OBJECTIVE--The SAVE study showed that captopril improves mortality in patients with left ventricular dysfunction after myocardial infarction and that this benefit occurred even in patients with no clinically overt heart failure. On the basis of this, it seems important to identify correctly which patients have left ventricular dysfunction after a myocardial infarction. The objective was to compare various methods of identifying patients with left ventricular dysfunction (left ventricular ejection fraction, LVEF, < or = 40%) after acute myocardial infarction. The methods compared were echocardiography (quantitative and qualitative visual assessment), clinical evaluation (subjective assessment and three clinical score methods), and measurement of plasma concentrations of cardiac natriuretic peptide hormones (atrial and brain natriuretic peptides, ANP and BNP). DESIGN--Cross sectional study of left ventricular function in patients two to eight days after acute myocardial infarction. SETTING--Coronary care unit of a teaching hospital. PATIENTS--75 survivors of a recent myocardial infarction aged 40 to 88 with no history of cardiac failure and without cardiogenic shock at the time of entry to the study. MAIN OUTCOME MEASURES--Sensitivities and specificities of the various methods of detecting left ventricular dysfunction were calculated by comparing them with a cross sectional echocardiographic algorithm for LVEF. RESULTS--Clinical impression was poor at identifying LVEF < 40% (sensitivity 46%). Clinical scoring improved this figure somewhat (modified Peel index sensitivity 64%). Qualitative visual assessment echocardiography was a more sensitive method (sensitivity 82%) for detecting LVEF < 40%. Plasma BNP concentration was also a sensitive measure for detecting left ventricular dysfunction (sensitivity 84%) but plasma ANP concentration was much poorer (sensitivity 64%). CONCLUSION--Left ventricular dysfunction is easily and reliably detected by echocardiographic measurement of LVEF and also by a quick qualitative echocardiographic assessment but is likely to be missed by clinical assessment alone. High concentrations of plasma BNP maybe another useful indicator of left ventricular dysfunction, particularly in hospitals where not all patients can be screened by echocardiography or radionuclide ventriculography after myocardial infarction.