Echocardiograms were acquired by trained study sonographers following a standardized protocol and using high-quality commercially available echocardiographs equipped with 3.0- to 3.5-MHz and 2.0- to 2.5-MHz probes. Standardized examinations included two-dimensional-guided M-mode recordings obtained in the parasternal long- and short-axis views below the mitral leaflet tips to the nearest millimeter and two-dimensional parasternal long- and short-axis (at the mitral leaflets, midpapillary, and apical levels) and apical (four-chamber, two-chamber, and long-axis) views. All measurements were performed off-line by one reader (R.S.) blinded to subjects' randomization status and clinical data and using a computerized review station (Freeland Systems, Cine View, Indianapolis, Indiana). A minimum of three cardiac cycles for patients in sinus rhythm and five cardiac cycles for patients in atrial fibrillation were captured and measured. The M-mode measurements included end-diastolic LV internal diameter, interventricular septal thickness and posterior wall thickness, and end-systolic and LV internal diameter. Cases where accurate M-mode measurements could not be obtained in the parasternal short axis were substituted by parasternal long-axis measurements. All M-mode measurements were performed using the American Society of Echocardiography (ASE) leading-edge convention (18). When optimal orientation of the LV imaging views could not be obtained, correctly oriented two-dimensional linear measurements were obtained using the ASE leading-edge convention. The LVM was calculated using the corrected ASE method (19), and LVMI was obtained by indexing for body surface area. Between-sonographer (on 83 duplicate examinations) and within-reader (on 80 duplicate readings) intraclass correlation coefficients for LVM were 0.84 and 0.85, respectively. Two-dimensional LVM, left ventricular end-diastolic volume (LVEDV) and left ventricular end-systolic volume (LVESV), and LVEF were calculated from two-dimensional recordings using the modified biplane Simpson's method (20). Wall motion score (WMS) and new wall motion abnormalities (NWMAB) were evaluated on two-dimensional echocardiograms using the 16-segment model (21). Between-sonographer and within-reader (on 20 duplicate readings) intraclass correlation coefficients for LVEF measurements were 0.70 and 0.75, respectively. Reproducibility of two-dimensional LVM measurements was lower than that of M-mode measurements and, therefore, the latter were used for the primary analysis of treatment effects on LVM and LVMI. An attempt was made to obtain all measurements from all echocardiograms of all 446 patients included in the final analysis. However, at times the image quality was suboptimal. Consequently, not all measurements were obtained in every patient. M-mode LVM calculations were possible in 99% of the patients; two-dimensional measurements of LVEDV, LVESV, LVEF, LVM, and WMS were obtained in 98%, 98%, 98%, 96%, and 98% of patients, respectively.