Meta-analysis to Assess the Impact of Centre’s Heart Transplant Status and Volume on in-hospital Outcomes Following Extracorporeal Membrane Oxygenation for Refractory Post-cardiotomy Cardiogenic Shock

Background: Postcardiotomy cardiogenic shock (PCS) that is refractory to inotropic support remains a major concern in cardiac surgery and is almost universally fatal unless treated with mechanical support. While reported mortality rates on ECMO vary from center to center, aim of the current report is assess if the outcomes differ between centres according to volume and heart transplantation status. Methods: A systematic search was performed according to PRISMA statement using PubMed/Medline databases between 2010 and 2018. Relevant articles were scrutinized and included in the meta-analysis only if reporting in-hospital/30-day mortality and heart transplantation status of the centre. Differences were assessed by means of subgroup meta-analysis and meta-regression. Results: Fifty-four studies enrolling N=4,421 ECMO patients were included. Of those, 6 series were performed in non-HTx centres (204 pts;4.6%). Overall 30-day survival (95% Confidence Intervals) was 35.3% (32.5-38.2%) and did not statistically differ between non-HTx: 33.3% (26.8-40.4%) and HTx centres: 35.7% (32.7-38.8%); Pinteraction=0.531. There was no impact of centre volume on survival as well: ßcoef=0.0006; P=0.833. No statistical differences were seen between HTx and non-HTx with respect to ECMO duration, limb complications, reoperations for bleeding, kidney injury and sepsis. There were however significantly less neurological complications in the HTx as compared to non-HTx centres: 11.9% vs 19.5% respectively; P=0.009; an inverse relationship was seen for neurologic complications in centres performing more ECMOs annually ßcoef=-0.0066; P=0.031. Weaning rates and bridging to HTx and/or VADs were higher in HTx facilities. Conclusions: There was no apparent difference in survival after ECMO implantation for refractory PCS according to centre’s ECMO volume and transplantation status. Potentially different risk profiles of patients in these centres must be

4 definite conclusions are drawn.
Despite growing worldwide utilization and experience in mechanical circulatory support (MCS), in particular, in-hospital outcomes while on ECMO have not shown substantial progress [6]; little is still known of who benefits most from ECMO not to mention cost effectiveness of such long and advanced therapy [7]. European Society of Cardiology guidelines therefore cautiously assigned ECMO class of recommendation IIb, level of evidence C for the management of cardiogenic shock in STEMI [8].
Unlike STEMIs, cardiac surgical patients are usually characterized by substantial pre-ECMO comorbidities and more advanced age [9]. All these factors, individually or in association, may inhibit the potential of myocardium to recover after the surgery and/or hamper favorable body response to prolonged MCS. Indeed, in some patients, prolonged MCS does not lead to improved cardiac function or organ integrity; clinicians are therefore forced to bridge the patient; since bridge to recovery is no longer an option, more advanced treatments, such as heart transplantation (HTx) or long-lasting ventricular assist devices (VADs) remain. Not all heart surgery centres perform HTx, and not all of them perform VADs.
We therefore, undertook current systematic review and meta-analysis to assess to which extent do the in-hospital outcomes differ across PCS-ECMO recipients in heart transplantation-as compared to non-transplant units to account for readability of ECMO teams and potentially shorter bridging times in these facilities. Additionally, we investigated on how are the in-hospital outcomes affected by centre's volume and annual ECMO institution rates to account for differences in between centres experience.

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) statement [10]. The PRISMA checklist is available as Appendix Selection criteria and quality assessment Studies were included if they met all of the following criteria: 1) human study; 2) studies assessing survival after ECMO instituted for postcardiotomy refractory cardiogenic shock; 3) study reporting institutional outcomes that for ECMO indication combined postcardiotomy and non-postcardiotomy cardiogenic shock, but postcardiotomy comprised majority of cases; 4) studies combining non-postcardiotomy and postcardiotomy cardiogenic shock patients but reporting outcomes of interest separately for the latter.
Studies were only excluded if: 1) paediatric and congenital heart surgery-related studies; 2) animal studies; 3) conducted in the setting of veno-venous ECMO for respiratory distress following heart surgery; and 4) studies not reporting survival/mortality rates.
Studies were only eligible if reporting the transplant status of the centre; whenever this was not retrievable from the individual study, institutional website was searched for information regarding range of procedures performed. Lack of clear indication whether the centre performs heart transplantation led to exclusion of the study. Similarly, registries incorporating multiple centres but not reporting the status for single facilities were not considered. Reviews and case reports were not considered as well.
Two independent reviewers (P.M. and K. Z.) selected the studies for inclusion, extracted studies, as well as patient characteristics of interest and relevant outcomes. Two authors (P.M. and K. Z.) independently assessed the trials' eligibility and risk of bias. Risk of bias at the individual study level was assessed using the ROBINS-I tool (Risk of Bias in Notrandomized Studies-of Interventions) [11]. Any divergences were resolved by a third reviewer (R. L.) and quantified using the approach of Cohen's kappa [12].

Endpoint selection
The primary endpoint was in-hospital survival. Secondary endpoints were in-hospital cerebrovascular events (CVE), limb complications, bleeding or reoperation for bleeding, sepsis and acute kidney failure w/wo continuous veno-venous hemofiltration (CVVH).
Bridging to VAD and/or HTx was analysed as well. 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 (Biostat, Englewood, NJ). The results are expressed as pooled untransformed proportions (eg. event rates (%) and means with their 95% confidence intervals (CI). Heterogeneity across studies was evaluated using the I 2 test. Where available, we digitised Kaplan-Meier curves using Engauge Digitizer 9.5 (Mark Mitchell, Torrance, CA) and reconstructed time-to-event data using the algorithm specified by Guyot et al. [13]. To control for the anticipated heterogeneity among observational studies, absolute values and means were pooled using random effects models. Studies were stratified a priori based on the centre status (HTx vs non-HTx performing centre); the interaction coefficient (Q-value) is provided for the comparison HTx vs non-HTx along with respective P interaction . Additionally, we investigated if HTx and non-HTx status had influence on ECMO duration, weaning rates, bridging to HTx/VAD rates; and further if ECMO duration and weaning rates in these centres correlated with bridging to HTx/VAD by means of meta-regression analyses [14]. Similarly meta-regression approach was used to determine whether annual ECMO institution rate for centre reporting such, affects the survival and remaining in-hospital outcomes. Annual ECMO institution rate was calculated by dividing n. of study subjects by study duration period. Sensitivity analyses were performed by excluding from analyses single studies, one at a time, and repeating the calculations. Subgroup analyses were performed for survival endpoint by dividing the studies into distinctive strata (by mean and median annual ECMO institution rate as well as in tertiles and quartiles) and reporting respective P interaction for between subgroup comparison. 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 [15].

Results
Initial search process yielded 22,609 records; of these, 183 abstracts were retrieved for scrutiny based on the item's title. Registries were excluded since they incorporated both HTx and non-HTx centres [16][17][18]. Following detailed assessment, 54 studies (N = 4,421 patients) [list of references to included studies] met inclusion criteria and entered quantitative analyses. PRISMA flow chart is available as Figure 1. Included studies were divided into HTx vs non-HTx centres subgroups: 48 studies including 4,217 (95.4%) patients were conducted in HTx-whereas 6 studies (N = 204) in non-HTx centres.
Prevalence of ECMO ranged from 0.26% [70] to 3.35% [28]. Patients receiving ECMO at HTx centres were significantly younger than their non-HTx counterparts 57.2±1.6 vs 64.2±1.6 P<0.001. CABG was most frequent procedure in both HTx and non-HTx centres 33.7% and 30.9% followed by valvular (25.1% and 21.1%) and combined surgeries (16.5% and 26.5%). Detailed characteristics of included studies as well as patients' baseline and surgical data are available in Table 1. Publication bias analysis along with reasons for bias risk increase is available as Appendix Table 2; studies were judged to me moderate to severe risk of bias as none previously compared directly HTx vs non-HTx centre performance; no signs of asymmetry were seen on visual inspection of funnel plot for primary endpoint (Appendix Figure 1).

ECMO strategy
In the studies that reported procedural details, ECMO was established during the initial cardiac surgery in 42.7% of cases because of circulatory instability during or immediately after weaning from cardiopulmonary bypass. ECMO was initiated in the OR in 56.5% of patients (50.1-62.7%), followed by ICU, cardiac catheterization laboratory, telemetry floor and emergency department. A statistical trend was observed for more common placement of ECMO in the OR than other locations in non-HTx -as compared to HTx centres: 64.5% (52.9-74.6%) vs 53.2% (45.6-60.7%); P = 0.108. Peripheral cannulation was preferred approach (69.0%) for ECMO institution. Median ECMO duration in the entire series was 5 (IQR: 3.3-6.0 days); without apparent differences between HTx (mean weighted average = 4.92 days) vs non-HTx-(5.04 days) centres. The details of procedural characteristics are available as Appendix Table 3. Successful weaning from ECMO was most often defined as decannulation after >48 hours. Overall, estimated 55.3% patients were weaned from ECMO with the weaning rates ranging from 31.4-100% in the entire series. There was a signal for higher weaning rates across HTx vs non-HTx centres (56.6% vs 50.4%; P = 0.118).

Discussion
To the best of our knowledge, the current meta-analysis represents the first attempt to address the differences in in-hospital outcomes of patients supported with VA-ECMO for refractory PCS between HTx and non-HTx centres. This research was aimed to investigate further factors other than the well known patients' clinical status and procedure type that may affect the final outcome in PCS-ECMO patient. The care center with experience in dealing with acute and chronic end stage heart failure with expertise and prompt resources availability (medium and long term mechanical circulatory support and heart transplantation) as factor potentially affecting this outcome was the primary hypothesis of our study.
VA-ECMO is increasingly used for cardiorespiratory support in patients affected by refractory cardiogenic shock or cardiac arrest after cardiac surgery. [2] Despite that growing worldwide utilization and experience, ECMO in-hospital outcomes have not shown substantial progress. Conversely, a trend towards worse survival rates, reaching a disappointing 15% has been recently reported in another analysis of the Extracorporeal Life Support Organization registry [2]. Patients undergoing heart surgery usually present with substantial pre-ECMO comorbidities, more advanced age and above all different stages of developed heart failure. All these factors, individually or in association influence the capability of the myocardium to recover after the surgery and thus preclude favorable body response to prolonged MCS. Unfortunately, in a considerable proportion of patients, the MCS regardless of its duration, does not prompt to improved cardiac function or organ integrity; in turn, clinicians are forced to bridging the patient to more advanced treatments, such as HTx or VADs. The insights from important recently available study by Distelmeier [20] are that prolonging of VA-ECMO duration is associated with a disproportionate mortality at early and later stages. In fact, lack of cardiac function improvement within 7 days post-op. was indicative of futile support in the analysis.
Consequently, this leads to conclusion that perhaps HTx or VADs in such ECMO-supported patients should be used much sooner, just in time to prevent life-threatening complications Such hypothesis led to conception of the current study which is the first to compare, although in indirect fashion, the outcomes between HTx and non-HTx performing centres in patients undergoing ECMO treatment for refractory PCS. The first consideration come from the study population of this meta-analysis: the majority of the patients (4,217 over 4,421) and number of reports (48 over 54) come from HTx centers suggesting that 1) cardiac surgery population may show similar or higher risk profiles in HTx centres thus, an increased baseline risk for developing PCS and 2) ECMO represents a tool routinely used for the treatment of refractory PCS in the HTx units; 3) there exists an unexplained underreporting from non-HTx centres with regard to the perioperative outcomes and in particular of patients undergoing ECMO treatment. Regardless, our main findings were that among patients with PCS no differences in 30 day/in-hospital mortality were observed between heart transplantation centres as compared to non-transplant units. This was also confirmed in a subgroup analysis. While neutral, this finding implies similar mortality rates among patients operated on in HTx and long-term assistance facilities as compared to lower risk cardiac surgery patients operated in non-transplant units, given their respective potentially higher and lower baseline risk.
Second, there was no difference between the centers type with respect to limb complications, reoperation for bleeding, sepsis and acute renal injury with or without dialysis, yet neurological complications occurred less frequently in HTx centres.
Neurologic complications are presumably a multifactorial entity with pre-ECMO illness severity and treatments, ECMO management, and post-ECMO events all contributing to CNS injury rates in these patients. Loss of cerebral autoregulation during severe arterial hypertension or hypotension, thromboembolic events, haemorrhage related to anticoagulation use, cerebral vasospasm, and secondary brain injury from reactive tissue oedema around an area of focal CNS injury have all been implicated in the genesis of brain injury in VA-ECMO patients. Although neurologic injury during VA-ECMO remains poorly defined in adult cohorts [16,[21][22][23], prior investigations comprehensively report neurologic complications occurrence in 6-17% in adults supported with VA-ECMO for postcardiotomy cardiorespiratory failure [21,24,25]. What seems even more illustrative, postmortem examination in adults supported with VA-ECMO has shown that neurologic injury may be clinically undetected in 23-50% of cases [26,27]. In the current analysis, we saw neurologic complications more frequent in non-HTx centres. While this could not be accounted for in that type of analyses, the "over-delay" to ECMO commencement in the institutions with lower experience with circulatory support systems (be that ECMO or VADs) may have played a role in the excess of strokes in this population.
Finally, patients on ECMO bridged to HTx or mechanical circulatory support are reported only in the HTx units. Single patients in non-HTx institutions died while waiting for referral to HTx hub after the decision to transplant was met. From the technical standpoint, is noteworthy to underline that the were no differences in the ECMO duration between the two centers, yet notice must me made of statistical trend for the different ECMO location placement and weaning rate; that is, ECMO was instituted in the OR more frequently in non-HTx centres but again since delay to ECMO was seldom reported, we cannot address the issue whether this might had affected patient outcome. On the other hand, this may further suggest an easier applicability of ECMO in locations other than OR in HTx-centres with possibly prompt ECMO team availability as compared to the non-HTx units.      Analysis of reoperations for bleeding following ECMO institution in HTx/VAD vs non-HTx/VAD centres. Remaining abbreviations as in Figure 1. Analysis of sepsis following ECMO institution in HTx/VAD vs non-HTx/VAD centres.
Remaining abbreviations as in Figure 1. Analysis of acute kidney injury following ECMO institution in HTx/VAD vs non-HTx/VAD centres. Remaining abbreviations as in Figure 1.

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.