3D dataset acquisition and analysis
Real-time 3DE image acquisitions of the LV from the 4-chamber view were performed using echocardiographic ultrasound platforms (GE Vivid E9/E95; Vingmed Ultrasound, Horten, Norway, iE33/EPIQ; Philips Medical Systems, Andover, MA, USA, and/or Siemens SC2000, Siemens Medical Solutions, Malvern, PA, USA). 3D datasets were excluded by the core lab if any portion of LV endocardial borders or LV apex were missing from the 3D dataset. Studies were also excluded if the vendor independent software (4D LV Analysis, TomTec Imaging Systems version 4.0, Unterschleissheim, Germany) was not able to automatically track the borders of the LV after three attempts. At each center (with the exception of center 5), the fully automated method was first used to generate LV end diastolic and end systolic volumes (EDV and ESV), stroke volumes (SV), and ejection fraction (EF) without editing the borders of the LV. Timing from the start to the end of the analysis was recorded. Then semi-automated analysis was performed by all centers on the same datasets, manually editing the contours to define the LV endocardial borders for the same variables. Timing from the start to the end of the analysis was again recorded. All LV EDV, ESV, SV, and EF data points from each center were entered into REDCap located at the core laboratory. Subsequently, all 3DE datasets were transferred to the core laboratory and had the LV volumes and EF re-measured using the fully automated and semi-automated methods described above. The core laboratory observer was blinded to the reported measures in REDCap to determine intercenter reproducibility. The core laboratory then repeated analysis for intraobserver and interobserver variability 6 months after initial analysis in 86 randomly selected studies at the core lab.