A Robust Multivariate Approach Unveils a ¹ H-NMR-Based Spectral Metabolomics Signature Predictive of Cardiovascular Disease in People with Type 2 Diabetes: A Prospective Di@bet.es Cohort Study

People with type 2 diabetes mellitus (T2DM) are at high risk of cardiovascular disease (CVD), which is often not explained by traditional risk factors such as low-density lipoprotein (LDL) cholesterol. Proton nuclear magnetic resonance (¹H NMR) metabolomics is a promising tool to help explain this residual risk. This study aimed to evaluate the incidence of CVD in people with T2DM via untargeted ¹H NMR data. The ¹H NMR raw spectra of 24 cases and 24 controls (with basal T2DM and with/without CVD at follow-up) matched by age, sex, body mass index, and LDL cholesterol from the Di@bet.es cohort were processed, and the peak-picked features (p = 269) were used in a partial least-squares discriminant analysis classification with repeated double cross-validation and validated against permuted data sets (AUC = 0.758; p-value = 0.011). For each feature, a stringent variable selection method analyzing the distributions of variable importance in projection scores and beta coefficients across all the repeated models was used, yielding a metabolomic signature composed of 16 selected features related to inflammation, triglycerides, muscular function, and HDL particles, together with features putatively arising from albumin, although further validation of the annotations is needed. In summary, untargeted ¹H NMR metabolomics can help assess cardiovascular risk in people with T2DM beyond LDL cholesterol.