A novel algorithm to predict incident type 2 diabetes mellitus (iT2DM) is presented considering data from a 20-year prospective study in a Southern Italy population. Eight hundred and fifty-eight out of 1,351 subjects (24-85 years range of age) were selected. Incident type 2 diabetes was diagnosed in 103 patients in a 20-year follow-up. The Finnish Diabetes Risk Score (FINDRISC) and the Framingham Offspring Study simple clinical model (FOS) have been used as reference algorithms. Two custom algorithms have been created using Cox parametric hazard models followed by PROBIT analyses: the first one (VHSRISK) includes all the study subjects and the second one (VHS95RISK) evaluates separately subjects with baseline fasting blood glucose (FBG) above/below 5.2 mmol/L (95 mg/dL). The 44 iT2DM cases below 5.2 mmol/L of baseline FBG were predicted by high LDL cholesterol, metabolic syndrome (ATPIII criteria), BMI > 30 kg/m(2), and high factor VII activity. The 59 cases above the FBG threshold were predicted by FBG classes, hypertension, and age. ROC areas for iT2DM prediction were: FINDRISC = 0.759, FOS = 0.762, VHSRISK = 0.789, and VHS95RISK = 0.803. In a Mediterranean population, the use of a custom generated algorithm evaluating separately low/high FBG subjects improves the prediction of iT2DM in subjects classified at lower risk by common estimation algorithms.
|Numero di pagine||7|
|Stato di pubblicazione||Published - 2012|
All Science Journal Classification (ASJC) codes