A method using a parameter from the field of nonlinear dynamics to quantify the variability of ventricular premature complexes (VPCs) is presented. One hundred patients with coronary artery disease and > or = 10 VPCs/hour were included in the study. The RR intervals were plotted in a three-dimensional artificial phase space, and the structures in phase space were quantified by the local scaling indices, alpha. In the frequency distribution histogram, n(alpha), for each patient, the maximum of the ventricular ectopies alpha VPC, adjusted to the VPC frequency, was assessed; alpha VPC was used as the risk indicator. Endpoints were total mortality and sudden cardiac death. During follow-up (mean 3.1 years), 28 out of 100 patients died, 16 suddenly; alpha VPC had a significant prognostic impact and was independent from other risk indicators, such as left ventricular ejection fraction (LVEF). Patients who died during follow-up were characterized by a high alpha VPC. The optimal discrimination of high risk patients and low risk patients occurred at alpha VPC = 3.0. After 4 years, the survival rate of patients with a alpha VPC > 3.0 was 59%, in contrast to 97% in patients with alpha VPC < or = 0.3. As to the sudden death mortality, the survival rates were 74% and 97%, respectively. The difference between the groups were significant for both endpoints. Patients with an increased VPC variability (i.e., alpha VPC > 3.0) were at enhanced risk of sudden death and total mortality risk; alpha VPC was independent from other risk indicators such as the LVEF or heart rate variability parameters.