Mechanical circulatory support in patients with cardiogenic shock not secondary to cardiotomy: a network meta-analysis

Cardiogenic shock (CS) is the most severe manifestation of acute heart failure [1]. The underlying pathophysiology is characterized by left ventricular (LV) dysfunction and elevated filling pressures, systemic hypoperfusion, and neurohormonal activation, triggering and perpetuating one another [2]. As such, treatment of CS is effective only if the cause is promptly removed and/or this vicious spiral is interrupted in the early phases. Indeed, early revascularization has cut mortality in post-myocardial infarction CS (MI-CS) [3]. On the contrary, the increase in cardiac oxygen consumption and afterload induced by inotropic and vasopressor drugs, respectively, outweighs the beneficial effects of these therapies and may eventually lead to higher mortality [4]. Mechanical circulatory support (MCS) improves hemodynamics without the detrimental consequences of pharmacological interventions; therefore, it is generally considered the best non-etiological therapy for CS. Following the introduction of the intra-aortic balloon pump (IABP) in clinical practice, several other devices for MCS, including micro-axial pumps (Impella, Abiomed, Danvers), centrifugal extracorporeal pumps (TandemHeart, LivaNova, London), and extracorporeal membrane oxygenation (ECMO), have become available [5]. Furthermore, emerging evidence suggests that the combination of different MCS is preferable to the use of a single device to support hemodynamics and decompress the LV in the meanwhile [6].

However, studies about the efficacy-safety profile of MCS in CS have produced discordant results, and the choice of the type of MCS remains largely based on individual expertise, costs, and local availability. The discrepancies in the literature about MCS in CS at least in part stem from the substantial differences existing between the two most common presentations of CS, i.e., MI-CS and post-cardiotomy CS, and from the variable duration of the follow-up, ranging from the length of the hospitalization to 2 weeks or 1 month. Clearly, the longer is the time window during which mortality is evaluated, the stronger are the conclusions on the effect of MCS on survival. These shortcomings have been carried over in the meta-analyses published so far on the topic [6‐12]. Furthermore, the statistical approach adopted for these meta-analyses did not allow drawing comparisons between different types of MCS when head-to-head primary studies were not available.

To overcome these limitations, we performed a network meta-analysis to obtain a comprehensive view of the data supporting the different MCS strategies in patients with non-post-cardiotomy CS.

This meta-analysis was registered on PROSPERO (CRD42020189859) and followed the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) recommendations [13] (Appendix Table S1).

Our search strategy yielded 4590 articles, of which 24 met the selection criteria and were thus analyzed (Appendix Figure S1 shows the PRISMA flowchart). Seven were RCTs and 17 observational studies; the list and corresponding references are given in Appendix Table S3. Among the included articles, 10 compared IABP vs. no MCS, 6 IABP vs. Impella, 3 IABP vs. TandemHeart, 2 ECMO vs. IABP, and 2 ECMO vs. ECMO + IABP, whereas each remaining direct comparison was based on a single study. The overall population amounted to 11,117 subjects, with 845 (7.6%), contributing with a total of 337 events, being from RCTs.

The number of patients treated with each MCS strategy is displayed in Fig. 1a. The main features of each study and the patients’ baseline characteristics are reported in Appendix Tables S4-S5. Median age, as calculated by the values provided in 21 (84%) of the selected articles, was 65 years. With the exception of the control arm of one investigation (20), male gender was always more frequent than female one. Information about body mass index, cardiovascular risk factors, and renal function was largely missing (Appendix Table S5). Remarkably, 20 studies (N = 9,864, 88.7% of total) included only MI-CS patients. Bias assessment is reported in Appendix Table S6. Overall, the risk of bias was low for RCTs, whereas it ranged from moderate to high for non-RCT studies.

The role played by MCS in the treatment of CS is still debated. From the pathophysiological standpoint, MCS is preferable to drugs with inotropic and vasopressor activity, since it compensates for the reduced output of the failing heart and maintains systemic perfusion [19]. However, the overall impact of MCS on hard clinical endpoints in RCTs has been disappointing [20‐22]. In fact, the use of IABP has been downgraded to a class IIa level of evidence (LoE) B and IIa LoE C recommendations, respectively, in the US and European guidelines [23, 24] following the results of the IABP-SHOCK II RCT [20, 25]. Nevertheless, it has been pointed out that at least some RCTs were not adequately powered to detect differences in mortality. It has also been argued that the population enrolled in RCTs was not representative of the patient phenotype that would actually benefit from MCS. RCTs might have missed the so called “sweet spot,” i.e. ,the best timing to implant a device for MCS before it is no longer effective in modifying the course of CS and, thereby, becomes futile [26, 27]. Therefore, a better estimate of the impact of MCS in CS may be obtained by pooling together the data from RCTs and observational studies. Here, we did so by performing a network meta-analysis. This methodology, which has only recently introduced in the field [28], allowed us to compare all the MCS strategies for CS on a solid statistical basis. By contrast, the classical meta-analytic approach does not measure the relative efficacy and safety of interventions that were not directly compared in prior studies.

Our work is also original for other aspects. We restricted the population of interest to the one affected by CS not secondary to cardiotomy, in order to avoid a selection bias in favor of ECMO, which is the most adopted MCS in the surgical setting. Moreover, we rigorously assessed mortality at 30 days, which in our opinion is a stronger efficacy outcome compared with in-hospital or 2-week mortality and gives a clear-cut measure of the effect that MCS may have in clinical practice.

Our meta-analysis suggests that ECMO reduces 30-day mortality of patients with CS, especially in combination with Impella or IABP. ECMO is a powerful MCS, which replaces the function of both ventricles and provides a flow of up to 4–6 l/min [29]. However, it injects oxygenated blood against the LV, increasing the afterload, worsening LV filling pressures, and favoring pulmonary congestion [29, 30]. This drawback can be overcome by surgical venting or by combining ECMO with another percutaneous MCS. Thus, unloading of the LV is a mechanistic explanation for the further decrease in mortality we found with the association of ECMO with Impella or IABP, as compared with ECMO alone. Indeed, Impella was shown to lower pulmonary capillary wedge pressure when added to ECMO (32), with hemodynamic and clinical results comparable to those of surgical venting [31]. Moreover, in a recent meta-analysis focused on the effects of combining IABP with ECMO, this approach significantly reduced short-term mortality in MI-CS [6]. A double-MCS strategy also offers the possibility of de-escalating the support to the LV, allowing a smoother transition toward weaning or destination therapy [32]. Unfortunately, we could not retrieve sufficient information about surgical venting in patients who received ECMO alone. Therefore, the superiority of ECMO + Impella or IABP might be due to inadequate venting in the arms treated with ECMO alone, rather than to the advantage of adding a second MCS device.

The present systematic review also revealed a significant publication bias, which must be taken into consideration when discussing the apparent efficacy of ECMO (with or without Impella or IABP), but not of the other types of MCS on mortality in CS. No RCT has evaluated ECMO and ECMO + Impella so far, and the benefit of these strategies for MCS was no longer significant after adjusting for study design.

It should be also noted that most of the studies included in the analysis were with Impella 2.5, when newer and more powerful Impella devices are available. Interestingly, observational investigations showed a similar survival with ECMO and Impella CP after adjusting for patients’ risk profile [33]. Hence, our results cannot be generalized to the whole Impella family. The ongoing Danish Cardiogenic Shock trial will shed light regarding the efficacy of Impella CP in MI-CS [34].

In this meta-analysis, there was a gradient in bleeding risk, with TandemHeart being the worst strategy and both Impella and IABP significantly increasing the odds of bleeding compared with no MCS. TandemHeart and Impella require large-bore arterial cannulation, which can be complicated by access-site bleeding. Furthermore, both TandemHeart and Impella require anticoagulation with unfractionated heparin, which increases the likelihood of bleeding. Although not mandatory, anticoagulation is also often adopted for IABP and this may be the reason why this device was also associated with a higher risk of bleeding. Due to the lack of data, we could not assess bleeding with ECMO. However, the considerations above hold true for this device, which has been associated with a significant and prognostically meaningful risk of bleeding [35]. Indeed, in 20 studies including 1866 subjects, the rate of ECMO-related bleeding was 41% [36].

CS is a complex syndrome, and our findings may not be valid for specific scenarios. For instance, we could not assess the efficacy of the different devices for MCS according the severity and duration of CS, or to the absence or presence of right ventricular failure. Patients treated with Impella and presenting with an INTERMACS class II have better survival than those in INTERMACS class I [37]. Moreover, Impella is a pure LV MCS and, as such, is not suitable in CS with biventricular failure [38]. Comorbidities and concomitant medications are other factors that influence the performance of MCS in CS, but we could not take into account.

Local expertise is another factor that is key for the successful management of CS, but we could not gauge. IABP is still widely used for incipient CS in centers where other types of MCS are not available or rarely employed [38]. Conversely, Impella and ECMO need dedicated personnel, who handle the devices in the optimal manner only after a learning curve, and a structured network of spoke and hub centers [2].

Favoring ECMO and, to a greater extent, its combinations with an unloading-dedicated device, our results prove that times would be ripe to test these strategies in adequately powered RCTs. Focusing on specific etiologies (e.g. non post-cardiotomy CS, as we did in the present work) and phenotypes (e.g., biventricular vs. LV isolated dysfunction) and standardizing the endpoints definitions would be necessary to make future studies more informative.

Data from RCTs and observational studies indicate that, in CS with various etiologies, ECMO significantly decreases short-term mortality as compared with other types of MCS or no support, especially when used in association with Impella or IABP. This finding should be considered hypothesis-generating and inform larger and adequately powered RCTs. The risk of bleeding is enhanced by MCS, but data on this outcome with ECMO are very limited.