Twenty-year coronary heart disease (CHD) incidence is analyzed for 279 men, CHD-free at the ages of 47 through 57 years and characterized by three examinations before 20 annual follow-up examinations. Sixty men developed CHD, 42 dying or suffering infarction. Among 20 entry characteristics, three had major predictive power, especially for CHD death or infarction: cold pressor test, a high level of serum cholesterol, and systolic blood pressure. Smoking and relative weight seemed unimportant. Major CHD incidence was proportional to cholesterol to the 3.4 power. Five-variable multiple logistic analysis permitted placement of men into deciles of estimated risk differing more than tenfold in CHD incidence. The combination of only cholesterol and systolic pressure was nearly as good. The multiple logistic coefficients applied to five-year data on 1,287 men among whom 182 CHD developed in 182 yielded satisfactory agreement between observed and predicted distributions of CHD cases in classes of estimated risk.