Agreement between visually estimated left ventricular ejection fraction on echocardiography and quantitative measurements using cardiac magnetic resonance

Objective: Visual estimation of left ventricular ejection fraction (LVEF) is still used in routine clinical practice. However, most of the studies evaluating the agreement between the visually estimated LVEF (ve-LVEF) and quantitatively measured LVEF (qm-LVEF) either have not used appropriate statistical methods or gold standard imaging modality. We aimed to assess the agreement between the ve-LVEF and qm-LVEF using contemporary statistical methods and Cardiac Magnetic Resonance imaging (CMRI). Methods: In 54 subjects who underwent 1.5-T CMRI, echocardiographic images were recorded after the CMRI procedure on the same day. ve-LVEFs were estimated by 2 independent observers on echocardiographic records in a random and blinded fashion. qm-LVEF was obtained by CMRI. Agreement between the ve-LVEF and qm-LVEF values, and intra/interobserver ve-LVEF estimations were assessed using intraclass correlation coefficient (ICC), Bland-Altman analysis, and kappa statistics. Results: There was a high agreement between the ve-LVEF and qm-LVEF (ICC 0.93, [95% CI 0.88-0.96]). Bland-Altman analysis also demonstrated a good agreement between the ve-LVEF and qm-LVEF; visually estimated LVEF was, on average, 0.6% lower than that obtained by CMRI (mean -0.6, limits of agreement -10.5 and +9.3). A good agreement was also observed for LVEF categories of ≤35%, 36-54%, and ≥55% (unweighted kappa 0.71, linearly weighted kappa 0.76); and LVEF of