To determine whether sudden versus non-sudden cardiac death could be predicted in high risk patients, 1157 medical patients were followed for an average of 46 months after a diagnostic coronary angiogram and 18 clinical, hemodynamic, and angiographic variables known to be associated with a high risk of mortality were analyzed. The total group of 141 deaths was classified into 3 subgroups: (1) 82 sudden deaths (less than 1 hour after onset of symptoms); (2) 46 deaths due to acute myocardial infarction with or without heart failure, and (3) 13 deaths unrelated to cardiac symptoms. In a subset of 64 patients, the duration of electrical systole (QTc) was calculated before angiography and before death. A comparison was made of QTc measurements at entry with QTc values of subjects with normal coronary arteries and normal left ventricular function. Deaths from cardiac causes could often be predicted from older age, male sex, history of myocardial infarction, unstable angina, congestive heart failure, abnormal cardiothoracic ratio, multivessel disease, abnormal left ventricular contraction, and abnormal ejection fraction. However, these variables did not discriminate between sudden and nonsudden cardiac deaths and both modes of death were characterized by depressed left ventricular function and multivessel coronary disease. During follow-up the incidence of acute myocardial infarction was not different in patients with cardiac and noncardiac deaths and in long-term survivors. However, patients dying from cardiac causes had a higher incidence of heart failure. Patients dying suddenly did not present new infarctions during follow-up whereas patients dying from acute myocardial infarction had a 13% incidence of prior infarction and a higher incidence of heart failure. In addition, QTc at entry was longer in nonsurvivors than in normal subjects (p less than 0.0001) and patients experiencing sudden death exhibited the highest incidence of QTc prolongation (greater than or equal to 440 ms) during follow-up (p less than 0.05). We conclude that: (1) although the severity of coronary disease and left ventricular dysfunction are closely related to cardiac mortality, they do not discriminate between sudden and nonsudden cardiac deaths; (2) patients experiencing sudden death are characterized by a low incidence of new myocardial infarction or congestive heart failure and prolongation of the QTc interval during follow-up.