Prediction of Coronary Heart Disease in Middle-Aged Adults With Diabetes

OBJECTIVE: To determine the 10-year probability of coronary heart disease (CHD) in diabetic adults and how well basic and novel risk factors predict CHD risk. RESEARCH DESIGN AND METHODS: We measured risk factors in 14054 participants (1500 with diabetes) initially free of CHD in the Atherosclerosis Risk in Communities study from 1987 to 1989 and followed them prospectively for CHD incidence through 1998. We used proportional hazards regression models and receiver operating characteristic (ROC) curves for CHD risk prediction. RESULTS: Based on our model using basic risk factors (age, race, total and HDL cholesterol, systolic blood pressure, antihypertensives, and smoking status), approximately 61% of diabetic women and 86% of diabetic men had a predicted 10-year CHD probability >or=10%. This CHD risk-prediction model had an area under the ROC curve of 0.72 in diabetic women and 0.67 in diabetic men. Novel risk factors or subclinical disease markers individually added only modest predictivity, but the addition of multiple markers (BMI, waist-to-hip ratio, Keys dietary score, serum albumin and creatinine, factor VIII, white blood cell count, left ventricular hypertrophy determined by electrocardiogram, and carotid intima-media thickness) increased the area under the curve by approximately 10%. CONCLUSIONS: Although all diabetic adults are at high risk for CHD, their variation in CHD risk can be predicted moderately well by basic risk factors. No single novel risk marker greatly enhanced absolute CHD risk assessment, but a battery of novel markers did. Our model can provide estimates of CHD risk for the primary prevention of this disease in people with type 2 diabetes.