1476-P: Metabolomic and Genetic Diabetes Predictors in the Diabetes Prevention Program (DPP)

Objective: Type 2 diabetes (T2D) can be prevented, but the risk is heterogeneous even among high-risk individuals. We aimed to create a metabolite T2D risk model and evaluate its utility compared to a genetic risk score (GRS) and traditional clinical factors in 1,635 participants with available genomics and metabolomics.

Methods: Predictors were picked using elastic net from 3,145 mass spectrometry measured metabolites in baseline fasting plasma. Predictive performance after 3.2 years follow-up was assessed using AUC and IDI of Cox models that included the metabolites, the 120-SNP GRS, and clinical factors.

Results: The composite monosaccharide measurement of glucose/fructose/galactose and a 50 carbon:1 double bond triacylglycerol (TG 50:1) were selected and minimally increased the AUC beyond the clinical model (AUC=0.71 vs AUC=0.69, Fig1A). There was no change with the addition of the GRS (AUC=0.69). When added to FPG, TG 50:1 provided the same increase in the integrated discrimination index (IDI) as with clinical triglycerides (Fig1B), but glucose/fructose/galactose provided a significant increase in IDI beyond clinical FPG.

Conclusion: Metabolites minimally improve DM prediction beyond clinical risk factors in a high-risk adult population. A glucose/fructose/galactose composite metabolite significantly improves IDI.

View largeDownload slide

View largeDownload slide

Close modal