2026/07/25 by Tim Mori, Christian Herder, Pedro Cardoso +3
Computer Science · Medicine · Biochemistry, Genetics and Molecular Biology · #Machine Learning in Healthcare #Diabetes, Cardiovascular Risks, and Lipoproteins #Genetic Associations and Epidemiology
paper · doi:10.1007/s00125-026-06802-6
Growing interest in precision diagnostics has stimulated efforts to classify type 2 diabetes into more granular subtypes. Recent research has focused on four pathophysiology-based subtypes: severe insulin-deficient diabetes (SIDD); severe insulin-resistant diabetes (SIRD); moderate obesity-related diabetes (MOD); and moderate age-related diabetes (MARD). While these subtypes provide an attractive framework for precision medicine, they also come with methodological challenges. This narrative review highlights three key issues: (1) uncertainty about the existence of four discrete subtypes; (2) low certainty in subtype assignment; and (3) limited evidence for prognostic and therapeutic value compared with other approaches. As an alternative to subtyping, we discuss individualised prediction models, which provide personalised treatment recommendations based on routine clinical features. These models avoid some of the limitations of discrete subtyping but they do not provide insights into a person's underlying pathophysiology. Thus, these two approaches can be viewed as complementary, each serving a different purpose in precision medicine.