OBJECTIVE--To evaluate the power of measurements of left ventricular size and function for predicting long term (82 month) mortality by performing echocardiography in 97 men who had survived an acute myocardial infarction. SETTING--University hospital specialising in cardiology. PARTICIPANTS--97 consecutive male patients who had survived a myocardial infarction. MAIN OUTCOME MEASURES--The additive prognostic value of functional measurements to that provided by primary risk factors (smoking habits and lipoprotein levels), radiological heart size, exercise capacity, and number of major coronary arteries with haemodynamically significant stenoses was evaluated. An echo index was calculated from three echocardiographic variables (yielding one score point each if: left ventricular diameter at the end of diastole (LVDD) > or = 5.7 cm, left ventricular fractional shortening < or = 24%, and E point-separation (EPSS) > or = 10 mm). MAIN OUTCOME--17 cardiac deaths occurred during follow up. RESULTS--Univariate analysis showed that treatment with loop diuretics for heart failure (P < 0.01), LVDD (P < 0.01), left ventricular diameter at the end of systole (LVDS) (P < 0.001), left atrial diameter (P < 0.001), fractional shortening (P < 0.05), and echo index (P < 0.001) were all associated with cardiac death. Angiographically determined regional wall motion disturbances (P < 0.005) and angiographic ejection fraction (P < 0.001) were also associated with cardiac death, as was the number of major coronary arteries with significant stenosis (P < 0.05). When all significant echocardiographic variables from univariate analysis were entered into Cox proportional hazards survival analysis, LVDS and left atrial diameter contributed independently to the prediction of cardiac death. If angiographic data were also entered into the model, the echo index made an independent contribution to the prediction of cardiac death. CONCLUSIONS--Among young male patients with a previous myocardial infarction, a simple M mode echocardiographic examination can identify high and low risk patients and improve the prediction of cardiac death made from clinical information, exercise test, chest x ray and angiographically determined ejection fraction.