Background: There is uncertainty about best selection criteria for extracorporeal cardiopulmonary resuscitation (eCPR) in the setting of refractory cardiac arrest. We aimed to identify factors associated with a favorable neurological outcome, and to build a score calculable at the time of ECMO insertion predicting the prognosis. Methods: Retrospective analysis of all patients who underwent eCPR between 2010 and 2017 in a single university hospital. Primary end point was survival with favorable neurological outcome at intensive care unit (ICU) discharge defined as a Cerebral Performance Category of 1 or 2.Results: Overall low-flow time of the 113 included patients was 84 minutes. Eighteen patients (16%) survived with a favorable neurological outcome. By multivariate logistic regression analysis, initial shockable rhythm, and arterial blood pH at the time of eCPR implantation ≥7.0, were independent predictors of survival with favorable neurological outcome. All of the patients presenting with both nonshockable rhythm and pH <7.0 at the time of eCPR implantation died in the ICU. Conclusions: At the time of eCPR start, only initial shockable rhythm and arterial pH ≥7.0 predicted neurological outcome. A selection of the patients who might benefit from eCPR, based upon initial rhythm and arterial pH rather than on low flow time, should be further evaluated.