Definition and management of right ventricular injury in adult patients receiving extracorporeal membrane oxygenation for respiratory support using the Delphi method: a PRORVnet study. Expert position statements

Abstract

Purpose:

Veno-venous extracorporeal membrane oxygenation (VV-ECMO) is an integral part of the management algorithm of patients with severe respiratory failure refractory to evidence-based conventional treatments. Right ventricular injury (RVI) pertaining to abnormalities in the dimensions and/or function of the right ventricle (RV) in the context of VV-ECMO significantly influences mortality. However, in the absence of a universally accepted RVI definition and evidence-based guidance for the management of RVI in this very high-risk patient cohort, variations in clinical practice continue to exist.

Methods:

Following a systematic search of the literature, an international Steering Committee consisting of eight healthcare professionals involved in the management of patients receiving ECMO identified domains and knowledge gaps pertaining to RVI definition and management where the evidence is limited or ambiguous. Using a Delphi process, an international panel of 52 Experts developed Expert position statements in those areas. The process also conferred RV-centric overarching open questions for future research. Consensus was defined as achieved when 70% or more of the Experts agreed or disagreed on a Likert-scale statement or when 80% or more of the Experts agreed on a particular option in multiple choice questions.

Results:

The Delphi process was conducted through four rounds and consensus was achieved on 31 (89%) of 35 statements from which 24 Expert position statements were derived. Expert position statements provided recommendations for RVI nomenclature in the setting of VV-ECMO, a multi-modal diagnostic approach to RVI, the timing and parameters of diagnostic echocardiography, and VV-ECMO settings during RVI assessment and management. Consensus was not reached on RV-protective driving pressure thresholds or the effect of prone positioning on patient-centric outcomes.

Conclusion:

The proposed definition of RVI in the context of VV-ECMO needs to be validated through a systematic aggregation of data across studies. Until further evidence emerges, the Expert position statements can guide informed decision-making in the management of these patients.

Introduction

Patients with severe forms of respiratory failure of any etiology are at high risk of developing right ventricular injury (RVI) which is characterized by abnormalities in the dimensions and/or function of the right ventricle (RV), and is associated with significant mortality [1]. One of the main pathophysiological mechanisms of RVI in patients with respiratory failure is pulmonary vasoconstriction caused by hypoxemia and/or hypercapnic acidemia leading to elevated RV afterload. Another potential contributor to increased RV afterload is elevated transpulmonary pressure caused by invasive mechanical ventilation [2].

Veno-venous extracorporeal membrane oxygenation (VV-ECMO) is increasingly being used as part of the management algorithm for patients with respiratory failure in whom evidence-based conventional ventilation strategies have failed to maintain adequate gas exchange at non-injurious levels of ventilation [3]. VV-ECMO supplies oxygen and removes carbon dioxide, enabling lung protective mechanical ventilation by facilitating gas exchange through a membrane lung allowing for a marked reduction in the intensity of ventilation [4, 5].

In patients with severe respiratory failure, VV-ECMO has the potential to ameliorate RVI through various mechanisms, including the alleviation of pulmonary arterial vasoconstriction, through reversal of hypercapnia, hypoxemia and acidemia, and reduction of transpulmonary pressure and alveolar overdistension through ultra-lung protective ventilation strategies [6, 7]. Despite this theoretical biophysical protection of the RV, RVI may persist or even worsen during VV-ECMO support and has been associated with high mortality [8]. Possible mechanisms include: (a) persistent pulmonary vascular dysfunction [9], (b) untreated pulmonary embolism, microvascular thrombosis or immunothrombosis [10], (c) partial reversal of metabolic derangement [11], and (d) reduction of positive end-expiratory pressure after VV-ECMO application which may cause atelectasis leading to high RV afterload (due to extra-alveolar vessels resistance overcoming alveolar vessels resistance) [7].

There is currently no universally accepted definition for RVI in patients receiving VV-ECMO. This may result in underdiagnosis of this clinical entity, heterogeneity in the patient inclusion for research, and clinical practice variations potentially affecting patient-centered outcomes. There is, therefore, a need to standardize diagnostic criteria as well as clarify therapeutic management. The objective of the current study is to generate consensus Expert position statements on the definition and management of RVI in adult patients receiving ECMO for respiratory support using a Delphi process.

Methods

A Delphi process was used to generate consensus [12–15]. We followed ‘Guidance on Conducting and Reporting Delphi Studies’ (CREDES) guidelines for presenting the Delphi study results (electronic supplementary file 1, Table 1) [15].

Table 1.

Final stability and consensus results of the Delphi process

Section 1: Definition of RVI	Agree (%)	Neutral (%)	Disagree (%)	Median (IQR)	χ2 p–value
1. There is a need for a consensus definition of RVI in patients receiving ECMO for respiratory support (veno–venous ECMO).	98	0	2	7(0)	0.05
2. RVI in patients receiving veno–venous ECMO can be defined as a clinical syndrome with a spectrum of abnormal RV biomechanics encompassing RV dilatation, RV dysfunction, RV failure or a combination thereof with or without circulatory (arterial and/or venous) and other organ system consequences.	100	0	0	7(1)	0.06
3. In patients with respiratory failure requiring veno–venous ECMO, RVI may develop at any time point.	100	0	0	7(0)	0.05
4. The definition of RVI in patients receiving veno–venous ECMO should be based on a multi–modal diagnostic approach.	96	2	2	7(1)	0.18
5. The definition of RVI in patients receiving veno-venous ECMO should include the following elements:	Imp. (%)	Neutral (%)	Not–Imp. (%)
• Pulmonary hemodynamic indices [e.g., systolic and diastolic pulmonary arterial pressure (PAP), pulmonary vascular resistance, PA pulsatility index]	78	6	16	5(1)	0.35
• Echocardiographic markers of RV function and structure	98	0	2	7(0–5)	0.64
• Markers of systemic venous congestion [e.g., central venous pressure >8 mmHg)]	82	12	6	6(1.5)	0.36
• Vasoactive inotropic score (e.g., VIS >60)	64	20	16	5(2)	0.85
• Level of VV ECMO support at the time of RV assessment	47	14	39	4(3)	0.52
• Presence of secondary organ injury (e.g., renal, liver)	82	10	8	6(1.5)	0.27
• Biomarkers (e.g., lactate, troponin I, N–terminal pro B–type natriuretic peptide)	68	8	24	5(2–5)	0.62
• Clinical examination findings suggestive of low flow state (e.g., capillary refill time >2 seconds, cool peripheries)	70	16	14	5(2)	0.32
• Clinical examination findings suggestive of systemic venous congestion [e.g., new peripheral edema, ascites, congestive nephropathy (cardiorenal syndrome)]	76	14	10	5(1.5)	0.17
6. The definition of RVI in patients receiving veno-venous ECMO may include the following echocardiographic metrics of abnormal RV biomechanics:	Imp. (%)	Neutral (%)	Not-imp (%)
• RVEDA/ LVEDA ratio >0.6	88	4	8	6(2)	0.06
• RVEDA/LVEDA ratio ≥1	84	6	10	6(2)	0.33
• Septal dyskinesia	82	10	8	6(2)	0.06
• Tricuspid annular plane systolic excursion <17mm	84	8	8	6(2)	0.18
• RV tricuspid annulus peak systolic velocity <9.5 cm/sec	63	22	14	5(2)	0.48
• RV Fractional area change <35%	74	18	8	5(2)	0.39
• RV Free wall strain >–20%	49	31	20	4(2)	0.72
• Pulmonary artery flow acceleration time <100 milliseconds	41	28	31	4(2)	0.93
• Pulmonary artery acceleration time/Right ventricular ejection time <0.29	31	32	37	4(2)	0.45
• Mid–systolic notching of pulmonary artery Doppler waveform	31	28	41	4(2)	0.47
7. The echocardiogram (transthoracic or transesophageal) for the diagnosis of RVI in patients receiving veno-venous ECMO may be performed and scored by:					0.95
• Echocardiography–accredited intensivist	94.1
• Echocardiography–accredited cardiologist	90.2
• Echocardiography–accredited cardiac physiologist (if available)	64.7
• Echocardiography–accredited sonographer (if available)	54.9
8. In patients receiving veno-venous ECMO who have evidence of abnormal RV biomechanics prior to initiation of ECMO:					0.35
• Echocardiography should be performed within one day after application of ECMO	44.9
• Echocardiography should be performed daily	2
• Echocardiography should be performed whenever there is an alteration in clinical status	85.7
• Echocardiography should be performed daily and whenever there is an alteration in clinical status	20.4
• None of the above	0
9. In patients receiving veno-venous ECMO, trends of standard hemodynamic indices of systemic venous congestion are more important than single time point values.	100	0	0	6(1)	0.55
10. Inclusion of invasive pulmonary hemodynamic parameters in the RVI definition in patients receiving veno–venous ECMO is not mandatory.	89.8	0	10.2	6(1)	0.05
11. In selected patients with RVI (e.g those with severe pulmonary hypertension, low flow state or lung transplant recipients) receiving veno–venous ECMO, invasive pulmonary hemodynamic monitoring may add diagnostic value.	96	2	2	7(1)	0.06
12. Standardization of the ECMO settings at the time of RV assessment is required to make a diagnosis of RVI in patients receiving veno–venous ECMO, but only if the clinical situation allows.	79.7	10.2	10.1	6(1)	0.65
13. Which biomarkers may be included in a definition of RVI in patients receiving veno–venous ECMO?					0.32
▪ Lactate	100
▪ Troponin I	40.8
▪ N–terminal pro B–type natriuretic peptide	75.5
▪ Liver function tests	75.5
▪ International normalized ratio	22.4
▪ None of the above	0
Section 2: RVI Severity	Agree (%)	Neutral (%)	Disagree (%)	Median (IQR)	χ2 p–value
1. In the context of routine clinical practice, there is a need for severity assessment of RVI in patients receiving veno–venous ECMO.	100	0	0	6(1)	0.57
2. In the context of trial enrolment, there is a need for severity assessment of RVI in patients receiving veno-venous.	100	0	0	7(1)	0.65
3. The following components may be included in RVI severity assessment, in the context of routine clinical practice:					0.99
• Echocardiography markers of RV dimensions	92.2
• Echocardiography markers of RV function	98
• Invasive pulmonary hemodynamics	60.8
• Evidence of systemic venous congestion	72.5
• Clinical examination	68.6
• Acute cor pulmonale score	47.1
• VIS	68.6
• Need for mechanical circulatory support	68.6
4. In the context of routine clinical practice, there is a need to categorize RVI for severity in patients receiving veno–venous ECMO.	95.9	4.1	0	6(0)	0.44
5. In the context of routine clinical care and research, RVI phenotyping in patients receiving veno–venous ECMO may be:					0.56
• Clinically–derived	72.5
• Physiologically–derived	82.4
• Echocardiography–derived	94.1
• Based on invasive pulmonary hemodynamics	49
• Based on phenotype clustering algorithms	25.5
6. RVI in patients receiving veno-venous ECMO may be classified into two major phenotypes based on whether the RV meets or does not meet flow demand*.	84.3	7.8	7.9	6(1)	0.14
7. RVI in patients receiving veno-venous ECMO may be classified into the following subphenotypes: Mild: RV dilatation, RV meets flow demand* Moderate: RV dilatation and impaired function, RV meets flow demand Severe RVI: RV dilatation and impaired function, RV does not meet flow demand	90.2	3.9	5.9	6(0)	0.11
8. The terms ‘RV injury’ and ‘RV dysfunction’ may be used interchangeably.	0	2	98	1(0)	0.70
9. The terms ‘RV injury’ and ‘RV failure’ may be used interchangeably.	0	4.1	95.9	1(0)	0.76
10. The terms ‘RV dysfunction’ and ‘RV failure’ may be used interchangeably.	0	4.1	95.9	1(1)	0.07
Section 3: Management Strategies	Agree (%)	Neutral (%)	Disagree (%)	Median (IQR)	χ2 p–value
1. There is a need for formal recommendations on the management strategies of RVI in patients receiving veno-venous ECMO.	100	0	0	7 (1)	0.21
2. The use of the following categories of pharmacological adjunct(s) may be beneficial in terms of improvement in markers of RV function in patients with RVI receiving veno–venous ECMO:	Imp. (%)	Neutral (%)	Not–Imp. (%)
• Positive inotropes (e.g., epinephrine, dobutamine)	96	2	2	6(1)	0.27
• Vasopressors (e.g., norepinephrine)	88	6	6	7(2)	0 07
• Inodilators (e.g., milrinone)	94	2	4	6(1.5)	0.22
• Inopressors (e.g., epinephrine)	94	2	4	6(1.5)	0.31
• Systemic pulmonary vasodilators (e.g., epoprostenol)	88	8	4	7(1.5)	0.14
• Diuretics (e.g., loop diuretics)	96	2	2	7(1)	0.06
• Thrombolytic therapy (systemic or catheter directed) in cases of confirmed pulmonary embolism	82	8	10	6(2)	0.54
3. The use of the following inhaled pharmacological adjunct(s) may be beneficial in terms of improved markers of RV function and RV–pulmonary artery coupling in patients with RVI receiving veno-venous ECMO:					0.94
▪ Inhaled nitric oxide	98
▪ Inhaled prostacyclin	88.2
▪ Inhaled iloprost	74.5
▪ Inhaled milrinone	39.2
4. In patients with respiratory failure and RVI receiving invasive ventilation and veno–venous ECMO, a low driving pressure should be targeted.	100	0	0	7(0)	0.58
5. In patients with respiratory failure and RVI receiving invasive ventilation and veno-venous ECMO, an acceptable range of ‘RV-protective’ driving pressure may be:					0.29
▪ <10 cmH2O	59.2
▪ 10–12 cmH2O	77.6
▪ 12–14 cmH2O	63.3
▪ 14–16 cmH2O	12.2
▪ 16–18 cmH2O	4.1
▪ >18 cmH2O (i.e., this is no need to have a low driving pressure)	2
6. In patients receiving veno-venous ECMO, in the setting of newly diagnosed RVI, the ECMO strategy at the time of RVI diagnosis should be targeted to the EOLIA† trial parameter settings (i.e., FDO2 adjusted to obtain PaO2 between 65 and 90 mmHg (8.6 and 12 kpa) and/or arterial oxygen saturation >90%, and sweep gas flow adjusted to maintain PaCO2 <45 mmHg (6 kPa) avoiding a large (>50%) relative decrease in PaCO2 in the first 24 hours	89.8	6.1	4.1	6(0)	0.95
7. Prone positioning in patients who develop RVI during veno-venous ECMO support may unload the RV.	84.3	11.8	3.9	6(1)	0.05
8. Prone positioning in patients who develop RVI during veno–venous ECMO support may reduce time–to–liberation from ECMO.	61.2	32.7	6.1	5(2)	0.37
9. Prone positioning in patients who develop RVI during veno–venous ECMO support may reduce time–to–liberation from invasive mechanical ventilation.	61.3	34.7	4	5(2)	0.31
10. Prone positioning in patients who develop RVI during veno–venous ECMO support may confer mortality benefit.	53.1	42.9	4	5(2)	1.0
11. Which of the following mechanical circulatory support options may be considered in patients receiving veno-venous ECMO support who develop RVI despite optimal ECMO, ventilatory and pharmacological measures?					0.65
▪ Veno–pulmonary ECMO (VP ECMO)	92.2
▪ Veno–veno arterial ECMO (VVA EcMo)	86.3
▪ Veno–arterial ECMO (VA ECMO)	52.9
12. Mechanical circulatory support may be considered in patients with respiratory failure who develop RVI:					0.81
▪ From the outset, as the configuration of choice (e.g., VP ECMO)	45.1
▪ When optimal respiratory ECMO support fails to improve pulmonary hemodynamics (pulmonary arterial pressure) and markers of RV biomechanics	41.2
▪ When the RV fails to meet flow demand despite optimal respiratory ECMO support, optimal ventilatory strategy and appropriate pharmacological measures	90.2
▪ When there is evidence of systemic venous congestion but RV meets flow demand	13.7

The statements which did not achieve consensus are highlighted in blue

ECMO extracorporeal membrane oxygenation, EOLIA ECMO to Rescue Lung Injury in Severe ARDS, LVEDA left ventricular end-diastolic area, FDO₂ fraction of delivered oxygen, IQR interquartile range, PaCO₂ partial pressure of carbon dioxide in arterial blood, PaO₂ partial pressure of oxygen in arterial blood, PAP pulmonary artery pressure, RV right ventricle, RVEDA right ventricular end-diastolic area, RVI right ventricular injury, VV-ECMO veno-arterial extracorporeal membrane oxygenation, VIS vasoactive inotropic score, VP-ECMO veno-pulmonary extracorporeal membrane oxygenation, VV-ECMO veno-venous extracorporeal membrane oxygenation, χ² Chi-square

Imp important, Not-Imp not important

^*meeting flow demand = end-organ and tissue perfusion is preserved with global oxygen delivery meeting oxygen consumption

^†EOLIA trial (https://doi.org/10.1056/NEJMoa1800385)

Delphi process

An international Steering Committee consisting of eight healthcare professionals involved in the management of patients receiving ECMO (VZ, MS, HY, AD, SD, KR, MVA, KS), and two Delphi methodologists (MJS and PN), was formed under the auspices of the Protecting the Right Ventricle network (PRORVnet). PRORVnet (www.prorvnet.com) is a not-for-profit international collaboration dedicated to conducting research investigating RV pathologies in critical illness. The Steering Committee convened an international panel of Experts involved in the management of patients with RVI, from different professional disciplines (intensive care medicine, anesthesiology, cardiac surgery, cardiology, pulmonary medicine or a combination thereof).

The Experts were screened through purposive sampling after screening through recent publications in the field of RVI and/or ECMO. The selection of Experts was accomplished through a self-administered survey based on the inclusion criteria: (1) working in an active extracorporeal life support organization (ELSO) member center or clinical experience in cardiorespiratory failure for more than 5 years, and (2) at least one peer-reviewed published original research article as a leading author, in the field of ECMO or cardiorespiratory failure. Consent for participation was obtained through the Delphi surveys and their identity was concealed until the end of the Delphi process. The study was granted exemption from the ethical review process because of the nature of the study by the Regional Research Ethics Committee, NMC Specialty Hospital, Dubai, the United Arab Emirates (Reference NMC/RREC/DXB2023/0012). The study protocol was registered with ClinicalTrials.gov (Identifier: NCT05948332).

Three Steering Committee members (SD, AD, VZ) systematically searched MEDLINE, Embase, and the Web of Science databases from inception to August 7, 2023, for original articles on RVI in patients receiving VV-ECMO (electronic supplementary file 2, Appendix). Based on the systematic search results, the Steering Committee identified a list of knowledge gaps pertaining to RVI definition, severity, and clinical interventions. These knowledge gaps were presented as clinical statements to the Expert panel through Google Forms either in a seven-point Likert-scale or multiple-choice question (MCQ) format. The Delphi round one survey had four domains: RVI definition, RVI severity, management strategies, and research priorities. The section on research priorities was removed from the subsequent surveys. The survey reports were analyzed by the Steering Committee after each round in a virtual meeting, and semantic changes to the clinical statements were made and statements were added or dropped based on the comments from the Experts, where deemed necessary. The statements were continued in the subsequent rounds of the Delphi process until stability of the responses was achieved. The Steering Committee did not participate in the voting process during the Delphi rounds to avoid influencing Experts’ opinion and instilling any potential bias, as their identity was not anonymized [13, 16].

Stability and consensus

Stability of the responses was checked to identify the need for further iteration [13, 14]. The statements were continued in the Delphi process, until no significant change in the responses occurred between two consecutive rounds. The stability of the responses was assessed from round two onwards with a non-parametric Chi-square (χ²) test (for multiple-choice format questions) or Kruskal–Wallis test (for Likert-scale questions). Stability was evaluated by comparing the responses of each statement in two consecutive rounds, with stability defined as p ≥ 0.05. A statement was considered unstable if p value was < 0.05 and continued in the Delphi process until stability was achieved.

Consensus was defined as achieved when ≥ 70% of Experts agreed (scores 5–7) or disagreed (scores 1–3) on a Likert-scale statement, or ≥ 80% of the Experts voted for an option of MCQ [12–14]. Median (Interquartile range; IQR) was used to express the central tendency and dispersion of responses.

The statements which achieved both stability and consensus were defined as consensus statements. The consensus was considered to be “strong” when a median score of ≥ 6 or ≤ 2 on the Likert-scale question or ≥ 90% votes for any MCQ option was achieved [12, 13].

Position statements

Expert position statements were drafted by the Steering Committee from the consensus statements. For the Expert position statements, the term “should” was used in position statements to reflect a strong consensus, measured by a median of ≥ 6 (for agreement) or ≤ 2 (for disagreement) in Likert-scale questions and > 90% for any option in MCQ. The term “may” was used in position statements when consensus was achieved without meeting the aforementioned criteria [12, 17].

A qualitative thematic analysis was performed for the questions on research priorities, Expert comments in rounds 2–4, and on clinical statements which failed to achieve consensus, to enlist recurring themes. The final results of the Delphi process, Expert position statements, research priorities, and the manuscript were reviewed and approved by the Experts prior to submission for publication.

Results

Of the invited 64 potential Experts, 53 met the selection criteria. A total of 52 Experts participated in the Delphi process, and 49 (94%) completed all 4 rounds. The median age of Experts was 46.8 years, and 13 (25%) were female (electronic supplementary file 3, Figs. 1, 2). Most Experts (92%) were affiliated with a university hospital and came from a diverse specialty background (electronic supplementary file 3, Table 2).

The Delphi process was conducted through four rounds between September 1 and November 26, 2023 (electronic supplementary file 4, Fig. 3 and electronic supplementary file 5, Round 1–4 reports). At the end of the Delphi process, 31 (89%) out of 35 clinical statements achieved consensus and stability (Table 1), from which 24 Expert position statements were conceived (Fig. 1). The thematic analysis of the research priorities and open questions in three broader themes (definition, phenotypes, and management) is enlisted in Fig. 2.

Fig. 1.

Position statements on RVI definition, severity, and management. ECMO extracorporeal membrane oxygenation, EOLIA ECMO to Rescue Lung Injury in Severe ARDS, LVEDA left ventricular end-diastolic area, PaCO₂ partial pressure of carbon dioxide in arterial blood, RV right ventricle, RVEDA right ventricular end-diastolic area, RVFAC right ventricular fractional area change, RVI right ventricular injury, RV TDI S’ right ventricular tricuspid annulus peak systolic velocity, TAPSE tricuspid annular plane systolic excursion, VP-ECMO veno-pulmonary extracorporeal membrane oxygenation, VVA-ECMO veno-veno-arterial extracorporeal membrane oxygenation, VV-ECMO veno-venous extracorporeal membrane oxygenation. For the Expert position statements, the term “should” was used in position statements to reflect a strong consensus, measured by a median of ≥ 6 (for agreement) or ≤ 2 (for disagreement) in Likert-scale questions and > 90% for any option in MCQ. The term “may” was used in position statements when consensus was achieved without meeting the aforementioned criteria [12, 17]. *meeting flow demand = end-organ and tissue perfusion is preserved with global oxygen delivery meeting oxygen consumption. † EOLIA trial (https://doi.org/10.1056/NEJMoa1800385)

Fig. 2.

Research priorities and open questions in each of the three topic domains (definition, phenotyping, and management). ECMO extracorporeal membrane oxygenation, PA pulmonary artery, PVR pulmonary vascular resistance, RV right ventricle, VA-ECMO veno-arterial extracorporeal membrane oxygenation, VP-ECMO veno-pulmonary extracorporeal membrane oxygenation, VV-ECMO veno-venous extracorporeal membrane oxygenation, VVA-ECMO veno-veno-arterial extracorporeal membrane oxygenation

Discussion

Domain: definition of RVI

Consensus definition: rationale and position statements

Patients with severe respiratory failure and abnormal RV biomechanics (described by abnormal RV morphology and/or function) during VV-ECMO support have high mortality [6, 8]. Heterogeneity remains in the structural and functional definition of RVI during VV-ECMO which may affect the way clinicians and researchers identify, and treat this disease process [6, 8, 12, 18]. Hence, a consensus definition of RVI in patients receiving VV-ECMO would clarify uncertainties around RVI and help standardize research and treatment (Fig. 1).

In the context of respiratory failure and elevated RV afterload, the RV may transition from normal morphology and function to refractory failure with systemic features (central, liver, and renal venous engorgement), and circulatory shock. There are different stages of abnormal RV biomechanics pertaining to structural changes and alterations in RV function: RV dilatation, RV dysfunction, and RV failure [19].

A wide heterogeneity has been found in the metrics and modalities used to characterize abnormal RV biomechanics and a composite of RV size and function metrics is being used in most studies [1, 18]. Current guidelines provide echocardiographic data elements for anatomical and functional RV assessment and diagnosis of abnormal RV biomechanics (e.g., RV dilatation and RV dysfunction) [20–22]. These abnormalities, however, are not always distinct or static, they may overlap, and have all been associated with increased mortality in patients with severe acute respiratory failure, particularly acute respiratory distress syndrome (ARDS) [1].

In a recent systematic review and meta-analysis, RVI in the context of ARDS was defined either as impaired RV function, impaired RV function with hemodynamic compromise, impaired RV function with shock (RV failure), or acute cor pulmonale (RV dilatation with septal dyskinesia), and was found to be significantly associated with mortality [1]. It would stand to reason that RVI in patients receiving VV-ECMO presents as a clinical syndrome encompassing different dynamic states such as RV dilatation, RV dysfunction, and RV failure (Fig. 3). However, these terms should not be used interchangeably (Fig. 1) [8, 23].

Fig. 3.

Key clinical concepts of RV injury in patients receiving VV-ECMO: major phenotypes, sub-phenotypes, and echocardiographic indices. The summary derived from the Expert consensus statements may be used to characterize the spectrum of abnormal RV biomechanics during VV-ECMO. The P–V loops demonstrate the RV pressure–volume relation at different loading conditions [19]. Ea arterial elastance, Es end-systolic elastance, LV left ventricle, LVEDA left ventricular end-diastolic area, PV pressure–volume, RV right ventricle, RVI right ventricular injury, RVEDA right ventricular end-diastolic area, RVFAC right ventricular fractional area change, TAPSE tricuspid annular plane systolic excursion, RV TDI S’ right ventricular tricuspid annulus peak systolic velocity, VV-ECMO veno-venous extracorporeal membrane oxygenation

Critically ill patients receiving VV-ECMO may have limited intrinsic RV contractile reserve required to compensate for the elevated RV load and as a result, RV dilatation ensues to preserve blood flow occasionally at the price of systemic venous congestion [24, 25]. Acute RV cavity dilatation may result in increased pericardial constraint and shifting of the interventricular septum toward the left ventricle (LV), impeding LV diastolic filling and reducing LV cardiac output (ventricular interdependence) [24, 25]. This negative diastolic interaction between the RV and LV may cause an acute decompensation and cardiogenic shock [25]. However, the chronology of these events during VV-ECMO support is uncertain and they may occur at any time point as well as before initiation of ECMO support (Fig. 1) [25].

Diagnosis: rationale and position statements

RVI during VV-ECMO may be subtle initially; however, isolated structural changes may progress to significant functional alterations culminating in systemic venous congestion, end-organ hypoperfusion, and multiple organ failure, despite VV-ECMO support [8, 19, 25, 26].

Hence, a multi-modal diagnostic approach to RVI should be performed (Fig. 1). In addition to echocardiography, this includes clinical assessment of systemic venous congestion or shock, markers of pulmonary vascular dysfunction and systemic venous congestion, and presence/absence of secondary organ injury (Fig. 1). The role of hyperlactatemia in the diagnostic approach is debatable as lactate levels lack sensitivity and specificity. Hyperlactatemia should be interpreted along with other markers and its etiology in patients receiving VV-ECMO could be multi-factorial and not necessarily due to RVI alone (Fig. 1) [25].

RVI may persist or worsen despite optimal respiratory ECMO support, occasionally requiring conversion to circulatory ECMO support [8, 27]. Therefore, VV-ECMO settings (fraction of delivered oxygen, extracorporeal blood flow, and sweep gas flow) need to be considered at the time of RV assessment and may have to be standardized to reliably assess the RV and make a diagnosis of RVI (Fig. 1) [8].

Echocardiographic indices: rationale and position statements

Right-to-left ventricular end-diastolic area ratio (RVEDA/LVEDA) with or without septal dyskinesia is the most common echocardiographic metric used to define RV dysfunction or acute cor pulmonale in the ARDS literature [18]. In studies including patients with ARDS receiving VV-ECMO, echocardiography markers used to assess the RV include both dimensional (RVEDA/LVEDA ratio, visual RV size, qualitative RV dilatation, RV wall thickness) and functional [tricuspid annular plane systolic excursion (TAPSE), RV tricuspid annulus peak systolic velocity (RV TDI S’), RV fractional area change (RVFAC), RV free wall longitudinal strain (RVFWLS)] measures of RV biomechanics [6, 8]. Abnormal values of the aforementioned parameters have been associated with increased mortality [8]. Due to the three-dimensional RV geometry, some of the aforementioned echocardiography markers may be subject to inter- and intra-observer variability [6]. The Experts, therefore, favored easily acquired RV anatomical and functional echocardiographic markers for the RVI definition (in isolation or combined), such as: RVEDA/LVEDA ratio > 0.6, RVEDA/LVEDA ratio ≥ 1, septal dyskinesia, TAPSE < 17 mm, RVFAC < 35%, and RV TDI S’ < 9.5 cm/s (Figs. 1, 3).

The optimal timing and frequency of echocardiography in patients with evidence of RVI prior to application of VV-ECMO are uncertain. However, given the non-linear RV adaptation to changes in loading conditions and the difficulties in identifying a temporal sequence of RV structural and functional alterations, echocardiography should be performed whenever changes in clinical status occur (e.g., increase in vasoactive drug requirements, new extrapulmonary organ dysfunction, clinical signs of hypoperfusion or evidence of worsening systemic venous congestion) (Fig. 1) [19, 24, 25].

Training, certification, and accreditation are required to perform and report an echocardiogram for the diagnosis of RVI, and there are different qualification pathways for procuring competency in critical care echocardiography [28, 29]. Depending on each country’s board regulations and policies, accreditation in critical care echocardiography is feasible within the critical care framework and cardiological practice [30]. The echocardiogram for RV assessment and RVI diagnosis may, therefore, be performed by a cardiologist, or an echocardiography-accredited intensivist who may be best placed to make this assessment and interpret it in the context of cardiac (vasoactive drugs) and respiratory (VV-ECMO, invasive mechanical ventilation) support (Fig. 1).

Systemic venous congestion indices: rationale and position statements

Systemic venous congestion may be assessed using surrogate static hemodynamic indices such as central venous pressure, right-sided venous flow patterns (hepatic, portal, and intra-renal venous flow), inferior vena cava size, and respirophasic variation assessed by echocardiography. Theoretically, VV-ECMO does not affect markers of systemic venous congestion in patients with normal RV biomechanics. In the context of respiratory failure with evidence of RVI, the application of VV-ECMO may decrease central venous pressure, henceforth organ congestion, through RV unloading [6, 7]. Static parameters of venous congestion may also be affected by interventions such as invasive mechanical ventilation and vasoactive drugs due to changes in pleural pressure and venous tone, respectively. Trends in heart–lung interactions, RV preload, and systemic venous congestion indices are, therefore, more important than single time point values in characterizing RVI (Fig. 1) [31, 32].

Invasive pulmonary hemodynamic indices: rationale and position statements

The use of pulmonary artery catheter (PAC) for invasive pulmonary hemodynamic monitoring has not been studied in the context of randomized controlled trials (RCTs) in patients with RVI receiving VV-ECMO. A recent systematic review and meta-analysis of non-randomized retrospective registry studies (n = 37,249) showed that in patients with cardiogenic shock requiring mechanical circulatory support, the lack of hemodynamic monitoring through PAC is associated with increased mortality [33]. This study highlighted the potential role of PAC in selected patient populations such as those with shock and severe pulmonary hypertension, and the need for prospective evaluation of its role in individual patient management. However, in the context of VV-ECMO, in particular femoro-jugular ECMO configuration, PAC insertion could be technically difficult. Moreover, the thermodilution technique used to measure cardiac output has not been validated in patients receiving ECMO. Inclusion of invasive pulmonary hemodynamic indices was, therefore, not deemed mandatory by the Experts for RVI definition in all patients receiving VV-ECMO (Fig. 1).

Domain: position statements on RVI severity

Assessment of severity: rationale and position statements

Evaluation and severity grading of RVI in the high-risk patient population receiving VV-ECMO may help in identifying patients with RVI and high mortality (prognostic enrichment), and selecting patients with certain RVI severity phenotypes who are likely to respond to RV-targeted therapies (predictive enrichment). Prognostic and predictive enrichment may increase the strength of future clinical trials to detect a beneficial RV-targeted treatment effect (Fig. 1) [18, 34].

Echocardiography is currently the preferred modality for assessing the RV in patients with ARDS receiving VV-ECMO [8]. Preliminary literature shows that different parameters of RV function (TAPSE, RVFAC, RVFWLS) and dimensions (RVEDA/LVEDA ratio, septal dynamics) may be used to quantify RVI severity even during VV-ECMO [35–37]. The Experts, therefore, agreed on the need to assess RVI severity in patients receiving VV-ECMO and to categorize RVI severity, using echocardiographic markers (Figs. 1, 3).

RVI phenotypes and sub-phenotypes: rationale and position statements

Patients with RVI receiving VV-ECMO may have distinct physiologic and echocardiographic features which highlight the possibility of phenotyping of this complex clinical entity. The aim of RVI phenotyping is to guide RV-targeted therapies, allow for risk stratification and inform patient selection for clinical trial recruitment [38]. Using machine learning algorithms, certain RVI phenotypes have been identified based on physiologic (e.g., systemic hemodynamic arterial and venous profile) or echocardiographic (e.g., RV dimensions and/or function) characteristics (Figs. 1, 3) [19, 38, 39].

The mechano-energetic coupling between the RV and pulmonary arterial (PA) circulation (RV–PA coupling, determined by end-systolic to PA elastance ratio), and the RV adaptation to different loading conditions determine RV forward flow output, interventricular relationships, and LV preload [19, 24, 25]. RVI may, therefore, be classified into two major physiologic phenotypes based on the intrinsic RV contractile function, and the mechanistic adaptive (homeometric and heterometric adaptations) or maladaptive RV response to PA load determining whether the RV flow output is sufficient to meet oxygen consumption and metabolic demand (Figs. 1, 3) [19, 24, 25, 40].

In patients with ARDS receiving VV-ECMO, structural (RVEDA/LVEDA > 0.6) and functional (TAPSE < 17 mm, RVFAC < 35%, and RVFWLS > − 20%) changes as the RV transitions from normal function to circulatory shock (RV failure) are associated with mortality [8]. The RV response to altered loading conditions is non-linear and despite RV dilatation, the homeometric RV adaptation can be maintained [24, 25]. In this context, isolated RV dilatation without systemic venous congestion may be considered a mild RVI sub-phenotype given that end-organ and tissue perfusion is preserved with global oxygen delivery meeting oxygen consumption (Figs. 1, 3) [19, 24, 25]. The critical level of transition from homeometric to heterometric adaptation of RV function to afterload during VV-ECMO has not been identified [23]. A maladaptive RV response to PA load characterized by RV dilatation and abnormal markers of RV function (RV dysfunction) while the RV flow output remains sufficient for tissue needs may be seen as moderate RVI (Figs. 1, 3). The latter may or may not be associated with elevated systemic venous pressure [23–25]. Worsening uncoupling between systolic RV function and the pulmonary circulation despite VV-ECMO may result in a severe form of RVI, where the RV fails to maintain sufficient flow output despite excessive use of the Frank–Starling mechanism, and resultant systemic venous congestion with shock (Figs. 1, 3) [23–25].

Domain: position statements on RVI management strategies

There is scarce evidence on effective management strategies of RVI during VV-ECMO. A recent scoping review on the effect of RV-specific clinical interventions on RV function in patients with ARDS noted a lack of standardization in treatment targets and patient-centered outcomes, and significant heterogeneity in pharmacological and non-pharmacological modalities used as well as the quality of evidence [41]. Our systematic search yielded similar results in patient populations with RVI receiving VV-ECMO (electronic supplementary file 2, Appendix). Therefore, the Experts agreed that formal consensus recommendations on RVI treatment strategies during VV-ECMO are needed (Fig. 1).

VV-ECMO settings: rationale and position statements

The significant reductions in tidal volume, respiratory rate, plateau pressure, and driving pressure, with an overall lower intensity of ventilation in the intervention (ECMO) arms of pivotal RCTs such as the conventional ventilatory support vs extracorporeal membrane oxygenation for severe adult respiratory failure (CESAR) and the more recent ECMO to Rescue Lung Injury in Severe ARDS (EOLIA) trials, is the likely mechanism of outcome benefit in patients receiving VV-ECMO [5, 42–44]. Given the deleterious effects that elevated driving pressure may have on the RV and pulmonary circulation, the EOLIA strategy may, therefore, confer RV-protective benefits [45]. However, it would be crucial to acknowledge that correction of hypercapnia should be approached with caution since a large relative reduction in partial pressure of carbon dioxide in the first 24 h following application of ECMO support has been associated with an increased risk of intracranial hemorrhage (Fig. 1) [46, 47].

Pharmacological management: rationale and position statements

There is limited data on the effect of parenteral vasoactive drugs on RV function in patients with respiratory failure receiving conventional invasive mechanical ventilation. A small-sample size RCT examining the effect of levosimendan, a calcium sensitizer with pulmonary vasodilatory properties, on hemodynamics in patients with ARDS and evidence of RVI showed that levosimendan improved RV performance and cardiac index [48]. Another small-sample size investigation enrolling patients with similar characteristics demonstrated that use of intravenous epoprostenol was associated with a reduction in mean PA pressure and an increase in RV ejection fraction [41, 49]. In a systematic review and meta-analysis of management strategies for critically ill patients with pulmonary vascular and RV injury (including patients with and without ARDS), a strong recommendation (moderate-quality evidence) was made that phosphodiesterase type III inhibitors reduce pulmonary vascular resistance and improve RV function [50]. In the majority of studies which enrolled patients with RVI receiving VV-ECMO, investigators have assessed mainly the effect of non-pharmacological interventions (e.g., mechanical circulatory support) on RV function (electronic supplementary file 2, Appendix) [8]. However, given their potential physiological benefit and hemodynamic rationale, certain systemic pharmacological therapies which theoretically can unload the RV and improve RV–PA coupling may be considered in patients with RVI receiving VV-ECMO (Fig. 1).

Acute RVI may develop late (up to 4–6 weeks) during VV-ECMO support. Possible contributing factors encompass pulmonary thromboembolic events, persistent hypoxemia, acidemia, the pathological evolution of ARDS, and sustained non-physiological flow to the RV [51, 52]. In these scenarios, pulmonary vasodilators such as inhaled nitric oxide may mitigate pulmonary vascular vasoconstriction, unload the RV, and restore RV–PA coupling [50, 53]. However, caution is warranted as their use might be associated with an elevated risk of renal dysfunction (Fig. 1) [54].

Mechanical ventilatory support and monitoring: rationale and position statements

Driving pressure defined by the ratio of tidal volume to respiratory system compliance is a surrogate of lung strain [55]. Elevated driving pressure may cause regional intra-alveolar vessel collapse resulting in elevated RV afterload [56, 57]. In patients with moderate to severe ARDS, a driving pressure value greater than 18 c mH₂O is an independent risk factor for the development of acute cor pulmonale, a RVI phenotype characterized by RV dilatation and septal dyskinesia [45]. Considering the general ARDS literature, a value below 15 cmH₂O is considered optimal [45, 55, 58, 59]. Therefore, aiming for a driving pressure below 15 c mH₂O, while ensuring it does not exceed 18 cmH₂O, may be part of a RV-protective mechanical ventilation strategy in ARDS during VV-ECMO support (Fig. 1) [45, 55, 58, 59].

In patients with severe respiratory failure receiving conventional invasive ventilation, in particular those with moderate to severe ARDS, prone positioning is associated with lower risk of death [60, 61]. Prone positioning in patients with severe ARDS is associated with RV unloading (reduction in RVEDA/LVEDA ratio and septal dyskinesia on transesophageal echocardiography) and an increase in cardiac preload and cardiac index by reducing pulmonary vascular resistance measured by PAC [62, 63]. Possible mechanisms of RV unloading in prone position include: reduction in pulmonary vascular tone and lung stress, and decreased strain through homogeneous lung aeration; improvement in gas exchange and reversal of hypoxemic/hypercapnic pulmonary arterial vasoconstriction; increase in central blood volume which may recruit the pulmonary microvasculature and reduce pulmonary vasoconstriction and RV afterload; and reduction in ventilator-induced lung injury [64]. The potential protective effect of prone positioning on the RV during VV-ECMO may be explained through a combination of ultra-protective lung ventilation facilitated by ECMO and the aforementioned mechanisms.

Mechanical circulatory support: rationale and position statements

In cases where optimal VV-ECMO, ventilatory, and pharmacological interventions are ineffective in addressing RVI, mechanical circulatory support becomes a potentially viable option. A veno-pulmonary ECMO mode (single-site dual lumen or dual-site single lumen cannulation) bypasses the injured RV and reduces systemic venous congestion and RV preload, decreasing RV wall stress and restoring RV–PA coupling [65–68]. A multi-modal management including early extubation, prompt corticosteroid therapy, high-dose anticoagulation, inhaled pulmonary vasodilators, diuretics, and the single-site dual lumen veno-pulmonary ECMO approach has demonstrated favorable outcomes in patients with coronavirus disease 2019 (COVID-19)-related ARDS [65–68]. Alternatively, conversion from VV-ECMO to veno-veno-arterial ECMO presents another option which may be considered for RV mechanical support (Fig. 1).

Dissensus among the Experts and open questions

Despite multiple iterative Delphi rounds, four clinical statements did not reach consensus. There was a lack of consensus among the Experts on the ‘RV-protective’ range of driving pressure during invasive ventilation on VV-ECMO. Small studies examining the effect of different combinations of driving pressure on ventilator-induced lung injury during VV-ECMO have indicated that efforts to lower driving pressure may yield outcome benefit [69–71]. Although the Experts agreed on a low driving pressure approach for RV protection, there was lack of consensus on a specific target driving pressure level that would potentially reduce the risk of further RVI. This discordance between Experts must be seen in the context of lack of data on the effect of different levels of driving pressure on RV biomechanics during VV-ECMO. There was also no consensus among the Experts on the effect of prone positioning on the duration of VV-ECMO support, duration of invasive ventilation or mortality in the context of RVI. Recent RCT data showed that although prone positioning was a safe intervention during VV-ECMO, it did not confer outcome benefit compared with supine positioning [72]. It should be noted, however, that patients with RVI were not identified in that study and this trial was conducted mostly in patients with COVID-19 [72]. In general, the VV-ECMO patient population with RVI is understudied which may explain the dissensus on the effect of prone positioning on outcomes. There is a need for research in these knowledge gaps for better prognostication and management of RVI during VV-ECMO (Fig. 2).

Limitations and strengths

Our study has several strengths. First, this is the first of its kind consensus definition of RVI in the context of VV-ECMO. The Expert position statements have a strong potential to guide clinicians in the bedside management of this very high-risk patient cohort, until evidence on these therapies evolves. Second, our panel of Experts provided position statements on phenotyping and overarching management principles for RVI during VV-ECMO in an area where quality evidence is lacking or ambiguous. Third, a rigorous Delphi process was followed to avoid bias from dominance and peer pressure, with Experts selected through explicit predefined selection criteria. The identity of the Experts was concealed from each other until the completion of the Delphi rounds. Fourth, we completed four Delphi rounds with an inclusive and diverse panel of 52 subject matter Experts, maintaining a stringent schedule of only 3 months and an attrition rate of only 6%. Finally, we identified research priorities in broader domains of definition, phenotyping, and RVI management in the setting of VV-ECMO.

Our study also has some limitations. The Delphi process ensured face validity for the consensus definition, but feasibility and reliability of the definition are lacking and are an area of future research. Although the position statements provided an overarching management principle, a more individualized approach might be required for specific patient populations and some clinical interventions. Inter-individual variations in the interpretation of echocardiography, training, and resource limitations may create challenges in implementing a few position statements. Despite a diverse panel of Experts from different professional disciplines involved in the management of patients with RVI, the interpretation of certain position statements might have influenced the opinion of the Experts. However, the iterative process of Delphi rounds, controlled feedback from the round reports, and assessment of the stability of the responses should have reduced this bias and ensured the fidelity of the responses. In this Delphi, we did not investigate domains pertaining to causes of ARDS that may independently affect RV function (e.g., sepsis) during VV-ECMO or interactions between VV-ECMO flow and native circulation. Finally, with ongoing research, the best practices and definition of RVI may change when evidence evolves.

Conclusion

Using Delphi methodology, consensus and stability were reached by a large international group of RV Experts on 31 clinical statements pertaining to RVI definition, severity, and management in the context of VV-ECMO. Expert position statements addressing important themes including RVI definition and diagnosis, echocardiographic indices, systemic venous congestion, pulmonary hemodynamic indices, RVI phenotyping, VV-ECMO settings, pharmacological, mechanical ventilatory, and mechanical circulatory support strategies were derived. The research priorities identified by the Experts cover open questions and key knowledge gaps in areas where clinical evidence is limited or absent. Addressing those priorities and clinical questions in high-quality prospective research may lead to the development of personalized RV-targeted therapies and improved patient-centered outcomes.

Data availability

‘Cumulative Delphi Round reports with Experts’ feedback are provided in the Online Supplementary material (Supplementary File 5). Further details can be accessed upon request to the corresponding author.