Several formulas exist for estimating left ventricular volumes and ejection fraction using conventional two-dimensional echocardiography from transthoracic views. Transesophageal imaging provides superior resolution of endocardial borders but employs slightly different scan planes. The estimation of left ventricular volumes by transesophageal echocardiography has not been validated in human patients. Therefore, the purpose of this study was to compare left ventricular volumes and ejection fraction derived from transesophageal short-axis and four-chamber images with similar variables obtained from ventriculography. End-diastolic and end-systolic volumes and ejection fraction were calculated using modified Simpson's rule, area-length and diameter-length models in 36 patients undergoing left ventriculography. Measurements of left ventricular length were obtained from the transesophageal four-chamber view and areas and diameters were taken from short-axis scans at the mitral valve, papillary muscle and apex levels. Data from transesophageal echocardiographic calculations were compared with end-diastolic volume (mean 172 +/- 90 ml), end-systolic volume (mean 91 +/- 74 ml) and ejection fraction (mean 52 +/- 15%) from cineventriculography using linear regression analysis. The area-length method (r = 0.88) resulted in a slightly better correlation with left ventricular end-diastolic volume than did Simpson's rule (r = 0.85) or area-length (r = 0.84) formulas. For end-systolic volume, the three models yielded similar correlations: Simpson's rule (r = 0.94), area-length (r = 0.93) and diameter-length (r = 0.95). Each of the methods resulted in significant underestimation of diastolic and systolic volumes compared with values assessed with angiography (p less than 0.003). Ejection fraction was best predicted by using the Simpson's rule formula (r = 0.85) in comparison with area-length (r = 0.80) or diameter-length (r = 0.73) formulas. Measurements of left ventricular length by transesophageal echocardiography were smaller for systole (mean 5.7 +/- 1.6 cm) and diastole (mean 7.7 +/- 1.2 cm) than values by ventriculography (mean 9.2 +/- 1.4 and 8.1 +/- 1.6 cm, respectively; p less than 0.0001), suggesting that underestimation of the ventricular length is a major factor contributing to the smaller volumes obtained by transesophageal echocardiography. In conclusion, currently existing formulas can be applied to transesophageal images for predicting left ventricular volumes and ejection fraction. However, volumes obtained by these models are significantly smaller than those obtained with angiography, possibly because of foreshortening in the transesophageal four-chamber view.