Non-invasive Risk Prediction Models in Identifying Undiagnosed Type 2 Diabetes or Predicting Future Incident Cases in...

Iran needs pragmatic screening methods for identifying those with undiagnosed type 2 diabetes or at high risk of developing it. The aim of this study was to assess performance of three non-invasive risk prediction models, i.e. the Finnish Diabetes Risk Score (FINDRISC), the Australian Type 2 Diabetes Risk Assessment Tool (AUSDRISK), and the American Diabetes Association Risk Score (ADA), for identifying those with undiagnosed type 2 diabetes (prevalent type 2 diabetes at baseline without any treatment) or those who would develop type 2 diabetes within 5 years of follow-up.

3467 participants aged ≥30 years without treated type 2 diabetes in the Tehran Lipid and Glucose Study (TLGS) were included in this study. The discrimination power of models was assessed by area under the curve (AUC), their calibrations were assessed by calibration plots and Hosmer-Lemeshow test, and their net benefits were assessed by decision curves.

430 participants had undiagnosed type 2 diabetes at baseline and 203 developed type 2 diabetes during 5 years of followup. AUSDRISK had the highest AUC (0.77) as compared to FINDRISC (0.75; P value: 0.014), and the ADA model (0.73; P value: <0.001). The original model for AUSDRISK and calibrated versions of FINDRISC and ADA models had acceptable calibration (Hosmer-Lemeshow chi-square <20) and these models were clinically useful in a wide range of risk thresholds as their net benefit was higher than no-screening scenarios.

The original AUSDRISK model and recalibrated models for FINDRISC and ADA are valid and effective tools for identifying those with undiagnosed or 5-year incident type 2 diabetes in Iran.