Gated SPECT (GSPECT) was evaluated for the measurement of left ventricular ejection fraction (LVEF) by comparing with equilibrium gated radionuclide ventriculography (RNVG). A total of 99 subjects underwent GSPECT and RNVG imaging. All studies were acquired in list mode with GSPECT studies processed to give 16- and 8-frames per R-R interval, and RNVG studies 24 frames per R-R interval. The Cedars-Sinai QGS software was used to calculate ejection fraction from GSPECT studies. RNVG studies were processed using a manually drawn single region of interest technique. Comparison of LVEF from GSPECT with RNVG yielded correlation coefficients of 0.82 and 0.81 for 16- and 8-frame GSPECT studies respectively. The mean 95% prediction interval was 33 +/- 11 percentage points for both 16- and 8-frame studies, indicating a great disparity between predicted ejection fraction values from GSPECT and actual RNVG values. Subgroup analysis of 29 patients with pathological Q-wave evidence of myocardial infarction demonstrated a poorer correlation coefficient of r = 0.69. Subgroup analysis of 32 patients with end-diastolic volumes < 100 ml demonstrated a poorer correlation coefficient of r = 0.32. Ejection fractions calculated from 16- and 8-frame studies showed a correlation of 0.99 with a mean 95% prediction interval of 8.7 +/- 0.04 percentage points. The 8-frame studies underestimated LVEF by 3.6 +/- 2.3% compared to the 16-frame studies. In conclusion, left ventricular ejection fractions calculated using the QGS algorithm from GSPECT studies are inadequate for use in clinical practice.