Plasma Lipid Metabolites, Clinical Glycemic Predictors and Incident Type 2 Diabetes

OBJECTIVE: Plasma metabolite profiling has uncovered several non-glycemic markers of incident type 2 diabetes (T2D). We investigated whether such biomarkers provide information about specific aspects of T2D etiology, such as impaired fasting glucose and impaired glucose tolerance, and whether their association with T2D risk varies by race. RESEARCH DESIGN AND METHODS: Untargeted plasma metabolite profiling was performed in the FINRISK 2002 cohort (N=7,564). Cox regression modeling was conducted to identify metabolites associated with incident T2D during 14 years of follow-up. Metabolites were clustered into pathways using Gaussian graphical modeling. Clusters enriched for T2D biomarkers were further examined for covariation with fasting plasma glucose (FPG), 2-hour post-challenge plasma glucose (2hPG), HbA1c, or fasting insulin. Validation analyses and tests of interaction with race were performed in the ARIC study. RESULTS: Two clusters of metabolites, representing diacylglycerols (DAGs) and phosphatidylcholines (PCs), contained the largest number of metabolite associations with incident T2D. DAGs associated with increased T2D incidence (HR 1.22, 95% CI 1.14-1.30) independent of FPG, HbA1c, and fasting insulin, but not 2hPG. PCs were inversely associated with T2D risk (HR 0.78, 95% CI 0.71-0.85) independent of FPG, 2hPG, HbA1c, and fasting insulin. No significant interaction between DAGs or PCs and race was observed. CONCLUSIONS: Fasting DAGs may capture information regarding T2D risk similar to that represented by 2hPG; PCs may capture aspects of T2D etiology that differ from those represented by conventional biomarkers. The direction-of-effect and strength of DAG and PC associations with incident T2D are similar across European and African Americans.