Acute myocardial infarction-related cardiogenic shock (AMI-CS) carries a dismal prognosis. Short-term mortality is in the range of 40–50% [1]. Until recently, only treatment of the culprit lesion by percutaneous coronary intervention (PCI) reduced mortality within randomized controlled trials (RCT) [1]. High expectations have been placed on venoarterial extracorporeal membrane oxygenation (VA-ECMO) and its use has risen by up to 40 times in the last decade despite a lack of relevant evidence from RCTs [2].

The concept is to provide temporary partial or complete circulatory and also respiratory support during the critical first days as a bridge-to-recovery, bridge-to-decision, bridge-to-durable left ventricular assist device (LVAD), or bridge-to-transplantation.

Evidence for VA-ECMO on mortality and secondary endpoints

Evidence regarding routine percutaneous VA-ECMO in AMI-CS is relatively robust with now 4 RCTs (ECLS-SHOCK I: n = 42: EURO SHOCK: n = 35 patients, ECMO-CS: n = 117 patients, ECLS-SHOCK: n = 420 patients) [3,4,5,6]. The only study powered for a mortality difference is the ECLS-SHOCK trial, which included 420 randomized patients with AMI-CS [6]. By study design, the included patients had more advanced CS, as lactate of > 3 mmol/L was defined as an inclusion criterion. There was no difference in 30-day mortality (49% control versus 47.8% VA-ECMO; relative risk 0.98; 95% confidence interval [CI] 0.80–1.19; p = 0.81) [6]. The neutral results in the primary endpoint were further supported by a lack of effect on secondary endpoints, such as lactate clearance, renal function, catecholamine use and duration (Fig. 1).

Fig. 1
figure 1

Current evidence of VA-ECMO in infarct-related cardiogenic shock, harm and possible limitations in evidence. Middle: current evidence for 30-day mortality of VA-ECMO versus control in infarct-related cardiogenic shock. Right panel: current evidence for harm of VA-ECMO versus control in infarct-related cardiogenic shock. Left panel: open issues and possible limitations of current evidence. AMI-CS acute myocardial infarction-related cardiogenic shock, ICU intensive care unit, ITT intention-to-treat, VA-ECMO venoarterial extracorporeal membrane oxygenation, OHCA out-of-hospital cardiac arrest, RCT randomized controlled trial, LVAD left ventricular assist device, HTx heart transplantation, PCI percutaneous coronary intervention

Full size image

The evidence for lack of benefit of VA-ECMO is further supported by an individual patient data (IPD) meta-analysis incorporating results from all 4 RCTs [7]. There was no significant 30-day mortality benefit for AMI-CS patients receiving routine VA-ECMO (45.7%) in comparison to control (47.7%), (odds ratio [OR] 0.92, 95% CI 0.66–1.29) [7].

Safety of VA-ECMO

VA-ECMO use was associated with a 23.4% rate of moderate to severe bleeding versus 9.6% in control (relative risk, 2.44; 95% CI 1.50–3.95) in ECLS-SHOCK [6]. This has been confirmed in the IPD meta-analysis (OR 2.44; 95% CI 1.56–3.84) [7]. Since bleeding is known to be associated with worse outcomes [8], these results indicate that VA-ECMO may even be harmful for those experiencing this complication.

Another typical drawback of VA-ECMO are peripheral ischemic complications. Despite a high rate (> 95%) of prophylactic antegrade perfusion cannulae applied in ECLS-SHOCK, ischemic complications occurred with an OR of 2.86 (95% CI 1.31–6.25) even further aggravated in the IPD meta-analysis (OR 3.53; 95% CI 1.70–7.34) [6, 7]. VA-ECMO modifications to enable LV-unloading, such as VA-ECMO + Impella or VA-ECMO + intraaortic balloon pump, should be further scrutinized, as despite potential benefit for LV recovery, they may increase bleeding risks even more.

Another side-effect of VA-ECMO was prolongation of mechanical ventilation time as well as intensive care unit stay by roughly 2 days which may have also caused more harm than benefit [6].

Limitations of current evidence

Negative or neutral trials trigger discussions in particular when results do not follow what the general perception is, i.e. VA-ECMO reduces CS mortality. A typical reflex is then to argue by registry data that RCTs are not valid [9].

Center experience

Some may argue that limited center experience has influenced the neutral results. However, for center participation, pre-study selection visits and pre-study qualification assessment to include experienced VA-ECMO centers only was performed. Based on this: 61.4% were tertiary university centers, 79.5% had a cardiac surgery department and 97.7% had a separate perfusionist department in ECLS-SHOCK.

Inclusion of resuscitated patients

The high rate of patients with successful resuscitation prior randomization (> 70%) in ECLS-SHOCK may have limited VA-ECMO results because, hypoxic brain injury cannot be positively influenced by mechanical circulatory support and shock/hypotension and elevated lactate after resuscitation may not be directly associated with prolonged decreased cardiac output to a similar extent as in CS without cardiac arrest. Notably, evidence for reduced cardiac output was not required in ECLS-SHOCK. As a result of the risk enhancement for inclusion, this rate was higher than in previous RCTs in AMI-CS. Furthermore, mortality in resuscitated patients was numerically even lower than in those without resuscitation [6]. In the IPD meta-analysis the number of resuscitated patients was lower and also no benefit was observed [7]. Importantly, exclusion of resuscitated patients would lead to less generalizability to all CS-like patients of any trial result.

VA-ECMO timing

Results derived from an observational meta-analysis for AMI-CS (IABP: n = 956, Impella: n = 203, VA-ECMO: n = 193), suggest, that initiation of VA-ECMO prior to PCI reduces mortality [10]. However, this was refuted in a subanalysis in ECLS-SHOCK and the IPD meta-analysis [6, 7].

There are also some other timing aspects. In ECLS-SHOCK, VA-ECMO was started routinely after randomization. It remains unclear whether there is any clinical benefit to a watch and wait strategy and to decide for or against VA-ECMO or other devices only in case of further clinical deterioration as this was the case in 26% of control patients.

Further patient selection aspects

In addition, to inclusion of resuscitated patients and timing aspects, the ECLS-SHOCK trial included patients with more advanced shock severity based on signs of tissue hypoperfusion. The SCAI shock classification was not in place at study start and the definition is dynamic which usually does not allow to perform immediate staging [11]. The distribution of the SCAI stages was, therefore, made retrospectively in ECLS-SHOCK using a modified post hoc definition [6]. Some argue therefore that SCAI C patients were not sick enough to benefit from VA-ECMO or on contrary SCAI stage E patients to be in a futile situation. Irrespective from these considerations no SCAI stage had a benefit from VA-ECMO.

The question remains whether specific patient subgroups in AMI-CS benefit from VA-ECMO, as current guidelines do not reflect upon patient selection [12]. Importantly, there was no signal for a survival benefit of VA-ECMO in any of the subgroups analyzed [6, 7]. If larger trials or artificial intelligence may identify specific subgroups needs to be shown in the future.

Low rate of left ventricular unloading

By its mode of operation, VA-ECMO increases afterload. Multiple unloading strategies have been developed but these also increase invasiveness and possibly complications. In ECLS-SHOCK, unloading criteria were predefined which led to a relatively low rate of approximately 6% active unloading. However, more patients in the VA-ECMO group were treated by dobutamine indicating medical unloading by increasing ventricular inotropy. When evaluating evidence for active unloading, it is also important to note that potential benefits were generated from registry studies only [13, 14]. A recent RCT comparing routine LV unloading by a transseptal left atrial cannula versus VA-ECMO alone showed no effect on mortality [15]. This evidence of VA-ECMO plus routine unloading must surely be generated before a further reflexive VA-ECMO plus routine Impella unloading is performed.

Low use of durable left ventricular assist devices or heart transplantation

In ECLS-SHOCK—similar to previous RCTs—the rate of patients getting a durable LVAD or heart transplantation was < 2%. Advanced heart failure specialists often argue that VA-ECMO is mainly considered as bridge-to-LVAD or transplantation and therefore the trial was doomed to fail [9]. Patients included in RCTs in AMI-CS are often in the upper age not being eligible for such treatment strategies. In addition, many of these patients have high rates of concomitant inflammation or infections precluding these advanced therapies.

In summary, for the vast majority of patients with AMI-CS, routine immediate VA-ECMO is not of benefit. If any subgroup is missing and if treatment modifications with complication reductions or LV unloading strategies may alter the outcome remains to be defined in future RCTs. In addition, it remains to be defined which patients may benefit from VA-ECMO in case of further clinical deterioration, where likely a relevant remaining group might be the refractory cardiac arrest.