Heart failure is one of the most important public health problems in western countries because of its frequent association with cardiac death and with rehospitalization. Patients with heart failure generally die of sudden arrhythmic death (SD) and progressive pump failure with a SD incidence inversely related to the severity of the underlying heart disease. SD occurs approximately in one half of the patients in New York Heart Association (NYHA) functional classes II-II/III and in one third of those in NYHA classes III/IV-IV, respectively. In the last decade, numerous studies have shown that the automatic implantable cardioverter-defibrillator (ICD) significantly reduces the incidence of SD in those patients who are identified as being at risk. Unfortunately, little is known on whether the ICD actually prolongs the survival of the subgroup of patients with most severe heart failure. The main reasons for such a paucity of information are the small number of available studies, the unavailability of randomized and controlled investigations and the difficulty in comparing the results of such studies owing to the lack of homogeneity. In patients with severe heart failure, both the perioperative mortality and morbidity related to transvenous ICD implantation are similar to those of patient subgroups with moderate or slight heart failure. The defibrillation threshold at implantation and the frequency of intractable ventricular arrhythmias during follow-up (2% of all ICD implantations) are slightly higher than in the patient subgroup with moderate or no heart failure. In the patient subgroup with severe heart failure or with a very poor left ventricular function, ICD implantation is unable to prolong the 1- and 2-year survival despite a clear reduction in the incidence of SD. On the other hand, in the patient subgroup with moderate left ventricular dysfunction, ICD implantation prolongs survival and reduces the incidence of SD. No information is available regarding the primary prevention of SD in patients with heart failure.