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

In the largest cohort with the longest follow up available in the literature, Rastan et al. reported results from 516 patients undergoing salvage ECMO for PCCS over a 12-year period from 1996 to 2008. They reported a survival to hospital discharge rate of 24.8% and a 13.7% 5-year survival [5]. In another large study Doll et al. reported results from 219 patients that underwent VA ECMO for PCCS from 1997 to 2002. They reported a 39% survival to hospital discharge and 17% 5-year survival [13].

In a multinational European study, Santarpino et al. collated data from 11 European cardiac surgical centers with cumulative results from 85 adult patients. Survival rate to hospital discharge was reported as 40% with a 1-year survival rate of 29.3% [14].

Li et al. reported on 123 adult patients who were salvaged with VA ECMO for refractory PCCS. They noted that 56% of patients were successfully weaned from VA ECMO and 34.1% survived to hospital discharge [15]. Elsharkawy et al. reported results from 233 patients with survival to hospital discharge was reported at 36% [16]. In another cohort study Wu et al. reported outcomes from 110 patients of whom 61% were successfully weaned and 42% were successfully discharged home [17].

Bakhtiary et al. reported a 55% survival to ECMO decannulation in their 45 patient cohort study, with a total in-hospital mortality rate of 71% for the cohort. In 3 years, 77% of the survivors were still alive [18]. Ko et al., analyzed the outcomes of 76 patients who underwent ECMO for intractable PCCS. Although 60.5% of patients survived to ECMO decannulation, 26.5% survived to hospital discharge. On 32+/−22 months follow-up, all survivors were of NYHA I-II functional status [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].

Major haemorrhage was the most commonly reported complication after institution of VA ECMO. It led to re-intervention 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, 20‐22]. 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].

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, 23‐25]. 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.

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 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).

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/m²) with haematocrit levels of around 20%, leading to lower than physiological systemic oxygen delivery (DO₂) [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 DO₂. 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 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 non-transplant 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.

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.