Skip to main content

Foundational Trials

ECLS-SHOCK

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Extracorporeal Life Support in Infarct-Related Cardiogenic Shock
  • Acronym: ECLS-SHOCK
  • Year: 2023
  • Journal published in: New England Journal of Medicine
  • Citation: Thiele H, Zeymer U, Akin I, et al. Extracorporeal Life Support in Infarct-Related Cardiogenic Shock. N Engl J Med. 2023;389:1286-97.

Context & Rationale

  • Background
    • Infarct-related cardiogenic shock carries very high short-term mortality despite urgent revascularisation, vasopressors/inotropes, and contemporary ICU care.
    • Physiologically, venoarterial extracorporeal life support (ECLS/VA-ECMO) can provide immediate circulatory support (and oxygenation/CO2 clearance), but increases afterload and exposes patients to major bleeding and limb/ischaemic complications.
    • Before ECLS-SHOCK, evidence for routine ECLS in infarct-related shock was largely observational and vulnerable to confounding, with wide practice variation in cannulation, anticoagulation, distal perfusion, and left ventricular (LV) unloading.
    • Guidelines and scientific statements had highlighted the need for adequately powered randomised trials of mechanical circulatory support strategies in cardiogenic shock.
  • Research Question/Hypothesis
    • Whether a strategy of routine early ECLS (initiated in the catheterisation laboratory around the time of revascularisation) reduces 30-day all-cause mortality versus standard care without routine ECLS.
    • Hypothesis: early ECLS would improve survival by limiting systemic hypoperfusion and end-organ injury in severe infarct-related shock.
  • Why This Matters
    • ECLS is resource-intensive, high-risk, and increasingly used; a null or harmful effect would have major clinical and systems implications.
    • If beneficial, routine early ECLS would represent a major escalation of standard care for a high-mortality condition and would drive network organisation and training requirements.

Design & Methods

  • Research Question: In adults with severe infarct-related cardiogenic shock (arterial lactate >3 mmol/L) undergoing planned revascularisation, does routine early ECLS plus standard care reduce 30-day mortality compared with standard care alone?
  • Study Type: Prospective, multicentre, investigator-initiated, randomised, open-label strategy trial conducted in catheterisation laboratories and ICUs at experienced centres within a cardiogenic shock network.1
  • Population:
    • Setting: Patients with acute myocardial infarction complicated by cardiogenic shock identified after coronary angiography; revascularisation planned (PCI or CABG).1
    • Key inclusion criteria (all required): AMI (STEMI or NSTEMI) with planned revascularisation; systolic blood pressure <90 mmHg for >30 minutes or catecholamines required to maintain systolic pressure >90 mmHg; signs of impaired organ perfusion (altered mental status or cold/clammy skin and extremities or oliguria <30 mL/h); arterial lactate >3 mmol/L; informed consent (stepwise process).1
    • Key exclusion criteria: Cardiopulmonary resuscitation >45 minutes; mechanical cause of cardiogenic shock; onset of shock >12 hours; severe peripheral artery disease precluding cannulation; age <18 or >80 years (age threshold amended during the trial); shock of another cause (e.g., sepsis, hypovolaemia); life expectancy <6 months from severe comorbidity; pregnancy; participation in another trial.1
  • Intervention:
    • Strategy: Routine early femoro-femoral ECLS (VA-ECMO) in addition to revascularisation and optimal medical care; ECLS insertion recommended before PCI provided no procedure delay >20 minutes; otherwise insertion could occur during/after revascularisation.1
    • Anticoagulation: Intravenous unfractionated heparin bolus; target activated clotting time >200 seconds; ongoing anticoagulation management per protocol/local practice parameters.1
    • Cannulation practice (as delivered): Antegrade distal limb perfusion was commonly used (per protocol recommendation; high utilisation in trial practice).
    • Weaning: Protocolised daily assessment; stepwise reduction of flows with haemodynamic and metabolic criteria (including lactate and venous saturation targets).1
  • Comparison:
    • Strategy: No routine ECLS; culprit lesion identification and revascularisation performed using standard technique.
    • Rescue/escalation: Cross-over to ECLS was discouraged; escalation to other devices (e.g., IABP, Impella, TandemHeart) permitted for clinical deterioration using pre-specified haemodynamic/metabolic criteria (e.g., impending collapse, lactate rise >3 mmol/L over 6 hours, or vasopressor escalation by 50% to maintain mean arterial pressure targets).1
  • Blinding: Unblinded (open-label) due to the nature of ECLS; primary outcome (death) is objective.
  • Statistics: A total of 394 patients were required to detect a reduction in 30-day mortality from 49% to 35% with 80% power at a two-sided final alpha of 0.048 (group-sequential design with one interim analysis); allowing for 6% drop-out, planned enrolment was 420; primary analysis was intention-to-treat.1
  • Follow-Up Period: 30 days for the primary endpoint; mid-term follow-up planned at 6 and 12 months (including vital status and rehospitalisation for heart failure).1

Key Results

This trial was not stopped early. One interim analysis was planned; enrolment proceeded to the planned sample size (420 randomised), with 417 included in the intention-to-treat analyses after post-randomisation exclusion for lack of consent.

Outcome Routine early ECLS Standard care Effect p value / 95% CI Notes
Death from any cause at 30 days (primary) 100/209 (47.8%) 102/208 (49.0%) RR 0.98 95% CI 0.80 to 1.19; P=0.81 No survival benefit; CI includes clinically meaningful benefit and harm.
Time until haemodynamic stabilisation (days) Median 3.1 (IQR 2.2–6.7) Median 3.1 (IQR 2.4–5.5) Hodges–Lehmann 0.27 95% CI −0.41 to 1.14; P=Not reported Key secondary outcome; no signal of earlier stabilisation.
Acute renal failure warranting renal-replacement therapy 17/209 (8.1%) 29/208 (13.9%) RR 0.58 95% CI 0.33 to 1.03; P=Not reported Trend towards fewer RRT events; CI crosses 1.
Recurrent myocardial infarction at 30 days 10/209 (4.8%) 7/208 (3.4%) RR 1.42 95% CI 0.55 to 3.66; P=Not reported Wide CI (imprecision).
Rehospitalisation for congestive heart failure at 30 days 3/209 (1.4%) 7/208 (3.4%) RR 0.43 95% CI 0.11 to 1.62; P=Not reported Few events.
Poor neurological outcome at 30 days (CPC 3–4 among survivors) 27/109 (24.8%) 24/106 (22.6%) RR 1.03 95% CI 0.88 to 1.19; P=Not reported Defined post hoc as CPC 3–4 (CPC 5 excluded from denominator); interpret cautiously.
Moderate or severe bleeding (BARC 3–5) (safety) 49/209 (23.4%) 20/208 (9.6%) RR 2.44 95% CI 1.50 to 3.95; P=Not reported Clinically important harm signal consistent with invasive support and anticoagulation.
Peripheral ischaemic vascular complications requiring surgical/interventional therapy (safety) 23/209 (11.0%) 8/208 (3.8%) RR 2.86 95% CI 1.31 to 6.25; P=Not reported Major limb complications increased with ECLS.
Stroke or systemic embolisation (safety) 8/209 (3.8%) 6/208 (2.9%) RR 1.33 95% CI 0.47 to 3.76; P=Not reported Numerically higher; imprecise estimate.
  • Mortality at 30 days was similar (47.8% with routine early ECLS vs 49.0% with standard care; RR 0.98; 95% CI 0.80 to 1.19; P=0.81).
  • Routine early ECLS increased major harms: moderate/severe bleeding (23.4% vs 9.6%; RR 2.44; 95% CI 1.50 to 3.95) and limb ischaemic complications requiring intervention (11.0% vs 3.8%; RR 2.86; 95% CI 1.31 to 6.25).
  • Resource utilisation was higher in the ECLS group (e.g., mechanical ventilation median 7 vs 5 days; ICU stay median 10 vs 8 days; hospital stay median 12 vs 10 days).

Internal Validity

  • Randomisation and Allocation: Central computerised, block-wise randomisation per site via an internet-based system; randomisation occurred immediately after diagnostic coronary angiography (reducing post-angiography selection but increasing “late shock” phenotype enrichment).1
  • Drop out or exclusions: 420 patients were enrolled; 3 were excluded from intention-to-treat analyses because informed consent could not be obtained (2 assigned to ECLS; 1 assigned to control), leaving 209 vs 208 analysed.
  • Performance/Detection Bias: Open-label delivery creates risk of differential co-interventions and thresholds for organ support; the primary endpoint (death) is objective and less vulnerable to ascertainment bias.
  • Protocol Adherence: ECLS was initiated during the index angiography in 192/209 (91.9%) in the ECLS group; ECLS was not initiated despite assignment in 17/209 (8.1%).
  • Baseline Characteristics: Groups were closely matched on key severity markers (median age 62 vs 63 years; 81% male; median lactate 6.8 vs 6.9 mmol/L; resuscitation before randomisation 162/209 [77.5%] vs 162/208 [77.9%]).
  • Heterogeneity: Multicentre delivery across experienced sites increases pragmatic relevance but permits practice variation (e.g., cannula selection, anticoagulation management, and LV unloading strategy).
  • Timing: The intervention was operationalised as “routine early ECLS” around angiography/revascularisation; ECLS was initiated before revascularisation in 44/201 (21.9%) vs 32/208 (15.4%) and after revascularisation in 104/200 (52.1%) vs 112/208 (53.8%) (reflecting a substantial proportion receiving ECLS after PCI even in the intervention arm).
  • Dose: Median ECLS duration was 2.7 days (IQR 1.5–4.8) in the ECLS group; oxygenator and cannula-related “dose” (flows, pressures, pulsatility) were not reported in the primary manuscript.
  • Separation of the Variable of Interest: Routine early ECLS achieved meaningful (but not complete) separation: ECLS not initiated in 17/209 (8.1%) assigned to ECLS; ECLS was used as rescue in 26/208 (12.5%) assigned to standard care.
  • Key Delivery Aspects: Distal limb perfusion was used in 182/191 (95.3%) of ECLS-group recipients, indicating high-quality cannulation practice; however, active LV unloading during ECLS was uncommon (11/190 [5.8%] in the ECLS group), limiting inference for “ECMO + routine unloading” strategies.
  • Crossover: Control-group rescue ECLS occurred in 26/208 (12.5%), mainly for clinical deterioration; this could dilute any mortality signal (bias towards the null) but also reflects ethically appropriate rescue in extreme shock.
  • Adjunctive therapy use: Other mechanical circulatory support (among patients without ECLS) was more frequent in standard care (28/208 [15.4%]) than in the ECLS group (8/209 [4.4%]); targeted temperature management was also more common in standard care (109/208 [52.4%] vs 82/209 [39.2%]).
  • Outcome Assessment: Death at 30 days is robust; several secondary outcomes are susceptible to provider decisions (duration of ventilation/ICU stay/catecholamine exposure), and neurological outcome used a post hoc definition (CPC 3–4 among survivors).
  • Statistical Rigor: The trial met the planned recruitment target, used a group-sequential design with prespecified alpha spending (final alpha 0.048), and analysed outcomes by intention-to-treat; the observed effect was markedly smaller than the effect size assumed in power calculations (RR near 1.0).

Conclusion on Internal Validity: Overall, internal validity is moderate: randomisation and objective primary outcome support causal inference, but open-label care, meaningful non-adherence/crossover (8.1% and 12.5%), and heterogeneity in ECLS adjunct strategies (notably low LV unloading) plausibly attenuated detectable treatment effects.

External Validity

  • Population Representativeness: The enrolled cohort reflects a very high-acuity phenotype (arterial lactate >3 mmol/L by design; ~78% resuscitated before randomisation; median lactate ~6.8–6.9 mmol/L), which may differ from “earlier” shock populations where the risk–benefit balance could be different.
  • Centre Capability: Participating sites were predominantly well-resourced and experienced (e.g., high prevalence of cardiac surgery and perfusionist services within participating centres in the supplemental site descriptions), which may limit generalisability to centres without mature ECLS programmes and shock-team infrastructure.1
  • Applicability: Findings apply most directly to infarct-related shock managed with early revascularisation and femoro-femoral VA-ECMO delivered in tertiary systems; extrapolation to non-AMI shock, prehospital ECLS, or mandated LV unloading (“ECMELLA” pathways) should be cautious.
  • Exclusions: Patients with prolonged resuscitation (>45 minutes), late shock onset (>12 hours), or vascular disease precluding cannulation were excluded, limiting applicability to these common real-world scenarios.

Conclusion on External Validity: Generalisability is moderate: results are most relevant to high-acuity infarct-related shock in specialised centres; transferability to earlier shock, lower-resource systems, and alternative VA-ECMO management paradigms (particularly routine unloading) is limited.

Strengths & Limitations

  • Strengths: Large, pragmatic randomised trial in a high-mortality condition; objective primary endpoint; prespecified interim monitoring and alpha control; delivered within an experienced shock network; rich reporting of clinically important harms (bleeding and limb ischaemia).
  • Limitations: Open-label design; incomplete intervention separation (8.1% non-initiation in ECLS arm; 12.5% rescue ECLS in control arm); heterogeneity in ECLS adjunct strategies with low LV unloading utilisation; several secondary outcomes influenced by clinician behaviour and competing risks; neurological outcome definition was post hoc.

Interpretation & Why It Matters

  • Routine early VA-ECMO is not a survival strategy
    In this population (lactate >3 mmol/L; ~78% post-resuscitation), routine early ECLS did not reduce 30-day mortality (47.8% vs 49.0%; RR 0.98; 95% CI 0.80 to 1.19).
  • Harm and resource burden are real
    ECLS increased major bleeding (23.4% vs 9.6%) and limb ischaemic complications requiring intervention (11.0% vs 3.8%), and was associated with longer ventilation and ICU/hospital stays.
  • Shifts the default to selective use
    The absence of benefit with excess harm argues against “routine early cannulate” pathways for all infarct-related shock; the therapeutic problem becomes selecting the minority who may benefit (and delivering ECLS in a way that minimises predictable complications).

Controversies & Subsequent Evidence

  • Physiology vs protocol reality (LV unloading): The trial evaluated routine VA-ECMO without mandated unloading; active LV unloading during ECLS occurred in 11/190 (5.8%) in the ECLS group, limiting applicability to contemporary “ECMO + unloading” strategies. Observational data suggest timing of active unloading during VA-ECMO may be associated with outcomes, but this was not tested here.5
  • Phenotype enrichment and the “too sick / too late” problem: High lactate (median ~6.8–6.9 mmol/L) and high prevalence of cardiac arrest before randomisation (~78%) may indicate a population where neurological injury and established multiorgan failure dominate mortality, potentially limiting the modifiable fraction of risk (highlighted in critical commentaries).3
  • Treatment separation and rescue ethics: Non-initiation of assigned ECLS (8.1%) and rescue ECLS in the control arm (12.5%) plausibly attenuate a mortality signal; however, these features reflect pragmatic care delivery and ethically necessary rescue in deteriorating shock.1
  • Secondary neurological outcome definition: “Poor neurological outcome” was defined post hoc as CPC 3–4 among survivors; this design choice affects interpretability (especially in a cohort with frequent pre-randomisation arrest) and underscores how outcome definition and competing risks matter in shock-ECLS trials.1
  • Editorial framing: The accompanying editorial emphasised that routine early ECLS adds substantial morbidity without clear survival benefit, reinforcing the need for careful patient selection and systems-level capability before deployment.2
  • Follow-up study: One-year outcomes have been published and are consistent with the short-term findings (no clear survival advantage with routine ECLS strategy).4

Summary

  • ECLS-SHOCK randomised 420 patients with severe infarct-related cardiogenic shock (arterial lactate >3 mmol/L) to routine early ECLS vs standard care.
  • Thirty-day mortality was similar with and without routine early ECLS (47.8% vs 49.0%; RR 0.98; 95% CI 0.80 to 1.19; P=0.81).
  • Routine early ECLS increased major bleeding (23.4% vs 9.6%; RR 2.44) and limb ischaemic vascular complications requiring intervention (11.0% vs 3.8%; RR 2.86).
  • ECLS was associated with greater resource utilisation (mechanical ventilation 7 vs 5 days; ICU stay 10 vs 8 days; hospital stay 12 vs 10 days).
  • The trial is practice-shaping: it argues against routine early VA-ECMO for all infarct-related shock and focuses attention on phenotype selection and safer delivery strategies.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • Where DOI or full bibliographic fields were not available in the provided trial documents, a journal/PubMed search link is supplied to enable rapid source verification without inventing citation details.
  • ECLS-SHOCK primarily informs practice for routine early VA-ECMO without mandated LV unloading in very high-acuity infarct-related shock (lactate >3 mmol/L; high prevalence of pre-randomisation arrest).

Overall Takeaway

ECLS-SHOCK is a landmark because it is among the first adequately powered randomised evaluations of a routine early VA-ECMO strategy in infarct-related cardiogenic shock and shows no 30-day survival benefit alongside substantially increased bleeding and limb complications. The trial shifts the default away from routine early cannulation towards selective, phenotype-driven use and safer delivery strategies (including consideration of unloading pathways that were not tested here).

Overall Summary

  • Routine early ECLS did not reduce 30-day mortality (47.8% vs 49.0%).
  • ECLS increased major bleeding (23.4% vs 9.6%) and limb ischaemic complications needing intervention (11.0% vs 3.8%).
  • Pragmatic separation was incomplete (8.1% non-initiation; 12.5% rescue ECLS), and LV unloading was uncommon (5.8%).

Bibliography