Mechanical circulatory support in cardiogenic shock from acute myocardial infarction: Impella CP/5.0 versus ECMO

Abstract

Background

Short-term mechanical circulatory support devices are increasingly used in cardiogenic shock after acute myocardial infarction. As no randomised evidence is available, the choice between high-output Impella or extra-corporeal membrane oxygenation (ECMO) is still a matter of debate. Real-life data are necessary to assess adverse outcomes and to help guide the treatment decision between the different devices. The purpose of this study was to compare characteristics and clinical outcomes of Impella CP/5.0 with ECMO support in patients with cardiogenic shock from myocardial infarction.

Methods

A retrospective, two-centre study was performed on all cardiogenic shock from myocardial infarction patients with Impella CP/5.0 or ECMO support, from 2006 until 2018. The primary outcome was 30-day mortality. Potential baseline imbalance between the groups was adjusted using inverse probability treatment weighting, and survival analysis was performed with an adjusted log-rank test. Secondarily, the occurrence of device-related complications (limb ischaemia, access site-related bleeding, access site-related infection) was evaluated.

Results

A total of 128 patients were included (Impella, N=90; ECMO, N=38). The 30-day mortality was similar for both groups (53% vs. 49%, P=0.30), also after adjustment for potential baseline imbalance between the groups (weighted log-rank P=0.16). Patients with Impella support had significantly fewer device-related complications than patients treated with ECMO (respectively, 17% vs. 40%, P<0.01).

Conclusions

Patients treated with Impella CP/5.0 or ECMO for cardiogenic shock after myocardial infarction did not differ in 30-day mortality. More device-related complications occurred with ECMO compared to Impella support.

Introduction

Cardiogenic shock (CS) in acute myocardial infarction (MI) remains a clinical challenge, with poor outcomes.¹ Mechanical circulatory support (MCS) is increasingly used as a treatment strategy in CS after MI. Several high-output MCS devices have been introduced into clinical practice for the treatment of CS. These short-term devices can be used as either a bridge to recovery, to a durable ventricular assist device (LVAD), to transplant or to decision strategy. MCS therapy increases cardiac output in patients with CS with the ultimate goal of improving end-organ tissue perfusion. The 2016 European Society of Cardiology (ESC) guidelines for the diagnosis and treatment of acute and chronic heart failure supports the use of MCS in CS patients who fail to stabilise or improve after initial pharmacological therapy.²

Current available high-output MCS devices, such as the Impella axial-flow pump devices and extracorporeal membrane oxygenation (ECMO), differ significantly in their physiological effects, insertion method, clinical management and applicability in patients with CS.³ However, there is no randomised evidence that favours the use of one high-output device over another in patients with CS after MI. As a result, the choice of the device that is used for circulatory support is still a matter of debate, and is generally based on local availability and experience. If we were able to compare the characteristics and clinical outcomes of patients treated with high-output MCS, this could help guide treatment decisions between the different devices in patients with CS after MI. In the absence of randomised trials, real-life data on the actual use of MCS and device-related complications are insightful.

In this study we focus on patients with CS after acute MI, who received high-output mechanical support. The aims of this study were to compare the characteristics and clinical outcomes of Impella CP/5.0 versus ECMO supported patients with CS after MI.

Methods

A retrospective, two-centre study was performed of all patients receiving high-output MCS in the setting of CS after MI until January 2018 at the Academic Medical Center (AMC), and the Erasmus Medical Center (MC). Both hospitals are tertiary referral centres that offer interventional and cardiac surgical care. The Erasmus MC is also a destination centre for patients requiring a long-term LVAD or heart transplantation. In the period of our study, the Impella was the only available high-output MCS device for the AMC, and ECMO was the only available high-output device for the Erasmus MC. Therefore, all patients in the Impella cohort were treated in the AMC, and all ECMO supported patients were treated in the Erasmus MC.

Patient selection

Patients were excluded from our analysis if the high-output MCS device was placed after revascularisation with coronary artery bypass grafting, and if patients only received passive or low-flow haemodynamic support (i.e. intra-aortic balloon pump (IABP) or Impella 2.5). Cardiac arrest patients in whom device insertion took place during ongoing cardiopulmonary resuscitation (i.e. without return of spontaneous circulation; ROSC) were also excluded from the cohort.

Data collection

Patient demographics, procedural characteristics, haemodynamics, laboratory values, complications and clinical outcomes were obtained by complete chart review. Follow-up data were gained by examination of discharge letters and inpatient and outpatient charts from the hospitals and referring centres.

Definitions

Baseline transthoracic echocardiography was evaluated prior to device placement. Left ventricular (LV) failure was classified as mild (40–49%), moderate (30–39%) or severe (<30%) and measured with the biplane Simpson method, estimated per 5–10%. Right ventricular (RV) failure was defined as a tricuspid annular plane systolic excursion less than 16 mm, or qualitatively assessed moderate or severe RV systolic dysfunction together with at least moderate tricuspid regurgitation and dilated inferior caval vein.⁴ Patients who met both criteria for severe LV and RV failure were classified as biventricular heart failure. MI was diagnosed and defined as dynamic electrocardiographic changes with a typical rise or fall in cardiac markers.

Successful weaning from a MCS device was defined as patients who survived at least 48 hours after device removal. Device-related vascular complications were defined as limb ischaemia (requiring surgical treatment or extraction of the device), access site infection (defined as catheter-associated bloodstream infections), or access site bleeding (defined as a bleed requiring interventional/surgical treatment or blood transfusion). Clinically relevant haemolysis was defined as haemolysis requiring transfusion or device extraction. Diagnosis of haemorrhagic or ischaemic stroke was confirmed by a neurologist and subsequent computed tomography scan.

The IABP-SHOCK II score was calculated for the cohort.⁵ This score is used for risk stratification in patients with CS after acute MI and consists of six variables that are independent predictors of 30-day mortality. One or two points were attributed to each variable, leading to classification in three risk categories: low (score 0–2), intermediate (score 3–4) and high (score 5–9). Patients with missing values were omitted from the calculation.

MCS policy

Patients with CS who displayed potential for recovery were considered for high-output MCS if their systemic perfusion did not recover with inotropes, vasopressors, IABP, and/or Impella 2.5 therapy. CS was defined as a clinical diagnosis made by the physician, based on the criteria for sustained hypotension (i.e. systolic blood pressure (SBP) ≤90 mmHg for at least 30 minutes or the need for vasopressors to maintain a SBP >90 mmHg) caused by heart failure, and other clinical signs of hypoperfusion (i.e. cold extremities, oliguria, altered mental state, elevated lactate levels, low mixed venous oxygen saturation; S(c)vO₂). The decision to insert a MCS device and the timing of placement (before or after revascularisation) were at the discretion of the treating physician. If the support level was insufficient and there was a need for escalation in the AMC (Impella cohort), patients could be upgraded to an Impella 5.0 (if Impella CP was implanted) or transferred to another centre for ECMO (in cases with profound biventricular or respiratory failure).

Clinical management

Dual antiplatelet therapy was prescribed in all patients post-percutaneous coronary intervention (PCI). All patients were treated with unfractionated heparin anticoagulation during MCS. For Impella supported patients the activated partial thromboplastin time was maintained at 45–60 seconds and heparin was continued until device removal. During ECMO support the activated partial thromboplastin time was maintained between 55 and 75 seconds and heparin was continued until 30 days post-ECMO.

Mechanical support systems

Impella CP and 5.0 devices

The Impella CP device (Abiomed Inc., Danvers, MA, USA) was inserted percutaneously by way of a 14 Fr introducer in the femoral artery. Measured in an ideal setting, this device provides up to 4.0 L/minute output at its maximum level. The Impella 5.0 provides a maximum output of 5.0 L/minute, but its introduction requires surgical insertion by way of a graft in the ascending aorta, subclavian artery or axillary artery. The Impella devices were placed in the catheterisation laboratory, under angiographic guidance, by the interventional cardiologists or cardiothoracic surgeon.

The need for Impella support continuation was evaluated daily by the responsible physician and initiated when inotropes and vasopressors could be reduced or on echocardiographic signs of LV recovery. The Impella CP device was removed percutaneously approximately one hour after heparin cessation, followed by 30 minutes of femoral compression. The Impella 5.0 device was removed surgically.

ECMO support

The ECMO system used was a PLS or Cardiohelp set (oxygenator and pump; Getinge Group, Wayne, NJ, USA) with blood flow set at 3.5–5.0 L/minute in order to reach stable haemodynamics (SvO₂ >60%, mean arterial pressure ≥60 mmHg, low lactate level, and diminished need for vasopressors) and regular aortic valve opening. In all cases, venoarterial (VA-)ECMO was placed by bifemoral cannulation by way of a percutaneous or surgical insertion, in which an antegrade cannula for leg perfusion was routinely implanted. Cannulation was performed in the catheterisation laboratory or operating room. Routine strategy for LV venting was the use of an IABP. Another (infrequently used) option for LV venting was percutaneous LV vent insertion.

Readiness for withdrawal of ECMO support was evaluated daily by haemodynamic and Doppler echocardiographic assessment.⁶ The ECMO cannulae were removed surgically in the operating room.

Statistical analysis

We reported continuous variables as mean with standard deviation (SD), or in the case of a non-normal distribution as median with interquartile range (IQR). Categorical variables were presented as count with percentages and were tested using the chi-square test or Fisher’s exact test. Differences between groups were analysed by Student’s t-test, one-way analysis of variance (ANOVA) or the Mann–Whitney test, when appropriate. A two-sided P<0.05 was considered statistically significant. The primary outcome was 30-day mortality. Kaplan–Meier survival analyses were calculated and the curves were compared by the log rank test. To adjust for potential confounders between the two groups at baseline we used the inverse probability of treatment weighting (IPTW) based on the propensity score (i.e. 1/score). The propensity score was calculated by using a multivariate logistic regression model with treatment (Impella/ECMO) as the dependent variable and all baseline covariates measured prior to treatment (as listed in Table 1, including year of insertion and time between revascularisation and device insertion) as the independent variables. To correct for potential instability induced by very large weights in the IPTW, we used stabilised weights (i.e. IPTW × marginal probability of receiving the actual treatment). We applied multiple imputation methods to estimate missing data of the baseline covariates (6.8% of total baseline data (patients × variables) were missing). For the adjusted survival analysis we used a weighted log rank test as described by Xie and Liu.⁷ Analyses were performed using SPSS statistics version 25.0 (SPSS, Inc., Chicago, IL, USA). The investigation conforms with the principles outlined in the World Medical Association Declaration of Helsinki. Approval and waived consent was obtained from both local medical ethics committees.

Results

During the study period, ranging from 2006 (first patient with Impella support) until January 2018, 273 patients were treated with Impella CP/5.0 in the AMC. During the same period, 206 patients received VA-ECMO support in the Erasmus MC. In total, 128 CS patients were eligible for inclusion in our study (N=90 for Impella, N=38 for ECMO). The flowchart for included patients in the study and the overall use of the MCS devices throughout the years are presented in Supplementary Appendices A and B.

Baseline characteristics before device placement

The groups were similar at baseline regarding sex distribution (75% male), cardiovascular risk factors, mean arterial pH, lactate value, creatinine value, number of patients with a cardiac arrest or biventricular failure, mechanical ventilation and inotropic/vasopressor therapy (Table 1). The patients in the Impella group were older, presented more often with an out-of-hospital cardiac arrest (OHCA) and had longer median ROSC times, compared to the patients in the ECMO group. The patients in the Impella group also had a higher mean arterial pressure and haemoglobin value before device placement. Implantation of low-output MCS devices prior to the initiation of high-output MCS was done more frequently in the ECMO group (74% vs. 14%, P<0.01). ECMO supported patients also had worse LV function (severe LV failure in 95% of ECMO patients vs. 54% in Impella patients, P<0.01). The procedural characteristics among Impella and ECMO supported patients are shown in Table 2.

There was a difference between the two groups in the time of device placement (P<0.01) (Table 3). The median time between revascularisation and device placement was one hour (IQR 0–7) for Impella supported patients, compared to 4 hours (IQR 1–26) for ECMO patients (P<0.01).

Clinical outcomes

The duration of device support was 3 days (IQR 2–6) for Impella versus 6 days (IQR 3–8) for ECMO (P<0.01). The median length of intensive care unit (ICU) stay was 6 days (IQR 3–14) for patients treated with Impella and 16 days (IQR 9–30) for patients treated with ECMO (P<0.01). During admission, additional IABP was employed in 4% of the Impella supported patients and 55% of the ECMO patients (P<0.01). The peak creatinine kinase-MB value was similar between the two groups.

The 30-day mortality was similar for both groups (53% vs. 49%, P=0.30) (Figure 1). Likewise, there was no difference in one-year mortality for patients treated with Impella or ECMO (P=0.62) (Figure 2). After adjusting for potential baseline imbalance between the groups using inverse probability weighting there was still no difference in 30-day or one-year mortality (respectively, weighted log rank P=0.16 and P=0.37). The causes of death are listed in Table 4.

Complications

Patients treated with Impella had significantly fewer device-related complications than patients treated with ECMO (respectively, 17% vs. 40%, P<0.01). There were no differences in the occurrence of stroke or other bleeding complications between the groups. Of the Impella supported patients, 63% received blood products versus almost all (97%) ECMO patients (P<0.01) (Table 3). We observed a trend in patients requiring renal replacement therapy (RRT) (42% vs. 26%, P=0.10).

One patient received a long-term LVAD after Impella support, and five patients after ECMO support (P<0.01). One-year mortality, after the exclusion of patients who received a long-term LVAD, did not differ between the groups (57% vs. 68%, P=0.77) (Supplementary Appendix C).

Treatment before/after the year 2014

There were no significant differences in 30-day or one-year mortality between the patients in the cohort that were treated before 2014 (n=64) or after 2014 (n=64) (Supplementary Appendix B). Patients treated after 2014 had more device-related complications, compared to patients treated before 2014 (31% vs. 16%, P=0.04).

Severity score

The IABP-SHOCK II severity score was applied to the cohort.⁴ The majority of patients qualified as low or intermediate risk and the score could not accurately predict the risk of mortality for the cohort (mortality rates for low, intermediate and high-risk were respectively; 43%, 56% and 50%) (Supplementary Appendix E).

Discussion

This study is the largest comparative study of CS patients requiring high-output mechanical support after acute MI, including both Impella and ECMO. The main finding of this study, comparing clinical outcomes of Impella CP/5.0 and ECMO in patients with CS after MI, is that patients with ECMO support experienced significantly more device-related complications. We found no difference between the groups in 30-day mortality. However, we observed that ECMO supported patients required more blood transfusions, additional IABP, long-term LVADs and had a longer duration of support, compared to the patients supported with Impella.

Comparison

While several studies (all observational, retrospective single-centre) previously compared Impella to ECMO support in CS patients of various aetiologies, our study focussed on patients with CS after acute MI exclusively.^8–10 Also, we compared two centres that used either Impella or ECMO support. In our opinion, the value of this comparison may be that our results are likely to be less affected by selection bias in comparison to other (single-centre) observational studies, in which both Impella and ECMO were available (for example, in the study recently conducted by Schiller et al.,¹⁰ Impella was the preferred choice; VA-ECMO was used in the presence of severe respiratory dysfunction/haemodynamic instability). Furthermore, in contrast to previous studies, the focus of our study was on high-output devices (we excluded patients who were only supported by Impella 2.5, which may offer insufficient support in profound CS) and our study reported on important device-related complications.

Impella or ECMO support was not associated with a difference in 30-day mortality, similar to previous reports. Lamarche et al. compared Impella (5.0/RD) to ECMO support in CS (n=61, 39% MI-related) and found no difference in 30-day mortality between the groups (38% vs. 44%, P=0.64).⁸ Likewise, when Impella (5.0) and Tandemheart devices were compared to ECMO support in CS (n=79, 58% MI-related) 30-day mortality was 50% versus 49%.⁹ Schiller et al. compared the mortality of Impella (2.5/CP/5.0/LD) supported CS patients to ECMO supported patients (n=94, 28% MI-related) and adjusted for disease severity through the survival after VA-ECMO (SAVE) score. Also, they found no difference in ICU mortality between patient groups (37% vs. 35%, P=0.95).¹⁰

Patients with Impella support had fewer device-related complications than patients treated with ECMO (respectively, 17% vs. 40%, P<0.01). In our cohort, access site-related bleedings occurred in 12 (13%) patients with Impella support versus seven (18%) patients with ECMO support. Impella 5.0 is expected to have a greater risk of vascular complications compared to Impella CP/2.5, as it requires surgical insertion. However, our results are not very different from a large cohort that reported 9% major vascular complications related to Impella CP/2.5 support.¹¹

Limb ischaemia occurred in two (2%) Impella supported patients, and two (5%) ECMO patients. The incidence of limb ischaemia in patients supported with ECMO is relatively low in our study in comparison to other ECMO cohorts. Muller et al. reported an incidence of 11% leg ischaemia in their study.¹² The low incidence in our cohort could be the result of standardly placing an additional cannula for leg perfusion in all patients with ECMO, or could be the result of under-registration of this important complication in the patient records.

The incidence of access site-related infections in the ECMO supported group was 16% vs. 1% in the Impella group. Other cohorts regarding ECMO supported CS patients report a femoral site infection rate of 12–17%.^12,13

Impella versus ECMO support

Based on our results and experience we believe that clinicians are more reserved with the placement of ECMO compared to Impella, due to the invasiveness and logistical difficulties of the device. In our cohort, this was shown by the fact that ECMO was mostly inserted in a separate procedure after revascularisation, almost all patients had an IABP placed prior to ECMO and patients had a relatively low haemoglobin value at baseline (possibly caused by fluid resuscitation). The median time between revascularisation and ECMO placement was 4 hours, compared to one hour for Impella. This possible delay in the initiation of high-output MCS could be detrimental in CS patients. A recent study found that the early initiation of MCS (prior to PCI, before escalating doses of inotropes and within 90 minutes of shock onset) was associated with improved survival in selected patients with CS after MI.¹⁴ While a benefit of IABP in CS patients has not been shown, its use could lead to potential harm. In the IABP-SHOCK II trial, the IABP group and the control group did not differ significantly in complication rates.¹⁵ However, the use of IABP may delay the decision for escalation to high-output support.

Patients who were supported with Impella had a shorter duration of support and length of ICU stay. It is unclear whether this difference can be explained by a quicker recovery with Impella support, indication bias, or other reasons. While both devices improved end-organ perfusion in a porcine model of profound CS, the Impella CP unloaded the left ventricle compared to VA-ECMO.¹⁶ Also, in experimental data, Impella improved myocardial recovery after MI by unloading of the left ventricle and reduction of ventricular wall stress.¹⁷ A clinical study showed an early and significant improvement of the LV ejection fraction with Impella support in patients with refractory CS.¹⁸ ECMO offers circulatory support, but this is not the same as cardiac support. LV overloading associated with ECMO support may lead to delayed recovery or worse outcomes. To prevent this, LV venting techniques are frequently used in patients with ECMO support. In our cohort 55% of the ECMO patients received additional IABP.

We observed a trend in patients requiring RRT (42% vs. 26%, P=0.10). As RRT is not a definitive or accurate measure for renal failure, it remains unclear whether differential flow rates or Impella/ECMO treatment were associated with the development of renal failure. We believe further studies are necessary to address this point.

In our opinion, rather than considering these devices as competing technologies, it is important to have a clear understanding of each device in order to benefit from their unique qualities in different scenarios. Furthermore, we should keep in mind that even though MCS devices enable stable haemodynamics, none of the devices has a confirmed effect on the outcomes of patients with CS after MI. Keeping the goals of high-output support in mind (improvement of end-organ tissue perfusion and myocardial recovery), we believe that Impella may be considered the first-line option for support in CS with isolated LV failure after MI. Impella support involves fewer device-related complications and is considered less invasive than ECMO, which facilitates a quick initiation of support. However, ECMO may be the preferred short-term high-output device in patients with biventricular failure and/or oxygenation problems.

Limitations

As with all observational studies, and especially as we examined the use of two devices in different centres, our study is prone to bias. A major difficulty in this study is selection bias, and we should be careful when comparing outcomes between the groups such as mortality, even after correction for possible confounders. Also, it is impossible to correct for (unknown) confounders that were not measured. However, data from randomised trials on CS patients who require MCS after MI is sparse and consists of a highly selected patient cohort. Therefore, real-life data on the characteristics and outcomes of patients that received high-output MCS, and especially the occurrence of complications, are insightful and important. Also, we compared different centres that used either Impella or ECMO support and other treatment differences may also affect outcomes. As not all complications are captured well in patient records it is possible that we underreported certain complications. Our definition of clinically relevant haemolysis was somewhat limited, because this complication is difficult to ascertain in retrospect and free haemoglobin levels were not available in both centres. Changes in time regarding treatment strategy, patient selection and device experience could also have influenced our results. To assess this, we compared the outcomes of patients treated before 2014 to patients treated after 2014, and found more device-related complications in the latter group. This makes it less likely that the device-related complications in our cohort are mainly related to the learning curve associated with device experience.

Conclusions

Patients treated with Impella CP/5.0 or ECMO for CS after MI did not differ in 30-day mortality, in this observational, non-randomised study conducted in two experienced centres in The Netherlands. More device-related complications occurred in patients with ECMO support. Further studies should examine whether high-output MCS improves the clinical outcome of CS after MI, and demonstrate the optimal treatment algorithm.

Conflict of interest

The authors declare that there is no conflict of interest.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Supplementary material

Supplementary material for this article is available online.

References