RESUMO
INTRODUCTION: Type 2 diabetes mellitus (T2D) is highly heterogeneous in disease progression and risk of complications. This study aimed to categorize Thai T2D into subgroups using variables that are commonly available based on routine clinical parameters to predict disease progression and treatment outcomes. RESEARCH DESIGN AND METHODS: This was a cohort study. Data-driven cluster analysis was performed using a Python program in patients with newly diagnosed T2D (n=721) of the Siriraj Diabetes Registry using five variables (age, body mass index (BMI), glycated hemoglobin (HbA1c), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C)). Disease progression and risk of diabetic complications among clusters were compared using the Χ2 and Kruskal-Wallis test. Cox regression and the Kaplan-Meier curve were used to compare the time to diabetic complications and the time to insulin initiation. RESULTS: The mean age was 53.4±11.3 years, 58.9% were women. The median follow-up time was 21.1 months (9.2-35.2). Four clusters were identified: cluster 1 (18.6%): high HbA1c, low BMI (insulin-deficiency diabetes); cluster 2 (11.8%): high TG, low HDL-C, average age and BMI (metabolic syndrome group); cluster 3 (23.3%): high BMI, low HbA1c, young age (obesity-related diabetes); cluster 4 (46.3%): older age and low HbA1c at diagnosis (age-related diabetes). Patients in cluster 1 had the highest prevalence of insulin treatment. Patients in cluster 2 had the highest risk of diabetic kidney disease and diabetic retinopathy. Patients in cluster 4 had the lowest prevalence of diabetic retinopathy, nephropathy, and insulin use. CONCLUSIONS: We were able to categorize Thai patients with newly diagnosed T2D into four clusters using five routine clinical parameters. This clustering method can help predict disease progression and risk of diabetic complications similar to previous studies using parameters including insulin resistance and insulin sensitivity markers.