Most sudden cardiac deaths are caused by fatal ventricular arrhythmias (ventricular tachycardia [VT] and fibrillation) in patients with and without known structural heart diseases. Given the large number of patients potentially at risk for developing ventricular arrhythmias, any strategy for treating them prophylactically requires efficient and effective risk stratification. Both non-invasive and invasive testing may be used for prognostic evaluation of patients with heart diseases. The optimal way to use them in the risk stratification for sudden cardiac death will depend in part on the goals of screening. At present risk markers perform better at identifying low-risk patients who may not need an implantable cardioverter-defibrillator (ICD), because all tests have a high negative predictive accuracy. In our opinion an electrophysiological test should not be performed and an ICD should not be implanted in post-myocardial infarction patients with moderate left ventricular dysfunction (left ventricular ejection fraction 30-40%) with a preserved autonomic balance and without non-sustained VT. In MADIT II-like patients electrophysiological testing does not seem necessary and an ICD could not be implanted only in patients with a negative T-wave alternans test. Most of the data available refer to patients with ischemic cardiomyopathy but the preliminary data on T-wave alternans suggest its usefulness in patients with non-ischemic cardiomyopathy too, although a large definitive study has not yet been completed in this important population.