This retrospective study was conducted in a tertiary hospital in Taipei City, Taiwan, with an emergency department with more than 1,15,000 visits annually and 220 intensive care unit (ICU) beds.

This study included patients with OHCA aged ≥ 18 years who were treated with ECPR between January 2012 and December 2020. Patients were excluded for the following reasons: 1) ECPR was initiated at another hospital and the patient was transferred after ECMO, 2) traumatic OHCA, and 3) OHCA with sustained ROSC (ROSC for more than 20 min), but ECMO was warranted for haemodynamic support.

The prehospital EMS system in the OHCA setting was reported in a previous study [17]. There were basic life support teams with defibrillation ability and advanced life support teams, which are qualified for intubation and intravenous epinephrine injection in the prehospital setting.

Advanced airway was established for 60 percent of patients. Chest compressions were performed with mechanical CPR devices for 85 percent of patients during transport, unless not feasible because of body size or other reasons. The dispatch center would inform the emergency department by telephone of the incoming OHCA patient information, including patient`s age, gender, prehospital DC shock times, prehospital intervention, and estimated arrival time.

If OHCA patients did not achieve ROSC after 10 min of standard Advanced Cardiac Life Support (ACLS), the emergency physician would discuss with the duty cardiovascular surgeon for ECPR eligibility. Patients were considered eligible for ECPR if they met all the following criteria: (1) age < 80 years, (2) witnessed collapse with no-flow time < 10 min, (3) pre-disease cerebral performance category (CPC) of 1–2 and no terminal malignancy, (if CPC category was not available, the pre-disease neurological status, as Glascow Coma Scale, would be used instead. All the eligible patients must have a GCS score of 15) and (4) presumed reversible cause (e.g. acute coronary syndrome or pulmonary embolism). The cardiovascular surgeon made the final decision on ECMO eligibility. The emergency department had a routine team structure while managing patients with OHCA. One nurse focused on blood sampling through puncturing femoral artery and the other nurse obtained venous access and epinephrine injection. The blood gas analysis was performed right after blood sampling and sent for point-of-care analysis using SIEMEMS RAPIDPoint 500 Systems in the resuscitation room. Arterial pH, PaCO2, and lactic acid levels were interpreted to guide further resuscitation.

The ECMO cannulation approach in our hospital is peripheral cannulation with open technique. Cannulation was performed under direct vision of femoral vessels with modified Seldinger method. All the cannulations were performed in the emergency department by the duty cardiovascular surgeon with the assistance of ECMO technicians. The ECMO components included a centrifugal pump and oxygenator (Medtronic, Anaheim, CA; Medos, Stolberg, Germany; Maquet, Rastatt, Germany). After ECMO, the patient underwent computed tomography to survey for possible OHCA causes, and the cardiologist evaluated the feasibility of coronary angiography. The patient was admitted to the ICU for post-resuscitation care.

Baseline characteristics and comorbidities were recorded in the medical records and retrospectively collected. Resuscitation variables were collected from the emergency medical service and hospital records. All time intervals were retrospectively calculated from the hospital records. Arrest-hospital time was defined as the time interval between cardiac arrest (CA) and hospital arrival. The arrest-ECMO time was defined as the time from CA to ECMO implementation. Arterial pH, PaCO2, and lactic acid levels were recorded from the first blood sample at the emergency department. Data on the intervention after ECPR, survival, and neurological outcomes at discharge were collected from the medical records.

The primary outcome was a favorable neurological outcome at discharge, defined as a Cerebral Performance Category score of 1 (good cerebral performance) or 2 (moderate cerebral disability). The secondary outcome was survival until hospital discharge.

Categorical variables are expressed as percentages and were compared using the chi-squared test. Continuous variables are expressed as means ± standard deviations, and t-tests were used to delineate differences. Statistical significance was set at p-value < 0.05. The predictive abilities of metabolic parameters (pH, PaCO2, and lactic acid) were tested using the area under the receiver operating characteristic (ROC) curve (AUC). We used a generalised additive model (GAM) to determine the relationship between PaCO2, FO, and survival. All variables with a p-value < 0.15 were included in multivariable logistic regression to determine the independent variables for predicting FO. All the time variables were treated as continuous variables in the multivariable regression model. A stepwise backward elimination method was used to select the final regression model. The selected model was then compared with current prognostic scores (including the TIPS 65 score [18], OHCA score [19], TTM score [20], and rCAST score [21, 22]). Subgroup analyses were performed to test the discriminative ability of PaCO2 in different subgroups: initial shockable cardiac rhythm versus non-shockable rhythm, hospital arrival time < 25 min versus > 25 min, and arrest-ECMO time < 60 min versus > 60 min. The p-value for the interaction was tested using an interaction test. All computations were performed using SPSS, version 16.0 (IBM Corp., Armonk, NY, USA).