Background A definition of metabolic syndrome (MetS) has been recommended as a tool to help identify individuals at risk of developing type 2 diabetes. However, an agreed protocol for defining MetS does not exist and some studies have shown MetS definitions to be inferior at predicting diabetes compared to a single measurement of fasting glucose. In this study we examined the ability of five proposed MetS definitions to discriminate incident cases in order to determine whether MetS more accurately predicts type 2 diabetes.

Methods This was a prospective study involving a random sample of 1,754 men and women aged 46–73 years. Receiver operating characteristic curve and net reclassification improvement (NRI) analyses were used to evaluate the ability of MetS definitions and components to accurately classify high-risk subjects.

Results A model including proposed MetS components displayed a significantly (P=0.02) higher area under the curve (AUC) to discriminate diabetes (AUC=0.90, 95% CI: 0.87–0.93) compared to fasting glucose alone (AUC=0.88, 95% CI: 0.83–0.92). Models using the European Group for the Study of Insulin Resistance MetS criterion, and which included glucose as a mandatory component, demonstrated significant overall NRI when compared to recommended and optimal fasting glucose cut-offs. A final model had a sensitivity of 0.91 and a specificity of 0.73.

Conclusion In this population there is evidence that a combination of MetS components may help predict diabetes beyond that which is measured by glucose alone. Proposed MetS definitions should include fasting glucose as a mandatory component.