This study used nuclear magnetic resonance (NMR)–based metabolomic measurements in a sample of 7,671 participants from the UK Biobank who had previously been classified into data-driven type 2 diabetes (T2D) subtypes. The authors aimed to determine whether metabolomic signatures differed across those subtypes and whether subtype-related metabolic patterns were associated with incident dementia, stroke, and all-cause mortality. All analyses reported here come from the preprint abstract and related sections of the source; full methods and additional details are available from the authors upon request according to the data availability statement.
Metabolites that differentiated the data-driven T2D subtypes were summarized using principal component analysis (PCA). The first five principal components (PCs) explained 76.7% of the variance in these subtype-related metabolites. According to the abstract, these PCs mainly reflected lipid and lipoprotein signatures rather than non-lipid metabolites. The PCA-derived axes were then used as exposure variables in time-to-event models examining incident clinical outcomes.
Cox proportional hazards models evaluated associations between the PCA-derived metabolomic axes and incident dementia outcomes after adjustment for T2D subtype and relevant confounders. A metabolomic axis described as an HDL-remodeling principal component was associated with higher risks of all-cause dementia (hazard ratio [HR] 1.17, 95% confidence interval [CI] 1.08–1.27) and vascular dementia (VaD) (HR 1.18, 95% CI 1.05–1.32). Another axis characterized as an LDL cholesterol–enriched profile showed an inverse association with all-cause dementia (HR 0.75, 95% CI 0.62–0.91), such that lower scores on this LDL-enriched axis were associated with increased dementia risk in the adjusted models.
No statistically significant associations between the PCA-derived metabolomic axes and incident stroke were reported in the analyzed sample. By contrast, several metabolomic PCs were associated with all-cause mortality. The HDL-remodeling PC was associated with higher all-cause mortality (HR 1.16, 95% CI 1.13–1.19). Lower scores on the LDL-enriched axis were also associated with increased all-cause mortality (HR 0.76, 95% CI 0.69–0.83). Additionally, a VLDL/LDL-enriched PC was inversely associated with mortality (HR 0.93, 95% CI 0.88–0.98). These effects indicate that specific lipid and lipoprotein patterns captured by PCA related to differential risks of death in this cohort.
The investigators performed attenuation analyses in which adjustment for the PCA-derived metabolomic patterns was applied to previously observed associations between T2D subtypes and clinical outcomes. The abstract reports that adjusting for the metabolomic PCs generally attenuated subtype–outcome associations, implying that metabolomic features partially mediate or explain differences in dementia and mortality risk between T2D subtypes. Exact magnitudes of attenuation and which subtype–outcome pairs were most affected are not detailed in the abstract.
To further probe causal relationships, the study used two-sample Mendelian randomization (MR). MR analyses identified 197 metabolite–outcome associations that remained statistically significant after false discovery rate (FDR) correction, suggesting that some metabolites may have causal roles in the outcomes examined. The abstract does not provide the full list of metabolites implicated, the genetic instruments used for each metabolite, or the detailed MR estimates; those elements are reported in the full preprint and supporting materials available from the authors.
In this UK Biobank–based analysis of 7,671 individuals with data-driven T2D subtypes, metabolic profiling revealed distinct lipid and lipoprotein signatures across subtypes. PCA-derived axes capturing HDL remodeling, LDL cholesterol enrichment, and VLDL/LDL patterns were associated with dementia subtypes and all-cause mortality, while no robust associations with stroke were observed in the reported analyses. Adjustment for metabolomic patterns attenuated subtype–outcome associations, and MR results identified numerous metabolite–outcome links that survived FDR correction, supporting potential causal relevance for some metabolite measures.
These findings—derived from a preprint that has not been peer reviewed—highlight lipid and lipoprotein remodeling as prominent metabolic features related to dementia and mortality risk in people with data-driven T2D subtypes. For full methodological details, complete lists of implicated metabolites, PCA loadings, and MR instrument information, readers should consult the full preprint and available supplementary materials, as the abstract does not report those specific details.