Risk factors for the 12-year incidence of definite coronary heart disease (CHD) among 3440 men who were middle-aged (51 to 59 years old) and 1419 men who were elderly (65 to 74 years old) at baseline examination were examined for differences in predictive values in terms of both relative risk and attributable (excess) risk of the highest versus the lowest quartile or appropriate categories. In multivariate models using Cox life-table regression procedures, serum cholesterol level, cigarette smoking, systolic blood pressure, and history of treatment for diabetes were significant predictors of incident CHD for both age groups. Alcohol consumption when modeled as drinker versus nondrinker showed a protective effect in both younger and older men. There was no dose relationship, however, among elderly drinkers. While the relative risks for the variables studied were similar between the two age groups, the excess risk was typically between 1.5 to 2.0 times higher for the older than the middle-aged men. In contrast, the detrimental effect of adiposity as measured by body mass index appeared to decline after age 65 for both measures of risk. This may partly be attributed to diminished adiposity overall in the older age group. The implications of these results are that serum cholesterol level, smoking, hypertension, diabetes, and possibly alcohol consumption continue to be important predictors for CHD when measured after age 65, and that the public health impact of these risk factors, in terms of excess risk, may be more important in the elderly.