Abstract:Background: Type 2 diabetes (T2D) is highly heterogeneous with varied patterns of clinical presentation. Machine learning, based on simple parameters, could help to classify T2D with its phenotypes. However, the association between profiled subgroups of T2D and diabetes-related outcomes, such as diabetic retinopathy, has not been outlined. Aim: We included newly diagnosed diabetic patients at baseline to identify novel T2D subgroups in Chinese general population based on cluster analysis with four routinely available and diabetes-related variables, that is, body mass index (BMI), waist circumference (WC), mean arterial pressure (MAP), and hemoglobin A1c (HbA1c). Furthermore, we intended to demonstrate the association of different clusters with diabetic retinopathy. Method: We developed a k-means clustering model in patients with newly diagnosed diabetes (N=1,910) from SENSIBLE-cohort at baseline in the Scikit‐learn library of Python. Clusters were based on four variables: BMI, WC, MAP, and HbA1c. Diabetic retinopathy was graded according to the Early Treatment Diabetic Retinopathy Study (ETDRS) scheme by two ophthalmologists, which was defined as ETDRS ≥ 20. 38.7% of the patients were followed at 3-year to explore the association between risk of diabetic retinopathy and different clusters by logistic regression. Results: Three clusters were identified. Cluster 0, which was featured by the lowest BMI, WC, MAP and HbA1c, accounted for the highest proportion of observations (48.2%); and was followed by cluster 2 (42.9%) featured by the highest BMI, WC and MAP; and cluster 1 (8.9%) featured by poor glycemic condition. The distribution pattern of the clusters differed across ethnic groups, and 26.1% of observations transitioned clusters between baseline and 3-year follow up point of measurement. Moreover, cluster 1 had the highest risk of diabetic retinopathy both at baseline (OR 6.248, 95%CI: 3.191-12.230, P <0.001) and follow-up (OR 9.158, 95%CI: 2.079-40.342, P =0.003) after adjustment. No significant association between the risk of diabetic retinopathy and cluster 2 was observed neither at baseline nor 3-year follow-up.Tabled 1Cluster 0Cluster 1Cluster 2PWC, cm80.8 ± 7.487.6 ± 9.796.5 ± 8.0<0.001BMI, km/m224.1 ± 2.525.9 ± 3.629.8 ± 3.3<0.001MAP, mmHg100 ± 12103 ± 14110 ± 13<0.001HbA1c, %6.2 ± 0.810.8 ± 1.86.5 ± 0.9<0.001 Open table in a new tab Discussion: Cluster analysis has proven to successfully identify patients with diabetes and link these clusters to clinically important outcomes. However, those clusters were based on sophisticated predictors, which are not available for every patient in rural areas or large population-based epidemiological surveys. In our study, novel cluster‐based subgroups of diabetes can be implemented in Chinese general population with T2D through simple and cost-effective indicators. These clusters could be used for risk stratification based on the underlying profile of patients with T2D. Furthermore, the first step towards precision medicine in diabetes and targeted early treatment to diabetic retinopathy can be approached based on the clustering profile at diagnosis. Patients featured by high blood glucose should pay more attention to the risk of diabetic retinopathy than those featured by high blood pressure and adiposity.

New clusters of serum electrolytes aid in stratification of diabetes and metabolic risk

Novel clusters of newly-diagnosed type 2 diabetes and their association with diabetic retinopathy: a 3-year follow-up study

Unsupervised cluster analysis of clinical and metabolite characteristics in patients with chronic complications of T2DM: an observational study of real data

IDF21-0141 Novel Clusters of Type 2 Diabetes and Their Association with Diabetic Retinopathy: a 3-year follow-up study

Circulating metabolite profile in young adulthood identifies long-term diabetes susceptibility: the Coronary Artery Risk Development in Young Adults (CARDIA) study

Clinical Characteristics and Risk of Diabetic Complications in Data-Driven Clusters Among Type 2 Diabetes

Serum Lipids in Association with Type 2 Diabetes Risk and Prevalence in a Chinese Population

Clinical characteristics and complication risks in data-driven clusters among Chinese community diabetes populations

Alterations of serum trace elements in patients with type 2 diabetes

High-Coverage Targeted Lipidomics Reveals Novel Serum Lipid Predictors and Lipid Pathway Dysregulation Antecedent to Type 2 Diabetes Onset in Normoglycemic Chinese Adults.

Serum lipidomics profiles reveal potential lipid markers for prediabetes and type 2 diabetes in patients from multiple communities

Cluster-based subgroups of prediabetes and its association with prediabetes progression and regression: a prospective cohort study

Identification of Two Novel Subgroups in Patients with Diabetes Mellitus and Their Association with Clinical Outcomes: A Two-Step Cluster Analysis

Novel Subgroups and Chronic Complications of Diabetes in Middle-Aged and Elderly Chinese:A Prospective Cohort Study

Metabolic Profiles of Type 2 Diabetes and Their Association with Renal Complications.

Clusters of carbohydrate-rich foods and associations with type 2 diabetes incidence: a prospective cohort study

1252-P: Differences in Cardiometabolic and Immune Biomarkers between Novel Subphenotypes of Individuals at Elevated Risk for Type 2 Diabetes

1389-P: Association Between Serum Metabolomic and Overall Mortality among Individuals with Type 2 Diabetes

Plasma Metabolomic Profiles in Association with Type 2 Diabetes Risk and Prevalence in Chinese Adults

Phenotyping lipid profiles in type 2 diabetes: Risk association and outcomes from the Cardiovascular Health Study

Circulating Metabolites and the Development of Type 2 Diabetes in Chinese Adults.