Hemodynamic-guided V-A ECMO management for cardiogenic shock: Insight from initial experience from TWEET study (to wean an ECMO trials)

Abstract

Background

Veno-arterial extracorporeal membrane oxygenation (V-A ECMO) is used for refractory cardiogenic shock, yet optimal weaning strategies and the role of left ventricular (LV) venting remain unclear. The TWEET-1 study was designed as an early feasibility investigation to define processes, timelines, and clinical parameters for future interventional trials evaluating Impella 5.5 for LV venting during V-A ECMO support.

Methods

This prospective, single-center observational study from 2021 to 2025 included patients who received standard-of-care V-A ECMO and Swan-Ganz monitoring. Patients were assigned to the ECPELLA + group (defined as the addition of Impella 5.5 to V-A ECMO to provide left ventricular unloading) when pulmonary artery (PA) diastolic pressure exceeded 20 mm Hg; other individuals served as controls. The primary outcome was survival to discharge. Secondary outcomes included V-A ECMO duration, complications, and SAVE score–adjusted survival.

Results

Among 117 screened patients, 43 were enrolled (25 ECPELLA +, 18 controls). Overall survival to discharge was 53% with no significant difference between groups. ECPELLA + patients had higher pulmonary pressures and longer V-A ECMO support but demonstrated appropriate selection for unloading. Both groups exceeded the SAVE score–predicted survival. Subgroup analysis suggested potential benefit of Impella venting in the most severe patients (SAVE −14 to −10).

Discussion

Hemodynamic-guided LV venting with Impella 5.5 during V-A ECMO was feasible and enabled prompt LV unloading (ECPELLA). While survival to discharge did not differ significantly between groups, patients with higher PA diastolic pressures and SAVE scores below −10 demonstrated a significant reduction in mortality with LV unloading.

Conclusion

Early hemodynamic-guided LV unloading may improve risk-adjusted V-A ECMO outcomes.

Keywords

V-A ECMO Impella 5.5 LV venting cardiogenic shock

Introduction

Veno-arterial extracorporeal membrane oxygenation (V-A ECMO) is widely used for refractory cardiogenic shock, despite a lack of clinical trial evidence.^1,2,3 While it is relatively easy to deploy and effective for reversing hemodynamic collapse and metabolic derangements, it may hinder left ventricular (LV) recovery⁴ and cause lung injury,⁵ access-related complications,⁶ and bleeding.^7,8 Several studies have identified risk factors, outcome predictors, and risk scores (SAVE and ENCOURAGE). A few small observational studies have described specific strategies for weaning from V-A ECMO.^1,2,9–13 However, as of November 2025, detailed weaning strategies after V-A ECMO initiation for refractory cardiogenic shock have not been uniformly adopted.

LV venting is considered standard of care on a case-by-case basis, but limited prospective data on criteria and timing have prevented the establishment of universal guidelines. In severe pulmonary edema, most physicians use surgical or percutaneous LV venting. In addition to upfront LV venting, the Impella device is used as a rescue strategy when V-A ECMO weaning has failed.

Because clinical investigations in V-A ECMO populations are challenging, the primary objective of this observational early feasibility study (TWEET-1) was to define the timeline, clinical data requirements, personnel, and logistics necessary for future interventional or randomized trials. Specifically, we evaluated Impella 5.5 (Abiomed) for LV venting in patients receiving V-A ECMO support (ECPELLA + group) versus conventional V-A ECMO without LV venting (control group) with respect to weaning and survival. The secondary objective was to compare clinical outcomes between groups.

Methods

We screened all V-A ECMO candidates between January 1, 2021, and January 1, 2025. The decision to initiate V-A ECMO was made on a case-by-case basis by our multidisciplinary team, comprising a board-certified critical care physician, cardiothoracic surgeon, interventional cardiologist, and heart failure physician, per our institutional V-A ECMO protocol. All patients were treated according to the current standard of care. To reflect real-world practice, we did not limit inclusion based on the underlying cardiogenic shock etiology, as the specific cause is often defined at the time of or after the decision to cannulate V-A ECMO.

The primary objective of this observational study was to design and collect an extensive prospective dataset to analyze V-A ECMO outcomes at a Platinum-tier Extracorporeal Life Support Organization (ELSO) center, with a particular focus on the effect of early LV venting using the Impella 5.5 before V-A ECMO weaning. The primary endpoint was survival to discharge. Secondary endpoints included duration of V-A ECMO support, length of hospital stay, rates of cardiac recovery (survival free from mechanical circulatory support [MCS]), heart transplantation or inotropic support at 30 days, prevalence of V-A ECMO-driven complications as defined by ELSO, SAVE scores, and survival adjusted to SAVE score (cumulative inotropic dose and inotropic score).

We adopted the ELSO-endorsed SAVE (Survival After Veno-Arterial ECMO) score¹² for V-A ECMO risk stratification. It includes 17 clinical, hemodynamic, and respiratory variables and predicts survival for adult patients undergoing V-A ECMO for refractory cardiogenic shock.¹² Demographic, clinical, and biochemical data needed to calculate individual SAVE scores were available for all patients. SAVE scores were calculated using the online calculator (https://www.elso.org/savescore/index.html).

All patients received a Swan-Ganz catheter at the time of V-A ECMO placement or within 24 h of admission. Pulmonary pressure was continuously monitored. A pulmonary artery (PA) diastolic pressure >20 mm Hg was used as the hemodynamic trigger for escalation to Impella 5.5, selected to prompt LV unloading before the development of pulmonary edema and lung injury as previously described.¹⁰ Time to Impella 5.5 placement was intentionally variable, with escalation initiated once PA diastolic pressure crossed the 20-mm Hg threshold to individualize the decision to each patient’s hemodynamic profile. Once diastolic pulmonary pressure was consistently >20 mm Hg, the patient was implanted with an Impella 5.5 and assigned to the ECPELLA + group. Patients who did not receive Impella 5.5 during V-A ECMO were included in the control group. PA diastolic pressure and insertion time were recorded.

The UTHealth Houston Institutional Research Board approved the study protocol (HSC-MS-21-0286).

Statistical analysis

We compared demographic, clinical, and outcome variables between the ECPELLA + and control groups, which were calculated using the non-parametric Wilcoxon signed-rank tests that do not have assumptions about the distributions of the variable, and chi-square tests were used for categorical variables.

To benchmark survival, we estimated the probability of surviving to discharge using the SAVE score, a logistic regression model with binary survival status at discharge as the dependent variable and the SAVE score as the independent variable. Paired t-tests were used to compare the probability of survival for the different levels of SAVE scores to the corresponding survival probability published by Schmidt M et al.¹² Results are presented in tables with medians and interquartile ranges for the survival probabilities.

All analyses were performed using Stata version 18.0 (College Station, TX), with p-values ≤0.05 considered statistically significant.

Results

TWEET-1 study phases and delayed definition of inclusion & exclusion criteria

To identify strengths and weaknesses after a prolonged enrollment period, the TWEET-1 pilot study concluded on January 1, 2025, before full enrollment was completed. A total of 117 patients were screened, and 43 patients (25 ECPELLA +, 18 controls) were enrolled.

Inclusion/exclusion criteria were initially modeled after a prior V-A ECMO with LV venting trial using a different device (NCT03431467). Early enrollment revealed that end-organ hypoperfusion inherent to the low cardiac output state caused systematically elevated MELD scores unrelated to underlying liver disease, resulting in excessive screen failures.¹⁴ Additional operational challenges—including complex baseline data collection logistics, research staff turnover, and variability in pulmonary catheter placement—necessitated multiple protocol revisions before the final inclusion/exclusion criteria were established (Tables 1 and 2).

Table 1.

Original and revised inclusion criteria.

Inclusion criteria (initial)	Inclusion criteria (final)
Adult patients (age >18 years old)	Adult patients (age >18 years old)
Patients initially cannulated for V-A ECMO support through peripheral (femoral) cannulation.	Patients initially cannulated for V-A ECMO support through peripheral (femoral) cannulation.
Patient is diagnosed with cardiogenic shock, including cases that are refractory to conventional therapy; cardiogenic shock is defined as (1) systolic BP <90 mm Hg, (2) cardiac index <1.8 OR a cardiac index <2.0 on moderate to high doses of inotropes and vasopressors for greater than 30 min, or (3) systemic signs of tissue hypoxia	Patient is diagnosed with cardiogenic shock, including cases that are refractory to conventional therapy; cardiogenic shock is defined as a systolic BP <90 mm Hg, AND cardiac index <1.8 OR a cardiac index <2.0 and VIS score of >10 OR systemic signs of tissue hypoxia
Grossly viable neurological status, including intact brainstem reflexes and moving extremities before enrolment	Grossly viable neurological status, including intact brainstem reflexes and moving extremities before enrolment

Abbreviations: BP, blood pressure; PA, pulmonary artery; V-A ECMO, veno-arterial extracorporeal membrane oxygenation; VIS, vasoactive-inotropic score.

Table 2.

Original and revised exclusion criteria.

Exclusion criteria (initial)	Exclusion criteria (final)
Recent significant pulmonary embolus	Recent significant pulmonary embolus within the last 3 months
Moderate to severe aortic valve insufficiency	Moderate to severe aortic valve insufficiency
Ongoing significant sepsis	Ongoing significant sepsis
Ongoing significant sepsis	Out-of-hospital cardiac arrest
Continuous CPR >40 min, except if neurological status is satisfactory	Continuous CPR >40 min, except if neurological status is satisfactory
Mechanical AVR	Mechanical AVR
Presence of LV thrombus	Presence of LV thrombus
Active malignancy	Active malignancy
Aortic dissection	Aortic dissection
Intracranial hemorrhage	Intracranial hemorrhage
Hypothermia	Hypothermia
Transfer from outside hospital on V-A ECMO (unless cannulation was done by faculty in our medical center and offsite data collection is possible)	Transfer from outside hospital on V-A ECMO (unless cannulation was done by faculty in our medical center and offsite data collection is possible)
MELD score greater than 25

Abbreviations: AVR, aortic valve replacement; LV, left ventricle; MELD, model for end-stage liver disease; V-A ECMO, veno-arterial extracorporeal membrane oxygenation.

Enrollment proceeded in two phases. During Phase 1 (October 2021–January 2023), 24 patients were screened under the original criteria, with three enrolled (88% screen failure). Following IRB approval of a revised protocol with modified exclusion criteria, Phase 2 (February 2023–January 2025) screened 93 patients and enrolled 40 (57% screen failure). The three Phase 1 patients met both original and revised criteria and were retained in the final analysis. Perfusionists were integrated into the data collection workflow to enable real-time prospective data capture during ECMO support, improving data quality and completeness.

Overall patient characteristics and outcomes

Table 3 describes the baseline clinical and demographic characteristics for all patients by venting strategy (ECPELLA + group, n = 25, vs control group, n = 18). Overall, 23 patients survived to hospital discharge (23/43, 53%); there were 13 in the ECPELLA + group (13/25, 52%) and 10 in the control group (10/18, 56%). No significant difference in survival was observed between groups at 30 days (68% vs 67%, p = 0.93) or at discharge (52% vs 56%, p = 0.82).

Table 3.

Baseline characteristics, primary endpoints, and secondary endpoints for the full TWEET-1 cohort (n = 43).

Factor	Level	Control	ECPELLA+	p-value
N		18	25
Age, mean (SD)		54 (14)	61 (13)	0.13
Sex		15 (83%)	16 (64%)	0.16
Length of stay, median (IQR)		12.5 (8, 30)	35.5 (20.5, 57)	0.014
Length of V-A ECMO, median (IQR)		2 (1, 8)	6 (2, 9)	0.12
Days to Impella, median (IQR)		-	2.28 (.78, 4.4)	-
Impella length (days), median (IQR)		-	6.1 (2.62, 9.2)	-
Alive at discharge	No	8 (44%)	12 (48%)	0.82
Alive at discharge	Yes	10 (56%)	13 (52%)	0.82
Alive at 30 days	No	6 (33%)	8 (32%)	0.93
Alive at 30 days	Yes	12 (67%)	17 (68%)	0.93
Outcome	LVAD implantation	0 (0%)	4 (16%)	0.13
Outcome	No LVAD or OHT	18 (100%)	21 (84%)	0.13
Complications
Cardiac arrhythmia	No	17 (94%)	20 (80%)	0.18
Cardiac arrhythmia	Yes	1 (6%)	5 (20%)	0.18
GI hemorrhage	No	15 (83%)	21 (84%)	1.00
GI hemorrhage	Yes	3 (17%)	4 (16%)	1.00
CNS infarction	No	18 (100%)	21 (84%)	0.13
CNS infarction	Yes	0 (0%)	4 (16%)	0.13
CNS hemorrhage	No	17 (94%)	22 (88%)	0.63
CNS hemorrhage	Yes	1 (6%)	3 (12%)	0.63
Impella CP prior	No	16 (89%)	16 (64%)	0.086
Impella CP prior	Yes	2 (11%)	9 (36%)	0.086
Type 2 diabetes	No	10 (62%)	8 (35%)	0.11
Type 2 diabetes	Yes	6 (38%)	15 (65%)	0.11
Hyperlipidemia	No	8 (50%)	5 (22%)	0.090
Hyperlipidemia	Yes	8 (50%)	18 (78%)	0.090
Indications for V-A ECMO	Cardiac arrest (assisted CPR)	6 (33%)	7 (28%)	0.71
	Cardiogenic shock/Severe cardiac failure	11 (61%)	17 (68%)
	Periprocedural support for high-risk percutaneous cardiac interventions	0 (0%)	1 (4%)
	Post cardiotomy: Inability to wean from cardiopulmonary bypass after cardiac surgery	1 (6%)	0 (0%)
In hospital cardiac arrest	No	5^a (31%)	7 (28%)	0.82
In hospital cardiac arrest	Yes	11 (69%)	18 (72%)	0.82
Total save score, mean (SD)		−6.44 (4.94)	−8.12 (5.66)	0.32

Abbreviations: CNS, central nervous system; GI, gastrointestinal; IQR, interquartile range; LVAD, left ventricular assist device; OHT, orthotopic heart transplantation; SD, standard deviation; V-A ECMO, veno-arterial extracorporeal membrane oxygenation.

^aTwo patients suffered a cardiac arrest at an outside facility.

When grouped by underlying etiology, there was no significant difference in indications for V-A ECMO, which included cardiac arrest, cardiogenic shock, acute decompensation, periprocedural support for high-risk interventions, and post-pericardiotomy syndrome. None of the patients in the control group underwent LVAD or transplant during the index hospitalization. In contrast, four patients (4/25, 16%) were transitioned from Impella 5.5 to LVAD as part of our current practice.

A significant fraction of patients (29/43, 67%) had in-hospital cardiac arrest. The rate did not differ between groups (p = 0.82): 18 patients (18/25, 72%) in the ECPELLA + group versus six patients (11/16, 69%) in the control group. Only two patients in the control group had an outside hospital witnessed arrest; all arrests in the ECPELLA + group occurred in-hospital.

Two patients (2/18, 11%) in the control group had Impella CP placed before V-A ECMO. In comparison, 9 patients (9/25, 36%) in the ECPELLA + group initially had an Impella CP that required an upgrade to V-A ECMO and, eventually, Impella 5.5 (p = 0.086).

Demographics, comorbidities, and inotropic score

On univariate analysis, demographic characteristics and comorbidities (sex, BMI, hypertension, diabetes, dyslipidemia) did not differ between groups. Mean SAVE scores were similar (−8.12 ± 5.66 in the ECPELLA + group vs −6.44 ± 4.94 in the control group; p = 0.32).

Length of V-A ECMO support

The median length of stay was longer in the ECPELLA + group (35.5 days, interquartile range [IQR] 20.5–57) than in the control group (12.5 days, IQR 8–30) (p = 0.014). Similarly, the median length of V-A ECMO support was longer in the ECPELLA + group (6 days, IQR 2–9) than in the control group (2 days, IQR 1–8) (p = 0.12). The median time to Impella 5.5 from V-A ECMO implant was 2.28 days (IQR 0.78–4.4). Once implanted, the median length of Impella support was 6.1 days (IQR 2.6–9.2).

Complications during V-A ECMO support

There was no statistical difference between the control and ECPELLA + groups for the four most prevalent complications patients may experience during V-A ECMO support: arrhythmias (control 6% vs ECPELLA + 20%; p = 0.18), gastrointestinal hemorrhage (control 17% vs ECPELLA + 16%; p = 1.00), central nervous system (CNS) infarction (control 0% vs ECPELLA + 16%; p = 0.13), and CNS hemorrhage (control 6% vs ECPELLA + 12%; p = 0.63) (Table 3).

Patients without cardiac arrest (no ECPR): Demographics, comorbidities, and outcomes

Table 4 describes the baseline clinical and demographic characteristics for patients who required V-A ECMO implantation as part of a worsening cardiogenic shock clinical picture but in the absence of cardiac arrest. Over the course of enrollment, this subgroup accounted for only a minority of enrolled patients (12/43, 28%). Results are presented based on venting strategy: ECPELLA + (n = 7) versus V-A ECMO only (control, n = 5). Overall, 10 patients survived to hospital discharge (10/12, 83%): 5 (71%) in the ECPELLA + group and 5 (100%) in the control group.

Table 4.

Primary and secondary outcomes in patients without cardiac arrest (no ECPR).

Factor	Level	Control	ECPELLA+	p-value
N		5	7
Length of stay, median (IQR)		24 (24, 31)	40 (21, 60)	0.58
Length of V-A ECMO, median (IQR)		1 (1, 6)	6 (2, 7)	0.28
Days to Impella, median (IQR)		-	6.76 (2.18, 7.17)	-
Impella length (days), median (IQR)		-	4.4 (2.18, 6.76)	-
Alive at discharge	No	0 (0%)	2 (29%)	0.19
Alive at discharge	Yes	5 (100%)	5 (71%)	0.19
Alive at 30 days	No	0 (0%)	2 (29%)	0.19
Alive at 30 days	Yes	5 (100%)	5 (71%)	0.19
Outcome	LVAD implantation	0 (0%)	1 (14%)	1.00
Outcome	No LVAD or OHT	5 (100%)	6 (86%)	1.00
Complications
Cardiac arrhythmia	No	5 (100%)	5 (71%)	0.19
Cardiac arrhythmia	Yes	0 (0%)	2 (29%)	0.19
GI hemorrhage	No	4 (80%)	5 (71%)	1.00
GI hemorrhage	Yes	1 (20%)	2 (29%)	1.00
CNS infarction	No	5 (100%)	7 (100%)
CNS infarction	Yes	0 (0%)	0 (0%)
CNS hemorrhage	No	5 (100%)	6 (86%)	1.00
CNS hemorrhage	Yes	0 (0%)	1 (14%)	1.00
Impella CP prior	No	5 (100%)	6 (86%)	1.00
Impella CP prior	Yes	0 (0%)	1 (14%)	1.00
Arterial hypertension	No	2 (40%)	1 (14%)	0.52
Arterial hypertension	Yes	3 (60%)	6 (86%)	0.52
Type 2 diabetes	No	4 (80%)	4 (57%)	0.58
Type 2 diabetes	Yes	1 (20%)	3 (43%)	0.58
Hyperlipidemia	No	4 (80%)	2 (17%)	0.080
Hyperlipidemia	Yes	1 (20%)	5 (83%)	0.080
Indications for V-A ECMO	Cardiogenic shock/Severe cardiac failure	4 (80%)	6 (86%)	0.68
	Periprocedural support for high-risk percutaneous cardiac interventions	0 (0%)	1 (14%)
	Post cardiotomy: Inability to wean from cardiopulmonary bypass after cardiac surgery	1 (20%)	0 (0%)
Total save score, mean (SD)		−3.8 (5.07)	−8.57 (3.87)	0.093

When grouped by underlying etiology, there was no significant difference in the indications for V-A ECMO (cardiogenic shock, periprocedural support for high-risk interventions, and post-pericardiotomy syndrome) between the groups. None of the patients in the control group and one patient in the ECPELLA + group underwent LVAD or transplant as part of the index hospitalization. Demographic characteristics and comorbidities (sex, BMI, hypertension, diabetes, dyslipidemia) were also similar. Mean SAVE scores were −8.57 ± 3.87 in the ECPELLA + group versus −3.8 ± 5.07 in the control group (p = 0.093).

Both the median length of stay (40 days, IQR 21–60 vs 24 days, IQR 24–31; p = 0.58) and length of V-A ECMO support (6 days, IQR 2-7 vs 1 day, IQR 1-6; p = 0.28) were longer in the ECPELLA + group than in the control group. The time to Impella 5.5 implantation from V-A ECMO implant was 6.7 days (IQR 2.2–7.2). The median length of Impella 5.5 support was 4.4 days (IQR 2.1–6.7). There was no significant difference between the groups in arrhythmias, gastrointestinal hemorrhage, CNS infarction, or CNS hemorrhage.

Patients with cardiac arrest (ECPR): Demographics, comorbidities, and outcomes

Table 5 summarizes baseline characteristics of patients who required V-A ECMO as part of the ECPR protocol (i.e., V-A ECMO was placed immediately/after cardiac arrest). Twenty-nine patients required ECPR in the TWEET-1 cohort: 18 in the ECPELLA + group and 11 in the control group. Both groups had high ECPR use (control group 11/18, 61%; ECPELLA + group 18/25, 72%; p = non-significant), underscoring the extreme clinical complexity of this population.

Table 5.

Primary and secondary outcomes in patients with cardiac arrest (received ECPR).

Factor	Level	Control	ECPELLA+	p-value
N		11	18
Length of stay, median (IQR)		10 (1, 12)	35.5 (19, 46)	0.003
Length of V-A ECMO, median (IQR)		2 (1, 8)	5.5 (2, 10)	0.10
Days to Impella, median (IQR)		-	1.26 (.53, 2.62)	-
Impella length (days), median (IQR)		-	6.0 (2.62, 10.3)	-
Alive at discharge	No	6 (55%)	10 (56%)	0.96
Alive at discharge	Yes	5 (45%)	8 (44%)	0.96
Alive at 30 days	No	6 (55%)	6 (33%)	0.26
Alive at 30 days	Yes	5 (45%)	12 (67%)	0.26
Outcome	LVAD implantation	0 (0%)	3 (17%)	0.27
Outcome	No LVAD or OHT	11 (100%)	15 (83%)	0.27
Complications
Cardiac arrhythmia	No	10 (91%)	15 (83%)	0.57
Cardiac arrhythmia	Yes	1 (9%)	3 (17%)	0.57
GI hemorrhage	No	9 (82%)	16 (89%)	0.62
GI hemorrhage	Yes	2 (18%)	2 (11%)	0.62
CNS infarction	No	11 (100%)	14 (78%)	0.27
CNS infarction	Yes	0 (0%)	4 (22%)	0.27
CNS hemorrhage	No	10 (91%)	16 (89%)	1.00
CNS hemorrhage	Yes	1 (9%)	2 (11%)	1.00
Impella CP prior	No	9 (82%)	10 (56%)	0.23
Impella CP prior	Yes	2 (18%)	8 (44%)	0.23
Type 2 diabetes	No	4 (44%)	4 (25%)	0.39
Type 2 diabetes	Yes	5 (56%)	12 (75%)	0.39
Indications for V-A ECMO	Cardiac arrest (assisted CPR)	6 (55%)	7 (39%)	0.47
Indications for V-A ECMO	Cardiogenic shock/Severe cardiac failure	5 (45%)	11 (61%)	0.47
Total save score, mean (SD)		−6.91 (4.25)	−7.94 (6.31)	0.64

Overall, 13 patients survived to hospital discharge (13/29, 45%); eight were in the ECPELLA + group (8/18, 44%) and 5 were in the control group (5/11, 45%). Although survival with ECPELLA + was higher at both 30 days (67% vs 45%, p = 0.26) and at discharge (45% vs 44%, p = 0.96), these differences were not statistically significant. Indications for V-A ECMO—cardiac arrest and cardiogenic shock—were similar between groups (Table 5).

None of the patients in the control group underwent LVAD or transplant during the index hospitalization, whereas 3 patients in the ECPELLA + group (3/18, 17%) were transitioned from Impella 5.5 to LVAD (p = non-significant).

Demographic characteristics and comorbidities (sex, BMI, hypertension, diabetes, dyslipidemia) were similar between groups with similar mean SAVE scores (−7.94 ± 6.31 in the ECPELLA + group vs −6.91 ± 4.25 in the control group; p = 0.64).

The median length of stay was longer in the ECPELLA + group (35.5 days, IQR 19–46) than in the control group (10 days, IQR 1-12) (p = 0.003). Similarly, the median length of V-A ECMO support was longer in the ECPELLA + group (5.5 days, IQR 2-10) than in the control group (2 days, IQR 1-8) (p = 0.10). The median time from ECMO to Impella placement in the ECPELLA + group was 1.26 days (IQR 0.53–2.62), and Impella support lasted a median of 6.0 days (IQR 2.6–10.3).

Rates of arrhythmias, gastrointestinal hemorrhage, CNS infarction, and CNS hemorrhage were similar between groups.

Pulmonary pressure and strategy selection

During enrollment, all patients received a Swan-Ganz at ECMO initiation, causing variation in initial PA pressure timing and patient selection for Impella. Over the full study period, 25 patients were enrolled in the ECPELLA + arm and 18 in the control arm. The mean PA pressure after V-A ECMO and before Impella deployment was 29 ± 12 mm Hg in the ECPELLA + arm versus 18 ± 5 mm Hg in patients who did not receive Impella 5.5 (p < 0.005). This demonstrates the appropriate selection of patients with higher mean PA pressure post V-A ECMO to receive LV venting.

SAVE score-adjusted survival

We compared observed survival to discharge with SAVE score-predicted survival. Because of the exploratory design, the current stage of the TWEET registry, SAVE score heterogeneity, and small group sizes, we clustered patients into three SAVE score groups: −14 to −10, −10 to −5, and > −5 (Figure 1). We compared survival to discharge in each observed SAVE score cluster against the predicted survival of patients with similar SAVE scores. Both the ECPELLA + and the control groups significantly outperformed the SAVE score match-predicted survival in all SAVE clusters. However, overall survival to discharge did not differ significantly between the ECPELLA + group (54.4%, confidence interval [CI] 41.9–60.6) and the control group (59.8%, CI 37.6–72.5) (p = 0.42) (Figures 2 and 3).

Figure 1.

SAVE adjusted survival at discharge in patients with ECPELLA versus SAVE predicted survival (reference).

Figure 2.

SAVE adjusted survival at discharge: Control versus SAVE predicted survival (reference).

Figure 3.

SAVE adjusted survival at discharge in patients with ECPELLA versus Control.

We also compared the survival of ECPELLA + patients with that of control patients across different SAVE score clusters to explore the possible differential impact of LV venting on patients with varying clinical severity grades, as graded by the SAVE score (Figure 3). Interestingly, we observed a significant difference in survival between the ECPELLA + group (39.9%, 95% CI 35.9–41.9) and the control group (30.3%, 95% CI 26.9–37.6) (p = 0.027) among patients with SAVE scores between −14 and −10. No significant difference was seen in the −10 to −5 cluster, whereas patients with SAVE scores greater than −5 (ECPELLA +: 62.6%, 95% CI 58.6–64.5; control: 69.1%, 95% CI 61.7–72.5; p = 0.077) exhibited a non-statistically significant trend towards worse outcomes.

Discussion

This study aimed to clarify the role of LV venting and identify patients who may benefit from full LV unloading after V-A ECMO is initiated for cardiogenic shock. This observational design evaluates a sequential-weaning approach in which early LV venting with an Impella 5.5 provides full unloading, followed by V-A ECMO weaning once metabolic clearance occurs. While limited by a small sample size, our preliminary findings signal that selective use of Impella 5.5 in selected patients on V-A ECMO may improve support and weaning outcomes in cardiogenic shock.

This study and its findings aim to address a growing disconnect between evidence and practice in ECMO and cardiogenic shock management,¹⁵ where the decision to escalate to LV venting frequently relies on center-specific experience rather than objective, patient-level criteria. This pilot investigated whether integrating the overall patient risk profile (SAVE score) with real-time hemodynamic status (PA diastolic pressure) could inform an individualized MCS escalation strategy. Although survival to discharge did not differ significantly between groups, subgroup analyses revealed that patients with higher PA diastolic pressures and SAVE scores below −10 demonstrated significantly greater mortality, suggesting that these combined parameters may identify a hemodynamic phenotype in which LV venting confers benefit.

Patient selection remains critical to the success of MCS. Most patients in this study were enrolled after cardiac arrest. While prompt V-A ECMO deployment can reverse metabolic derangements resulting from cardiogenic shock, it cannot prevent further anoxic brain injury; thus, proper selection criteria are imperative to avoid futile interventions. The enrollment criteria developed during the initial study phase balanced the inclusion of appropriate candidates, including patients with cardiac arrest, while avoiding those with irreversible injury. Additionally, concomitant Swan-Ganz placement at V-A ECMO initiation enabled continuous hemodynamic monitoring and timely device escalation before refractory hypoxia or lung injury developed due to increased afterload, depressed myocardial function, or escalating pressor and inotrope use. While the non-randomized protocol design and well-established V-A ECMO practice in our center may have pre-emptively influenced patient selection, patients with higher mean pulmonary pressure were preferentially assigned to the ECPELLA + group, as initially designed in the current protocol.

The median time from V-A ECMO implantation to Impella placement was 2.28 days. This delay reflected rising pulmonary pressures after V-A ECMO initiation and logistical challenges due to transferring patients on V-A ECMO from outside hospitals. Generally, once patients arrived at our center, Impella 5.5 placement or upgrade occurred within 12 h if their PA diastolic pressure remained consistently above 20 mm Hg. Importantly, although not statistically significant, we observed 11 upgrades from Impella CP to Impella 5.5 (11/43, 25%) due to hemolysis or limited LV unloading, defined as persistent PA diastolic greater than 20 mm Hg.

Although none of the patients in the control group required transition to surgical LVAD or heart transplantation, four ECPELLA + patients were successfully bridged to LVAD. While not statistically significant—likely due to a small sample size—this may explain the longer length of stay in the ECPELLA + group. Nonetheless, the high SAVE-adjusted survival seen in the ECPELLA + group may reflect the ability to fast-track patients from temporary to surgical LVAD while using Impella 5.5 for pre-habilitation¹⁴ during recovery from V-A ECMO.

Several limitations warrant consideration. First, although SAVE score adjustment was performed, this approach does not fully account for time-dependent confounding or treatment indication bias. The small sample size precludes more robust analytical methods such as propensity score matching or marginal structural models, and the non-randomized allocation strategy based on PA diastolic pressure inherently introduces selection bias that cannot be entirely mitigated in an exploratory study of this size. Second, complication rates were low in both the control and ECPELLA + groups; however, the longer duration of ECMO support observed in the ECPELLA + group may bias complication comparisons, and the limited sample size precludes meaningful adjustment for duration of support.

Finally, despite the limited size of the TWEET-1 cohort, the protocol yielded SAVE score-adjusted observed outcomes that significantly outperformed the SAVE-predicted outcome across a wide range of risk profiles (SAVE −14 to 0). Notably, the granular nature of the dataset also suggests a preferential role for ECPELLA + in patients with more compromised clinical conditions (SAVE −14 to −10), who may be candidates for upfront Impella 5.5 deployment. Given its widespread adoption and ELSO endorsement, the SAVE score provides a practical and validated framework for extending these TWEET-1 findings into larger, prospective, multicenter trials. While these results remain hypothesis-generating, the integration of SAVE score and PA diastolic pressure—both well-established parameters within the ECMO community—identified signals that may inform the design of adequately powered randomized trials to establish definitive cutoff values for MCS escalation decisions.

Conclusions

In this limited, prospective, single-center study, a combination of specific patient selection criteria, uniform, continuous hemodynamic monitoring, and hemodynamic-guided LV unloading with the Impella 5.5 outperformed SAVE-predicted survival in patients on V-A ECMO across a wide range of SAVE scores.

Limitations

Given this study’s single-center, exploratory design, conclusions drawn from this relatively small sample size should be confirmed in larger multicenter trials.

We acknowledge that SAVE score adjustment does not fully account for time-dependent confounding or treatment indication bias, and that the small sample size precludes more robust methods such as propensity score matching or marginal structural models.

Complication rates were low in both the control and ECPELLA + groups, and the limited sample size precludes meaningful adjustment for duration of support.

Outcomes presented here are also subject to the initial triage process for offering V-A ECMO to individual patients. This process was performed by an experienced multidisciplinary team at an ELSO Platinum-level center; therefore, outcomes may differ at other centers.

Footnotes

Acknowledgements

The authors would like to thank all clinical personnel of the Heart & Vascular Institute at the Memorial Hermann Texas Medical Center Hospital in Houston for their relentless and devoted assistance in caring for these critically ill patients. The authors also acknowledge Jennifer Peterson, PhD, ELS, and Jessica Moody, PhD, ELS, for their editorial assistance.

ORCID iDs

Angelo Nascimbene

Igor D. Gregoric

Ethical considerations

The UTHealth Houston Institutional Research Board has reviewed and approved the study protocol (HSC-MS-21-0286).

Author contributions

AN, ST, SN, HD, IS, JX, MJ, JP, MP, JW, IG, BK. AN, ST, SN, HD, JW, LJ, KK —Performed the research. AN, JW—Formal Analysis. AN, ST, SN, HD, IS, JX, JP, MP, IG, BK —Investigation. AN,ST, JW—Data Curation. AN,JW, SN, BK—Wrote the manuscript. All authors read and approved the final manuscript.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: A.N. is supported by the National Institutes of Health (NIH), National Heart, Lung, and Blood Institute (NHLBI) grant 1R01HL163549-01A1. S.N and A.N. received research grants from Abiomed.

Declaration of conflicting interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: IAS has research-related financial interests in Abiomed.

Data Availability Statement

The datasets generated during and/or analyzed during the current study are not publicly available due to patients’ confidentiality but are available from the corresponding author on request.*

References

Thiele

Zeymer

Akin

, et al. Extracorporeal life support in infarct-related cardiogenic shock. N Engl J Med 2023; 389(14): 1286–1297. https://doi.org/10.1056/NEJMoa2307227

Zeymer

Freund

Hochadel

, et al. Venoarterial extracorporeal membrane oxygenation in patients with infarct-related cardiogenic shock: an individual patient data meta-analysis of randomised trials. Lancet 2023; 402(10410): 1338–1346. https://doi.org/10.1016/S0140-6736(23)01607-0

Bogerd

Griffioen

Bunge

JJH

, et al. Clinical use and impact of mechanical circulatory support for myocardial infarction-related cardiogenic shock in the Netherlands: a registry-based propensity-matched analysis. Open Heart 2025; 12(1): e002846. https://doi.org/10.1136/openhrt-2024-002846

Burkhoff

Sayer

Doshi

, et al. Hemodynamics of mechanical circulatory support. J Am Coll Cardiol 2015; 66(23): 2663–2674. https://doi.org/10.1016/j.jacc.2015.10.017

Teijeiro-Paradis

Gannon

Fan

. Complications associated with venovenous extracorporeal membrane oxygenation-what can Go wrong? Crit Care Med 2022; 50(12): 1809–1818. https://doi.org/10.1097/CCM.0000000000005673

Zangrillo

Landoni

Biondi-Zoccai

, et al. A meta-analysis of complications and mortality of extracorporeal membrane oxygenation. Crit Care Resusc 2013; 15(3): 172–178.

Nunez

Gosling

O'Gara

, et al. Bleeding and thrombotic events in adults supported with venovenous extracorporeal membrane oxygenation: an ELSO registry analysis. Intensive Care Med 2022; 48(2): 213–224. https://doi.org/10.1007/s00134-021-06593-x

Aubron

DePuydt

Belon

, et al. Predictive factors of bleeding events in adults undergoing extracorporeal membrane oxygenation. Ann Intensive Care 2016; 6(1): 97. https://doi.org/10.1186/s13613-016-0196-7

Keebler

Haddad

Choi

, et al. Venoarterial extracorporeal membrane oxygenation in cardiogenic shock. JACC Heart Fail 2018; 6(6): 503–516. https://doi.org/10.1016/j.jchf.2017.11.017

10.

Truby

Takeda

Mauro

, et al. Incidence and implications of left ventricular distention during venoarterial extracorporeal membrane oxygenation support. ASAIO J 2017; 63(3): 257–265. https://doi.org/10.1097/MAT.0000000000000553

11.

Saito

Tateishi

Toda

, et al. Complications and outcomes of impella treatment in cardiogenic shock patients with and without acute myocardial infarction. J Am Heart Assoc 2023; 12(17): e030819. https://doi.org/10.1161/JAHA.123.030819

12.

Schmidt

Burrell

Roberts

, et al. Predicting survival after ECMO for refractory cardiogenic shock: the survival after veno-arterial-ECMO (SAVE)-score. Eur Heart J 2015; 36(33): 2246–2256. https://doi.org/10.1093/eurheartj/ehv194

13.

Salas De Armas

Patel

Akay

, et al. Impella CP to impella 5.5 uninterrupted upgrade using a double device technique. ASAIO J 2022; 68(7): e113–e115. https://doi.org/10.1097/MAT.0000000000001587

14.

Blandino Ortiz

Lamanna

Antonucci

, et al. Altered liver function in patients undergoing veno-arterial extracorporeal membrane oxygenation (ECMO) therapy. Minerva Anestesiol 2017; 83(3): 255–265. https://doi.org/10.23736/S0375-9393.16.11442-7

15.

Lusebrink

Binzenhöfer

Hering

, et al.

Scrutinizing the role of venoarterial extracorporeal membrane oxygenation: has clinical practice outpaced the evidence?

Circulation 2024; 149(13): 1033–1052. https://doi.org/10.1161/CIRCULATIONAHA.123.067087