Type 2 diabetes shows clinical heterogeneity which cannot be fully captured by glycemic metrics, highlighting the need to better define patient phenotypes and progression pathways. This study aims to elucidate the heterogeneity of type 2 diabetes and risks of major associated diseases, as well as underlying proteomic mechanism. We applied discriminative dimensionality reduction with trees (DDRTree) algorithm to construct a tree from seven clinical variables (body mass index, high density lipoprotein cholesterol, triglyceride, HbA1c, systolic blood pressure, diastolic blood pressure, and total cholesterol). Disease risks were assessed using competing risk models. This study included 6406 individuals with newly diagnosed type 2 diabetes from the UK Biobank. All seven clinical variables formed a gradient distribution across the DDRTree-derived phenotypic structure, revealing three distinct disease risks patterns: participants with adiposity, hypertension and dyslipidemia exhibited elevated risks of macrovascular complications, diabetic kidney disease, Parkinson’s disease and non-alcohol fatty liver disease; those with hyperglycemia and dyslipidemia had higher risks of myocardial infarction, diabetic neuropathy, diabetic retinopathy, depression and chronic obstructive pulmonary disease; while elevated total cholesterol and high-density lipoprotein cholesterol were associated with increased risks of cancer and Alzheimer’s disease. Proteomic analyses identified pattern-specific pathways: metabolic dysregulation and extracellular matrix remodeling in the first pattern, inflammatory activation and lipid metabolism alterations in the second, and immune activation with chemical carcinogenesis in the third. Furthermore, sensitivity to lifestyle factors were phenotypic-specific. These patterns were similar in the Dongfeng-Tongji Diabetes cohort, though cardiovascular disease and retinopathy risks were strongly associated with hypertension. An online tool was provided for individual risk prediction. Our findings reveal distinct spatial distributions of clinical features and associated disease risks in type 2 diabetes, with subgroups exhibiting unique proteomic signatures and differential lifestyle responses, underscoring the importance for personalized management for diabetes care.
Che et al. (Mon,) studied this question.