We sought to compare the reliability and prognostic implications of left ventricular (LV) ejection fraction (EF) measurements obtained in routine clinical practice. We retrospectively selected from our clinical database a group of 422 patients with known or suspected ischemic heart disease, studied by two-dimensional echocardiography, gated single-photon emission computed tomography (SPECT) and left ventriculography (LVG) for clinical purposes. In each diagnostic procedure LVEF was measured as done routinely. The LVEF values obtained by the three methods were similar and closely related. The correlation coefficient r was equal to 0.83 between echocardiographic and LVG, to 0.75 between gated SPECT and LVG and to 0.81 between echocardiographic and SPECT. During follow-up (median 41 months), 31 patients died. The values of LVEF obtained by echocardiography, gated-SPECT and LVG were all powerful predictors of all-cause mortality: chi(2) = 12.3 for echocardiography, 14.4 for gated SPECT and 14.5 for LVG. However, including LVEF values into a model based on patient age, sex, history of angina, evidence of previous infarction and number of stenotic coronary arteries, the ability to predict patient survival significantly increased only including LVEF values measured by gated SPECT (chi(2) = 40.6, P = 0.039). Thus, in a large cohort of unselected patients with known or suspected ischemic heart disease, the values of LVEF routinely measured by echocardiography, gated SPECT and LVG were closely correlated, and provided a powerful prognostic information, that was incremental to clinical variables for gated SPECT.