Scientific Programme
Abstract
Register
Abstract
Year 2021
September 2021

SHBC1300

Abstract Title
Machine learning-based Clustering in Two Prospective Cohorts Reveals Distinct Clinical and Lipidomic Profiles in Multi-Ethnic Asians with Recent-Onset Type 2 Diabetes Mellitus: Impacts on Cardiorenal Outcomes
Authors

J.WANG1, J.J.LIU1, A.MC.MOH1, S.KW.WONG2, Y.M.SHAO1, K.K.L.ANG1, J.H.CHING3, JP.KOVALIK3, F.TORTA4, A.YL.LIU5, S.TAVINTHARAN5, C.F. SUM5, S.C.LIM1

Institutions

Khoo Teck Puat Hospital1, NHG Polyclinics2, Duke-NUS Graduate Medical School3, National University of Singapore4, Admiralty Medical Centre5

Background & Hypothesis

Partitioning adult-onset type 2 diabetes (T2D) population into relatively homogenous subgroups is under-studied in Asians, who tend to develop diabetes at younger age with lower BMI, and have higher incidences of complications than other ethnic populations. This study aims to cluster multi-ethnic Asians with recent-onset T2D, and investigate clinical and lipidomic profiles of the derived clusters.

Methods

A total of 1402 T2D subjects with diabetes duration ≤ 5 years were screened from two prospective T2D cohorts. K-means algorithm was run to partition the subjects based on baseline age of onset, BMI, HbA1c, β-cell function and insulin resistance­­. Up to 316 lipid species were quantified from plasma samples by targeted mass spectrometry. Clinical endpoints included all-cause mortality and cardiorenal diseases (heart failure, MACE and progressive DKD).

Results

Three clusters with distinct clinical profiles were identified: cluster 1 (19.0%) with younger onset, obese, severe insulin resistant (IR) and decompensated β-cell response; cluster 2 (44.8%) with classical T2D phenotypes; and cluster 3 (36.2%) with older onset, relatively lean and insufficient beta-cell function. Cluster 3 had high risk of DKD. Cluster 1 had higher risks of both DKD and heart failure. Besides, C18-ceramides were increased in cluster 1, suggesting lipotoxicity-induced IR and beta-cell dysfunction. Elevated glycerophospholipids with plasmalogens may function as potential precursors of inflammatory mediators in both cluster 1 and 3.

Discussion & Conclusion

The disparate clinical and lipidomic characteristics in the derived T2D clusters may suggest dissimilar pathogenesis and a need for subtype-specific interventions to improve lipid metabolism and prevent cardiorenal disease progression.

Top