BACKGROUND: Cardiopulmonary resuscitation of cardiac arrest has poor outcomes. Extracorporeal life support (ECLS) could represent a salvage option. This study aimed to analyze the outcomes of ECLS used for refractory cardiac arrest. METHODS: In this observational analysis, patients were divided into an in-hospital cardiac arrest group (IHCA) and an out-of-hospital (OHCA) cardiac arrest group. The primary end point was survival to hospital discharge with good neurologic outcome. Both groups were compared after propensity score matching. Risk factors were searched with multivariate analyses. RESULTS: From January 2007 to December 2016, study investigators performed 131 ECLS procedures (IHCA, n = 45, 34.4%; OHCA, n = 86, 65.6%). The mean age of patients was 43.2 years, and 71.8% were male. Baseline characteristics were comparable between both groups except mean no-flow duration (0.2 minutes vs 2.5 minutes; p \textless 0.001) and low-flow duration (46.9 minutes vs 85.3 minutes; p \textless 0.001), which were significantly shorter in the IHCA group. A total of 103 (82.4%) patients died during ECLS (IHCA, 79.1% vs OHCA, 84.1%; p = 0.479). The complication rate during ECLS was comparable between both groups. Twenty (16%) patients were successfully weaned from ECLS (IHCA, 18.6% vs OHCA, 14.6%; p = 0.565) after a mean support period of 6.7 days. Survival to hospital discharge with good neurologic outcome was not different between the two matched groups (odds ratioOHCA vs IHCA 1.3; 95% confidence interval, 0.023 to 74.902; p = 0.9). Presence of shockable rhythm was associated with a better outcome (odds ratioshockable vs nonshockable 6.674; 95% confidence interval, 1.078 to 41.336; p = 0.044). CONCLUSIONS: Patients in the IHCA and OHCA groups experienced the same survival with good neurologic outcome after ECLS support. A better selection of patients with IHCA is mandatory to avoid futile support.