Study design
We conducted a retrospective cohort study between January 2012 and December 2016 at Samsung Medical Center (a 1979-bed tertiary referral hospital with tertiary-level intensive care units) in Seoul, South Korea. All patients 18 years of age or older for whom ECMO support was required for severe acute respiratory failure were enrolled in the study. A total of 136 ECMO runs in 133 patients were identified during this period. Twenty patients who were transported to our facility after initiation of ECMO in other hospitals were excluded because the decision regarding whether or not the patient was a suitable candidate for ECMO and initial management were not made by our ECMO team. The remaining 116 eligible ECMO runs were divided into the pre-ECMO team period (before January 2014, n = 70) and the post-ECMO team period (after January 2014, n = 46), according to the date of ECMO initiation (Fig. 1).
The institutional review board of the Samsung Medical Center approved this study and waived the requirement for informed consent because of the observational nature of the study. In addition, patients’ information was anonymized and deidentified prior to analysis.
ECMO team and management of ECMO
In our institution, ECMO support has been available since 2004. In the first few years, veno-arterial ECMO was primarily used in patients with cardiac failure, with veno-venous ECMO used for less than five cases per year. The incidence of ECMO runs has gradually increased, with more than 100 cases currently performed annually. The application of veno-venous ECMO for severe respiratory failure has also grown and currently represents up to 20–30% of all ECMO runs.
Before 2014, there were standard criteria for indications and contraindications of ECMO (Additional file 1), but decisions about initiation and decannulation were mostly left to the physicians that oversaw patients. Cannula- or circuit-related issues were treated through elective consultation with cardiothoracic surgeons who had experience with ECMO. In 2014, our hospital adopted a multidisciplinary ECMO team consisting of cardiovascular surgeons, cardiologists, critical care physicians, and an ECMO specialist who is a cardiovascular perfusionist trained to manage the ECMO system and clinical needs of the patients on ECMO under supervision of ECMO-trained physicians. This ECMO team was responsible for every issue related to ECMO in the hospital. Protocols for indications and contraindications, management of patients and equipment, and weaning of patients from ECMO were revised. In addition, the ECMO team was charged with educating all medical personnel including bedside nurses caring for patients on ECMO at our institution. When a patient was deemed eligible for ECMO, the final decision to initiate ECMO was made by the treating intensivist and ECMO team, consisting of two more critical care physicians who are board certified in pulmonary and critical care medicine and cardiovascular surgeon, after a comprehensive assessment based on our protocol outlining the indications and contraindications. The primary cannulation strategy for adult respiratory ECMO was the veno-venous mode. The veno–veno-arterial mode was considered if the patient needed additional support due to hemodynamic failure. Cannulation was performed by the attending cardiothoracic surgeons using either the percutaneous method with the Seldinger technique or the surgical method at the bedside or in the operating room. Cannulation sites and cannula sizes were selected at the discretion of the cardiothoracic surgeons. Usually, a 20–28-Fr cannula was used for venous drainage via the common femoral vein, and a 14–18- or 20–24-Fr cannula was used for venous return via the internal jugular or the common femoral vein, respectively. The Prolonged Life Support System (Quadrox PLS, Maquet Inc., Rastatt, Germany) and the Capiox Emergency Bypass System (Capiox EBS; Terumo, Inc., Tokyo, Japan) were used. Pump blood flow and sweep gas flow rates were adjusted to maintain a target oxygen saturation and carbon dioxide removal rate. The mechanical ventilation (MV) strategy during ECMO was adapted from the study protocol of the CESAR trial [1], providing assisted pressure-controlled ventilation while limiting the peak inspiratory pressure to 25 cmH2O and applying positive end-expiratory airway pressure of 10 cmH2O, and respiratory rate of 10 breaths/min, on inspired oxygen fraction of 30%. Once the patients were stabilized and lightly sedated, spontaneous ventilation with pressure support mode was considered. In all patients, arterial catheterization was performed for continuous hemodynamic monitoring. Our ECMO team performed daily rounds and assessed the state of the ECMO circuit, development of ECMO-associated complications, and the possibility of weaning. An ECMO-trained physician provided 24-h on-call coverage, and an ECMO specialist participated in all intra-hospital transport of patients on ECMO. If a patient was considered ready to be weaned off ECMO support, decisions regarding decannulation were assessed through a protocolized weaning trial. Cannulae were removed at the bedside by cardiothoracic surgeons.
Data collection and clinical outcomes
The clinical and laboratory data from patients who were treated with ECMO have been prospectively registered in the ECMO database of our hospital since 2012. For this study, these data were supplemented with a retrospective review of all hospital medical records. Demographic data, including age, sex, comorbidity, immune state, history of cardiac arrest, diagnosis, acute physiology and chronic health evaluation (APACHE) II score, and sequential organ failure assessment (SOFA) score, were recorded at admission to the intensive care unit (ICU). Presence of an artificial airway, use of MV, ventilator setting immediately before ECMO initiation, use of rescue and adjunctive treatment before ECMO, worst values from arterial blood gas and lactate tests within 6 h before ECMO initiation, respiratory extracorporeal membrane oxygenation survival prediction (RESP) score [11], predicting death for severe ARDS on VV-ECMO (PRESERVE) score [10], ECMO mode, and cannulation site were recorded on the first day of ECMO support.
The primary outcome in this study was in-hospital mortality. Secondary outcomes were ICU mortality, rate of weaning from ECMO, duration of ECMO support, adverse events during ECMO, rate of weaning from MV, duration of MV before weaning, ICU and hospital lengths of stay, and 1-year mortality after ECMO initiation. Adverse events during ECMO were defined as follows: cannula-related (vessel perforation with hemorrhage, arterial cannulation, malposition requiring repositioning, or accidental decannulation), other ECMO-related, hematological (gastrointestinal bleeding, cannula site bleeding, surgical site bleeding, plasma hemoglobin level > 50 mg/dL, or disseminated intravascular coagulation), neurological (brain death, seizure, cerebral infarction, or brain hemorrhage), cardiovascular (inotrope or vasopressor use, myocardial stunning, arrhythmia, cardiac tamponade, or cardiac arrest), pulmonary (pneumothorax or pulmonary hemorrhage), renal (serum creatinine level > 1.5 mg/dL or continuous renal replacement therapy), and infection (white blood cell count < 1500 × 103 cells/mm3, culture-confirmed new infection, or ECMO-associated wound infection). Clinical outcomes were identified through medical record review. The Korean National Database, which uses citizen registration numbers, was used to obtain information about patient death at 1 year after ECMO initiation.
Statistical analysis
To compare characteristics and clinical outcomes between the two periods, we analyzed categorical variables using χ2 tests or Fisher’s exact tests, when applicable, with data presented as numbers and percentages. Continuous variables were presented as medians with interquartile ranges, and Mann–Whitney U test was used for analysis. To adjust for potential confounding factors in the association between implementation of a multidisciplinary ECMO team and in-hospital mortality, multiple logistic regression analysis was used. Data are presented as adjusted odds ratios (ORs) with 95% confidence intervals (CI). Survival curves were constructed using the Kaplan–Meier method and compared with a log-rank test. For all analyses, a two-tailed test with a P value less than 0.05 was considered statistically significant. Statistical analyses were performed using SPSS 18.0 for Windows (Chicago, IL, USA).