidisciplinary teams

2ECPR programs should regularly review morbidity and mortality, including physical and psychological impacts on survivors and their families. Follow‐up should extend to least to 6 months post ECPR survivors where possible

General clinical management

3All patients should undergo urgent cardiac catheterization post initiation of ECMO, unless clear concern of a noncardiac cause of arrest

4Sedation should be minimized and tailored to the individual, with reduction initiated once the patient is stable enough for neuroprognostication. Short‐acting sedatives and anxiolytics are preferred where possible

5Antibiotics should be considered post cannulation based on the individual risk of infection

Ventilation, oxygenation, and carbon dioxide management

6Lung‐protective strategies (eg, tidal volumes <6 mL/kg) should be used post ECPR initiation. Higher positive end‐expiratory pressure strategies (eg, >10 cm H2O) should be used in cases of pulmonary edema, atelectasis, or left ventricular dilatation

7Extreme hyperoxemia (PaO2 >300 mm Hg) should be avoided where possible. Use of a blender to titrate fraction of inspired oxygen is recommended, targeting a right radial PaO2 of 75–100 mm Hg or peripheral oxygen saturation of 94%–98%. Postoxygenator monitoring should be performed routinely to avoid hypoxemia

8Aim for normal PaCO2 levels as per standard cardiac arrest guidelines (35–45 mm Hg). Rapid drops in PaCO2 (>20 mm Hg in 24 h) should be avoided by gradual titration of the sweep gas flows based on regular arterial blood gases.

9Regular (or continuous) monitoring of both right radial and postoxygenator gases is important. Oxygen and carbon dioxide management should take into account the risk of dual circulations and differential gas tensions

Hemodynamic management

Mean arterial pressure

10Specific MAP, pulse pressure, and cardiac index targets should be used to assist the hemodynamic management of the patient. An initial MAP target of >65 mm Hg is recommended

11In patients with low MAP, prioritizing increased ECMO flows is recommended over excessive dose vasopressors, fluids, inotropes

ECMO flows and monitoring

12ECMO blood flow should target adequate systemic perfusion, guided by markers such as lactate clearance and other markers of end organ perfusion. A pulse pressure of at least 10 mm Hg is a reasonable initial target. Inotropes are an important adjunct to aid aortic valve opening and enhance perfusion as needed

13A multimodal approach is recommended to assess ECMO flow adequacy, combining clinical signs, laboratory values (eg, lactate clearance), echocardiography, and hemodynamic data. Pulmonary artery catheters or continuous monitors may be considered but are not required in all ECPR patients

14As myocardial function recovers, ECMO flow should be reduced provided systemic perfusion remains adequate

Aortic valve opening LV unloading/venting

15Maintaining consistent AV opening is an important goal post ECPR. However intermittent or absent AV opening within the first 24 h is not necessarily a sign of futility

16Failure of AV opening or the presence of spontaneous echo contrast in the LV or aortic root should prompt reassessment and optimization of LV ejection, hemodynamics, and anticoagulation

17Mechanical LV unloading/venting should be considered when there is persistent AV closure, worsening LV dilation, pulmonary edema, spontaneous echo contrast, or low pulse pressure (<5 mm Hg) despite optimization

18Invasive LV venting is not required in all ECPR patients, but centers should have a defined strategy for its use. The optimal venting method/device remains uncertain

19Interventions that improve AV opening (eg, increased inotropes, ECMO flow reduction or afterload reduction) should be balanced against maintaining adequate systemic and cerebral perfusion

20Consider afterload reduction if MAP >80 mm Hg with signs of LV distension, reduced pulse pressure or impaired AV opening

Coagulation and hematological management

21ECPR centers should establish protocols for anticoagulation and bleeding management, with anticoagulation testing performed at least twice daily or more frequently when adjustments are made. Bleeding events should be defined in accordance with international guidelines (eg, Bleeding Academic Research Consortium, International Society on Thrombosis and Haemostasis)

22ECMO flow rates should be maintained at sufficiently high levels (eg, >2.0 L per min) if anticoagulation is withheld to minimize thrombosis risk

23In nonbleeding patients, a red blood cell transfusion if hemoglobin level is <80 g/L and a platelet transfusion if the level is <50 000/mL is reasonable

Temperature and complication management

24Fever (>37.5 °C) should be avoided. ECPR‐specific guidelines for target temperature management should be established

25Routine head‐to‐pelvis CT imaging should be performed post resuscitation to assess complications or identify the cause of cardiac arrest. Early neuroprognostication on CT imaging should be avoided unless catastrophic injury is present

26Echocardiography should be performed within 24 h of intensive care unit admission, with ongoing assessments thereafter. Pericardial effusions may lack classic signs of tamponade on ECMO and should be closely monitored and drained if they impede LV ejection or ECMO weaning

27A distal perfusion cannula should be considered for all ECPR patients. Near‐infrared spectroscopy should not be solely relied upon for diagnosing limb ischemia. Postdecannulation deep vein thrombosis surveillance (eg, ultrasound screening) is recommended

28There should be consideration for increased nurse‐to‐patient ratios compared with typical intensive care in the first 24 h post ECPR initiation. Pressure area care and turning of ECPR patients should follow standard intensive care unit practices

Weaning from ECMO support

29Weaning from ECMO should follow a standardized protocol, incorporating echocardiography and weaning studies

30A postdecannulation strategy should be defined in advance, including clear criteria for potential reinitiation of ECMO support

31Failure to wean from ECMO should lead to consideration of ventricular assist devices, cardiac transplantation, or palliation, depending on the patient’s condition, prognosis, and values

32Routine therapeutic anticoagulation is not required post decannulation unless clinically indicated

Withdrawal of life support and neuroprognostication

33Avoid early prognostication post ECPR initiation, as multiorgan dysfunction and severe myocardial impairment post ECPR may be reversible. The presence of these alone should not be used for early prognostic decisions

34ECPR patients may take longer to regain consciousness than non‐ECPR OHCA patients. In the immediate period post ECPR decisions based on early neurological findings (eg, Glasgow Coma Scale score or lack of brainstem reflexes) should be avoided

35Brain imaging should be part of multimodal neuroprognostication, as recommended in OHCA neuroprognostication guidelines

36Additional lifesaving interventions (eg, surgery) should not be routinely withheld on the basis of presumed poor prognosis, and should be considered on a case‐by‐case basis

37Neuroprognostication should be delayed until at least 72 h in ECPR patients (as per existing guidelines) who do not meet brain death criteria or have a clear catastrophic injury. Neuroprognostication should follow existing multimodal guidelines for OHCA. ECPR‐specific guidelines should be developed in the future

38Somatosensory evoked potentials and electroencephalograms can be used in ECPR patients, following existing guidelines for OHCA

39Brain death determination is possible on ECMO and should follow local guidelines. Performance of a reliable apnea test on venoarterial ECMO requires simultaneous measurement of both right radial and postoxygenator blood gases. Confirmatory brain death testing (eg, imaging, nuclear perfusion scan) should be used when apnea testing is challenging

40Organ donation can proceed after brain death or circulatory death. ECPR should not be considered a barrier to organ donation, and should be conducted according to local guidelines

41Support and follow‐up (>1 year) should be provided for ECPR survivors where possible, as they can show improvements in functional status

42Life support should not be withdrawn prematurely in ECPR patients without clear evidence of poor prognosis. ECPR specific guidelines for withdrawal of life support should be developed. Early discussions with family members about patient expectations are essential and decisions should next into account patients, and next of kin/family wishes

AV indicates aortic valve; CT, computer tomography; ECMO, extracorporeal membrane oxygenation; ECPR, extracorporeal cardiopulmonary resuscitation; LV, left ventricle; MAP, ean arterial pressure; OHCA, out‐of‐hospital cardiac arrest; PaCO₂, partial pressure of carbon dioxide; and PaO₂, partial pressure of oxygen.

Table 2.

Key Statements Not Meeting Consensus and Identified Future Research Priorities

Initial ECMO flow rate	Targeted temperature management for ECPR patients
Specific ECMO flow rate targets	ECPR‐specific neuroprognostication and withdrawal of life support guidelines
Postoxygenator partial pressure of oxygen targets	Anticoagulation strategies
Initial sweep (fresh gas flow) settings
Nursing to patient care ratios
The use of neuronal specific enolase in prognostication

ECMO indicates extracorporeal membrane oxygenation; and ECPR, extracorporeal cardiopulmonary resuscitation.

Discussion

Our modified Delphi process that included a diverse, multinational range of ECPR experts reached consensus on 126 statements across key postresuscitation domains providing 42 summarized consensus statements to guide postresuscitation management of the ECPR patient.

Domain: General Management Considerations

The management of patients after successful implementation of VA ECMO in the context of ECPR requires substantial human and material resources and expertise. Patients after ECPR often experience additional problems and complications, they have an uncertain prognosis, and mortality is high. Preliminary evidence based on observational data suggested improved outcomes in higher volume ECPR centers ¹⁹ , ²⁰ with more experienced and structured programs. ²¹ , ²² Consequently, previous expert statements recommend ECPR patients be managed in expert centers. ²³ However, ECPR events are infrequent, and the feasibility of transporting ECPR patients to dedicated ECPR hospitals within requisite time frames is not always possible. ²⁴ , ²⁵ A recent multicenter randomized controlled trial of ECPR compared with continuation of conventional CPR demonstrated challenges with variable ECPR experience across receiving sites. ²⁶ Our expert panel strongly agreed on the need for common postresuscitation protocols for trials or services operating across multiple receiving hospitals, the need for a multidisciplinary approach to such management, and strong governance and quality assurance of ECPR programs. The development and publishing of postresuscitation protocols may also improve and align the provision of care and enable comparison of ECPR outcomes across jurisdictions by reducing practice variability. In keeping with existing non‐ECPR guidelines, sedation and analgesia for ECPR patients should be preferred to short‐acting agents, tailored to individual patients and reduced where feasible.

Domain: Ventilation, Oxygenation, and Carbon‐Dioxide Management

Hyperoxemia has been consistently associated with poorer outcomes in observational studies of patients experiencing OHCA resuscitated with conventional CPR. ²⁷ However, randomized controlled trials in non‐ECPR populations with OHCA have not confirmed a benefit from restrictive oxygen targets. ²⁸ In VA ECMO, supraphysiological oxygen partial pressures are common—often exceeding those seen with invasive mechanical ventilation—and have similarly been linked to increased mortality in observational data. ²⁹

Oxygen “blending,” which involves mixing air with oxygen to titrate delivery through the membrane lung, enables more precise control of oxygenation but may increase the risk of hypoxemia, itself associated with adverse outcomes. ³⁰ The BLENDER (Blend to Limit Oxygen in ECMO) trial, published shortly before the initiation of this Delphi process, evaluated 300 VA ECMO patients in Australia and New Zealand. It compared a restrictive oxygen strategy (peripheral oxygen saturation 92%–96%) with a liberal strategy (peripheral oxygen saturation 97%–100%) and found no significant difference in intensive care unit free days. Although not powered specifically for ECPR, a subgroup analysis of 108 ECPR patients similarly showed no benefit from a restrictive oxygen approach and revealed more protocol violations, primarily due to hypoxemia. ³¹ Given conflicting observational and trial data, these recommendations should be interpreted as expert consensus rather than evidence‐based directives.

Despite these findings, the expert panel reached strong consensus on several key principles: the importance of avoiding both extreme hyperoxemia and hypoxemia, the routine use of oxygen blenders in all ECPR patients, and the need for close monitoring of oxygenation. This includes simultaneous assessment of right radial and postmembrane oxygen levels, with a recommended right radial peripheral oxygen saturation target >94%.

The interim ELSO guidelines, which suggest maintaining oxygen saturations between 94% and 98% or partial pressure of oxygen (PaO₂) 75–100 mm Hg while avoiding hypoxemia (PaO₂ <60 mm Hg), were considered a reasonable framework for oxygen management in ECPR. ³² Although consensus on a specific postmembrane oxygen target was not achieved after 3 Delphi rounds, current ELSO guidance recommends a postmembrane PaO₂ of approximately 150 mm Hg. ³³ Notably, near‐consensus (69% agreement) was reached on a postmembrane oxygen target range of 150 to300 mm Hg, suggesting this may be a clinically acceptable range pending further evidence.

Partial pressure of arterial carbon dioxide (PaCO₂) levels and variations in PaCO₂ post ECPR have also been associated with increased mortality in some, but not all, observational studies of ECPR patients. ³² , ³⁴ In a retrospective analysis of the ELSO registry, reduced complications (including intracranial hemorrhage and death) were reported in patients with mild hypercapnia (PaCO₂ 45–54 mm Hg), compared with higher levels of carbon dioxide, before ECPR cannulation. ³⁵ Recently, a multicenter, retrospective observational study using the Japanese Association of Acute Medicine—Out‐of‐Hospital Cardiac Arrest Registry reported that at both 6 and 24 hours, post‐ECMO initiation for OHCA, in comparison with high normocapnia (PaCO₂ 40–50 mm Hg), hypocapnia, low normocapnia, mild hypercapnia, and moderate to severe hypercapnia were all associated with a reduced probability of good functional outcomes. ³⁶ Aiming for eucapnia post ECPR, was supported by the expert panel, in keeping with non‐ECPR OHCA guidelines. ¹⁰

Significant and rapid reductions in PaCO₂ are achievable with ECMO and have been associated with increased mortality in some studies. ²⁶ , ³² Although our expert panel agreed on targeting eucapnia (35–45 mm Hg), in keeping with non‐ECPR OHCA guidelines and emphasized the importance of avoiding rapid PaCO₂ reductions and implementing protocolized regular blood gas monitoring, the optimal initial sweep gas flow could not reach consensus.

Domain: Hemodynamic Management

Mean Arterial Pressure and ECMO Flow Management

The optimal mean arterial pressure (MAP) target is not yet established for ECPR patients. The possible empirical benefit of maintaining high MAP for promoting cerebral perfusion must be weighed against the risk of left ventricle (LV) distension leading to LV stasis, thrombosis, and pulmonary edema. ³⁴ Current ELSO guidelines recommend maintaining MAP between 60 and 80 mm Hg by vasopressor titration while maintaining blood flow at 3 to 4 L/min. ³² Despite a wide range of minimum MAP targets identified in Round 1, a high degree of consensus was gained on maintaining MAP ≥65 mm Hg, achieved with a preference for increasing blood flow, when possible, over the use of vasoactive medications. Intervening on higher MAPs (eg, >80 mm Hg) in the presence of LV distension and pulmonary congestion was deemed reasonable. The expert group emphasized the need to individualize MAP targets and to be willing to review targets during an ECMO run, all the while maintaining adequate systemic and cerebral perfusion.

The setting, monitoring, and maintenance of ECMO blood flow may be conceptualized as initial ECMO settings and ongoing flow management. In contrast to existing guidelines, ³² a high level of consensus was not able to be gained on initial flow rate nor how to estimate adequate perfusion to titrate initial flow (eg, based on body surface area or cardiac index). Aside from lactate clearance and the importance of daily echocardiography, no individual marker or measurement tool of perfusion met high consensus to guide ECMO blood flow titration and a multimodal approach to flow delivery was preferred.

Aortic Valve Opening

Failure or only intermittent aortic valve opening is a known complication often occurring in the early period after ECPR initiation and low pulse pressure after ECPR initiation has been associated with worse clinical outcomes. ³⁵ Although this is clinically challenging and needs to be addressed promptly, ³⁵ our expert panel did not feel failure of or only intermittent aortic valve opening was an indicator of futility especially when present in the first 24 hours post cardiac arrest. Further, according to the expert panel, it may be tolerated in the short term in the absence of LV distension, clot, or pulmonary edema if systemic perfusion requirements are being met. Commonly used management of intermittent or nonaortic valve opening includes commencement or increasing inotropic agents, ³⁴ reduction of ECMO flows, ³⁴ judicious volume replacement and implementation of an LV venting strategy ³⁶ (see next section). According to expert consensus, reduction of ECMO flows to encourage aortic valve opening is commonly implemented but should not occur at the expense of adequate systemic and cerebral perfusion. Lower ECMO flows have been associated with lower cerebral oxygen delivery and cerebral hypoxemia in already dysregulated cerebral vascular bed ³⁷ and careful consideration is required before implementing such strategy.

Left Ventricular Unloading/Venting

Peripheral implementation of VA ECMO via the femoral vessels results in retrograde aortic blood flow. It is traditionally conceived that VA ECMO increases afterload on the LV, ³⁸ thereby increasing LV end‐diastolic pressure, which in turn may lead to LV distension, increased wall stress, reduced coronary perfusion, risk of LV stasis, and thrombus formation and pulmonary edema. ³⁹ Although emerging real world human data conflict with these findings, reporting lower LV end‐diastolic pressure and LV energetic consumption at higher ECMO blood flow rates, with a modest increase in afterload, ⁴⁰ concerns over the potential sequelae of increased afterload and raised LV end‐diastolic pressure have led to adjunctive strategies to decompress the LV. These include LV unloading (during which interventions are focused on reducing LV work) or LV venting (during which interventions reduce LV filling pressures but do not necessarily reduce LV work). Unloading and or venting of the left ventricle during ECPR is most often completed by using an intra‐aortic balloon pump or microaxial flow device (eg, Impella) ³⁹ but may include other cannulation strategies such as venous drainage combined with pulmonary arterial or left atrial reinfusion.

In patients with cardiogenic shock treated with ECMO, retrospective observational studies and a meta‐analysis have suggested a survival benefit for the use of an LV unloading device with VA ECMO ⁴¹ , ⁴² and a recent meta‐analyses of unadjusted observational studies of ECPR patients treated with a microaxial flow device or intra‐aortic balloon pump also proposed a survival benefit. ⁴³ , ⁴⁴ However, propensity‐matched studies have not reported a benefit from microaxial flow devices and findings in intra‐aortic balloon pump are inconsistent. ⁴³ , ⁴⁵ Moreover, a recent target emulation study using >3000 ELSO registry patients and 621 matched pairs reported LV unloading was not associated with increased survival in ECPR patients and may increase complication rates calling into question the utility of venting or unloading the LV. ⁴⁶

Our expert panel has provided some guidance on LV venting and unloading. First, not all ECPR patients require prophylactic LV venting or unloading and the optimal LV venting strategy has yet to be determined. Second, when considering the decision to vent or unload the LV, multiple factors should be assessed rather than a single variable. Third, sites should have a defined LV venting or unloading strategy should it be deemed necessary. Finally, the presence of a closed aortic valve, progressive LV dilation, pulmonary edema, pulse pressure <5 mm Hg, or spontaneous echo‐contrast in the aortic root or LV—despite optimization (eg, ventilator, inotropes)—should prompt consideration for implementing an LV unloading strategy.

Domain: Coagulation and Hematological Management

Although specific anticoagulants and anticoagulation protocols were not assessed, the need for a predefined anticoagulation protocol met consensus, as did the need for anticoagulation monitoring tests at least twice daily if stable or more often when commencing or changing anticoagulation. Routine therapeutic anticoagulation post resuscitation is not required and should be implemented when other indications for anticoagulation are present. Bleeding and thrombosis in ECPR patients are common, ⁴⁷ , ⁴⁸ with major bleeding seen in up to 22% to 68% ⁴⁷ , ⁴⁹ of patients and thrombosis in 24%. ⁴⁷ Substantial heterogeneity in antithrombotic practices among ECMO centers exists, potentially contributing to variable device–associated bleeding and thrombotic complications. ⁵⁰ In our study, there was a wide range of approaches to surveillance and management of bleeding and transfusion practices. The use of thromboelastic tests (TEG and ROTEM) in addition to standard coagulation tests was not thought to add value either in monitoring of patients’ coagulation profile or in management of bleeding with an individualized approach being favored. Guidelines on thresholds for red blood cell and platelet transfusion were provided; however, consensus on fibrinogen and antithrombin replacement targets was not met.

Domain: Temperature and Complication Management

The optimal temperature control strategy in ECPR patients remains unknown. The 2023 American Heart Association guidelines for non‐ECPR‐treated OHCA recommended an expanded target temperature range of 32 °C to 37.5 °C. However, ECPR patients are generally characterized by prolonged low‐flow times, prolonged cerebral hypoperfusion and hypoxia, increasing the risk of hypoxic brain damage. Immediate cerebral blood flood on initiation of ECMO theoretically may increase the risk of reperfusion injury in ECPR patients. ⁵¹ Therefore, it is unclear to what extent data and clinical experience from target temperature management in non‐ECPR patients are applicable to patients after ECPR.

ECMO circuits can provide a rapid and stable method of intravascular cooling, which in theory may contribute to better neurological protection, though no benefit in survival has been demonstrated. ⁵² Our expert panel’s opinion on targeted temperature management in ECMO patients varied considerably. However, consensus was gained on the need to avoid fever (>37.5 °C), that the optimal strategy is not known, and that further research is a priority for ECPR patients with a need to develop ECPR‐specific guidelines for targeted temperature management. In the interim, targeting normothermia is reasonable.

Trauma

Traumatic complications of both prolonged CPR and initiation of ECMO are common in ECPR patients. ⁵³ Early computed tomography (CT) after ECPR can identify pathology related to the cardiac arrest or sequalae from prolonged CPR but carries risks of intrahospital transfer of a critically ill patient. As yet, limited nonrandomized retrospective data do not demonstrate a survival benefit of routine full‐body CT scans. ⁵⁴ However, early CT does allow for early identification and treatment of concomitant injuries, ⁵⁵ with 1 study reporting similar outcomes between patients with chest wall injury from CPR to those with no chest wall injuries ⁵⁶ when an early imaging strategy was employed. Preemptive surveillance for complications as part of a bundle of care was identified with a high degree of consensus as critical for postresuscitation practice. This included routine “CT pan‐scan” early after initiation of ECPR and coronary angiography, routine and regular echocardiography (at least daily), surveillance of deep vein thrombosis by ultrasound upon decannulation, and routine implantation of distal perfusion cannulae in all ECPR patients, where possible. Routine insertion of distal perfusion cannulae is reported to reduce ischemic events, ⁵⁷ whereas the utility of near‐infrared spectroscopy as a sole method to monitor for distal limb perfusion (without distal perfusion cannulae) is less clear. A recent systematic review reported only moderate pooled sensitivity and specificity for detection of ischemia. ⁵⁸ Although near‐infrared spectroscopy is a useful adjunct, the expert panel did not believe its use obviates the need for distal perfusion cannulae and clinical assessment.

Nursing Workload

Studies on nursing workloads have reported that significantly increased nursing hours are required when managing ECMO/ECPR patients ⁵⁹ and a 1:1 nurse‐to‐patient ratio for ECPR patients has been proposed. ⁶⁰ In non‐ECMO intensive care patients, safe nurse staffing levels have been associated with a 14% reduction in hospital mortality, shorter intensive care unit lengths of stay, and a 20% improvement in infection prevention. ⁶¹ Existing ELSO ECPR Interim Guidelines ³² do not prescribe explicit nurse‐to‐patient ratios for ECPR cases, instead emphasizing the importance of appropriate personnel, comprehensive training, competency maintenance, and adequate programmatic resources. Our expert group acknowledged the substantial nursing burden an ECPR patient may demand, especially in the first 24 hours post cannulation. However, there was no consensus on a particular nurse‐to‐patient ratio for ECPR patients, likely owing to the substantial differences in nursing resources across different health systems.

Domain: Weaning From ECMO

Systematic and protocolized ECMO weaning protocols may reduce rates of recannulation and weaning failure ⁶² and help predict weaning success. ⁶³ Consistent with existing ELSO guidelines, ⁶⁴ the panel strongly agreed on the need for a standardized approach to ECMO weaning, with a preemptive plan before decannulation in the event of a failed decannulation.

Domain: Neuroprognostication and Withdrawal of Life Support Therapies

Prognostication, including neuroprognostication, and decisions about withdrawal of life sustaining therapies (WLST) in ECPR patients are challenging. Thus far, 2 predictive ECPR survival scores have been published and validated, reporting only modest predictive value; the TiPS65 score for OHCA ⁶⁵ (area under the curve 0.72) and the Resuscitation Using ECPR During In‐Hospital Cardiac Arrest score for in‐hospital cardiac arrest ⁶⁶ (area under the curve 0.68). Thus far, neuroprognostication in ECPR is reliant on guidelines for non‐ECPR out‐hospital cardiac arrest. ⁵ The utility of neuroprognostication tests and guidelines for ECPR, which tend to have longer low flow times and more profound derangement of physiology, are only beginning to be elucidated. Premature WLST on ECPR ⁶⁷ is common, with over half (55.5%) of patients in one ELSO registry analysis having experienced WLST within the first 72 hours of ECMO support. ⁶⁸ A recent substudy of the SAVE J‐II (Study of Advanced Cardiac Life Support for Ventricular Fibrillation With Extracorporeal Circulation in Japan) study reported ∼30% of patients underwent WLST, nearly 80% of which were attributed to perceived unfavorable neurological prognosis or perceived unfavorable cardiopulmonary prognosis. ⁶⁹

Throughout this modified Delphi process and the virtual online meeting, the expert panel acknowledged the challenges in prognostication and WLST decisions in ECPR patients. Understanding the patient’s prior wishes, considering the next of kin’s perspective, and maintaining clinicians’ ability to make decisions about futility were discussed and viewed as important. The rxpert panel reinforced the critical need to avoid early presumptions on prognosis in ECPR patients, early neuroprognostication and premature WLST. Further, additional interventions, including surgery, if required, should be provided unless there are clear reasons to withhold such interventions.

The timing of neuroprognostication, although subject to debate, was agreed to be at least 72 hours, consistent with international guidelines ⁵ and seminal OHCA trials. ⁷⁰ , ⁷¹ Experts agreed on the need for multi‐modal assessments in keeping with these guidelines, including electroencephalogram, somatosensory evoked potentials, and CT imaging, but did not agree on neuronal specific enolase (NSE) despite it being recommended as part of multimodal assessments in cardiac arrest guidelines. ⁵ Despite NSE being prone to hemolysis (as may occur with ECMO), its utility in prognostication of ECPR arrests is reported. ⁷² The Prague‐OHCA biomarker substudy reported utility of NSE in both CPR and ECPR groups (area under the curve 0.89 versus 0.78; 0.9 versus 0.9; 0.91 versus 0.9) in predicting neurological outcome. The optimal cutoff points of NSE were higher in ECPR when compared with conventional CPR. ⁷³ The availability of NSE is less common in the United States, and on sensitivity analysis a substantial difference was seen in North America (deemed NSE unimportant) versus Europe (deemed NSE as important). ECPR‐specific neuroprognostication guidelines were recommended to be considered by the expert panel.

Brain death testing and organ donation post ECPR were both thought by the expert panel to be possible. ⁷⁴ Organ donation in ECPR patients is increasing, ⁷⁵ , ⁷⁶ in line with international consensus statements ⁷⁷ , ⁷⁸ supporting such practices.

For surviving patients, the expert panel reinforced the need for longer‐term follow. The current, albeit limited, data report that functional status and health‐related quality of life of ECPR patients are similar to conventionally treated cardiac arrests ⁷⁹ or other cardiac interventions ⁸⁰ and improve over time. ⁸¹ Further long‐term data as to the physical, psychological, and social impacts of ECPR survivorship are required.

Limitations

This study has several important limitations. First, ECPR patients represent a highly heterogeneous population, with diverse clinical presentations and trajectories. This variability may limit the consistency and generalizability of the position statements across all patient groups. Second, although the consensus statements offer overarching principles for post‐ECPR management, individualized approaches may be necessary for specific subpopulations—a nuance repeatedly emphasized by the expert panel.

Third, despite efforts to recruit a broad and internationally representative panel, substantial global variation persists in access to ECPR, availability of equipment, and clinical expertise. These disparities may influence the applicability of certain recommendations across different health care systems.

Additionally, due to the extensive scope of topics and the length of the survey process, we did not assess the stability of responses over time. Finally, as the field continues to evolve, emerging evidence may refine or challenge current best practices, underscoring the need for ongoing research and periodic reassessment of consensus guidelines.

Conclusions

This modified Delphi study, informed by a diverse international panel, identified key domains of post‐ECPR care, and resulted in consensus‐based position statements addressing key aspects of care. In the absence of robust empirical evidence, these expert‐derived recommendations offer a valuable framework to support clinical decision‐making and the development of standardized post‐ECPR resuscitation protocols.

Appendix

The members of the CARES Delphi Expert group:

Marta Velia Antonini (Anesthesia and Intensive Care Unit, Bufalini Hospital—AUSL della Romagna, Cesena, Italy and Department of Biomedical, Metabolic and Neural Sciences, University of Modena & Reggio Emilia, Modena, Italy, Eduard Argudo (Intensive Care Medicine department, Vall d’Hebron Barcelona Hospital Campus, Barcelona, Spain and School of Medicine, University of Barcelona, Barcelona, Spain), Jan Bělohlávek (2nd Department of Medicine—Department of Cardiovascular Medicine, First Faculty of Medicine, Charles University in Prague and General University Hospital, Prague, Czech Republic), Yih‐Sharng Chen (Cardiovascular Center National Taiwan University Hospital, Taipei, Taiwan and Division of Cardiovascular Surgery, Department of Surgery National Taiwan University Hospital and College of Medicine Taiwan), Sung‐Min Cho (Department of Surgery, Division of Cardiac Surgery, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA and Department of Neurology, Neurosurgery and Anesthesiology and Critical Care Medicine, Neuroscience Critical Care Division, Johns Hopkins University school of Medicine, Baltimore, Maryland, USA), Genex Correa, Dirk Donker (Cardiovascular and Respiratory Physiology, TechMed Center, University of Twente, Enschede and The Netherlands and Intensive Care Center, University Medical Center Utrecht, Utrecht, The Netherlands), Vadim Gudzenko (Division of Critical Care Medicine, Department of Anesthesiology and Perioperative Medicine, David Geffen School of Medicine at University of California‐Los Angeles, Los Angeles, California), Toru Hifumi (St. Luke’s International Hospital, Department of Emergency and Critical Care Medicine, 9–1 Akashi‐cho, Chuo‐ku, Tokyo 104–8560, Japan), Cindy Hsu Department of Emergency Medicine, University of Michigan, Ann Arbor, Michigan, USA and The Weil Institute for Critical Care Research and Innovation, Ann Arbor, Michigan, USA), Shingo Ichiba (Department of Anesthesiology, Nippon Medical School, Tokyo, Japan and Department of Surgical Intensive Care Medicine, Nippon Medical School Hospital, Tokyo, Japan, Akihiko Inoue (Department of Emergency and Critical Care Medicine Hyogo Emergency Medical Center Kobe Japan), Jan Matthias Kruse (Department of Nephrology and Medical Intensive Care, Charité‐Universitäts‐Medizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt‐Universität zu Berlin, Berlin, Germany), Keibun Liu (Non for Profit Organization ICU Collaboration Network (ICON) Tokyo Japan and Critical Care Research Group The Prince Charles Hospital, Chermside, Queensland Australia), Roberto Lorruso (Cardio‐Thoracic Surgery Department, and Cardiovascular Research Institute Maastricht, Maastricht, The Netherlands), Jonathan Marinaro (Department of Emergency Medicine, Division of Prehospital Care, University of New Mexico School of Medicine, Albuquerque, New Mexico, USA and UNM Center for Adult Critical Care, University of New Mexico School of Medicine, Albuquerque, New Mexico, USA), Christian Meuwese (Department of Intensive Care Medicine, Erasmus Medical Center, Rotterdam, the Netherlands; Department of Cardiology, Erasmus Medical Center, Rotterdam, the Netherlands), Dinis Dos Reis Miranda (Helicopter Emergency Medical Services, Trauma Centre Zuid‐West Nederland and Erasmus University Medical Centre, Rotterdam, The Netherlands and Department of Intensive Care, Erasmus University Medical Centre, Dr. Rotterdam, The Netherlands), Hiromichi Naito (Department of Emergency, Critical Care, and Disaster Medicine, Okayama University Faculty of Medicine, Graduate School of Medicine, Dentistry, and Pharmaceutical Sciences, Okayama, Japan), Nchafatso Obonyo (IDeAL, KEMRI‐Wellcome Trust Research Programme, Kilifi, Kenya and The Wellcome Trust Centre for Global Health Research, Imperial College London, London, UK), Yohei Okada (Department of Preventive Services, School of Public Health/Graduate School of Medicine, Kyoto University, Yoshida‐Konoe‐Cho, Sakyo, Kyoto, Japan and Health Services and Systems Research, Duke‐NUS Medical School, National University of Singapore, Singapore, Singapore), Marcel Van de Poll (Department of Intensive Care Medicine, Maastricht University Medical Centre, Maastricht, the Netherlands; Department of Surgery, Maastricht University Medical Centre, Maastricht, the Netherlands and School for Nutrition and Translational Research in Metabolism, Maastricht University, Maastricht, the Netherlands)., Michael Poppe (Department of Emergency Medicine, Medical University of Vienna, Vienna, Austria), Carla Richardson (University Hospitals Birmingham, England, UK), Sacha Richardson (The Alfred Hospital, Melbourne, Australia and Department of Public Health and Preventive Medicine, Monash University, Melbourne, Australia), Claudio Sandroni (Department of Intensive Care, Emergency Medicine and Anaesthesiology—Fondazione Policlinico Universitario A. Gemelli, IRCCS, Italy; Catholic University of the Sacred Heart, Rome, Italy), Mara Scandroglio (Department of Anesthesia and Intensive Care, IRCCS San Raffaele Scientific Institute, 20 132 Milan, Italy), Zachary Shinar (Department of Emergency Medicine, Sharp Memorial Hospital, San Diego, CA, USA), Ben Singer (Barts Health NHS Trust, London’s Air Ambulance, United Kingdom), Justyna Swol (Department of Respiratory Medicine, Paracelsus Medical University, Nuremberg, Germany and Cardiac Surgery and Extracorporeal Life Support, Department of Cardio‐Thoracic Surgery, ECLS Program, Heart & Vascular Centre MUMC+, Cardiovascular Research Institute Maastricht (CARIM), Maastricht, Netherlands), Fabio Taccone (Department of Intensive Care, Erasme Hospital, Université Libre de Bruxelles, Brussels, Belgium), Joseph Tonna (Division of Cardiothoracic Surgery, Department of Surgery and Department of Emergency Medicine, University of Utah Health, Salt Lake City, USA), Georg Trummer (Department of Emergency Medicine, Medical Center‐University of Freiburg, Medical Faculty‐University of Freiburg, Freiburg, Germany), Charudatt Vaity (Critical Care, Fortis Hospital, Mulund, India), Leen Vercaemst (Department of Perfusion, University Hospitals Leuven, Herestraat, Belgium), Melissa Ann Vogelsong (Division of Cardiac Anesthesia, Department of Anesthesiology, Stanford University, Stanford, California, USA), Chih‐Hsien Wang (Department of Surgery, National Taiwan University Hospital, National Taiwan University College of Medicine, Taipei, Taiwan), Ryo Yamamoto (Department of Emergency and Critical Care Medicine Keio University School of Medicine Tokyo Japan), Shoji Yokobori (Department of Emergency and Critical Care Medicine, Nippon Medical School, Bunkyo, Japan), Bishoy Zakhary (Division of Pulmonary and Critical Care Medicine, Oregon Health and Science University, Portland, Oregan, USA).

University of Minnesota, Minneapolis, Minnesota, USA: Alejandra Gutierrez Bernal, Adam Gottula, Jeff Dellavolpe, Andrea M. Elliot, Demetris Yannopolous.

Department of Emergency Medicine, Sapporo Medical University, Sapporo, Japan: Naofumi Bunya.

Harefield Hospital, Part of Guy’s and St Thomas’ NHS Foundation Trust, London, United Kingdom: Clara Hernandez Caballero, Ana Sofia Dacostapinto.

Department of Internal Medicine II, University Hospital; Regensburg, Regensburg D‐93053, Germany: Alexander Dietl, Thomas Muller.

SAMU de Paris, Necker University Hospital, Assistance Publique‐Hôpitaux de Paris, Paris, France: Alice Hutin, Lionel Lamhaut.

Scott D. Weingart (Department of Emergency Critical Care, Nassau University Medical Center, East Meadow, NY, USA).

Sources of Funding

Mark Dennis is supported by National Heart Foundation Future Leader Fellowship (#107174) and the New South Wales Health, Early and Senior Career Researcher Scheme. Aidan Burrell is supported by a National Health and Medical Research Council Emerging Leader grant (#2010110).

Disclosures

None.

Supporting information

Data S1–S2

Tables S1–S4

Figure S1