Abstract 13600: Performance of a Multi-Metabolite Score Added to a Risk Prediction Model for Incident Diabetes in Young Adults: The CARDIA Study

Abstract

Introduction: The Finnish Diabetes Risk Score (FINDRISC) is a risk prediction tool used to estimate absolute risk for type 2 diabetes mellitus (DM), a critical risk factor for cardiovascular disease. The Diabetes Risk Index (DRI) is a multi-metabolite DM risk score developed by combining the Lipoprotein Insulin Resistance Index (LP-IR), calculated from six lipoproteins, and the branched chain amino acids, valine and leucine. Although the DRI is independently associated with future DM in middle-aged adults, the performance of this novel biomarker when added to FINDRISC to predict incident DM in young adults is not known. Hypothesis: The addition of DRI to FINDRISC will improve the 10-year(y) risk prediction of DM across all ages in the Coronary Artery Risk Development in Young Adults (CARDIA) study. Methods: We measured the DRI using NMR spectroscopy in CARDIA participants across different age categories: 20-30y (n=1,911), 30-40y (n=2,716), 40-50y (n=2,232). We used Cox proportional hazards regression to evaluate the association of the DRI with incident 10-y DM risk. We assessed FINDRISC model performance using C-statistic and continuous Net Reclassification Index (NRI) with and without the addition of the DRI as a covariate. Results: DRI was associated with 10-y DM risk across all age categories. Addition of the DRI modestly improved the C-statistic in the 30-40-y age category. Addition of DRI to FINDRISC improved the overall NRI in the 30-40- and 40-50-y age categories. The non-event NRI was significant across all 3 age categories and reclassified 25-36% of participants who did not develop DM to a lower estimate of incident DM risk (Table). Conclusions: Among young and middle-aged adults, DRI scores were associated with 10-y risk of DM. The addition of DRI to FINDRISC led to improvements in reclassification of DM risk, as the addition of DRI to FINDRISC correctly classified up to 36% of participants who did not develop DM (over 10y) to a lower estimate of incident DM risk.