Background: Prediabetes (PD) is a high-risk state for type 2 diabetes (T2D), but the risk of progression from PD to T2D can be prevented by early diagnosis and intervention. Metabolomics studies suggest that metabolic biomarkers can shed insight into etiological mechanisms and improve the accuracy of prediction of disease onset.

Objectives: We aimed to identify serum metabolic biomarkers and verify their predictive performance for PD in the Korean population, as compared to the performance of known clinical risk factor (CRF) as well as previously reported metabolites (PRM) in other population studies.

Methods: A targeted metabolomics was carried out to quantify serum metabolites for 1,723 participants from the Korea Association REsource (KARE) cohort from which 500 individuals were followed-up to 6-years. PD-related metabolites were identified at baseline by statistical methods including multivariable regression analysis, and tested their association to incidence of PD during follow-up.

Results: The 12 metabolites were significantly altered in PD (adjusted P<4.07E-04). These metabolites predicted incidence of PD with an area under curve (AUC) of 0.71. The AUC of the metabolic markers was significantly higher than that of the two PRM models (0.56 and 0.64, P<0.005) and the CRF (0.64, P<0.01). The performance of the metabolic markers compared to glucose model was significantly higher among the obese (BMI >= 25 kg/m2, 0.79 vs. 0.58, P<0.001), significantly lower among the female (0.73 vs. 0.85, P<0.002) and the lean (BMI<25 kg/m2, 0.68 vs. 0.78, P<0.01), and not significantly different among the age and the male. The full model with metabolic markers, CRF, glucose yielded the best prediction (AUC=0.84).

Conclusion: Our results revealed that the novel metabolic markers are not only associated with the risk of PD, but also improve the prediction performance in combination with classical approaches. These findings may help understanding causality of diabetes progression and developing preventive strategies for T2D.


H. Lee: None. T. Park: None. B. Kim: None.

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at