Development and validation prediction models for risks of coronary artery disease in patients with type 2diabetes

Prior studies established risk score prediction models for coronary artery disease of patients with type 2 diabetes in Western ethnicity, but few in Asian population. The aim of the study is to develop a prediction model for coronary artery disease by considering the variation of fasting blood glucose, and traditional risk factors in Chinese patients with type 2 diabetes from the database of Taiwan National Diabetes Care Management Program. A total of 27,756 patients aged 30-84 years were randomly allocated into derivation and validation sets with a ratio of 2:1. During a mean follow-up of 7.7 years, 25.51% of patients developed CAD in the derivation set. In validation set, 25.01% of patients developed CAD during a mean 7.8 years of follow-up. Cox model was used to determine risk scores of CAD. Calibration was assessed using by Hosmer-Lemeshow test, and discrimination ability was examined by the areas under receiver operating curve (AUROCs). This study found age, gender, smoking, alcohol drinking, obesity, blood pressure, high-density lipoprotein, HbA1c, total cholesterol, triglyceride, variation of fasting blood glucose, hypertension, diabetes retinopathy, anti-diabetes medications, cardiovascular medications and statins were significantly associated with CAD. The risk score ranged from -10 to 62. Calibration was excellent in two sets (P=0.13 and 0.91). The AUROCs of 5-year CAD risk was 0.631 in the derivation set, and was 0.643 in the validation set. In conclusion, this risk score model can be used by the clinicians to stratify Chinese patients with type 2 diabetes into groups at varying risk of CAD.