Contrast administration reduces interobserver variability in determination of left ventricular ejection fraction in patients with left ventricular dysfunction and good baseline endocardial border delineation.

Echocardiographic contrast agents improve endocardial border delineation in patients with technically difficult baseline studies. With medical and device therapy for heart failure increasingly based on left ventricular (LV) ejection fraction (EF) partition values, the accurate and reproducible assessment of LV function is necessary. It was hypothesized that routine contrast enhancement would significantly reduce interobserver variability in the determination of LVEFs in a cohort of patients with LV dysfunction and good baseline endocardial delineation. All patients underwent baseline noncontrast studies followed by contrast-enhanced imaging using Definity. Two experienced echocardiographers, blinded to the clinical data, determined LVEFs using 4 different techniques: noncontrast estimated (NCE), noncontrast calculated (NCC), contrast estimated (CE), and contrast calculated (CC). Using a mixed-model procedure that allows for fixed and random events, the variance due to error and that due to the patient was obtained (interclass correlation). The proportion of variation due to the reader was calculated as 1--interclass correlation. Mean standardized percentage differences ([reader 1 EF--reader 2 EF]/mean EF) were also calculated for each method. The proportion of variation due to the reader was smallest in the CC group and largest in the NCC group (NCE = 0.21, NCC = 0.33, CE = 0.25, CC = 0.11). The results were similar when only patients with NCE EFs >or=20% and <or=50% were analyzed (NCE = 0.29, NCC = 0.47, CE = 0.33, CC = 0.15). The mean standardized percentage difference between readers was smallest in the CC group for all patients and for those with NCE EFs >or=20% and <or=50%. In conclusion, contrast administration reduces interobserver variability in LVEF assessment, even in patients with good baseline endocardial border delineation, and should be used routinely regardless of the perceived adequacy of baseline images, especially if the LVEF is >or=20% and <or=50%.