The association between epicardial fat and coronary artery disease (CAD) might be affected by general adiposity. We aimed to determine whether the percentage of epicardial adipose tissue (%EAT), defined as the mass ratio of epicardial fat to body fat, could improve prediction of asymptomatic CAD. We consecutively enrolled 846 adults who underwent coronary computed tomography angiography as part of a health check-up and assessed their coronary stenosis severity and epicardial fat mass. Body fat mass was measured by bioelectrical impedance analysis. Subjects with CAD history, hyperthyroidism, pitting edema, or subjects taking diuretics or thiazolidinedione were excluded. Obstructive CAD was defined as at least one coronary artery with 50 % or greater obstruction, and severe CAD was defined as 70 % or greater obstruction. The %EAT had the maximum area under the curve for predicting the presence of CAD and superior discriminative performance to EAT and other EAT-indexed parameters. Multivariable logistic regression analysis revealed that %EAT >0.41 % was a predictor of obstructive CAD [odds ratio 3.59 (95 % confidence interval 2.28–5.64)], and %EAT >0.47 % was a predictor of severe CAD [4.01 (2.01–7.99)] after adjustment for calcium score and Framingham risk score. This prediction was more pronounced in subjects with higher body fat percentage (≥25 % for men and ≥35 % for women), Framingham risk score (≥10 %), or calcium score (≥100). A spillover of body fat at epicardium over a critical threshold is associated with significant coronary stenosis. This association was independent of obesity, coronary calcium burden, and Framingham risk factors.