Extracorporeal membrane oxygenation (ECMO) is a rescue therapy used to support patients with severe cardiac and/or pulmonary dysfunction refractory to conventional treatment. It was first used successfully in adults over 40 years ago, but its use has only recently become widespread, with technological advances in the available devices and with growing evidence of its effectiveness.1,2

The basic elements of an ECMO circuit are two cannulas (inflow and outflow), a centrifugal pump and an oxygenator. The latter is a gas exchanger containing a semipermeable membrane separating two chambers, one for blood and the other for gas. Deoxygenated blood is drained by the external pump, passes through the oxygenator, in which carbon dioxide is exchanged for oxygen, and is returned to the patient. When the blood is both drained from and returned to the venous system, the circuit is termed venovenous (VV) ECMO and only provides respiratory support; if it is drained from the venous system and returned via an artery, it is termed venoarterial (VA) ECMO and provides both respiratory and circulatory support.2

ECMO devices are light and portable, and so patients undergoing the treatment are relatively mobile, facilitating transport within and between hospitals, and the cannulas can be placed percutaneously or centrally, depending on clinical circumstances; the technique can be used in cases of cardiac arrest. Although ECMO provides only temporary support and has limitations, its versatility makes it useful in a variety of clinical situations, from clinical stabilization leading to complete recovery to situations requiring a bridge to decision, whether for long-term ventricular support or transplantation, or for suspension of support.3 The latest European Society of Cardiology guidelines on heart failure recommend VA-ECMO for patients in Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) levels 1 or 2.4

The aim of the present study is to describe patients’ demographic characteristics, indications, comorbidities, and physiological variables before ECMO; treatment (duration, modality, and complications); and mortality (on ECMO and in-hospital) of a population treated by ECMO in a single tertiary hospital. We also analyze possible predictors of mortality in this population.

This was a retrospective observational cohort study that included all patients supported with ECMO in two different units – general intensive care (ICU) and cardiac care (CCU) – of a tertiary hospital, from the first patient cannulated in April 2011 up to October 2016. Patients in whom ECMO was used following cardiotomy were excluded from the analysis as they were treated according to different implantation and follow-up protocols.

During this period a total of 48 ECMO devices were placed in these units, 29 for cardiac failure and 19 for respiratory failure.

Relevant variables were selected on the basis of the published literature and patient data (demographic characteristics, comorbidities, and indications for ECMO) were collected from medical records. The values of these variables before ECMO considered for analysis were the worst in the first 24 h following admission to the unit. The Modification of Diet in Renal Disease (MDRD) formula was used to estimate glomerular filtration rate (GFR). ECMO-related complications and duration of ECMO support and of stay in the unit were recorded. Complications were classified as bleeding (access site, gastrointestinal or cerebral), ischemic, or access-related (thrombosis, dissection, aneurysm, pseudoaneurysm, or fistula). Acute renal failure (ARF) was defined according to the RIFLE classification. Hematological dysfunction was defined as platelets <100000/μl or hematocrit <35%, and liver dysfunction as transaminases more than three times the upper normal limit. Data were collected on mortality during and after ECMO and during hospital stay. Successful weaning was defined as that occurring with the intention to survive or as a bridge to another type of support. Comparisons were performed between the VA-ECMO and VV-ECMO groups and between survivors and non-survivors in each group. Survival after VA-ECMO was predicted by the SAVE-score, which uses a scale of -35 to 17 points based on diagnosis, age, weight, other acute or chronic dysfunction, ventilation parameters, duration of intubation before ECMO, pre-ECMO cardiac arrest, and hemodynamic parameters. The SAVE-score stratifies patients in five risk classes according to expected survival: class I (score >5) 75% survival; class II (1 to 5) 58%; class III (-4 to 0) 42%; class IV (-9 to -5) 30%; and class V (≤-10) 18%.5

All ECMO devices were placed percutaneously, either in the unit or in the catheterization laboratory, by a multidisciplinary team composed of intensivists, cardiologists, perfusionists and nurses. The Cardiohelp system (Maquet, Hirrlingen, Germany) was used in all cases except one. Femoral access was used in most cases of VA-ECMO and one femoral and one jugular access for VV-ECMO. A distal reperfusion cannula was used routinely when the indication for ECMO was myocardial infarction (MI), and for other indications when there was clinical evidence of poor distal perfusion, under echographic guidance in all cases. Anticoagulation was achieved by a bolus of unfractionated heparin at the time of cannulation (50-100 U/kg, maximum 5000 U), followed by perfusion adjusted to maintain a target activated partial thromboplastin time of 1.5 times the upper normal limit.

Surgical ventricular assistance was used when pharmacological and mechanical attempts to maintain adequate left ventricular venting were unsuccessful. Patients were referred for cardiac transplantation (not available in our center) when their etiology or clinical course indicated that left ventricular function was unlikely to recover. We recently established a referral protocol with one of the largest national transplantation centers.

Over a five-year period, 48 patients underwent ECMO: 29 for cardiovascular failure and 19 for respiratory failure. Table 1 shows that disease markers, at least initially, in patients receiving VA-ECMO were more severe, with higher serum lactate and AST levels, greater prevalence of moderate to severe renal dysfunction, and lower serum bicarbonate. The shorter time between diagnosis and ECMO placement and the greater frequency of previous IMV in this group are explained by the greater severity of clinical presentation. Almost half of the VA-ECMO group were transferred in cardiogenic shock from another hospital, and were thus more likely to have been suffering shock for longer and to have more advanced organ dysfunction.

Percutaneous left ventricular decompression techniques were used in around 30% of patients in our cohort. Other measures, including reduction of afterload, use of arterial vasodilators and inotropic drugs, fluid restriction and hemofiltration can help prevent ventricular distension, and were applied systematically in all cases.6–8 This complication, which occurs in 10-60% of patients undergoing VA-ECMO and is more frequent in newborns and children,7 hampers recovery and leads to pulmonary edema.

Anticoagulation is a challenge in these patients, as contact between the blood and the extracorporeal circuit leads to activation and consumption of pro- and anticoagulant factors, meaning a delicate balance must be struck between risk of thrombosis and of bleeding.9 The recommended anticoagulant is unfractionated heparin.9,10 All patients in our population underwent continuous heparin infusion and only one was diagnosed with heparin-induced thrombocytopenia, which necessitated replacing heparin with lepirudin.

Although ECMO is increasingly used to support patients in cardiac or respiratory failure, it is often associated with complications, some of which can have a significant impact on the quality of life of survivors.6 Bleeding complications occurred in 10 patients (21%) in our series. The most frequent bleeding site was the point of insertion of the cannulas. In our cohort, bleeding and quantity of transfused blood products were not predictors of mortality, although this has been reported in the literature.3,11–13 This is probably due to the small sample size, which meant these parameters did not reach statistical significance. Neurological complications, the incidence of which in the literature ranges between 15% and 37%,14,15 were rare in our series, in only one patient (spontaneous intracerebral bleeding). This low figure is probably due to underuse of cerebral computed tomography and the lack of postmortem anatomopathological studies.

Acute lower limb ischemia occurred only in the VA-ECMO group (6/29 patients). Use of a distal reperfusion cannula and contralateral venous cannulation are essential to reduce the incidence of this complication.16,17 Three patients required urgent fasciotomy (10.3% of the VA-ECMO group) and one underwent lower limb amputation. These results are similar to those in the literature, in which lower limb ischemia affects 10-30% of patients and compartment syndrome affects 8-14%.8,12,16,18–20

As in a meta-analysis by Cheng et al.,20 ARF was the complication most frequently associated with ECMO, affecting over 75% of those undergoing VA-ECMO and nearly 40% in those receiving VV-ECMO. Pump speed and red cell distribution width have recently been identified as predictors of ARF.21 In our population, 10 patients undergoing VA-ECMO and one receiving VV-ECMO required RRT. Although need for RRT reflects inadequate renal perfusion and/or direct injury to the kidneys by the support technique through hemolysis, transfusions, or rhabdomyolysis, it was not significantly associated with greater mortality in either of our groups. This may be explained by the small sample size, since several series have demonstrated that ARF and RRT are independent predictors of mortality in individuals receiving ECMO.6,11,19,22,23 In-hospital mortality was 36.8% in the VV-ECMO group and 62.1% in the VA-ECMO group. These figures are consistent with those observed in the international ELSO registry, in which in-hospital mortality in adults was 42% in VV-ECMO and 59% in VA-ECMO.24 In our series, 69% of patients undergoing VA-ECMO and 78.9% of those receiving VV-ECMO were successfully weaned.

CESAR was the first randomized trial to show clear benefits of VV-ECMO. This study randomized 180 patients with severe respiratory failure to a conservative strategy or to transfer to a center with ECMO, 63% of whom survived, as opposed to 47% of controls.25 The H1N1 pandemic in 2009 led to greater experience with ECMO in many centers,26 and in our center, five patients with ARDS secondary to H1N1 infection all survived.

In this study, univariate analysis revealed no predictors of mortality in VV-ECMO, whereas in published series previous pulmonary disease, advanced age, previous barotrauma and number of days on IMV before ECMO were all independent predictors of mortality.13,27,28

In VA-ECMO, a greater number of inotropic drugs used during hospitalization predicted death. In other series advanced age, female gender, high body mass index, diabetes, elevated serum lactate, number of red blood cell units transfused, and placement of ECMO during cardiac arrest were all independent predictors of mortality.3,6,18,23,29,32

Mortality in our sample was 77%. Analysis of mortality in terms of reasons for device placement shows that cardiogenic shock following MI and shock following cardiotomy has the highest mortality, most series reporting figures of 40-75%.18,29–32 In these series, ECMO placement in such situations occurred mainly before or during percutaneous revascularization, while in our population it was only placed as ventricular support when a patient remained in refractory cardiogenic shock after a successful percutaneous intervention, probably due to pump failure and thus probably with less likelihood of recovery.

In view of the high cost and high short-term mortality of ECMO, careful patient selection is vital. In 2015 Schmidt et al. published the SAVE-score, which is designed to predict survival after VA-ECMO.5 In our population, the mean SAVE-score was -4.2±4.9, median -4.5 (IQR -0.25/-8). Thus most of our patients were in risk class III or IV, with estimated survival of 30-42%, and actual survival was 37.9%. This score still needs to be validated but could be a useful tool in the future.5

The limitations of our study include its design as a retrospective cohort study without a control group. The sample size was small, and certain variables which the authors consider important could not be included due to gaps in the recording of certain data. There may therefore have been biases that could have influenced the results presented. In addition, this is a single-center analysis and hence generalization of the results is limited.

Despite these limitations, the study provides real-world results of the use of ECMO in a tertiary hospital.

Patients with cardiovascular failure placed on VA-ECMO have much higher mortality than those with respiratory failure placed on VV-ECMO. In VV-ECMO, later device placement appears to be associated with a worse prognosis.

Complications were relatively common, particularly vascular complications. Given the frequency of hematological dysfunction and the difficulty of achieving proper hemostasis, ways of minimizing the consequences of vascular complications are a priority in the management of these patients.

The need for more inotropic drugs was a predictor of mortality in VA-ECMO.

In conclusion, despite the small sample size, our results are similar to those in most published series. This is the first published record of the overall experience with ECMO in a Portuguese tertiary hospital.

The authors have no conflicts of interest to declare.