This study was designed to assess the interference by closed-chest cardiopulmonary resuscitation (CPR) on the ventricular fibrillation (VF) ECG signal in a porcine model of cardiac arrest and to elucidate which variable of VF spectral analysis reflects best myocardial blood flow and resuscitation success during CPR. Fourteen domestic pigs were allocated to receive either 0.4 U/kg vasopressin (n = 7) or 10 ml saline (n = 7) after 4 min of VF and 3 min of CPR. Using radiolabeled microspheres, myocardial blood flow was determined during CPR before, and 90 s and 5 min after, drug administration. Using spectral analysis of VF, the median frequency, dominant frequency, edge frequency and amplitude of VF were determined simultaneously and before the first defibrillation attempt. Using filters in order to specify frequency ranges, stepwise elimination of mechanical artifacts resulting from CPR revealed that at a frequency bandpass of 4.3-35 Hz, median fibrillation frequency has a sensitivity, specificity, positive and negative predictive value of 100% to differentiate between resuscitated and non-resuscitated animals. The best correlation between myocardial blood flow and fibrillation frequency was found at a median frequency range of 4.3-35 Hz. We conclude that spectral analysis of VF can provide reliable information relating to successful resuscitation. In this model after elimination of oscillations due to mechanical CPR, median fibrillation frequency best reflects the probability of resuscitation success.