Extra-corporeal membrane oxygenation for refractory cardiogenic shock after adult cardiac surgery: a systematic review and meta-analysis

Background Postcardiotomy cardiogenic shock (PCCS) refractory to inotropic support and intra-aortic balloon pump (IABP) occurs rarely but is almost universally fatal without mechanical circulatory support. In this systematic review and meta-analysis we looked at the evidence behind the use of veno-arterial extra-corporeal membrane oxygenation (VA ECMO) in refractory PCCS from a patient survival rate and determinants of outcome viewpoint. Methods A systematic review was performed in January 2017 using PubMed (with no defined time period) using the keywords “postcardiotomy”, “cardiogenic shock”, “extracorporeal membrane oxygenation” and “cardiac surgery”. We excluded papers pertaining to ECMO following paediatric cardiac surgery, medical causes of cardiogenic shock, as well as case reports, review articles, expert opinions, and letters to the editor. Once the studies were collated, a meta-analysis was performed on the proportion of survivors in those papers that met the inclusion criteria. Meta-regression was performed for the most commonly reported adverse prognostic indicators (API). Results We identified 24 studies and a cumulative pool of 1926 patients from 1992 to 2016. We tabulated the demographic data, including the strengths and weaknesses for each of the studies, outcomes of VA ECMO for refractory PCCS, complications, and APIs. All the studies were retrospective cohort studies. Meta-analysis of the moderately heterogeneous data (95% CI 0.29 to 0.34, p < 0.01, I 2 = 60%) revealed overall survival rate to hospital discharge of 30.8%. Some of the commonly reported APIs were advanced age (>70 years, 95% CI −0.057 to 0.001, P = 0.058), and long ECMO support (95% CI −0.068 to 0.166, P = 0.412). Postoperative renal failure, high EuroSCORE (>20%), diabetes mellitus, obesity, rising lactate whilst on ECMO, gastrointestinal complications had also been reported. Conclusion Haemodynamic support with VA ECMO provides a survival benefit with reasonable intermediate and long-term outcomes. Many studies had reported advanced age, renal failure and prolonged VA ECMO support as the most likely APIs for VA ECMO in PCCS. EuroSCORE can be utilized to anticipate the need for prophylactic perioperative VA ECMO in the high-risk category. APIs can be used to aid decision-making regarding both the institution and weaning of ECMO for refractory PCCS.


Background
Postcardiotomy cardiogenic shock (PCCS) refractory to inotropic support and intra-aortic balloon counter pulsation (IABP) is an infrequent but almost universally fatal condition without mechanical circulatory support (MCS) [1][2][3][4][5]. Veno-Arterial (VA) extra-corporeal membrane oxygenation (ECMO) has been utilized as a salvage MCS after cardiac surgery for almost 50 years [6,7]. The decisions surrounding when best to institute or withdraw such invasive and resource-intensive therapy remains controversial and there are no universally agreed upon guidelines on the indications for this therapy. VA ECMO in the context of refractory PCCS is mainly instituted as a temporizing measure as a "bridge to recovery" [5,8,9]. However, it has also been utilized as a "bridge to decision" and "destination therapy" with long-term implantable devices (e.g. left ventricular assist device, LVAD), and more rarely in the UK, "bridge to orthotopic heart transplantation (OHT)" [8][9][10]. Nevertheless, ECMO carries with it a significant morbidity rate, often associated with prolonged hospital stays and poor quality of life for the survivors after hospital discharge [5,11].
In this systematic review and meta-analysis we have looked at the survival rate following VA ECMO for intractable PCCS in adults and some of the most commonly and consistently reported adverse prognostic indicators (API) in this group of patients.

Methods
A comprehensive literature search was performed in January 2017 using OVID/MEDLINE (PubMed) for all articles published in the English language in peerreviewed journals. The inclusion criterion was postcardiotomy VA ECMO in adults. We excluded all articles pertaining to VA ECMO for paediatric cardiac surgery as well as non-surgical indications for the use of this therapeutic modality e.g. myocarditis or cardiomyopathy. We also excluded case reports, review articles, expert opinions, and letters to the editor.
The search was performed, with no limit to the year of publication, using the following three search strategies: Search 1: "(((Extracorporeal membrane oxygenation) AND thoracic surgery) AND cardiac surgical procedures) AND shock, cardiogenic", The first search strategy yielded 179 articles, the second search strategy yielded 149 articles and the third search strategy yielded 103 articles. The authors assessed the abstracts in all three searches. The "preferred reporting items for systematic reviews and metaanalyses" ("PRISMA") guideline [12] was followed.
When it was not possible to ascertain article suitability from its abstract alone we obtained the full text article for further assessment.
As this was a systematic review, ethical board review was not required. The primary outcome measure was determining the survival benefit following the institution of VA ECMO for refractory PCCS. The secondary outcome measure was identifying the most commonly reported APIs. The demographic data, APIs that had reached statistical significance (P < 0.05) in each study and crude data on outcome and follow-up, where available, were collated from each study. We then performed meta-regression to assess whether the APIs identified were significantly associated with survival.

Results
We identified 24 retrospective cohort studies that met the inclusion criteria from 1992 to 2016 in the English language literature. These studies as well as their strengths and weaknesses, are summarized in Table 1. In this review a cumulative total of 1926 patients required VA ECMO for PCCS.

Survival and follow-up
In the largest cohort with the longest follow up available in the literature, Rastan [19]. In a smaller cohort conducted by some of the authors of this systematic review, 35% survival to hospital discharge was observed and all survivors were still alive at 12 months with NYHA class I-II [20].

Complications
Major haemorrhage was the most commonly reported complication after institution of VA ECMO. It led to reintervention in almost half of the patients in a few good quality studies [3,5,13,14,19,20]. Factors such as the insult of the original operation, heparin infusion and heparin coated circuits are known to be the primary causes of bleeding [13]. Renal failure requiring renal replacement therapy (RRT) was the next most commonly reported complication [3,5,13,[20][21][22]. Stroke and sepsis also followed amongst the commonly encountered complications (see Table 1). Distal limb ischaemia is one of the most dreaded complications because if often leads to major morbidity. The largest study in this systematic review reported that almost 20% of patients with peripheral VA ECMO developed a degree of distal limb ischaemia. However, the use of a polyethylene terephthalate tube graft (e.g. Dacron R ) as a sewn side arm to the femoral or axillary arteries or the distal leg perfusion cannula reduced distal limb ischaemia and fasciotomy for compartment syndrome rate by almost 40% [5].

Adverse prognostic indicators
APIs that reached statistical significance and were directly associated with mortality in each study are summarized in Table 1. Advanced age (typically >70 years), although perhaps not in isolation [5,15,17,21,23], was the most commonly reported API. Development of renal failure, whilst on VA ECMO that required RRT had a strong association with mortality [5,16,[23][24][25]. While this common complication in the context of VA ECMO for refractory PCCS is usually multifactorial. It is imperative to determine the most likely underlying aetiology/ies for renal dysfunction as early as possible. The following potential aetiologies should be identified and corrected as early as possible postoperatively: renal hypoperfusion due to poor forward flows whilst on VA ECMO, acute tubular necrosis due to prolonged hypotension preceding institution of support or syndrome of inappropriate anti-diuretic hormone secretion (SIADH) [26]. It is worth noting that the use of loop diuretics to address fluid overload or poor urine output leads to worsening renal dysfunction [26]. Rising serum lactate whilst on VA ECMO [5,15,21] has also been reported as a strong predictor of mortality. One study [5] advocated use of sodium bicarbonate infusion at an early stage to reduce metabolic acidosis and the organ damage that might ensue. Diabetes mellitus [5,23], obesity [5,23], gastrointestinal complications whilst on ECMO, high EuroSCORE (>20) and protracted ECMO support (>48 h) were also amongst the commonly reported APIs [17,21,26,27]. It must however be acknowledged that prolonged ECMO support might reflect the complexity of the original operation and the patient's poor clinical.
Although EuroSCORE was widely reported in the studies, it was quoted in its three versions (i.e. the additive EuroSCORE, logistic EuroSCORE and EuroSCORE II) over nearly 3 decades covered by this systematic review. This variation made meta-analysis for EuroSCORE as an API not possible.

Meta-analysis of survival to hospital discharge
We found that of the pooled total of 1926 patients, 594 (30.8%) survived to hospital discharge. We performed a meta-analysis on the outcome of interest i.e. the proportion of survivors. The software R was used with the package "metaprop"**. Meta-analysis is conducted to estimate the true unknown success rate of a procedure, by combining the results of several studies. Table 2 summaries the variables included in the meta-analyses. Figure 1 shows the forest plot describing each of the study's proportion of survivors with their 95% confidence interval (CI).
Heterogeneity is present as defined by a statistically significant I 2 = 60%, which represents the percentage of variability in the effect estimates due heterogeneity rather than random error. A value of I 2 between 30 and 60% is considered as a moderate heterogeneity. By looking at the forest plot, one can observe that study 11 is well out from the rest. By removing study 11, the I 2 value dropped to 59%. The overall proportion of survivors is 31% (95% CI 0.29 to 0.34, p < 0.01, I 2 = 60%), but when considering random effect it raises to 33%.

Meta-regression of the adverse prognostic indicators
Meta-regression on the effects of moderators such as age, usage of pre-VA ECMO IABP support and duration of VA ECMO support was performed as these were the most commonly and consistently reported variables by most of the studies. Analysis showed that heterogeneity had been observed due to random sampling and variation in the methods used for the studies (Table 3). In order to account for at least part of the heterogeneity, mixed effects models are used by including moderators. This was also carried out using the package "metafor" in R software. By including the moderators the heterogeneity statistic I 2 , which represents presence of variation in an inter-study effect size (proportion of survivors) dropped to 52.2% from the model without moderators (previously 60% see forest plot, Fig. 1), which is classified as moderate. We found that none of the coefficients of the moderators are statistically significant (Table 3), which means that there is no moderating effect of the mean age, usage of pre-VA ECMO IABP and the mean number of days on ECMO on the effect size. One of the reasons of the lack of power for the test on the coefficients is the small sample sizes used in the studies.
Furthermore, we looked at the possibility of publication bias in the meta-analysis. Our analysis showed no departure from symmetry in the funnel plot (Fig. 2), hence absence of bias. This claim is supported by the Egger's test (p = 0.556).

Discussion
Our systematic review is the first of its kind to be published, analyzing the efficacy of VA ECMO as a salvage modality for refractory PCCS.
Refractory PCCS typically transpires at the end of a complex and prolonged operation [28]. It can also occur in an otherwise routine operation [5,29] that has encountered an unexpected technical difficulty e.g. an iatrogenic injury to a vital structure during the course of an operation. In either case the patient cannot be weaned from cardiopulmonary bypass (CPB) with survivable haemodynamic and arterial blood gas parameters, despite multiple inotropic support agents and IABP [30]. Furthermore, adding to the predicament, such cases typically encroach into "out-of-hours" [31] when the operating theatre team is fatigued and there is a relative lack of availability of technical assistance and experienced advice.

Why VA ECMO?
VA ECMO for refractory PCCS is typically established centrally i.e. arterial line through the ascending aorta and the venous line through the right atrium as part of the continuum with CPB at the end of the operation [5,29]. Although in principal the conduct of CPB and VA ECMO are similar, there are several important advantages in using VA ECMO in the context of refractory PCCS [32]. The CPB achieves full VA bypass at lower flow rates (2-2.4 L/min/m 2 ) with haematocrit levels of around 20%, leading to lower than physiological systemic oxygen delivery (DO 2 ) [32]. A large dose (300-400 Units/kg) of unfractionated heparin (UH) is required to run CPB, as it is an "open circuit" with a venous reservoir (stagnant blood). In addition, full VA bypass is said to cause stasis of blood in the cardiac chambers and the pulmonary circulation warranting higher activated clotting times (ACT). VA ECMO on the other hand is more physiological. It uses partial VA bypass through a "closed circuit" (without a venous reservoir) and shorter tubing, typically in normothermia, with normal haematocrit levels, aiming for near normal DO 2 . As VA ECMO allows venous return to the heart, it allows cardiac ejection and less risk of thrombosis, thereby requiring minimal doses of UH as compared to CPB. This in-turn leads to fewer rates of postoperative bleeding complications requiring re-exploration whilst on MCS. VA ECMO is more versatile and is more easily manageable in the intensive care unit (ICU) setting for often-prolonged periods of time (i.e. days-weeks), unlike CPB, which is geared more towards short-term support (i.e. a few hours) [32]. The VA ECMO line tubing can be tunneled through the skin to allow chest closure. This maybe significant in that "incomplete sternal closure" has been reported as an independent predictor of mortality, as identified in a study by Unosawa et al. [27]. However, line change over to other sites, with axillary arterial, femoral arterial and femoral venous cannulation sites have previously been reported [5,13]. Occasionally combined right atrial and femoral venous cannulation have been used to improve venous drainage [5]. Vascular access can be established with either surgical cut down or by Seldinger techniques. Some sew a side arm tube graft to the artery or use a distal perfusion cannula to reduce the risk of distal limb ischaemia and compartment syndrome (see complications) [5,13]. Although ECMO is a valuable salvage modality, it is expensive [33]. It is very resource intensive due to its high demand on the ICU staff. It requires skilled staff performing high frequency monitoring for its safe application and maintenance [33]. These factors can pose a significant burden particularly on small and intermediate sized cardiac surgical centers, where the work force and  Table 2, describing each studies proportion of survivors (CI 95%) bed capacity are both limited [30]. ECMO patients, on average, require more prolonged ICU stay than elective cases thereby leading to increased cancellation of elective and sometimes urgent cardiac surgical operations. As such, indiscriminant use of VA ECMO can result in major disturbance to an already stretched service [30]. Some centers in the UK rely on transplant centers for advice regarding whether or not to place PCCS patients on VA ECMO. If these patients are commenced on ECMO, they are transferred over to the larger transplant centers for further management [30,34]. Critics had argued that, if only transplant units are funded to provide ECMO for PCCS support, similar patients at nontransplant cardiac surgical centers are denied of a life saving therapy [35]. In reality the use of VA ECMO for PCCS is not formally commissioned by the National Health Service (NHS) in the UK and the cost of treatment has to be absorbed by individual hospitals.
Kashani et al. [36], published an abstract on a systematic review covering only 11 case series with a cumulative pool of 1328 patients whereas our review covers 24 studies with a cumulative pool of 1926 patients. The survival rate to hospital discharge reported by Kashani et al. of 31.48% was comparable to our report of 30.8%. In their systematic review, they found similar APIs to our systematic review such as advanced age, elevated serum lactate after initiation of ECMO and renal failure [36].
Our meta-regression of APIs such as mean age, pre-VA ECMO use of IABP, effect of renal failure and mean ECMO duration however showed no statistically significant correlation between these parameters and survival, mainly due to the small sample sizes and presence of wide heterogeneity amongst study populations (Table 3).
This topic remains a controversial one with ethical and financial considerations for any cardiac surgical service. The decision as to whether or not to institute VA ECMO in the setting of PCCS remains difficult. Institution of VA ECMO for PCCS is usually not planned and the aetiology of the patient's lack of progress may not be immediately apparent [19]. A study of 100 trans-catheter aortic valve implantation (TAVI) patients, reported that anticipation for ECMO and institution of this mode of support prophylactically in the "high-risk" category on the EuroSCORE scale would potentially prevent the need for salvage VA ECMO, in a less optimal and controlled clinical setting, and carry better outcomes. In this study all-cause mortality occurred in none of the high-risk patients undergoing prophylactic VA ECMO (although p > 0.05) [37].
The literature demonstrates overwhelming evidence pointing towards reasonable survival rate to hospital discharge for patients undergoing ECMO for an otherwise universally fatal clinical condition. Furthermore reasonable intermediate-term and long-term survival rates as well as good quality of life have been reported in a few studies (Table 1) mainly for the survivors that do not manifest the APIs (see results). Major life threatening ECMO complications are common. We advocate a multidisciplinary team (MDT) approach to decision making for institution of ECMO in the context of refractory PCCS [29]. We recommend that given the ethical and cost implications, the surgeon, the anaesthetist, the on-call intensivist, the on-call perfusionist and a cardiac surgeon not involved in the operation (e.g. the on-call cardiac surgeon) should be involved in the decision making in whether or not to institute ECMO for refractory PCCS on a case-by-case basis [20]. We believe that due to the complexity of such cases protocols may not be an adequate substitute for an MDT approach to decision-making and management of such complex patients [5,20]. Finally, in order to both preserve the patients' autonomy and aid decision-making in the event of encountering refractory PCCS, we advocate that the possibility of the need for VA ECMO along with its pros and cons should be discussed with high-risk patients preoperatively and informed consent should be obtained in this regard.

Conclusions
We believe that VA ECMO provides a survival benefit for a significant proportion of patients with refractory PCCS, which is invariably a fatal clinical state. For hospital survivors, a reasonable intermediate and long-term functional outcome can be expected albeit at the expense of prolonged and often ridden hospital stay. We identified advanced age, renal failure and prolonged VA ECMO support to be commonly reported APIs. This claim however could not be supported by metaregression due to small patient numbers and heterogeneity of the patient populations. Risk stratification tools such as the EuroSCORE can be utilized to anticipate the need for prophylactic perioperative VA ECMO in the high-risk category. The reported APIs should be taken into consideration before institution of and ongoing treatment with this laborious, invasive and expensive therapeutic modality.

Limitations
Most of the evidence available in the literature, including our analysis, pertains to older studies from the 1990s. Due to the nature of PCCS, randomization would not be appropriate and the studies typically constitute heterogeneous patient populations leading to skewing of the data. As for the pre-VA ECMO IABP application subgroup analysis, a few studies had not reported this data. In such studies we assumed that, in principal, all patients would have had IABP pre-VA ECMO institution. Hence the outcome of data analysis errs on the side of IABP usage.
Small number of studies and heterogeneity of the patient populations meant that the statistical tests were not strong enough to detect statistical significance in our study. The data on EuroSCORE, rising lactate whilst on ECMO and the rate of obesity was not reported by enough studies to constitute a more objective analysis, hence deriving any solid conclusions regarding these indicators was difficult.