Epicardial adipose tissue (EAT) volume is reported to be associated with coronary plaques. We evaluated whether non-invasive measurement of EAT thickness by echocardiography can predict high-risk coronary plaque characteristics determined independently by coronary computed tomography (CT) angiography. We enrolled 406 patients (mean age 63 years, 57 % male) referred for 64-slice CT. EAT was measured on the right ventricle free wall from a parasternal long-axis view at the end of systole. High-risk coronary plaques were defined as low-density plaques (<30 Hounsfield units) with positive remodeling (remodeling index >1.05). Patients were divided into thin or thick EAT groups using a cutoff value derived from receiver operator characteristic curve analysis for discriminating high-risk plaques. The receiver operator characteristic cutoff value was 5.8 mm with a sensitivity of 83 % and specificity of 64 % (area under the curve 0.77, 95 % confidence interval 0.70–0.83, p < 0.01). Compared with the thin EAT group, the thick EAT group had a high prevalence of low-density plaques (4 vs. 24 %, p < 0.01), positive remodeling (39 vs. 60 %, p < 0.01), and high-risk plaques (3 vs. 17 %, p < 0.01). Multiple logistic analysis revealed that thick EAT was a significant predictor of high-risk plaques (odds ratio 7.98, 95 % confidence interval 2.77–22.98, p < 0.01) after adjustment for covariates, including conventional risk factors, visceral adipose tissue area, and medications. The measurement of EAT thickness by echocardiography may provide a non-invasive option for predicting high-risk coronary plaques.