Improving the quality of preclinical research echocardiography: observations, training, and guidelines for measurement

Informal training in preclinical research may be a contributor to the poor reproducibility of preclinical cardiology research and low rates of translation into clinical research and practice. Mouse echocardiography is a widely used technique to assess cardiac structure and function in drug intervention studies using disease models. The interobserver variability of clinical echocardiographic measurements has been shown to improve with formalized training, but preclinical echocardiography lacks similarly critical standardization of training. The aims of this investigation were to assess the interobserver variability of echocardiographic measurements from studies in mice and address any technical impediments to reproducibility by implementing standardized guidelines through formalized training. In this prospective, single-site, observational cohort study, 13 scientists performing preclinical echocardiographic image analysis were assessed for measurement of short-axis M-mode-derived dimensions and calculated left ventricular (LV) mass. Ten M-mode images of mouse hearts acquired and analyzed by an expert researcher with a spectrum of LV mass were selected for assessment and validated by autopsy weight. After the initial observation, a structured formal training program was introduced, and accuracy and reproducibility were reevaluated. Mean absolute percentage error for expert-calculated LV mass was 6 ± 4% compared with autopsy LV mass and 25 ± 21% for participants before training. Standardized formal training improved participant mean absolute percentage error by ~30% relative to expert-calculated LV mass ( P < 0.001). Participants initially categorized with high-range error (25-45%) improved to low-moderate error ranges (