Aim: The lack of longitudinal metabolomics data and the statistical techniques to analyse them has limited the understanding of the metabolite levels related to type 2 diabetes (T2D) onset. Thus, we carried out logistic regression analysis and simultaneously proposed new approaches based on residuals of multiple logistic regression and geometric angle-based clustering for the analysis in T2D onset-specific metabolic changes.
Materials and methods: We used the sixth, seventh and eighth follow-up data from 2013, 2015 and 2017 among the Korea Association REsource (KARE) cohort data. Semi-targeted metabolite analysis was performed using ultraperformance liquid chromatography/triple quadrupole-mass spectrometry systems.
Results: As the results from the multiple logistic regression and a single metabolite in a logistic regression analysis varied dramatically, we recommend using models that consider potential multicollinearity among metabolites. The residual-based approach particularly identified neurotransmitters or related precursors as T2D onset-specific metabolites. By using geometric angle-based pattern clustering studies, ketone bodies and carnitines are observed as disease-onset specific metabolites and separated from others.
Conclusion: To treat patients with early-stage insulin resistance and dyslipidaemia when metabolic disorders are still reversible, our findings may contribute to a greater understanding of how metabolomics could be used in disease intervention strategies during the early stages of T2D.
Keywords: T2D onset-specific metabolic changes; longitudinal metabolomic study; metabolomics; type 2 diabetes (T2D).
© 2023 The Authors. Diabetes, Obesity and Metabolism published by John Wiley & Sons Ltd.