OBJECTIVE - To determine the association of coronary heart disease (CHD) incidence with diabetes, fasting serum glucose, and insulin in a biracial cohort of middle-aged men and women. RESEARCH DESIGN AND METHODS - We examined a population-based sample (n = 13,446 free of baseline CHD) from four U.S. communities in 1987-1989. We defined diabetes on the basis of baseline fasting glucose concentration (≤7.8 mmol/l), medical history, and current medications. A central laboratory measured fasting insulin with a nonspecific radioimmunoassay. After 4-7 years, 209 men and 96 women developed CHD. RESULTS - After adjustment for sociodemographic characteristics, smoking status, ethanol intake, sports participation, and hormone replacement therapy, the relative task of CHD for people with diabetes versus those without diabetes was 3.45 (95% CI 2.16-5.50) among women and 2.52 (1.78- 3.56) among men. Relative risks of CHD with diabetes were somewhat lower in blacks than non-blacks, but because diabetes was more than twice as prevalent in blacks, the percentage of CHD cases attributable to diabetes (population attributable risk) was 27% for black women, 15% for non-black women, 8% for black men, and 12% for non-black men. Among people without diabetes, fasting glucose was not independently associated with CHD incidence. Among women without diabetes, there was a positive association between fasting insulin and CHD; multivariable adjusted relative risks of CHD across quintiles of fasting insulin were 1.00, 0.76, 2.08, 2.08, and 2.82 (P for linear trend = 0.02). However, among men without diabetes, fasting insulin and CHD were not associated. CONCLUSIONS - Diabetes conveys a high risk of CHD in black and non-black middle-aged men and women. Fasting insulin, however, is a CHD risk factor only among women in this cohort.