Study population
This single-center retrospective study included all patients treated with VA-ECMO for confirmed or suspected cardiogenic shock between 1 January 2014 and 31 December 2020 in the multipurpose intensive care unit (ICU) of Felix Guyon Hospital in Reunion Island. The diagnosis of cardiogenic shock could not be confirmed in patients who received VA-ECMO during an in-hospital cardiac arrest of suspected cardiac origin.
Patients were excluded if they met any of the following criteria: age under 18 years at VA-ECMO initiation, presence of refractory out-of-hospital cardiac arrest, and missing data on ICU mortality and/or time to death from initiation of VA-ECMO.
The study was approved by the Ethics Committee of the French Society of Anesthesia and Intensive Care Medicine (SFAR) (IRB 00010254-2022-096).
Indications for venoarterial extracorporeal membrane oxygenation
The decision to initiate VA-ECMO was taken during multidisciplinary consultations between intensivists, cardiac surgeons, and referring physicians. Patients were implanted only if a feasible therapeutic strategy was available: i.e., VA-ECMO as a bridge to recovery, long-term left ventricular assistance, or heart transplantation. Note that the initiation of VA-ECMO as a bridge to long-term left ventricular assistance or heart transplantation required transfer to a specialized center in metropolitan France.
Data Collection
Data were collected retrospectively by reviewing patients’ medical records. All data were anonymized before statistical analysis.
The following data were collected:
Patient characteristics at ICU admission:
- Demographic characteristics: age, sex.
- Body mass index.
- Scores: simplified acute physiology score II (SAPS II), survival after VA-ECMO (SAVE) score, modified SAVE score.
- Comorbidities: chronic heart disease, chronic lung disease, chronic renal failure, occlusive arterial disease of the lower limbs, arterial hypertension, diabetes, smoking, chronic alcoholic use, cancer, immunosuppression.
Patient characteristics at VA-ECMO initiation:
- Diagnoses associated with cardiogenic shock: acute myocardial infarction, decompensation of cardiomyopathy, myocarditis, cardiotropic poisoning, pulmonary arterial hypertension, infective endocarditis, post-cardiotomy cardiogenic shock, heart rhythm disorders.
- Invasive mechanical ventilation.
- Vasoactive inotropic scores: dobutamine levels, adrenaline levels, noradrenaline levels.
- Prothrombin time.
- Creatinine levels.
- Hemoglobin levels.
- Platelet count.
- Lactate levels.
- Renal replacement therapy.
- Cardiac arrest.
VA-ECMO-related complications:
- Neurological complications: acute ischemic stroke, hemorrhagic stroke.
- Hemorrhagic complications, defined by the need for surgical exploration and/or immediate transfusion of erythrocyte concentrates.
- Circuit thrombosis/membrane exchange.
- Pneumonia, bacteremia, fungemia, or cannula-related infections occurring at least 48 hours after implantation and up to 72 hours after explantation. In the case of canula-related infections, the time frame was extended to 30 days after explantation, in accordance with the definition of surgical site infection (11,12).
- Acute limb ischemia.
- Acute pulmonary edema.
- Acute mesenteric ischemia.
Patient outcome:
- Duration of mechanical ventilation.
- Duration of VA-ECMO support.
- Length of stay in ICU.
- Death in ICU.
Venoarterial extracorporeal membrane oxygenation initiation, management and weaning
All patients were implanted with a drainage cannula and a return cannula. A limb reperfusion cannula was used in cases of peripheral femoral cannulation. The VA-ECMO was used with a centrifugal pump [Rotaflow® (Maquet/Getinge)], an oxygenator, and a heat exchanger. Some patients were also supported by an intra-aortic balloon pump.
Implantation was performed by a cardiac surgeon using surgery or percutaneous puncture. Management of VA-ECMO was provided by the intensivist in charge of the patient, in daily consultation with a cardiac surgeon. Blood and gas flows were adjusted based on clinical and biological parameters and hemodynamic monitoring.
The decision to wean patients from VA-ECMO was made in the presence of favorable hemodynamic parameters. Weaning was achieved with progressive filling of the heart through a reduction in pump flows (13,14).
Study endpoints
The primary endpoint was death within 72 hours of VA-ECMO implantation. Patients who died within that time frame formed the early mortality group, while those who lived beyond it formed the survival group. A cut-off time of 72 hours was selected because this is the time needed to initiate resuscitative measures and to assess neurological prognosis in patients with cardiac arrest (15).
Statistical analyses
Pre-implantation characteristics were compared between patients who died within 72 hours of VA-ECMO initiation (early mortality group) and those who lived beyond that time (survival group). Categorical variables were compared using the Chi-square test or Fisher’s exact test, as appropriate. Continuous variables were compared using the nonparametric Mann-Whitney test. Results were expressed as number (%) for categorical variables and as median [25th-75th percentiles] for continuous variables. The significance level was set at p < 0.05.
Factors significantly associated with early mortality in univariate analysis were entered into the multivariate logistic regression model. In cases of collinearity, only the most clinically relevant factors were entered in the model. Model calibration was assessed using the Hosmer-Lemeshow goodness-of-fit test. Statistical analyses were performed using SAS statistical software (8.2, Cary, NC, USA) and the pvalue.io interface of R software (16).
Score construction
Regression coefficients generated by the logistic regression model were multiplied by 10 and subsequently converted to the nearest integer. The number of points per coefficient was attributed after identifying a common denominator across all coefficients. The score ranging from 0 to 19 was the sum of the points corresponding to each variable. Score performance was evaluated by bootstrapping (2000 iterations) using the area under the receiver operating characteristic curve (AUC). The score, named “IMPACT” (predIction of early Mortality associated with VA-ECMO using Pre-implAntation CharacTeristics), was developed, validated, and reported following the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) recommendations (17). SAS statistical software (8.2, Cary, NC, USA) was used for score construction.