Comparing Surrogate Indexes for Insulin Resistance as Predictors of Type 2 Diabetes (T2D)

CONTEXT: Decreased insulin sensitivity (insulin resistance) is associated with increased type 2 diabetes (T2D) risk. Identifying reliable surrogate indexes that use standard clinical assays might improve the prevention of T2D. OBJECTIVE: The objective was to compare the ability of 18 surrogate indexes of insulin resistance to predict T2D. DESIGN: Cohort study of indigenous Americans from the southwest United States. SETTING: Community-based longitudinal study. PATIENTS AND OTHER PARTICIPANTS: We studied T2D risk in 2260 people followed for up to 14.5 years (509 cases of T2D). We estimated correlation coefficients (r) of all indexes with sensitivity (M) measured by hyperinsulinemic-euglycemic clamp in 286 persons. We calculated indexes obtained from oral glucose tolerance tests (OGTTs) using fasting and 2-hour glucose and insulin measures. INTERVENTIONS: None. MAIN OUTCOME MEASURES: Predictive performance was assessed by the hazard ratio per SD and by the area under the receiver operating curve (AUC). RESULTS: Indexes calculated from OGTTs performed best; Matsuda had the highest r with M (0.691); Gutt and Cederholm had the highest AUCs (0.728 and 0.728, P < .05 compared with all other indexes). In indexes that only used fasting insulin levels, the quantitative insulin sensitivity check index and Homeostatic Model for Insulin Resistance had the best performance with equal r (0.644) and AUC (0.701). Among indexes that did not require insulin measurement, the metabolic score for insulin resistance (METS-IR) and single-point insulin sensitivity index had the highest r (0.597 and 0.595), while METS-IR had the highest AUC (0.688). CONCLUSION: Surrogate measures based on OGTTs are best for predicting T2D, while measures based on fasting insulin perform better than those that do not measure insulin.