Aims/hypothesis: The aim of this study was to construct a model for predicting CHD and cardiovascular disease (CVD) risk in patients with newly diagnosed type 2 diabetes in a southern European region. External validation of two other cardiovascular risk models and internal validation of our model were assessed.
Methods: We studied 65,651 people attending a primary care setting in the Basque Country Health Service. A 10-year prospective population-based cohort study was performed with 777 patients newly diagnosed with type 2 diabetes older than 24 years in a Sentinel Practice Network. Cardiovascular risk factors, CVD events and mortality were registered. Coefficients for the significant predictors of CHD and CVD were estimated using Cox models. We assessed the discrimination and calibration of the UK Prospective Diabetes Study risk engine (UKPDS-RE), the Framingham Risk Score-Regicor Study (FRS-RS) and the cardiovascular risk model we developed.
Results: The incidence rate per 1,000 patients/year was calculated for microvascular and cardiovascular complications, and death. Age, the ratio of non-HDL- to HDL-cholesterol, HbA1c, systolic blood pressure and smoking were significant predictors of cardiovascular events. A risk model was developed using these predictors. The UKPDS-RE and FRS-RS showed inadequate discrimination (Uno's C statistics 0.62 and 0.58, respectively) and calibration (24% overestimation and 51% underestimation, respectively) for predicting CHD risk. The internal discrimination and calibration of the developed model were acceptable for predicting fatal/non-fatal 2- and 5-, but not 10-year CHD and CVD risk.
Conclusions/interpretation: This study is the first southern European validated population-derived model for predicting 5-year fatal/non-fatal CHD and CVD risk in patients with newly diagnosed type 2 diabetes.