Cardiorenal Risk Profiles Among Data-Driven Type 2 Diabetes Sub-Phenotypes: A Post-Hoc Analysis of the China Health and Nutrition Survey.

Background and Aim: Evidence about recently proposed data-driven clusters of type 2 diabetes (T2D) is mainly about its prognostic effects and Western populations. We tested the applicability of this clustering approach among the Chinese population. We further investigated the cardiorenal risk profiles among different T2D sub-phenotypes cross-sectionally and before diabetes diagnosis. Methods: With the use of data from the China Health and Nutrition Survey (1989-2009), 6,728 participants with available fasting blood samples and completed questionnaires in the 2009 survey were included. Glycemic statuses (normoglycemia, prediabetes, and new-onset T2D) were defined according to the 2020 American Diabetes Association criteria. Data-driven cluster analysis was conducted among new-onset T2D based on five variables: age at onset, body mass index (BMI), hemoglobin A1c, homeostasis model estimates of ß-cell function, and insulin resistance. Linear regression models were used to cross-sectionally examine the differences of cardiorenal risk factors (body fat distribution, blood pressure, lipid profiles, and kidney function) between glycemic statuses. Mixed-effects models were used to explore a maximum of 20-year trajectories of cardiovascular risk factors (body fat distribution and blood pressure) before diabetes diagnosis. Results: Among 557 (8.3%) new-onset T2D, four sub-phenotypes were found, with 57 (10.2%) assigned to the severe insulin-resistant diabetes (SIRD), 72 (12.9%) to the severe insulin-deficient diabetes (SIDD), 167 (30.0%) to the mild obesity-related diabetes (MOD), and 261 (46.9%) to the mild age-related diabetes (MARD). People clustered within different T2D sub-phenotypes had different cardiorenal risk profiles. Three T2D sub-phenotypes (SIRD, SIDD, and MOD) had worse cardiorenal abnormalities, while the risk burden in the MARD sub-phenotype was similar to that in prediabetes. Compared with people with other T2D sub-phenotypes, people in the MOD sub-phenotype had a faster increment in BMI, waist, upper arm circumference, and triceps skinfold up to 10 years before diagnosis. Blood pressure was less distinct in different T2D sub-phenotypes; however, SIDD and MOD clusters had higher blood pressure levels before diabetes diagnosis. Conclusions: Data-driven T2D sub-phenotyping is applicable in the Chinese population. Certain sub-phenotypes such as MARD only have a minor cardiorenal risk burden, and distinct cardiovascular risk development occurs long before diabetes diagnosis. Our findings can help improve early prevention and targeted treatment for diabetes.