Timely extracorporeal membrane oxygenation assist reduces mortality after bypass surgery in patients with acute myocardial infarction

Patients with acute myocardial infarction (AMI) are at high risk when undergoing emergency coronary artery bypass graft (CABG)‐surgery. Their outcome remains poor despite increased use of extracorporeal membrane oxygenation (ECMO). We investigated the impact of timing for perioperative ECMO‐support in these patients.

Lactate-increase greater than 4 mmol/L seems to be a helpful threshold to trigger the timely onset of ECMO-therapy, providing better survival.

K E Y W O R D S
acute myocardial infarction, cardiogenic shock, circulatory support, coronary artery bypass graft, extracorporeal membrane oxygenation 1 | INTRODUCTION Emergent revascularization is necessary to preserve myocardial function in patients with acute myocardial infarction (AMI). Especially ST-elevation myocardial infarction (STEMI) often causes cardiogenic shock (CS) with hemodynamic instability and low cardiac output syndrome (LCOS). 1 This leads to multiple organ dysfunction and is associated with 30-day mortality-rate of more than 40% irrespectively of the procedure used for revascularization-percutaneous intervention (PCI) or coronary artery bypass graft (CABG)surgery. 2,3 The concept of early revascularization led to improved long-term outcome in patients with CS after AMI 4 and established primary PCI as the dominant therapy option during the past decade in Europe. 1 In patients wtih STEMI, PCI is currently recommended for treatment of culprit lesion only. 1,5 The treatment of remaining lesions could either be accomplished using PCI or CABG depending on the extent of the disease and complexity of stenosis at a later time-point.
Still, CABG can provide good results in patients with STEMI without CS, 1,6 especially since CABG results in a more complete revascularisation than PCI. 1 Even more, emergency CABG may be necessary in patients with AMI, which are anatomically unsuitable for PCI or with mechanical complications during PCI. 1 Mortality of either PCI or CABG in this situation remains high (>40%), 7 but STEMI-patients with CS, initially treated with PCI, showed a better survival when subjected to subsequent CABG instead of PCI. 3,7,8 Emergency-CABG is associated with a significant operative risk in patients with AMI because of the risk of developing CS. 9 In these patients, mechanical circulatory support (MSC) is an option to support the left ventricle (LV) to preserve cardiac function and provide a better perfusion of organs particularly vulnerable to hypoxia, eg, liver, kidneys, or brain. The "IABP-SHOCK-II-Trial" showed no significant short-term 10 or long-term survival benefit of intra-aortic balloon counter pulsation (IABP) in AMI-patients with CS, as recently shown by the long-term 6-year outcome data. 11 Still the use of other MSC-systems, eg, Impella or extracorporeal membrane oxygenation (ECMO), emerged during the last decade. An investigation from the Impella-EUROSHOCK-registry evaluated 120 AMIpatients with CS and showed reduced lactate-concentration suggesting a better organ-perfusion, but no effect on 30-day-mortality of greater than 60%. 12 Also, TandemHeart did not reduce 30-daymortality when compared with IABP in small randomized controlled trials (RCTs). 13 On the other hand, primary PCI has been performed with veno-arterial (VA)-ECMO-support in hemodynamically preserved AMI-patients with beneficial outcome and lower 30-daymortality. 14,15 A meta-analysis described a survival benefit for patients with CS during cardiac surgery if treated with VA-ECMOassist preoperatively. 16 The few studies examining VA-ECMO therapy in patients with AMI subjected to CABG were either based on a low patient number (<10), 17 or excluded patients with preoperative ECMO therapy. 18 Therefore, these studies were unable to investigate optimal timing of ECMO-start. At the same time, several studies investigated ECMO-support before or during emergency PCI, and showed that only early onset of ECMO therapy leads to significantly better clinical outcome. In support of this, experimental data from a porcine model with delayed reperfusion suggested that early "unloading" of the LV, using an intracorporeal axial flow-catheter, is associated with myocardial protection. 19 Hence, it seems plausible that preoperative or intraoperative relief of the LV with ECMO could be a potential new strategy for patients with AMI subjected to emergency-CABG. Still, one must bear in mind that ECMO does not lead to unloading of the LV, but rather supports the circulation and can even cause LV-overload. 20 We hypothesized, that preoperative or intraoperative LV-support with VA-ECMO is associated with better survival in patients with AMI subjected to emergency-CABG. The aim of this study was to compare outcome, complications, and predictive value of risk-scores after preoperative, intraoperative, and postoperative ECMO-assist.

| Patients
In this retrospective double-center study, we investigated 201 consecutive patients with AMI who underwent emergency isolated CABG-surgery between January 2008 and December 2017. Indications for CABG were left main or three-vessel disease, unsuitable anatomy for PCI, unsuccessful PCI, or angiographic accident. All cases were emergency-admissions from the heart catheterization laboratory of our university hospitals or were assigned from the catheter labs of surrounding smaller hospitals without cardiothoracic surgery departments. The heart-team decided whether to perform CABG immediately or at a later stage. This decision was based on the patients' symptoms, progress in heart enzymes-levels, and signs of inotropesand vasopressors-refractory CS. The study was in accordance with the declaration of Helsinki. The ethical review-board of our institutions waived the need for patient-consent in this retrospective study.

| Indication for ECMO-implantation
The indication for ECMO therapy was supported by perioperative diagnostic tools: transoesophageal echocardiography (TEE), pulse contour cardiac output or pulmonary artery catheter-analysis.
Regarding these, CS was defined as a state of LCOS with systolic arterial hypotension (<90 mmHg) and/or low cardiac index (<2.2 L/min/m 2 ). Other indicators of LCOS were signs of organ hypoperfusion (oliguria <0.5 mL/kg/h), signs of anabolic metabolism with lactate greater than 4.0 mmol/L despite high dosage of vasopressors and inotropic agents (norepinephrine >0.1 µg/kg/min, any dobutamine or epinephrine dose) as well as a shock-index (systolic blood pressure/heart rate) less than 1.

| ECMO-management and weaning criteria
In all six pre-ECMO patients, cannulation was performed via femoral vessels. In ECC-ECMO patients the system was connected intraoperatively to the central cannulas in the ascending aorta and right atrium during full heparinisation, and after protamine infusion (50% of the initial heparin-dose) activating clotting time was kept between 150 and 200 seconds. In these patients, the chest remained open during ECMO-support, except for two patients, where the thorax was closed after cannulating the right subclavian artery and femoral vein. In post-ECMO patients, ECMO was installed on ICU using peripheral cannulation of the femoral vessels except in two cases with arterial cannulation via the right subclavian artery. In all patients who received peripheral arterial ECMO cannulation via the superficial femoral artery, a reperfusion cannula was inserted distally into the superficial femoral artery to provide lower limb perfusion. After bleeding cessation on the ICU, heparin was given targeting a partial thromboplastin time of 50 to 60 seconds. Cardiocirculatory recovery was evaluated by stepwise reduction of ECMO flow (0.5-1.0 L/min every 24 hours) and assessment of myocardial function in TEE, as well as metabolic and organ-perfusion related parameters (SvO 2 , lactate, diuresis, creatinine, bilirubin). ECMO-weaning was started when a stable cardiopulmonary condition was reached under only a moderate inotropic support (adrenaline <0.2 µg/kg/min, dobutamine <3 µg/kg/ min, and/or milrinone <0.4 µg/kg/min) and a FiO 2 less than 0.7 at a maximum ECMO flow of 1 to 2 L/min. In case of inadequate recovery, full ECMO flow was reinstalled for another 24 hours before the next weaning attempt. Successful weaning from ECMO was defined as survival for greater than 48 hours after ECMO-explantation. expressed as mean ± SEM. If two groups were compared, the unpaired t test was performed. If data from more than two groups were compared over time, two-way analysis of variance (ANOVA) with Bonferroni multiple comparison-statistics was applied when appropriate, and if significant, changes between the groups at each time-point were evaluated using ANOVA with the Tukey post-hoc analysis. P < .05 was considered statistically significant.

| Patient characteristics, preoperative riskscores, and perioperative results
Evaluation of preoperative data showed STEMI in all pre-ECMO, but only in 74.1% of post-ECMO patients ( Table 1 and Table S1).
Cardiopulmonary resuscitation (CPR) was most often necessary in pre-ECMO patients, and CABG was preceded by PCI within 2 days before surgery most often in this group. Longest door-to-operating room (OR)-time and longest time from PCI-to-OR were seen in pre-ECMO patients. Fastest PCI-to-OR-time was observed in ECC-ECMO patients. HAMIKO ET AL.

| 1245
At admission, stdEuroSCORE, logEuroSCORE, APS-, and APACHE-II-scores were significantly higher in pre-and ECC-ECMO when compared with No-ECMO patients. Also, shock-index was ≥1 in these groups.
Surgery duration was significantly longer in post-ECMO compared with No-ECMO group Table 2. Post-ECMO group showed the longest aortic cross clamping time. Longest CPB-time was observed in ECC-ECMO patients. No difference was observed concerning the number of peripheral bypass anastomoses. A significantly higher amount of intraoperative blood product transfusions was needed in all ECMO groups when compared with the No-ECMO group, while no difference was observed between the ECMO groups (Table S2).
Post-ECMO patients showed lowest 30-day-survival, while earlier ECMO-start was associated with higher survival rates ( Figure 1A). Causes for mortality are displayed in Figure 1B.
We observed a significantly higher lactate concentration in pre-ECMO and ECC-ECMO groups on admission and preoperatively when compared with No-ECMO group ( Figure 1C). Intraoperative lactate levels increased significantly in the ECC-ECMO group, while  (Table S1).
Ejection fraction (EF) was significantly reduced on admission in pre- No-ECMO-group and stayed reduced until 2 days after. Both values were more stable, and systolic pressure even increased before surgery in the pre-ECMO group, most likely due to early preoperative ECMO-assist ( Figure 2B and 2C). Pressure-decrease was accompanied by deteriorated EF (Figure 2A) and associated with lactate-increase in all ECMO groups perioperatively ( Figure 1C).
Maximum lactate of greater than 8 mmol/L was observed 1 hour after CABG in pre-and ECC-ECMO group, but also after 6 hours in post-ECMO group.
Perioperative catecholamine-therapy showed significantly higher norepinephrine-doses in the ECC-ECMO and post-ECMO group, while the pre-ECMO group showed no increase in vasopressor-need compared with No-ECMO patients ( Figure   1SA). Pre-ECMO and ECC-ECMO groups showed highest preoperative dobutamine-dosage, while highest intraoperative dosage was observed in post-ECMO group ( Figure 1SB). In pre-ECMO patients, epinephrine-doses were comparable to No-ECMO group, but it was significantly higher in ECC-and post-ECMO groups ( Figure S1C). Its maximum was observed at the end of CPB in pre-ECMO group, which might have triggered the surgeons' decision to switch from CPB to ECMO. Milrinone was administered in comparable doses in all ECMO groups.
The pre-ECMO group even showed a significant bilirubin-increase perioperatively. The first bilirubin-downtrend was observed after
postoperative blood product transfusions were needed in all ECMO patients but with no difference between the ECMO groups. Rethoracotomy-rate was highest in the ECC-ECMO group, while wound-healing disorders were most common in post-ECMO group.
Hemofiltration was most often needed in post-ECMO patients; also, the lowest diuresis, the highest 24 hours fluid-balance, and the highest incidence of pneumonia were observed in this group.
The incidence of postoperative delirium was significantly higher in the No-ECMO group than in other groups.

| Evaluation of risk-scores
Patients in the post-ECMO and ECC-ECMO group revealed significantly longer ventilation-time, ICU-, and hospital-stay and higher nursing-workload-index TISS-10 when compared with the No-ECMO group (Table 3 and Figure S3). Comparison of all three ECMO groups showed no significant difference in APS-and APACHE-IIscore, while they had a higher score when compared with No-ECMO group after surgery. The difference in APS-and APACHE-II-score between No-ECMO and pre-ECMO group was not significant, probably due to the small number of patients. Simplified acute physiology-score-I-and quick sepsis-related organ failure assessment-score were significantly higher in the pre-ECMO when compared with the No-ECMO group (Table 3 and Table S3).
Therefore, these scores did not provide discrimination in regard to mortality-risk of the ECMO groups.
Post-ECMO group showed the longest time from admission to ECMO-implantation, while the pre-ECMO group had the longest duration of ECMO therapy. Before ECMO-implantation, survival after VA-ECMO (SAVE)-score was significantly lower in post-ECMO patients when compared with both other ECMO groups. ENCOURAGE-score was significantly lowered in the post-ECMO group when compared with the pre-ECMO group (Table 4 and Table S4). No difference was seen regarding successful ECMO-weaning, ECMO-replacement by another LVAD and ECMO withdrawal due to poor prognosis.

| DISCUSSION
In the last few years, ECMO became a standard therapy for postcardiotomy-patients with CS, 16 while our institutions also established it as early LV-support in AMI-patients with poor hemodynamic conditions. Our study investigated the impact of ECMO therapy timing and shows a significantly higher survival-rate when ECMO-implantation is not delayed and performed even before CABG-procedure. On the basis of our data, an increase in lactateconcentration greater than 4 mmol/L seems to be a reasonable trigger for ECMO-implantation, and only ECMO-related SAVE-score provided a reliable survival estimation.
Patients with AMI need emergency revascularization to preserve LV-function. In cases of STEMI, hemodynamic impairment is frequently encountered and this may lead to CS. This critical condition is still associated with high mortality, independent of the procedure used-PCI or CABG. 2,3 Several small RCTs investigated the outcome in this situation using Impella and Tandem-Heart, but failed to show improvement in 30-day-mortality when compared with IABP. 10,13,21 In acute CS, Impella-2.5-assist was also not associated with a better 30-day-survival when compared with medical therapy without assist. 12 Still, ECMO was able to decrease 30-day-mortality in patients with primary PCI for STEMI with CS. 14,15 Another study showed a better outcome in a heterogenic patient-cohort with CS, which also included AMI-patients, when Impella was combined with VA-ECMO-support. 22 A meta-analysis revealed a survival-benefit of perioperative VA-ECMO-use in postcardiotomy-patients with CS, which show similar clinical conditions as STEMI-patients with CS. 16 In a recent retrospective multicentre study, only one-third of patients who received VA-ECMO due to respiratory or cardiac failure after CABG survived to discharge. 18 These data supported our initial clinical observation that earlier perioperative ECMO-assist may be associated with better survival in patients with AMI subjected to emergency-CABG. To our knowledge, no study has investigated patients with ECMO-implantation before CABG. Since this extent of therapy involves a large amount of infrastructure and resources, Switch from ECMO to assist device, n (%) 0 (0) 5/66 (7.6) 2 (7.4) .784 Note: Values are expressed as mean ± SD or as number and percentage (in brackets). For categorical variables the Pearson χ 2 test was performed as appropriate. When describing changes of metric values, ANOVA with Tukey post-hoc analysis was performed, and P < .05 was considered statistically significant (displayed in italics). * and **, P < .05 vs. G1; Ψ, P < .05 vs. G2; #, P < .05 vs. G3; †, P < .05 vs. G4. For additional data see Table S4. Abbreviations: ANOVA, analysis of variance; ECMO, extracorporeal membrane-oxygenation; SAVE, survival after VA-ECMO.
which are mostly limited, it is of utmost importance to identify patients which may benefit from early ECMO-support. In this sense and based on suggestions that EuroSCORE can be a useful predictor for prophylactic perioperative ECMO in patients with postcardiotomy, 16 we also evaluated different risk-scoring-systems.
Our data show that AMI-patients with ECMO-start after CABG had the lowest 30-day-survival of 40.7%, which is comparable to other studies. 18 Our analysis of different timing for ECMO therapy revealed that preoperative and intraoperative ECMO-start was associated with significantly better survival reaching up to 66.7%. Interestingly, experimental work in swine showed that prior "LV-unloading" with an extracorporeal axial flow-catheter while delaying coronary reperfusion provided myocardial protection. 19 But, the Impella-support alone was not able to improve 30-day-survival in AMI-patients with CS. 12 Still, timely LV-support using ECMO seems to be beneficial in our patients. This is also supported by the time-course of bilirubin-concentration, which influences survival in patients with ECMO-support. 23 higher shock-index was found on admission in patients from the preand ECC-ECMO group, which also reflected frequent CPR before admission in these groups. This finding was supported by significantly higher logEuroSCORE and stdEuroSCORE, APS-, and APACHE-IIscore when compared with the No-ECMO group preoperatively. But, the differential analysis of these scores between the three ECMO groups showed no significant difference, except for the stdEuro-SCORE, when comparing the pre-and post-ECMO group. These findings reflect the lack of impact ECMO therapy has on the abovementioned scores. We therefore also analysed the value of the ECMO related SAVE-and ENCOURAGE-score. The SAVE-score has been shown to predict survival for patients receiving ECMO for refractory CS. 30 We found lowest SAVE-scores in the post-ECMO group, correctly reflecting the lowest survival-rate. The ENCOU-RAGE-score has been developed to predict in-ICU mortality of VA-ECMO, 31 with a highest value of ≥28 being associated with a survival-probability of only 7%. Our analysis showed the highest ENCOURAGE-score in the pre-ECMO group, while ECC-and post-ECMO groups had comparable values, corresponding to an estimated survival probability of ≈25%. Therefore, our results are in contrast to the published survival-possibilities, and this may indicate the potential limitation of this score in the setting of AMI-patients with CS. Taken together, only the SAVE-score was able to correctly predict the 30-day-mortality rate in our patients. Unfortunately, the assessment of SAVE-score at admission would not have been able to correctly discriminate the patients needing ECMO, since no increase in shock-index or other parameters was observed in the post-ECMOgroup. Therefore, we still do not have an appropriate scoring-system respecting pre-emptive ECMO-timing.

| LIMITATIONS
The presented study is a retrospective study and only two major centers for cardiac surgery were included. Furthermore, the number of six patients treated with preoperative ECMO-assist does not permit normal distribution. Also, a selection bias for ECMOtreatment in regard to the time point cannot be excluded. In this retrospective study, the answer to the question how the decision was made to use additional methods of left ventricular unloading (eg, IABP or Impella) remains speculative.

| CONCLUSION
Our study provides evidence for a novel concept of timely ECMO support for the failing ischemic myocardium based on association between early ECMO-start and a better survival in patients with AMI undergoing emergency CABG. While decision-making still remains complex in regard to correct timing for the ECMO-start, our findings support the use of lactate-increase above 4 mmol/L in an impaired hemodynamic situation as a valuable trigger for ECMO-implantation, as proposed in the organization chart ( Figure 4). In these patients the analysed risk-scores assessment tools only gave limited support in estimating survival-probability correctly (we found only a limited support of different risk-score assessment tools for correct estimation of survival-probability