Unsupervised Cluster Analysis of Diabetes Mellitus: A Systemic Review from Eastern Indian Population

Devendra Prasad Singh, M Kamran Khan

Abstract: Background: Diabetes mellitus, a complex metabolic disorder, presents significant challenges in patient management due to its heterogeneity. Unsupervised cluster analysis has emerged as a promising approach for unraveling this complexity. This systematic review evaluates the effectiveness of unsupervised cluster analysis in identifying diabetes phenotypes, assessing complication risks, and differentiating treatment responses. Methods: We explored Embase, PubMed, and Scopus, evaluating 38 pertinent studies. Furthermore, a cross-sectional study was performed using K-means cluster analysis on real-world clinical data from 625 patients with diabetes. Results: The analysis consistently identified five reproducible clusters (MOD, MARD, SAID, SIDD, SIRD) across diverse populations, spanning various ethnicities and patient origins. The MOD (mild obesity-related diabetes) and MARD (mild age-related diabetes) clusters were most prevalent, while SAID (severe autoimmune diabetes) was least common. Subgroup analysis by ethnicity showed a higher prevalence of SIDD (severe insulin-deficient diabetes) among individuals of Asian descent. These clusters shared similar phenotypic traits and complication risk profiles, with variations in distribution and key clinical variables, such as glycemic control, lipid metabolism, and renal function. Notably, the SIRD (severe insulin-resistant diabetes) subtype was strongly associated with diverse kidney-related outcomes. Alternative clustering techniques may reveal additional clinically relevant subtypes. Our cross-sectional study identified five subgroups with distinct profiles in glycemic control, lipid metabolism, blood pressure, and kidney function. Conclusions: Unsupervised cluster analysis demonstrates significant potential for identifying clinically meaningful diabetes subgroups with distinct characteristics, complication risks, and treatment responses, which may remain undetected using conventional methods.

Keywords: Diabetes mellitus, unsupervised cluster analysis, K-means clustering, diabetes phenotypes, MOD, MARD, SAID, SIDD, SIRD

How to Cite?: Devendra Prasad Singh, M Kamran Khan, "Unsupervised Cluster Analysis of Diabetes Mellitus: A Systemic Review from Eastern Indian Population", Volume 14 Issue 9, September 2025, International Journal of Science and Research (IJSR), Pages: 841-845, https://www.ijsr.net/getabstract.php?paperid=SR25918163728, DOI: https://dx.doi.org/10.21275/SR25918163728