2371-P: Dimensionality Reduction with DDRTree Identifies Onset Age as the Key Clinical Phenotype and Supports Age-Related Endotypes in Type 1 Diabetes

Introduction and Objective: Type 1 diabetes (T1D) exhibits substantial heterogeneity, yet the key clinical phenotypes remain unclear. We applied DDRTree machine learning to identify the key phenotype characterizing T1D heterogeneity and validate the existence of age-related endotypes. Methods: We performed DDRTree dimensionality reduction analysis using data on 13 phenotypes (age at onset, BMI, blood pressure, ALT, HbA1c, creatinine, triglycerides, HDL-C, total cholesterol, GADA, IA-2A, ZnT8A) from 879 patients with T1D to identify the most important clinical phenotype. We then evaluated the rationality of classifying patients into age-related endotypes (<7, 7-13, ≥13 years) and compared clinical, immunological, and genetic profiles across groups to confirm the presence of age-related endotypes. Finally, we validated the existence of these endotypes in an external UK Biobank cohort and investigated longitudinal cardiovascular outcomes associated with age-at-onset phenotypes. Results: Age at onset emerged as the most important phenotype for T1D heterogeneity, explaining 64.6% of the variance in the DDRTree dimensionality reduction. Classification into age-related endotypes showed significant differences in DDRTree dimensional coordinates (p = 1.61×10-54), supporting the rationality of this grouping. Autoantibody positivity (IA-2A, ZnT8A: p < 0.001), HLA genetic risk (p < 0.001), and metabolic characteristics differed significantly, further confirming age-related endotypes. Moreover, phenotypic heterogeneity by onset age attenuated with disease duration in both the internal and UK Biobank cohorts. Regarding long-term outcomes, older-onset T1D patients exhibited higher cardiovascular risk (sHR = 2.93, 95% CI: 1.49-5.78 for ≥13 vs. <7 years). Conclusion: DDRTree analysis identified age at onset as the key phenotype of T1D heterogeneity. Dimensionality reduction demonstrated excellent discriminative performance, supporting the existence of age-related endotypes. Disclosure S. Chang: None. H. Tan: None. T. Yue: None. Y. Ding: None. Z. Gu: None. Y. Shi: None. L. Pan: None. C. Guo: None. J. Weng: None. X. Zheng: None. Funding Foundation programs supporting included Noncommunicable Chronic Diseases-National Science and Technology Major Project (2023ZD0509100; 2023ZD0509102), the National Natural Science Foundation (8247087), and Program for Innovative Research Team of The First Affiliated Hospital of University of Science and Technology of China (CXGG02)