Potential Novel Serum Metabolic Markers Associated With Progression of Prediabetes to Overt Diabetes in a Chinese Population

Abstract

Background: Identifying the metabolite profile of individuals with prediabetes who turned to type 2 diabetes (T2D) may give novel insights into early T2D interception. The purpose of this study was to identify metabolic markers that predict the development of T2D from prediabetes in a Chinese population.

Methods: We used an untargeted metabolomics approach to investigate the associations between serum metabolites and risk of prediabetes who turned to overt T2D (n=153, mean follow up 5 years) in a Chinese population (REACTION study). Results were compared with matched controls who had prediabetes at baseline [age: 56 ± 7 years old, body mass index (BMI): 24.2 ± 2.8 kg/m2] and at a 5-year follow-up [age: 61 ± 7 years old, BMI: 24.5 ± 3.1 kg/m2]. Confounding factors were adjusted and the associations between metabolites and diabetes risk were evaluated with multivariate logistic regression analysis. A 10-fold cross-validation random forest classification (RFC) model was used to select the optimal metabolites panels for predicting the development of diabetes, and to internally validate the discriminatory capability of the selected metabolites beyond conventional clinical risk factors.

Findings: Metabolic alterations, including those associated with amino acid and lipid metabolism, were associated with an increased risk of prediabetes progressing to diabetes. The most important metabolites were inosine [odds ratio (OR) = 19.00; 95% confidence interval (CI): 4.23-85.37] and carvacrol (OR = 17.63; 95% CI: 4.98-62.34). Thirteen metabolites were found to improve T2D risk prediction beyond eight conventional T2D risk factors [area under the curve (AUC) was 0.98 for risk factors + metabolites vs 0.72 for risk factors, P < 0.05].

Interpretations: Use of the metabolites identified in this study may help determine patients with prediabetes who are at highest risk of progressing to diabetes.

Keywords: clinical; diabetes; metabolites; prediabetes; prediction.