Secondary Logo

Journal Logo

Adult Circulatory Support

Prognostic Factors for Survival After Extracorporeal Membrane Oxygenation for Cardiogenic Shock

Huang, Mingjie; Ong, Boon Hean; Hoo, Anne Ean Ean; Gao, Fei; Chao, Victor Tar Toong; Lim, Chong Hee; Tan, Teing Ee; Sin, Kenny Yoong Kong

Author Information
doi: 10.1097/MAT.0000000000000984


Cardiogenic shock is an often fatal condition with poor survival rates. Conventional treatment for cardiogenic shock consists of invasive monitoring, vasopressor agents, and intraaortic balloon pump support.1 There has been growing interest in the use of veno-arterial extracorporeal membrane oxygenation (VA-ECMO) as a form of salvage therapy for patients in cardiogenic shock refractory to conventional treatment. However, there is a low rate of survival to weaning or bridging therapy in this group of patients (42–56%).2,3,4,5 Other considerations include the high costs required for intensive care, and the high rate of complications such as bleeding (39%) and renal failure (31%)6 which develop while the patient is on VA-ECMO. Hence, it is useful to prognosticate which patients are likely to survive to weaning or bridging therapy. To this end, we analyzed our institution’s experience with patients in cardiogenic shock requiring VA-ECMO support over the past 15 years to determine whether any patient characteristics before the insertion of VA-ECMO, as well as complications sustained while on VA-ECMO, would be predictive for mortality while on VA-ECMO.

Materials and Methods

Study Population and ECMO Protocol

We conducted a retrospective study of patients who received VA-ECMO for cardiogenic shock between January 2003 and December 2017. We excluded patients who required VA-ECMO for cardiopulmonary resuscitation and patients who had VA-ECMO after cardiac surgery. A total of 127 patients aged 21–70 years old were included in our cohort.

In addition to providing ECMO support for patients that present to our hospital, our cardiothoracic unit also serves as a tertiary referral center for various nearby hospitals. Physicians in nearby hospitals consult our hospital when a patient may require ECMO support, and our ECMO team then proceeds on site to assess and institute ECMO support as required. Activation time is defined as the interval between ECMO activation and ECMO cannulation. Biochemical markers were measured within 2 hours before ECMO initiation.

VA-ECMO was considered whenever there was cardiogenic shock refractory to inotropes or intraaortic balloon pump. Cardiogenic shock was defined as having persistent hypotension with systolic blood pressure below 90 mmHg, on a background of acute or chronic heart failure.

Our adult ECMO circuit uses a centrifugal pump (Bio-Pump BP-80, Medtronic Bio-Medicus, Eden Prairie, MN, USA; SP45 Pump, Terumo Co., Tokyo, Japan; and Rotaflow Pump, Jostra, Maquet Cardiopulmonary AG, Hirrlinger, Germany) together with an oxygenator (SciMed Life Systems, Minneapolis, MN, USA; Emergency Bypass System, Terumo Co.; QuadroxD Polymethylpentene, Jostra, Medizintechnik AG). The ECMO tubing is coated [Bioline, Maquet Cardiopulmonary AG; phosphorylcholine (PC-coated) PH.i.S.i.O., Dideco s.r.l., Mirandola, Italy; Carmeda Bio-Active Surface, Minneapolis, MN, USA; and X-coating, Terumo Co. Tokyo, Japan]. Our femoral arterial cannulas are typically 17 Fr to 19 Fr (Bio-Medicus Cannula, Medtronic Inc.) and our femoral venous cannulas are 19 Fr to 21 Fr in size (Bio-Medicus Cannula, Medtronic Inc.). Cannulae size is selected based on the patient’s body surface area. All 127 patients received peripheral femoral artery or femoral vein cannulation via percutaneous Seldinger technique whenever possible, although an open cut-down was performed where required.

Once VA-ECMO was commenced, the pump flow was adjusted to achieve a cardiac index of at least 2.4/l/min2. After institution of ECMO, a continuous heparin infusion was started to maintain an activated clotting time between 180 and 200 seconds. A hemofiltration unit was connected to the VA-ECMO circuit if dialysis was indicated.

The decision to wean from VA-ECMO was guided by invasive hemodynamic monitoring and transthoracic echocardiography to determine if there was evidence of recovery of cardiac function. If a patient was deemed to have sufficiently recovered from cardiogenic shock, explant of the VA-ECMO was performed in the operating theater by gradual reduction of ECMO flows under transesophageal echocardiographic guidance. If the patient remained hemodynamically stable for at least 30 minutes with an ECMO pump flow of less than 0.5 l/min, decannulation was performed and the femoral artery was repaired. If the unilateral femoral vein had been cannulated, it was repaired as well. Otherwise, the venous cannula was removed from the femoral vein and manual compression was applied over the site of puncture. Patients who could not be weaned from ECMO were considered for left ventricular assist device (LVAD) implantation or heart transplant. Decision for LVAD was made by a multidisciplinary team and considerations include a left ventricular ejection fraction < 25%, progressive end-organ dysfunction due to hypoperfusion despite adequate ventricular filling pressure (pulmonary capillary wedge pressure ≥ 20 mmHg and systolic blood pressure ≤ 80–90 mmHg or cardiac index ≤ 2 L/min). Contraindications to LVAD include severe right ventricular dysfunction with severe tricuspid regurgitation, severe sepsis, severe renal, pulmonary or hepatic dysfunction, or poor neurologic status.

ECMO duration is defined as the duration between ECMO cannulation and ECMO explant. We measured four complications while on VA-ECMO—hemorrhage, stroke, requirement for dialysis, and limb ischemia. A hemorrhagic complication was defined as any bleeding requiring either medical therapy or blood transfusion, including epistaxis and gastrointestinal bleeding. Limb ischemia was diagnosed when there was pallor, pulselessness, or a cool limb on palpation. A distal perfusion cannula was inserted in the ipsilateral superficial femoral artery in all patients diagnosed with lower limb ischemia. Patients who were not diagnosed with lower limb ischemia did not receive a distal perfusion cannula. Clinical suspicion of stroke was confirmed with a computed tomography scan of the brain.

Data Collection

The following data were collected—general demographics, time from activation to insertion of VA-ECMO, duration of VA-ECMO support, indication for VA-ECMO, past medical history, laboratory parameters within 2 hours before VA-ECMO insertion, complications while on VA-ECMO, and survival to weaning or bridging to a ventricular assist device. All variables were compared between patients who died on ECMO versus patients who were able to wean off ECMO or bridge to other therapy.

Statistical Analysis

Statistical analysis was performed using SPSS for Windows (version 20.0, Spss Inc., Chicago, IL, USA). The Mann–Whitney U test was used for comparison of continuous variables in this dataset, and the χ2 or Fischer exact test was used for categorical variables. Statistical significance was established at alpha = 0.05. All factors with a p value of <0.05 were included for logistic regression. A multivariate forced entry logistic regression method was used to identify independent risk factors of mortality on VA-ECMO.


A total of 127 patients underwent VA-ECMO insertion for cardiogenic shock. Seventy-six (60%) had acute myocardial infarction, 34 (27%) had viral myocarditis, and the remaining 17 (14%) had other conditions including pulmonary embolism, dilated cardiomyopathy, acute mitral regurgitation, and thyrotoxicosis (Table 1). A χ2 test for independence indicated no significant association between etiology and survival to weaning or bridging therapy, x2 (2, n = 127) = 1.58, p = 0.455, phi = 0.111. Forty-four (35%) out of 127 patients died while on ECMO. Of the 83 (65%) patients that survived, nine (7%) were bridged to a LVAD, and 74 (58%) were weaned off ECMO (Figure 1). The type of pump used was not found to be significantly correlated with mortality (Table 1). There was also no significant difference in mortality between patients who had to undergo an ECMO circuit change versus those who did not require one (Table 1). The year of ECMO support was not found to be significantly related to mortality (p = 0.376).

Table 1
Table 1:
Univariate analysis comparing aetiology of cardiogenic shock, ECMO circuit and ECMO complications between survivors and non-survivors
Figure 1
Figure 1:
Outcome of patients grouped by etiology.

Table (Supplemental Digital Content 1, compares the baseline characteristics of the nonsurvivors and the patients who survived to weaning or bridging therapy. Only 51 (40%) of patients had ECMO instituted in our hospital, while the remaining 76 (60%) had ECMO instituted at a nearby hospital before being transferred to our hospital. Both the site of ECMO initiation (p = 0.656) and activation time (p = 0.691) were not significantly correlated with mortality on ECMO. Table 1 illustrates the frequency of complications in survivors and nonsurvivors.

Mann–Whitney U Test revealed the following to be associated with mortality: a shorter ECMO duration (U = 1,219, z = −3.083, p = 0.002, r = 0.273), a higher body mass index (BMI) (U = 589, z = −2.336, p = 0.019, r = 0.207), and a lower serum albumin(U = 923, z = −2.252, p = 0.024, r = 0.200). The r values suggest a small to medium effect for these factors.

χ2 test for independence (with Yates Continuity Correction) showed preimplantation serum creatinine > 100 mmol/l (x2 [1, n = 127] = 3.15, p = 0.047, phi = −0.179), the development of stroke (x2 [1, n = 127] = 7.92, p = 0.005, phi = −0.273), and limb ischemia (x2 [1, n = 127] = 6.25, p = 0.012, phi = 0.241) to be significantly associated with mortality. The phi values for preimplantation serum creatinine > 100 mmol/l, stroke, and limb ischemia suggest a small to medium effect using Cohen’s criteria.

There was no significant association between ECMO duration and the development of hemorrhage (p = 0.241), stroke (p = 0.680), limb ischemia (p = 0.051), or requirement for dialysis (p = 0.199) (Table 3).

Logistic regression was performed to assess the impact of ECMO duration, BMI, preimplantation serum creatinine > 100 mmol/l, preimplantation serum albumin, stroke, and limb ischemia on mortality (Table 2). Following univariate analysis, variables with p < 0.05 were included in multivariate analysis to obtain adjusted odds ratios and p values. Stroke was excluded from multivariate analysis because of its low incidence—only 18 patients developed stroke on ECMO. BMI was excluded from the model as well as it was not available for 33 out of 127 patients before ECMO was initiated. As shown in Table 2, only ECMO duration and preimplantation serum albumin were statistically significant after multivariate analysis.

Table 2
Table 2:
Logistic Regression Predicting Likelihood of Mortality While on ECMO
Table 3
Table 3:
Univariate analysis for ECMO duration and complications

Patients with lower preimplantation serum albumin and shorter ECMO duration were more likely to die. The adjusted odds ratio for preimplantation serum albumin was 1.08, indicating that for every 1 mmol decrease in serum albumin, patients are 1.08 times more likely to die (or 1.47 times more likely to die for every 5 mmol decrease in serum albumin). The adjusted odds ratio for ECMO duration was 1.15, indicating that for every 1 day less of ECMO support patients are 1.15 times more likely to die.


Although VA-ECMO has been used in patients with cardiogenic shock for more than 2 decades, survival to weaning or bridging therapy ranges from 42% to 56%.2,3,4,5 In our cohort, 65% of patients were successfully weaned off ECMO or bridged to a ventricular assist device. Although myocarditis has been reported to have a better prognosis by some authors,7 this was not reflected in our own experience. However, this may be due to the small number (34) of myocarditis cases in our cohort of patients. VA-ECMO instituted to support cardiopulmonary resuscitation for cardiac arrest is known to have a higher mortality rate,7 and as such we have excluded this group of patients from our analysis.

While cardiogenic shock stems from a decrease in contractility, hypoperfusion leading to multiorgan dysfunction syndrome ultimately impacts mortality.8 Our results suggest this as well—elevated creatinine, hypoalbuminemia, and the development of limb ischemia and stroke were all predictors of mortality. These findings suggest that apart from the restoration of circulation and blood pressure alone, careful attention to multiorgan dysfunction is important in optimizing survival in this group of patients. It is important to initiate ECMO before irreversible organ failure occurs. However, initiating ECMO too early may result in high healthcare costs and inappropriate resource consumption, whereas late ECMO initiation may result in a worse prognosis.8

Hypoalbuminemia may be seen in hepatic synthetic dysfunction, systemic inflammation, or malnutrition. Several studies in mechanical circulatory devices have found serum albumin to be a useful predictor of mortality and morbidity. Cotts et al. found hypoalbuminemia to be a significant predictor of hospital length of stay following LVAD implantation.9 Lietz et al. demonstrated that an albumin of 3.3 g/dl conferred an odds ratio of 3.8 (p < 0.001) for 90 day in-hospital mortality in patients receiving a pulsatile pump for destination therapy.10 Jellinge et al. studied 5,894 patients admitted acutely and multivariate logistic regression showed hypoalbuminemia to be increased with mortality (OR: 1.95 [95% CI: 1.31–2.90]). The exact mechanism with which hypoalbuminemia may cause an increased mortality is unclear.11

Duration of ECMO support and timing of weaning have been examined in other studies. Early withdrawal of ECMO support may result in mortality due to insufficient myocardial recovery, whereas a prolonged duration of ECMO support may result in an increased risk of ECMO-related complications. Other studies have found longer duration of ECMO support to be a predictor of in-hospital mortality.2,5 In our series, patients who died while on VA-ECMO had a shorter duration of ECMO support than those who survived to weaning or bridging therapy (median duration of support 4 days versus 7 days, p = 0.002). However, there was no significant association found between the development of complications and ECMO duration (Table 3). In particular, stroke and limb ischemia, which were significantly associated with mortality in our analysis, were not found to be significantly associated with ECMO duration.

To shorten the delay in instituting ECMO, our center has a mobile ECMO team that can institute ECMO off site. In our cohort, intrahospital referrals accounted for 51 (40%) of patients, whereas the remaining 76 (60%) of patients were from nearby hospitals. There was no significant difference in mortality between patients within our center and patients who had ECMO instituted off site. Median activation time was 49 minutes (10–92). The incidence of ECMO-associated complications was not significantly different in interhospital referrals versus intrahospital referrals (Table 4).

Table 4
Table 4:
Comparing Complications Between Intrahospital Referrals and Interhospital Referrals

The VA-ECMO complication rate reported by other studies is similar to those seen in our cohort. In a meta-analysis of 1,866 patients, 17% had lower extremity ischemia, 13% had neurologic complications, 46% required dialysis, and 41% had hemorrhagic complications.12 Another study of 23,951 patients reported the incidence of ischemic stroke and intracranial hemorrhage to be 8%.13 Our cohort of patients had a similar rate of complications—24% had lower limb ischemia, 14% had stroke, 42% required dialysis, and 40% had hemorrhagic complications. Stroke and limb ischemia were significantly associated with mortality. This finding is also supported by Lan and colleagues—multivariate analysis showed that the occurrence of stroke significantly increased the risk of death (OR: 4.94 [95% CI: 1.65–14.80]).14

More research should be focused on how to prevent neurologic complications such as ischemic stroke and intracranial hemorrhage, given its strong association with mortality. However, whether ECMO is a risk factor for neurologic complications, or if patients with these specific neurologic conditions are just more likely to require ECMO, is yet to be ascertained.13 Hesham et al. found that pre-ECMO lactic acid greater than 10 mmol/l independently predicts the occurrence of ischemic stroke during ECMO support. They suggest early and aggressive treatment of the causes of significantly elevated lactic acid, maintaining adequate tissue perfusion and trending lactic acid levels. Most of the ischemic stroke cases in this cohort also had activated partial thromboplastin time and activated clotting time within the therapeutic reference ranges, suggesting that the ischemic stroke probably occurred because of brain insults that are not due to embolic events from a thrombosed ECMO circuit.15

Acute kidney injury (AKI) is associated with mortality in patients receiving ECMO support. Sung et al. studied 322 adult patients receiving ECMO and found Kidney Disease Improving Global Outcomes stage 3 AKI to be associated with in-hospital mortality, with a hazard ratio of 2.690 (95% CI: 1.472 –4.915) compared with non-AKI (p = 0.001). Initial preimplantation serum creatinine was also a significant predictor of mortality.6,16,17

Acute kidney injury arises in patients with cardiogenic shock because of an interplay of decreased renal perfusion, exposure to nephrotoxic agents, as well as high intrathoracic pressures from mechanical ventilatory support. Acute kidney injury can lead to oliguria, exacerbating fluid overload, and worsening tissue oxygenation and oxygen transport further. This eventually results in multiorgan dysfunction involving the heart, brain, and lungs. Fluid overload also increases preload, worsening circulatory failure and this may further aggravate in-hospital mortality.18 Reperfusion injury after restoration of circulation with ECMO may worsen AKI.19 Univariate analysis of our cohort showed preimplantation serum creatinine > 100 mmol/l to be significantly associated with mortality. Given the above findings, clinicians should take into consideration the preimplantation serum creatinine when deciding whether to institute VA-ECMO.

In conclusion, preimplantation hypoalbuminemia, stroke, limb ischemia, preimplantation creatinine > 100 mmol/l, elevated BMI, and shorter ECMO duration are significantly associated with mortality while on VA-ECMO. These factors should be considered where instituting and withdrawing VA-ECMO support. Although multivariate analysis only showed ECMO duration and preimplantation hypoalbuminemia to be significantly associated with mortality, these results may be limited by the moderate number of cases.


The findings of this study are limited by its retrospective design and moderate number of cases. Cases are also reflective of a single center’s experience. Further multicenter studies involving a larger number of cases would be helpful in delineating prognostic factors in this group of patients.


We would like to thank the perfusion unit in National Heart Centre Singapore for their assistance with data collection.


1. Tarzia V, Bortolussi G, Bianco R, et al. Extracorporeal life support in cardiogenic shock: Impact of acute versus chronic etiology on outcome. J Thorac Cardiovasc Surg 2015.150: 333–340.
2. Lee SH, Chung CH, Lee JW, Jung SH, Choo SJ. Factors predicting early- and long-term survival in patients undergoing extracorporeal membrane oxygenation (ECMO). J Card Surg 2012.27: 255–263.
3. Schaheen BW, Thiele RH, Isbell JM. Extracorporeal life support for adult cardiopulmonary failure. Best Pract Res Clin Anaesthesiol 2015.29: 229–239.
4. Sakamoto S, Taniguchi N, Nakajima S, Takahashi A. Extracorporeal life support for cardiogenic shock or cardiac arrest due to acute coronary syndrome. Ann Thorac Surg 2012.94: 1–7.
5. Pappalardo F, Pieri M, Arnaez Corada B, et al. Timing and strategy for weaning from venoarterial ECMO are complex issues. J Cardiothorac Vasc Anesth 2015.29: 906–911.
6. Gray BW, Haft JW, Hirsch JC, Annich GM, Hirschl RB, Bartlett RH. Extracorporeal life support: experience with 2,000 patients. ASAIO J 2015.61: 2–7.
7. Combes A, Leprince P, Luyt CE, et al. Outcomes and long-term quality-of-life of patients supported by extracorporeal membrane oxygenation for refractory cardiogenic shock. Crit Care Med 2008.36: 1404–1411.
8. Schmidt M, Burrell A, Roberts L, et al. Predicting survival after ECMO for refractory cardiogenic shock: the survival after veno-arterial-ECMO (SAVE)-score. Eur Heart J 2015.36: 2246–2256.
9. Cotts WG, McGee EC Jr, Myers SL, et al. Predictors of hospital length of stay after implantation of a left ventricular assist device: an analysis of the INTERMACS registry. J Heart Lung Transplant 2014.33: 682–688.
10. Lietz K, Long JW, Kfoury AG, et al. Outcomes of left ventricular assist device implantation as destination therapy in the post-REMATCH era: implications for patient selection. Circulation 2007.116: 497–505.
11. Jellinge ME, Henriksen DP, Hallas P, Brabrand M. Hypoalbuminemia is a strong predictor of 30-day all-cause mortality in acutely admitted medical patients: a prospective, observational, cohort study. PLoS One 2014.9: e105983.
12. Cheng R, Hachamovitch R, Kittleson M, et al. Complications of extracorporeal membrane oxygenation for treatment of cardiogenic shock and cardiac arrest: a meta-analysis of 1,866 adult patients. Ann Thorac Surg 2014.97: 610–616.
13. Nasr DM, Rabinstein AA. Neurologic complications of extracorporeal membrane oxygenation. J Clin Neurol 2015.11: 383–389.
14. Lan C, Tsai PR, Chen YS, Ko WJ. Prognostic factors for adult patients receiving extracorporeal membrane oxygenation as mechanical circulatory support–a 14-year experience at a medical center. Artif Organs 2010.34: E59–E64.
15. Omar HR, Mirsaeidi M, Shumac J, Enten G, Mangar D, Camporesi EM. Incidence and predictors of ischemic cerebrovascular stroke among patients on extracorporeal membrane oxygenation support. J Crit Care 2016.32: 48–51.
16. Lee SW, Yu MY, Lee H, et al. Risk factors for acute kidney injury and in-hospital mortality in patients receiving extracorporeal membrane oxygenation. PLoS One 2015.10: e0140674.
17. Demondion P, Fournel L, Golmard JL, Niculescu M, Pavie A, Leprince P. Predictors of 30-day mortality and outcome in cases of myocardial infarction with cardiogenic shock treated by extracorporeal life support. Eur J Cardiothorac Surg 2014.45: 47–54.
18. Chen YC, Tsai FC, Fang JT, Yang CW. Acute kidney injury in adults receiving extracorporeal membrane oxygenation. J Formos Med Assoc 2014.113: 778–785.
19. Kilburn DJ, Shekar K, Fraser JF. The complex relationship of extracorporeal membrane oxygenation and acute kidney injury: causation or association? Biomed Res Int 2016.2016: 1094296.

Supplemental Digital Content

Copyright © 2020 by the American Society for Artificial Internal Organs