Postprandial Metabolite Profiles and Risk of Prediabetes in Young People: A Longitudinal Multicohort Study.

Abstract

Prediabetes in young people is an emerging epidemic that disproportionately impacts Hispanic populations. We aimed to develop a metabolite-based prediction model for prediabetes in young people with overweight/obesity at risk for type 2 diabetes. In independent, prospective cohorts of Hispanic youth (discovery; n = 143 without baseline prediabetes) and predominately Hispanic young adults (validation; n = 56 without baseline prediabetes), we assessed prediabetes via 2-h oral glucose tolerance tests. Baseline metabolite levels were measured in plasma from a 2-h postglucose challenge. In the discovery cohort, least absolute shrinkage and selection operator regression with a stability selection procedure was used to identify robust predictive metabolites for prediabetes. Predictive performance was evaluated in the discovery and validation cohorts using logistic regression. Two metabolites (allylphenol sulfate and caprylic acid) were found to predict prediabetes beyond known risk factors, including sex, BMI, age, ethnicity, fasting/2-h glucose, total cholesterol, and triglycerides. In the discovery cohort, the area under the receiver operator characteristic curve (AUC) of the model with metabolites and known risk factors was 0.80 (95% CI 0.72-0.87), which was higher than the risk factor-only model (AUC 0.63 [0.53-0.73]; P = 0.001). When the predictive models developed in the discovery cohort were applied to the replication cohort, the model with metabolites and risk factors predicted prediabetes more accurately (AUC 0.70 [95% CI 40.55-0.86]) than the same model without metabolites (AUC 0.62 [0.46-0.79]). Metabolite profiles may help improve prediabetes prediction compared with traditional risk factors. Findings suggest that medium-chain fatty acids and phytochemicals are early indicators of prediabetes in high-risk youth.