Left Ventricle Unloading with Veno-Arterial Extracorporeal Membrane Oxygenation for Cardiogenic Shock. Systematic Review and Meta-Analysis.

During veno-arterial extracorporeal membrane oxygenation (VA-ECMO), the increase of left ventricular (LV) afterload can potentially increase the LV stress, exacerbate myocardial ischemia and delay recovery from cardiogenic shock (CS). Several strategies of LV unloading have been proposed. Systematic review and meta-analysis in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement included adult patients from studies published between January 2000 and March 2019. The search was conducted through numerous databases. Overall, from 62 papers, 7581 patients were included, among whom 3337 (44.0%) received LV unloading concomitant to VA-ECMO. Overall, in-hospital mortality was 58.9% (4466/7581). A concomitant strategy of LV unloading as compared to ECMO alone was associated with 12% lower mortality risk (RR 0.88; 95% CI 0.82-0.93; p < 0.0001; I2 = 40%) and 35% higher probability of weaning from ECMO (RR 1.35; 95% CI 1.21-1.51; p < 0.00001; I2 = 38%). In an analysis stratified by setting, the highest mortality risk benefit was observed in case of acute myocardial infarction: RR 0.75; 95%CI 0.68-0.83; p < 0.0001; I2 = 0%. There were no apparent differences between two techniques in terms of complications. In heterogeneous populations of critically ill adults in CS and supported with VA-ECMO, the adjunct of LV unloading is associated with lower early mortality and higher rate of weaning.

A well recognized limitation of the retrograde aortic flow while on VA-ECMO is the increase of left ventricular (LV) afterload [10], which can potentially lead to high LV stress and may exacerbate myocardial ischemia thus delaying recovery from cardiogenic shock [11]. Elevated LV pressure can also promote LV dilatation and trigger ventricular arrhythmias, or, secondarily, increase left atrial pressure causing pulmonary edema [12]. Ultimately, a reduced flow across the aortic valve can induce formation of thrombus in the LV or the aortic root [13].
Several LV unloading strategies have been described and proposed in order to minimize the risk of these complications [14], however, the available evidences are still conflicting whether these techniques are safe and useful adjuncts to VA-ECMO in patients with cardiogenic shock [15][16][17][18].
The aim of this study was to comprehensively assess the impact on early outcomes of different strategies of LV unloading in patients undergoing VA-ECMO and sustaining advanced cardiogenic shock by various etiologies.

Data Sources and Search Strategy
This systematic review and meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [19]. The PRISMA checklist is available as Appendix A Table A1. Research of relevant studies was limited to the period January 2000-March 2019, through PubMed, EMBASE, CINAHL, the Cochrane Register of Controlled Clinical Trials (CENTRAL) and Google Scholar. Abstracts were eligible for detailed assessment if available online and reporting outcomes of interest. The search terms were: "extracorporeal membrane oxygenation" and "extracorporeal life support". No language restrictions were imposed. References of original articles were reviewed manually and cross-checked for other relevant reports. Authors of individual studies were contacted for missing data.

Selection Criteria and Quality Assessment
Human studies were included if they assessed survival after VA-ECMO or weaning from VA-ECMO support instituted for refractory cardiogenic shock. Research centers were checked to avoid potential overlapping patients and those reporting on smaller samples of patients were excluded. Reviews and case reports were not considered. Two independent reviewers (M.K. and K.Z.) selected the studies for inclusion, extracted studies, as well as patient characteristics of interest and relevant outcomes. Two authors (M.K. and K.Z.) independently assessed the trials' eligibility and risk of bias. Risk of bias at the individual study level was assessed using the Risk of Bias in Not-randomized Studies-of Interventions (ROBINS-I) tool [20]. Any divergences were resolved by a third reviewer (G.R.) and quantified using the approach of Cohen's kappa [21].

Endpoint Selection
The primary endpoints were in-hospital/30-day survival and weaning from VA-ECMO. Secondary endpoints were in-hospital cerebrovascular events (CVE), brain death, limb complications, reoperation for bleeding, sepsis and acute kidney failure w/wo continuous veno-venous hemofiltration (CVVH). Outcome definitions were the ones adopted by the investigators of the included studies.

Statistical Analysis
Statistical analyses were performed in Comprehensive Meta-Analysis, v. 2.0 (Biostat, Englewood, NJ, USA) and Review Manager 5.3 (The Nordic Cochrane Centre, Copenhagen, Denmark). The results are expressed as pooled untransformed proportion risk ratios (RR) with their 95% confidence intervals (CI). Heterogeneity across studies was evaluated using the I 2 test. To control for the anticipated heterogeneity among observational studies, absolute values and means were pooled using inverse variance random effects models. The primary endpoints were assessed in relation to the specific setting according to etiology of cardiogenic shock which included: (1) postcardiotomy shock (PCS), (2) acute myocardial infarction (AMI), (3) myocarditis and (4) mixed cohort of different etiologies including both postcardiotomy shock and AMI and other etiologies. Number needed to treat (NNT) was calculated for these subgroups. Secondary analysis focused on specific left ventricular unloading strategy: (1) intra-aortic balloon pump (IABP), (2) LV venting with cannula in left atrium or ventricle or (3) percutaneous ventricular assist device (Impella, Abiomed, Danvers, MA, USA). We performed separate analysis of studies with propensity score matching or presenting propensity score adjusted odds ratio (OR) of primary endpoints. We investigated if use of different unloading strategies had influence on complication rates, ECMO duration and weaning rates by means of meta-regression analyses [22]. Similarly, we addressed the impact of hypertension, diabetes, age and gender on mortality outcome. Sensitivity analyses were performed by excluding from analyses single studies, one at a time, and repeating the calculations. Publication bias was assessed (1) by visual approach plotting log event rate against standard error in the funnel plot; and (2) by linear regression approach [23].

Study Selection
The study selection process and reasons for exclusion of some studies are described in Figure 1. A systematic search of the online databases allowed us to screen based on title and collect 271 potentially eligible records that were retrieved for scrutiny. Of those, 204 were further excluded because they were not pertinent to the design of the meta-analysis or did not meet the explicit inclusion criteria based on their content. To avoid potential double inclusion of patients' populations, 5 studies were excluded (Supplementary Material: Part 1) since they were conducted in the same institution in overlapping time frames. Sixty-three series of patients from 62 observational studies (Supplementary Material: Part 2) that enrolled 7581 patients eventually were included in the analysis. Patients were divided into 2 groups: those undergoing LV unloading concomitant to VA-ECMO and those undergoing VA-ECMO alone; (3337; 44.0%) vs. (4244; 56.0%). Patients undergoing VA-ECMO had a mean age of 57.8 years and 71.0% were male. Follow-up across the studies varied between 30-day and in-hospital survival. Appendix Table A2 details about  studies and Appendix Table A3 about patients' characteristics. Risk of bias for each study across each of the seven risk of bias domains is presented in Appendix Table A4. Overall, the studies reported either moderate or serious risk of bias. Given the overall high risk of bias along with the limited number of studies, all articles were retained for the purposes of this review. Most commonly, biases arose from (1) selection of participants for the study, and (2) subjective distribution of the participants within the study arms.
Populations included patients on VA-ECMO support for cardiogenic shock secondary to mixed etiologies (23 series, 4204 patients), PCS (22 series, 2324 patients) and AMI (14 series, 950 patients); VA-ECMO was employed for myocarditis in 4 series enrolling 103 patients. Patients undergoing VA-ECMO had a mean age of 57.8 years and 71.0% were male. Follow-up across the studies varied between 30-day and in-hospital survival. Table A2 details about studies and  Table A3 about patients' characteristics. Risk of bias for each study across each of the seven risk of bias domains is presented in Table A4. Overall, the studies reported either moderate or serious risk of bias. Given the overall high risk of bias along with the limited number of studies, all articles were retained for the purposes of this review. Most commonly, biases arose from (1) selection of participants for the study, and (2) subjective distribution of the participants within the study arms.

Sensitivity Analyses
Analyses were repeated as sensitivity for primary endpoints mortality and weaning from ECMO this time included only studies that reported effect estimates for propensity matched cohorts only: 5 studies (Supplementary Material: Part 3) provided propensity adjusted estimates of mortality; pooled together, LV unloading on top of ECMO was associated with over 25% statistically significant reduction in the odds of mortality as compared to ECMO alone: OR (95%CIs): 0.74 (0.60-0.91); p = 0.004; I 2 = 42%; Figure 5a.

Sensitivity Analyses
Analyses were repeated as sensitivity for primary endpoints mortality and weaning from ECMO this time included only studies that reported effect estimates for propensity matched cohorts only: 5 studies (Supplementary Material: Part 3) provided propensity adjusted estimates of mortality; pooled together, LV unloading on top of ECMO was associated with over 25% statistically significant reduction in the odds of mortality as compared to ECMO alone: OR (95%CIs): 0.74 (0.60-0.91); p = 0.004; I 2 = 42%; Figure 5a.
Weaning rates for comparison LV unloading + ECMO and ECMO alone adjusted for propensity were reported in 4 studies (Supplementary Material: Part 4); again, LV unloading on top of ECMO was associated with over 75% significantly higher odds to wean from ECMO: OR (95%CIs): 1.78 (1.40-2.28); p < 0.001; I 2 = 0%; Figure 5b. Sensitivity analyses performed by deleting each study, one at a time, and repeating the calculations did not change the direction nor magnitude of the treatment effect, suggesting absence of big-study effect.
Sensitivity analyses performed by deleting each study, one at a time, and repeating the calculations did not change the direction nor magnitude of the treatment effect, suggesting absence of big-study effect.

Discussion
VA-ECMO is an established treatment able to provide a mechanical circulatory support for patients in cardiogenic shock, aiming a bridge to decision or to myocardial recovery [1][2][3][4][5][6][7][8][9]. Improvements in technology have mitigated the interaction between artificial surfaces of ECMO circuits and blood [24]. However, other adverse effects, known as "flow-related dynamic", are strictly associated, both in central and peripheral ECMO configuration, with the retrograde direction of the flow towards a dysfunctioning left ventricle. Two major issues have been longer debated by the scientific community: the first is the difference in outcomes and hemodynamic support between the central and peripheral cannulation; the second is the clinical impact of the left ventricle unloading and the strategy to achieve a safe and effective ventricular decompression. The first issue has been already addressed by our group [25]; aim of the current meta-analysis is to address the question whether myocardial unloading is beneficial or, by raising the complexity of ECMO management, futile or potentially detrimental to patients' outcomes.
ECLS institution increases the left ventricle afterload with a rise in LV end-systolic volume and reduction in LV stroke volume. If peripheral resistance and LV contractility are fixed, increase in LV end-diastolic volume is the only way to overcome the afterload via the Frank-Starling mechanism. In this case, higher levels of VA-ECMO flow cause a progressive rise in LV end-diastolic pressure, LA pressure, pulmonary capillary wedge pressure, that are associated with a further reduced LV stroke volume [26][27][28]. High afterload situations with inability of LV to manage the transpulmonary blood flow, inadequate response to inotropes, complete cardiac arrest with incomplete venous drainage and aortic valve incompetence are the commonest risk factors for LV distension. Patients with severely impaired LV function and/or right ventricular dysfunction are more prone to develop an ineffective LV unloading [29]. LV overload increases wall stress, myocardial oxygen consumption and induce sub-endocardial ischemia and ventricular arrhythmias, jeopardizing ventricular recovery particularly in the presence of ischemia-induced myocardial impairment. The consequence of the pressure overload may ultimately account for pulmonary congestion and edema.
If the overload is extreme and LV contractile impairment significant, the LV is unable to provide a sufficient flow against the increased afterload and the aortic valve may remain closed even during systole, causing blood stasis in the left ventricle, left atrium and aorta, and accounting for intracardiac thrombosis which has been reported in up to 6% of the cases [30,31]. The LV dilatation may further induce annular dilatation and mitral valve leaflet tethering with severe functional regurgitation, thus, particularly in in patients with a history of chronic heart failure and LV dysfunction with a dilated LV, worsening the pulmonary congestion [32].
Definition of LV distension during VA-ECMO is lacking in the literature. Truby et al. [33] attempt to classify and grade the LVD according to the evidence of pulmonary edema on chest radiography and increased pulmonary artery diastolic blood pressure (>25 mmHg). The latter was a surrogate of the wedge pressure evaluated in the "Should we emergently revascularize occluded coronaries for cardiogenic shock" (SHOCK) trial [34]. Clinical evidence of LV distension requiring immediate decompression was inversely related to the chance of myocardial recovery. Meani et al. [32] defined and graded the severity of LV loading during VA-ECMO according to hemodynamic parameters, chest X-ray and echocardiogram findings.
These differences in definitions and assessments may account for the high variability of LV distension rate in the literature. Camboni et al. [35] reported need for LV decompression in 2% of the cases in more than 600 patients. A strict and longer afterload reduction (> 24 hours), targeted lower ECMO flow and a restrictive fluid management were the strategy adopted in this large series.
In Truby et al. [33] the clinical and subclinical (not warranting immediate decompression) LV distension occurred in 7% and 22% of patients, respectively. Among 184 peripheral VA ECMO in the series of Meani et al. [36], 5.4% required IABP placement because of a protracted closure of the aortic valve.
Drugs administration is the first line treatment of left ventricle distension. Inotropes can be administered to increase LV contractility while vasodilators may reduce the peripheral resistances and decrease left ventricle afterload. A careful fluid balance (diuretics/fluid restriction) avoiding fluid overload can reduce the risk of pulmonary edema. Ventilatory optimization, including higher PEEP, prolonged expiration time and lower tidal volume, may further improve the venous drainage.
When medical treatment is not successful, the non-pharmacological management of LV distension, acting with a "direct" or indirect" mechanism, can be obtained through a surgical or percutaneous strategy ( Figure 6).

IABP
IABP has been the most used technique to unload the left ventricle during ECMO support [37]. The IABP acts with several "indirect" mechanisms reducing both the LV afterload (enhanced systolic ejection) and the LV end-diastolic pressure (enhanced left atrial and pulmonary venous unloading). The IABP induces the aortic valve opening [36], improves coronary and abdominal circulation [38], allows pulsatility in end organ capillary bed [39], it is easy to implant and has contained costs. In animal studies the role of counterpulsation in VA-ECMO support seems controversial. Zobel [40] and Sauren [41] showed that IABP has beneficial effects on LV performance. Instead, Belohlávek et al. [42] showed that the combination of femoral VA-ECMO and IABP could impair coronary perfusion. In clinical practice the combination ECMO/IABP was associated with improvement in hemodynamics parameters [43,44], weaning rate [43,45] and survival [45,46].

ECPELLA
The use of Impella in combination with VA-ECMO (also known as ECPELLA/ECMELLA) has been shown to provide improved weaning and survival rates compared to ECMO alone strategy and to established risks scores [47][48][49][50]. The addition of a continuous flow vent reduces LV volumes and pressures. The LV stroke volume progressively decreases as pump flow increases, with the raise of systemic blood pressure and reduction of LA and pulmonary capillary wedge pressures. Despite the aortic valve does not open, there is no risk of blood stasis in the LV and the aortic root. The uncoupling of LV and aortic pressure is a sign of an effective unloading of the ventricle. In this situation a flat systemic pressure line is a sign of maximal unloading. Secondary changes in myocardial contractility and peripheral resistance may further enhance the LV unloading [26,50]. The Impella can also reduce RV afterload and facilitate RV output and pulmonary blood flow with improvement in gas exchange [51,52]. Alongside these hemodynamic features, the use of an axial flow pump may provide a circulatory support while weaning from VA-ECMO. The possibility of reducing the duration of ECLS has been reported by Scharge et al. [50], however, in the experience of Pappalardo et al. [48], the association of Impella and VA-ECMO prolonged the time of support but provided a successful recovery of patients who might not have survived under VA-ECMO treatment alone. The use of Impella has been associated with a significant risk of severe bleeding, vascular complications and cerebral stroke [53,54]. In patients receiving the dual treatment with VA-ECMO, a higher occurrence of hemolysis has been reported [48], however, no difference was generally found in terms of risk of major and minor bleeding, and cerebral stroke compared to VA-ECMO alone [47,49]. These initial results seem to support an expanding use of Impella for LV unloading. Despite the evidences are still limited and coming from retrospective studies, most of the patients who underwent ECPELLA therapy were in cardiogenic shock with severely impaired LV function, were upgraded to VA-ECMO while on axial flow pump due to a progressive deterioration, or needed the implantation of Impella following significant and complicated LV distension.

Other Techniques
Other unloading strategies have been reported in the literature and address the endpoints of this meta-analysis (Table A5). Briefly, the left atrium can be drained surgically by a cannula in the left atrial roof or in the right superior pulmonary vein or percutaneously [32,55] by an interatrial septostomy (septostomy usually with ballooning or stent) or a cannula attached to the ECMO venous return or to device like TamdemHeart ® ). Direct left ventricle unloading can be also achieved or by a surgical cannulation of the ventricle apex [56,57] and through the mitral valve from the left atrium [56,58] or percutaneously by a catheter across the aortic valve. The surgical or percutaneous pulmonary artery cannulation [56,57], increasing the right-side blood drainage, will indirectly reduce the pulmonary venous return and left cardiac chamber loading. The experiences with these last unloading strategies include small populations, however, these studies found a positive impact of these adjuncts on patients' survival.
Hemodynamic responses to ECMO are different among patients and are affected by clinical presentation, associated comorbidities and the cardiovascular system coupling. This high variability may explain the difficulties in driving robust conclusions in terms of efficacy and safety of LV unloading during VA-ECMO.
Up to date and to the best of our knowledge other two meta-analysis have been published on LV unloading strategy [30,37]. In 2015, Cheng et al. [30] reported the impact of IABP on survival among 1517 patients (16 studies

Limitations
As analysis of only non-randomized studies, our analysis shared similar limitations with these reports which included experiences with small populations and lacked some critical information about the timing of ECMO institution, the timing of LV unloading adjunct, or the weaning protocols. Most importantly, none of the studies report exact criteria for therapy escalation e.g., addition of IABP or Impella device to ECMO. In addition, observational nature of these studies promotes selection bias. However, compared to previous meta-analyses, that present certain methodological flaws (e.g., Russo by applying the very same search strategy included 17 studies and 3997 patients), the current study, including 62 studies and more than 7500 patients, represents the first comprehensive approach addressing LV unloading strategies during ECMO support.
We found that, regardless the strategy (IABP, Impella, others) and the etiology (PCS [59][60][61], AMI, other), LV unloading has a positive impact in patients' weaning, without adding any further risk of CVE, sepsis, acute renal injury requiring dialysis, limb complications and reoperation for bleeding. We have also provided a separate analysis of propensity-score matched and adjusted studies, trying, in the absence of prospective randomized data, to address the high heterogeneity of the included experiences due to different baseline populations' characteristics. This further analysis confirmed these findings favoring LV unloading techniques during VA-ECMO.
Despite the expected different flow patterns and afterload increase by central and peripheral cannulation, these two strategies were not significantly associated with a higher odds ratio risk of mortality considering the adjunct or the absence of LV unloading. However, we found a tendency in the association of higher odd ratio risk and progressively higher percentage of patients receiving peripheral cannulation, this finding couples the non-significant difference in outcomes in the PCS populations that have received a central VA-ECMO in almost 30% of the cases (less than 10% in the mixed populations, 0% in AMI patients), and suggests, within the limitations of this analysis, a more pronounced positive impact of LV unloading in the peripheral VA-ECMO setting.
The analysis of weaning, additionally included as a sensitivity analysis, might give presumptive underlying evidence of true reasons for improved survival after VA-ECMO support. The possibility of providing an adequate oxygen delivery associated with the reduction of myocardial injury and the relief of pulmonary congestion, thus enhancing arterial oxygenation and reducing pulmonary complications, may explain the higher rate of survival in patients who received an adjunct treatment able to prevent or solve left ventricular distension during VA-ECMO support.

Conclusions
During veno-arterial extracorporeal membrane oxygenation, the increase of left ventricular afterload can negatively impact the recovery from cardiogenic shock. In this meta-analysis including 7581 patients on VA-ECMO support, the adjunct of left ventricular unloading was associated with 35% higher probability of weaning and 12% lower risk of mortality.   Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, outcomes and study design (PICOS). 3

Protocol and registration 5
Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number.

Eligibility criteria 6
Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.

3
Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched. 3 Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated. 3 Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis). 3 Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators. 4

Data items 11
List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.

3-4 Risk of bias in individual studies 12
Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.

Summary of evidence 24
Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users and policy makers).

11-13
Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias). 12

Conclusions 26
Provide a general interpretation of the results in the context of other evidence, and implications for future research.

11-13
Funding Funding 27 Describe sources of funding for the systematic review and other support (e.g., supply of data); role of funders for the systematic review. 13                                        Figure A12. Mortality by device.