Objective: To develop and prospectively validate a machine learning-based, multi-modal personalized therapy model (Perso-Endo-T2D) to guide the selection of initial glucose-lowering medication in treatment-naïve patients with Type 2 Diabetes (T2D). Methods: A prospective, randomized controlled trial was conducted across four endocrine centers. 450 drug-naïve T2D patients were randomized (1:1) to either: (1) Standard Care (following ADA/EASD guidelines, typically Metformin first) or (2) Model-Guided Care. The Perso-Endo-T2D model was trained on historical data, integrating pharmacogenomic data (25 known drug-response variants), clinical phenotyping (HOMA-IR, HOMA-B, C-peptide, BMI), and 30-day baseline digital activity data. The model provided a ranked list of predicted drug efficacy (Metformin, SGLT2-i, GLP1-RA) for the Model-Guided arm. The primary outcome was the mean reduction in HbA1c at 6 months. Results: Baseline characteristics were matched. The Model-Guided group achieved a significantly greater mean HbA1c reduction compared to the Standard Care group (-1.95% vs. -1.38%; mean difference -0.57%; 95% CI -0.71 to -0.43; p<0.001). Furthermore, 76.0% (n=171/225) of the Model-Guided group reached the target HbA1c <7.0%, compared to 55.1% (n=124/225) in the Standard Care group (p<0.001). The model successfully identified a "Metformin Low-Responder" subgroup (based on specific SNPs) and routed them to SGLT2-i therapy, avoiding 3-6 months of suboptimal control. Conclusion: A multi-modal, machine learning-driven personalized therapy model is significantly more effective at achieving glycemic control than the standard, guideline-based approach. This represents a paradigm shift from reactive, stepwise care to precise, individualized endocrine management, with broad implications for other metabolic diseases.

Personalized medicine, endocrinology, Type 2 Diabetes (T2D), machine learning, clinical decision support, pharmacogenomics, polygenic risk score (PRS), multi-modal data, precision medicine.

Ahlqvist, E., Storm, P., Käräjämäki, A., Martinell, M., Dorkhan, M., Carlsson, A., ... & Groop, L. (2018). Novel subgroups of adult-onset diabetes and their association with outcomes: a data-driven cluster analysis of six variables. The Lancet Diabetes & Endocrinology, 6(5), 361-369.

Davies, M. J., Aroda, V. R., Collins, B. S., Gabbay, R. A., Green, J., Maruthur, N. M., ... & Buse, J. B. (2022). Management of hyperglycemia in type 2 diabetes, 2022. A consensus report by the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD). Diabetes Care, 45(11), 2753-2786.

Global Burden of Disease (GBD) 2019 collaborators. (2020). Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. The Lancet, 396(10258), 1204-1222.

Zhou, K., Yee, S. W., Seiser, E. L., van Leeuwen, N., Tavendale, R., Bennett, A. J., ... & Pearson, E. R. (2016). Variation in the glucose transporter gene SLC22A1 is associated with metformin response in patients with type 2 diabetes. Nature Genetics, 48(9), 1055-1059.

2. Skripkin Y.K., Butov Y.S. (2009). Clinical Dermatovenereology: A Guide for Physicians (Vol. 2). Moscow: GEOTAR-Media.

3. Arifov S.S., Tursunov B.S. (2017). The role of immune disorders in the pathogenesis of chronic pyoderma and the ways of their correction. European Science Review, (1-2), 55-57.

4. Pinegin B.V., Khaitov R.M. (1996). Immunomodulators and some problems of modern immunotherapy. Klinicheskaya Meditsina, (8), 13-18.