Predictive Models for Type 2 Diabetes Mellitus in Han Chinese with Insights into Cross-Population Applicability and Demographic Specific Risk Factors.

Abstract

Background: The rising global incidence of type 2 diabetes mellitus (T2DM) underscores the need for predictive models that enhance early detection and prevention across diverse populations. This study aimed to identify predictors of incident T2DM within a Han Chinese population, assess their impact across various age and sex demographics, and explore their applicability to European populations.

Methods: Using data from about 65,000 participants in the Taiwan Biobank (TWB), we developed a predictive model, achieving an area under the receiver operating characteristic curve of 90.58%. Key predictors were identified through LASSO regression within the TWB cohort and validated using over 4 million records from Taiwan's Adult Preventive Healthcare Services (APHS) program and the UK Biobank (UKB).

Results: Our analysis highlighted 13 significant predictors, including established factors like glycosylated hemoglobin (HbA1c) and blood glucose levels, and less conventionally considered variables such as peak expiratory flow. Notable differences in the effects of HbA1c levels and polygenic risk scores between the TWB and UKB cohorts were observed. Additionally, age and sex-specific impacts of these predictors, detailed through APHS data, revealed significant variances; for instance, waist circumference and diagnosed mixed hyperlipidemia showed greater impacts in younger females than in males, while effects remained uniform across male age groups.

Conclusion: Our findings offer novel insights into the diagnosis and management of diabetes for the Han Chinese and potentially for broader East Asian populations, highlighting the importance of ethnic and demographic diversity in developing predictive models for early detection and personalized intervention strategies.