Although fatty liver predicts ischemic heart disease, the incidence and predictors of fatty liver need examination. The objective of this study was to determine fatty liver incidence and predictive variables. Using abdominal ultrasonography, we followed biennially through 2007 (mean follow-up, 11.6±4.6 years) 1635 Nagasaki atomic bomb survivors (606 men) without fatty liver at baseline (November 1990 through October 1992. We examined potential predictive variables with the Cox proportional hazard model and longitudinal trends with the Wilcoxon rank-sum test. In all, 323 (124 men) new fatty liver cases were diagnosed. The incidence was 19.9/1000 person-years (22.3 for men, 18.6 for women) and peaked in the sixth decade of life. After controlling for age, sex, and smoking and drinking habits, obesity (relative risk (RR), 2.93; 95% confidence interval (CI), 2.33-3.69, Po0.001), low high-density lipoprotein-cholesterol (RR, 1.87; 95% CI, 1.42-2.47; Po0.001), hypertriglyceridemia (RR, 2.49; 95% CI, 1.96-3.15; Po0.001), glucose intolerance (RR, 1.51; 95% CI, 1.09-2.10; P¼0.013) and hypertension (RR, 1.63; 95% CI, 1.30-2.04; Po0.001) were predictive of fatty liver. In multivariate analysis including all variables, obesity (RR, 2.55; 95% CI, 1.93-3.38; Po0.001), hypertriglyceridemia (RR, 1.92; 95% CI, 1.41-2.62; Po0.001) and hypertension (RR, 1.31; 95% CI, 1.01-1.71; P¼0.046) remained predictive. In fatty liver cases, body mass index and serum triglycerides, but not systolic or diastolic blood pressure, increased significantly and steadily up to the time of the diagnosis. Obesity, hypertriglyceridemia and, to a lesser extent, hypertension might serve as predictive variables for fatty liver.