Visual Abstract
In persons without diabetes, the product of estimated insulin secretion and sensitivity, the disposition index (DI), is used as an index of beta cell function and a predictor of T2D. However, its use in predicting T2D disregards differences in insulin secretion and sensitivity among persons with the same DI. We studied diabetes incidence in 420 persons without diabetes followed for a median of 9.9 years during which time 146 (35%) developed diabetes. Insulin secretion was measured 3-5 min after a 25gm glucose bolus (AIR) and sensitivity (M) by the hyperinsulinemic-euglycemic clamp. The Figure plots AIR by M and shows bands representing a range of lowest to highest incidence. Different (hyperbolic) lines of constant DI are superimposed. A single DI line crosses several risk zones; subjects with the same DI but lower M had higher T2D incidence, suggesting that DI is not optimal for prediction. We also compared the performance of different Cox models in predicting T2D using the area under the receiver operating characteristic curve (AUC). In prediction by separate secretion and sensitivity variables, AUC=0.72, higher (p<0.01) than for DI (AUC= 0.66). Results were similar using OGTTs in a longitudinal population study and using other measures of discrimination, and in persons with normal glucose regulation. Separate measures of insulin secretion and sensitivity are better than DI in predicting T2D.
E. Vazquez arreola: None. R. L. Hanson: None. C. Bogardus: None. W. C. Knowler: None.
