The relationship between selected baseline risk factors and subsequent coronary heart disease (CHD) death and total mortality among participants in the Multiple Risk Factor Intervention Trial (MRFIT) was studied in order to determine whether the three risk factors used to identify high-risk men for the trial were associated with CHD death; whether other risk factors measured at baseline, especially lipoprotein cholesterol levels, were associated with CHD and total mortality; and whether there were any differences between special intervention (SI) and usual care (UC) participants in the relationship of the specific levels of risk factors to CHD or total mortality. The three main risk factors (blood cholesterol, cigarette smoking, and diastolic blood pressure) and age were significantly associated with CHD mortality; age, diastolic blood pressure, and cigarette smoking were associated with total mortality. The risk score based on the multiple logistic equation developed from the Framingham Study was also strongly associated with CHD mortality. When the joint associations of selected baseline risk factors with CHD and total mortality were considered, age, diastolic blood pressure, cigarette smoking, and low- and high-density lipoprotein cholesterol were significantly associated with CHD mortality; age, cigarette smoking, and low-density lipoprotein cholesterol were positively associated with total mortality. Systolic blood pressure significantly improved the prediction of CHD mortality for SI and UC men when it was added to a regression model that included age, diastolic blood pressure, cigarettes smoked per day, body mass index, and lipoprotein levels, but improved the prediction of total mortality only for SI men. In similar analyses, serum thiocyanate improved the prediction of both CHD and total mortality for UC men. Among SI men the improved prediction gained by considering serum thiocyanate was less pronounced and not significant for CHD death. This latter finding may be due in part to the changes made in smoking behavior by SI participants during the course of the study. The estimated regression coefficients for CHD and total mortality endpoints were not significantly different between the SI and UC groups.