Asian Journal of Research and Reports in Endocrinology

Diabetes mellitus is no longer understood as a single disease but as a heterogeneous collection of conditions unified by chronic hyperglycaemia and diverging in aetiology, pathophysiology, clinical trajectory and treatment response. Precision medicine seeks to exploit this heterogeneity by matching diagnostic categories, risk predictions and therapeutic choices to the biological and clinical characteristics of the individual rather than to the population average. This critical narrative review evaluates the state of precision medicine across diabetes, synthesising evidence on monogenic diabetes, data-driven and genetic subclassification of type 2 diabetes, pharmacogenetic and phenotype-guided treatment selection, staging and immunotherapy in type 1 diabetes, technology-enabled phenotyping through continuous glucose monitoring and machine learning, precision nutrition, heterogeneity in gestational diabetes, polygenic risk prediction, and artificial intelligence in complication screening.

Literature was identified through structured searching of biomedical and multidisciplinary databases together with citation tracking of major consensus statements, restricted to sources whose bibliographic identity and DOI could be independently verified. The review finds that precision medicine has achieved unambiguous clinical translation only in monogenic diabetes, where molecular diagnosis directly redirects treatment with well-replicated benefit. In type 2 diabetes, cluster-based and genetically informed subtyping schemes have consistently identified reproducible axes of heterogeneity related to insulin secretion, insulin resistance and autoimmunity, and phenotype-guided treatment-selection models, including a prospectively validated crossover trial and a large observational prediction model, show modest but real improvements in glycaemic outcomes. Evidence for precision approaches in type 1 diabetes staging and immunotherapy, technology-enabled phenotyping, nutrition and gestational diabetes is more preliminary, frequently derived from single-centre or homogeneous-ancestry cohorts, and constrained by weak external validation.

Across domains, three recurring limitations restrict clinical translation: the transferability of genetic and phenotypic prediction tools across ancestrally diverse populations remains poor; prospective, outcome-based validation is rare relative to the volume of descriptive subtyping studies; and the infrastructure, cost and workforce requirements for routine implementation are seldom evaluated. The review distinguishes well-supported precision applications from promising but preliminary ones and proposes prioritised directions for future research, emphasising prospective trials of treatment stratification, diversification of genomic and phenotypic reference datasets, and formal evaluation of implementation feasibility in resource-constrained settings. Precision medicine in diabetes has moved from a conceptual aspiration towards selective, evidence-based clinical application, but its broad realisation across the diabetes spectrum remains incomplete.