We performed a prospective anatomic validation study to determine the accuracy of left ventricular (LV) mass estimates from clinical two-dimensional echocardiographic (2-D echo) studies. In 21 subjects, antemortem 2-D echo LV mass determinations were compared with anatomic LV weight by postmortem chamber dissection. Major cardiac diagnoses included anatomic LV aneurysm in four, status post aneurysmectomy in one, transmural myocardial infarction in seven, congestive cardiomyopathy in five, rheumatic mitral disease in two, chronic severe mitral or aortic regurgitation in three, amyloid heart in two, and normal heart in three. Marked right-heart dilatation was present in 11 patients and LV thrombus in four. Regression equations derived in vitro for each 2-D echo instrument were used to correct LV mass estimates based on a short-axis, area-length method: uncorrected LV mass = 1.055 x k x 5/6 (AtLt - AcLc) + b, where At = total short-axis LV image area at the high papillary muscle level, Lc = endocardial LV length, k = an instrument-specific regression slope and b = an instrument-specific intercept. LV mass by 2-D echo correlated extremely well with actual LV weight (r = 0.93 slope = 0.85, SEE = 31 g, range 77-454 g). In contrast, M-mode echocardiographic LV mass estimates were less reliable (r = 0.86, SEE = 59 g) in these markedly distorted hearts. These 2-D echo LV mass results compare favorably with reported results from biplane angiography and M-mode echocardiography in more symmetric hearts. Thus, regression-corrected 2-D echo may be the method of choice for determining LV mass in man.