Skip to main content

Foundational Trials

LEVOECMO

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Levosimendan to Facilitate Weaning From Extracorporeal Membrane Oxygenation in Patients With Severe Cardiogenic Shock: The LEVOECMO Randomized Clinical Trial
  • Acronym: LEVOECMO
  • Year: 2026
  • Journal published in: JAMA
  • Citation: Combes A, Saura O, Nesseler N, et al; LEVOECMO Trial Group; International ECMO Network (ECMONet). Levosimendan to facilitate weaning from extracorporeal membrane oxygenation in patients with severe cardiogenic shock: the LEVOECMO randomized clinical trial. JAMA. 2026;335(1):60-69.

Context & Rationale

  • Background
    • VA-ECMO is increasingly used for refractory cardiogenic shock as a bridge to recovery, durable mechanical support, or transplantation, but prolonged runs expose patients to bleeding, thrombosis, infection, limb ischaemia, stroke, and resource-intensive ICU care.
    • Successful and timely liberation from VA-ECMO is clinically and operationally important, yet weaning is frequently limited by persistent left ventricular dysfunction, stunned myocardium, or inadequate pulsatility.
    • Levosimendan (a calcium-sensitiser with KATP-channel opening effects) has an inodilator profile and a long-acting active metabolite, creating biological plausibility for improving myocardial performance and facilitating separation from VA-ECMO.
    • Before LEVOECMO, evidence in VA-ECMO populations was dominated by retrospective studies (including attempts at causal emulation) reporting associations with higher weaning success and/or improved survival, but with substantial risk of confounding by indication and centre-level practice differences.45
    • Systematic reviews and meta-analyses suggested benefit (higher likelihood of weaning; possible survival signals), but pooled estimates were largely based on non-randomised datasets with heterogeneity and variable co-interventions (unloading strategies, weaning protocols, thresholds for transplantation/durable support).23
  • Research Question/Hypothesis
    • Does early levosimendan infusion, compared with placebo, reduce time to successful VA-ECMO weaning in adults with severe but potentially reversible cardiogenic shock supported with VA-ECMO?
    • Hypothesis: levosimendan increases the cumulative incidence of successful weaning (accounting for competing risks of death and weaning failure).
  • Why This Matters
    • A pharmacological adjunct that reliably accelerates liberation from VA-ECMO could reduce complications, ICU/hospital length of stay, and costs in a high-mortality, high-resource patient population.
    • A neutral or harmful result is equally practice-changing, given levosimendan’s cost, variable availability, and plausible arrhythmogenic/vasodilatory risks in critically ill patients.

Design & Methods

  • Research Question: In adults with severe but potentially reversible cardiogenic shock supported with VA-ECMO initiated within the prior 48 hours, does early levosimendan (vs placebo) reduce time to successful VA-ECMO weaning within 30 days?
  • Study Type: Randomised, multicentre, double-blind, placebo-controlled, parallel-group trial conducted across 11 ICUs in France (investigator-initiated; International ECMO Network [ECMONet]).
  • Population:
    • Setting: ICU patients receiving VA-ECMO for refractory cardiogenic shock; recruitment August 27, 2021, to September 10, 2024; final follow-up November 10, 2024.
    • Key inclusion: Adults with acute cardiogenic shock refractory to conventional therapy, with VA-ECMO started in the preceding 48 hours; shock considered potentially reversible (etiology strata: acute myocardial infarction, myocarditis, postcardiotomy, other).
    • Key exclusions (high-yield): VA-ECMO >48 hours; cumulative low-flow >30 minutes in prior 48 hours; irreversible neurological pathology; end-stage cardiomyopathy with no expectation of recovery; mechanical complication of myocardial infarction; aortic regurgitation > grade II; heart transplant recipients; moribund at randomisation; SAPS II >90; recent torsades de pointes; additional exclusions detailed in the published eAppendix.
    • Consent: surrogate consent where available; emergency consent permitted with deferred patient/representative consent once feasible.
  • Intervention:
    • Levosimendan continuous infusion for 24 hours without bolus.
    • Starting rate: 0.15 μg/kg/min; increased to 0.20 μg/kg/min after 2 hours if no rate-limiting adverse effects.
    • Discontinuation criteria included anaphylaxis, severe hypotension (protocol-defined), or intractable arrhythmias.
  • Comparison:
    • Placebo continuous infusion for 24 hours without bolus, using a polyvitamin (riboflavin-containing) preparation to match the yellow appearance of levosimendan.
    • Delivered with the same titration schedule and discontinuation rules as the intervention.
    • Usual ICU care (sedation, anticoagulation, haemodynamic and ECMO management) in both groups; VA-ECMO initiation and co-interventions at clinician discretion within standard practice.
  • Blinding: Double-blind (participants, clinicians, and outcome assessors); designated preparation pathway for indistinguishable infusates; allocation concealed via central internet randomisation.
  • Statistics: Sample size 206 planned to detect a subdistribution hazard ratio of 1.75 (corresponding to an increase in successful weaning from 50% to 70%) with 80% power at a two-sided α=0.05; primary analysis by intention-to-treat using a Fine–Gray competing-risk model (competing events: death and weaning failure); prespecified sensitivity analyses included cause-specific modelling.
  • Follow-Up Period: 60 days after randomisation (primary endpoint window: 30 days; “successful weaning” required survival without ECMO/other mechanical support or transplant for 30 days after ECMO removal).

Key Results

This trial was not stopped early. A total of 205/206 planned patients were enrolled; final follow-up was completed 60 days after the last randomisation (November 10, 2024).

Outcome Levosimendan Placebo Effect p value / 95% CI Notes
Successful VA-ECMO weaning within 30 days (primary; competing risk) 69/101 (68.3%) 71/104 (68.3%) sHR 1.02 95% CI 0.74 to 1.39; P=0.92 Risk difference 0.0% (95% CI −12.8% to 12.7%)
Death before VA-ECMO weaning (competing event) 15/101 (14.9%) 12/104 (11.5%) sHR 1.32 95% CI 0.62 to 2.79; P=0.47 Competing-risk component of primary analysis
VA-ECMO weaning failure (competing event) 15/101 (14.9%) 21/104 (20.2%) sHR 0.72 95% CI 0.37 to 1.38; P=0.32 Includes death, need for further circulatory support (eg, another VA-ECMO run/LVAD), or cardiac transplant within 30 days of separation
All-cause mortality at day 30 26/101 (25.7%) 23/104 (22.1%) RR 1.16 95% CI 0.71 to 1.90 Risk difference 3.6% (95% CI −8.9% to 15.8%)
All-cause mortality at day 60 28/101 (27.7%) 26/104 (25.0%) RR 1.11 95% CI 0.70 to 1.75; P=0.78 Risk difference 2.7% (95% CI −9.0% to 15.3%)
Duration of VA-ECMO support (days) up to day 30 Median 5 (IQR 4–7) Median 6 (IQR 4–11) Median difference −1 day 95% CI −2 to 1; P=0.53 Restricted mean-based comparison used for time-to-event with death as competing risk
Days alive without VA-ECMO by day 30 Median 14 (IQR 0–20) Median 13 (IQR 0–19) Median difference 1 day 95% CI −1 to 5 Support-free days; no multiplicity-adjusted p value reported
ICU length of stay (days) up to day 60 Median 23 (IQR 12–39) Median 24 (IQR 14–39) Mean difference −1 day 95% CI −5 to 3; P=0.42 Restricted mean-based comparison accounting for death
Hospital length of stay (days) up to day 60 Median 28 (IQR 18–48) Median 35 (IQR 22–56) Mean difference −7 days 95% CI −12 to −2 No multiplicity-adjusted p value reported
Organ failure–free days by day 30 Median 12 (IQR 0–18) Median 14 (IQR 0–19) Median difference −2 days 95% CI −10 to 4 Organ failure defined by SOFA subcomponents 0–1
Kidney replacement therapy by day 30 29/101 (28.7%) 38/104 (36.5%) RR 0.79 95% CI 0.53 to 1.17 Absolute difference −7.8% (95% CI −21.4% to 4.7%)
Major adverse cardiovascular events by day 60 36/101 (35.6%) 39/104 (37.5%) RR 0.95 95% CI 0.65 to 1.40 Composite: death, transplant, escalation to permanent LVAD, stroke, or HF re-hospitalisation
Cardiac transplant by day 30 0/101 (0.0%) 5/104 (4.8%) Risk difference −4.8% 95% CI −9.6% to −1.0% Interpret cautiously (low event counts; decision-driven endpoint)
Ventricular arrhythmia (VF/VT/torsades) during study drug exposure 18/101 (17.8%) 9/104 (8.7%) RR 2.06 95% CI 0.97 to 4.37 Absolute difference 9.2% (95% CI −0.4% to 18.0%)
Suspected adverse drug event leading to treatment cessation 12/101 (11.9%) 4/104 (3.8%) RR 3.09 95% CI 1.07 to 8.93 Absolute difference 8.0% (95% CI 0.6% to 15.4%)
  • Primary outcome was identical between groups: 68.3% achieved successful weaning by day 30 (sHR 1.02; 95% CI 0.74 to 1.39; P=0.92).
  • No mortality signal at day 30 (25.7% vs 22.1%) or day 60 (27.7% vs 25.0%; P=0.78); secondary clinical and organ-support outcomes were broadly concordant with neutrality.
  • A safety signal was observed: ventricular arrhythmias occurred more often with levosimendan (17.8% vs 8.7%), and treatment cessation due to suspected adverse drug events was more frequent (11.9% vs 3.8%).

Internal Validity

  • Randomisation and allocation: Centralised 24-hour internet randomisation with stochastic minimisation (probability 0.8 after the first 20 patients), stratified by centre and shock aetiology; allocation concealment was appropriate.
  • Blinding and outcome ascertainment: Double-blind design with indistinguishable infusates and masked clinicians/outcome assessors; the primary endpoint (time to successful weaning with competing risks) is objective in principle but operationally depends on weaning decisions (mitigated by protocolised weaning testing).
  • Losses, exclusions, and missingness: 205 randomised; 2 patients in each group did not receive study drug; 2 were lost to follow-up for day 60 outcomes (1 before the primary endpoint); only 1 patient remained on VA-ECMO at day 30 and was censored for the primary analysis.
  • Protocol adherence and treatment separation: Dose was escalated to 0.20 ± 0.01 μg/kg/min in 93% (levosimendan) vs 96% (placebo); treatment interruption was more common in the levosimendan group (definitive interruption 10.9% vs 2.9%; cessation due to suspected adverse drug event 11.9% vs 3.8%), which could dilute efficacy while simultaneously evidencing harm-related intolerance.
  • Baseline comparability: Groups were broadly similar (median age 59 vs 58 years; postcardiotomy 38.6% vs 38.5%; LVEF median 15% in both), with chance imbalances including SAPS II (median 43 vs 38), invasive ventilation at baseline (86.1% vs 76.9%), and prior PCI (25.7% vs 11.5%).
  • Heterogeneity: Etiologic heterogeneity (acute MI, myocarditis, postcardiotomy, other) was prespecified and used for minimisation; subgroup analyses showed no statistically significant interaction (P=0.74), but precision was limited for smaller strata (eg, myocarditis).
  • Timing: Randomisation occurred early after VA-ECMO initiation (median 24 hours vs 27 hours), consistent with the mechanistic intent to support early myocardial recovery rather than late-stage weaning failure.
  • Dose: 24-hour infusion at conventional ICU dosing without bolus (to mitigate hypotension); the design tests a pragmatic “single-course” strategy aligned with levosimendan’s long-acting metabolite, but may be insufficient if benefit requires repeated dosing or different timing.
  • Key delivery aspects (weaning protocol): Daily weaning testing was protocolised and echocardiography-guided (eg, stepwise reduction of VA-ECMO flow and physiologic criteria for weanability), which likely improved standardisation and reduces centre-to-centre decision variability.
  • Crossover and co-interventions: Crossover was not a design feature; durable LVAD by day 30 was 4.0% vs 3.8%, but cardiac transplantation by day 30 differed (0% vs 4.8%), highlighting that downstream pathway decisions can influence composite “weaning success” constructs.
  • Statistical rigour: Competing-risk methodology (Fine–Gray) matched the endpoint structure; the trial was powered for a large effect (target sHR 1.75), so the confidence intervals exclude only large benefits and remain compatible with modest benefit or harm.

Conclusion on Internal Validity: Overall, internal validity appears strong given concealed randomisation, double-blinding, prespecified competing-risk analysis, and minimal missing data; precision is limited for modest effects because the trial was powered for a large benefit and treatment interruptions were more frequent with levosimendan.

External Validity

  • Population representativeness: Adults with VA-ECMO for acute cardiogenic shock initiated within 48 hours, treated in experienced French ICUs; the cohort included common real-world aetiologies (postcardiotomy 38.5%, acute MI 27.3%, myocarditis 13.7%).
  • Illness severity spectrum: Entry after VA-ECMO initiation (median 24–27 hours) means patients were, by definition, those who survived and remained eligible after initial stabilisation; findings may not extrapolate to peri-cannulation instability, very high lactate phenotypes, or later “difficult weaning” scenarios beyond 48 hours.
  • Centre capability dependence: Results are most applicable to high-resource centres with protocolised weaning, echocardiography expertise, and access to escalation pathways (LVAD/transplant); generalisability to lower-volume or resource-limited systems is uncertain.
  • Therapy availability: Levosimendan is not universally available across jurisdictions; where available, practice impact depends on local thresholds for use and monitoring for arrhythmias/hypotension.

Conclusion on External Validity: Generalisability is moderate: the population and VA-ECMO indications are common in tertiary practice, but outcomes may differ in less experienced centres, in later-stage weaning failure, or where co-interventions (unloading/transplant pathways) are structured differently.

Strengths & Limitations

  • Strengths: Multicentre randomised double-blind placebo-controlled design; clinically meaningful endpoint incorporating durability of separation; appropriate competing-risk methods; protocolised weaning evaluation; high follow-up completeness with small loss to follow-up.
  • Limitations: Powered for a large effect (target sHR 1.75), limiting inference about modest benefit; heterogeneous aetiologies with limited power for subgroup inference; downstream decisions (eg, transplant) can influence “weaning success” constructs; treatment interruptions were more frequent with levosimendan and ventricular arrhythmias were increased; conduct modifications late in the study introduced potential (though unproven) threats to blinding.

Interpretation & Why It Matters

  • Clinical practice
    • Routine early levosimendan to facilitate VA-ECMO liberation is not supported: successful weaning was identical (68.3% vs 68.3%), with no shortening of VA-ECMO duration (median 5 vs 6 days).
    • Given higher ventricular arrhythmias (17.8% vs 8.7%) and more drug cessation (11.9% vs 3.8%), any off-trial use should be highly selective and accompanied by rigorous rhythm and haemodynamic monitoring.
  • Mechanism vs outcome
    • Physiological plausibility and observational associations did not translate into improved weaning in a blinded RCT, reinforcing the unreliability of non-randomised effect estimates in high-intensity supportive care pathways.
  • Research direction
    • Future trials may need phenotype enrichment (eg, myocarditis/stunning-dominant presentations), alternative timing (pre-ECMO vs immediately post-cannulation), and integration with unloading strategies and standardised escalation pathways to detect clinically relevant but modest effects.

Controversies & Subsequent Evidence

  • Sample size assumptions did not match observed event rates: the design targeted a very large benefit (increase in successful weaning from 50% to 70%), whereas observed overall weaning failure was substantially lower than anticipated, limiting power to exclude modest benefit or harm.1
  • Blinding integrity was potentially challenged late in the trial when placebo procurement and preparation processes changed (site-level preparation by a designated nurse), creating a theoretical pathway to unblinding; this is most concerning for subjective outcomes, but the trial was neutral on objective primary and most secondary endpoints.1
  • Subgroup estimates suggested divergent point estimates by aetiology (eg, myocarditis sHR 2.16; 95% CI 0.97 to 4.80), but tests for heterogeneity were negative and subgroup strata were small, making these findings hypothesis-generating rather than actionable.1
  • Pre-LEVOECMO observational evidence (including causal emulation approaches) reported associations consistent with improved weaning and/or survival; LEVOECMO’s null findings substantially weaken causal interpretations of those observational signals in similar populations.45
  • Meta-analyses pooling largely retrospective VA-ECMO studies report higher weaning success with levosimendan and variable survival signals, but heterogeneity and residual confounding remain major limitations; LEVOECMO provides the highest-level counterweight to those pooled observational estimates.23
  • A second double-blind RCT (WEANILEVO) was terminated early for insufficient funding after enrolment of 82 patients, leaving LEVOECMO as the definitive randomised evidence base for this specific indication to date.1

Summary

  • In 205 adults with severe but potentially reversible cardiogenic shock on VA-ECMO, early levosimendan did not improve time to successful weaning vs placebo (68.3% vs 68.3%; sHR 1.02; 95% CI 0.74 to 1.39; P=0.92).
  • There was no mortality benefit at day 30 (25.7% vs 22.1%) or day 60 (27.7% vs 25.0%; P=0.78), and no consistent improvement in organ-support–free outcomes.
  • VA-ECMO duration was not shortened (median 5 vs 6 days; median difference −1 day; 95% CI −2 to 1; P=0.53).
  • Levosimendan increased ventricular arrhythmias (17.8% vs 8.7%) and was associated with more treatment cessation due to suspected adverse drug events (11.9% vs 3.8%).
  • LEVOECMO substantially recalibrates the evidence base away from routine levosimendan use to facilitate VA-ECMO weaning, despite prior observational signals.

Overall Takeaway

LEVOECMO is a landmark because it provides the first adequately sized, double-blind, placebo-controlled randomised test of levosimendan to facilitate VA-ECMO weaning in refractory cardiogenic shock and demonstrates no benefit on its clinically anchored primary endpoint. The trial also identifies a clinically meaningful harm signal (ventricular arrhythmias and higher treatment cessation), supporting a shift away from routine levosimendan use for VA-ECMO weaning in similar patients.

Overall Summary

  • Levosimendan did not shorten time to successful VA-ECMO weaning (68.3% vs 68.3%; sHR 1.02; P=0.92).
  • No improvement in 30- or 60-day mortality, VA-ECMO duration, or organ-support–free outcomes.
  • Higher ventricular arrhythmias and more treatment cessation with levosimendan, supporting avoidance of routine use for this indication.

Bibliography