An analysis recently published in Cardiovascular Diabetology used a cluster-based approach to quantify the risk for developing type 2 diabetes (T2D) among older adults (aged 61-82). Building on a previously published proof-of-concept study, 843 participants in the Kora F4 study were assigned to one of six predefined phenotype clusters, with clusters determined by a number of metabolic and clinical characteristics shown to be predictive of diabetes incidence, mortality, and prediabetes complications in a middle-aged population.
Age, BMI, waist and hip circumference, fasting and 2-hour glucose and insulin were among the variables, with the 6 clusters roughly defined as follows (in order from 1 to 6): low risk, very low risk, beta-cell failure, low risk obese, high risk insulin-resistant fatty liver, and cluster 6, high risk visceral fat nephropathy. Additionally, a 73-analyte proteomics panel that included pro- and anti-inflammatory cytokines, chemokines, and other biomarkers of inflammation was included for each participant.
Clustering was associated with differences in T2D incidence as well as cardiometabolic risk, the incidence and prevalence of kidney disease and distal sensorimotor polyneuropathy, inflammatory load, and the prevalence of cardiovascular disease. For example, an odds ratio of at least 32 for T2D was found when comparing cluster 5, which had the highest incidence (53%), to cluster 2 with the lowest incidence (3.2%). The difference in inflammatory load was also the greatest between these two clusters. Notably, clusters 3 and 6 had an increase in the incidence of T2D, but did not have an elevation in inflammatory markers, pointing to potentially different pathways driving the diabetic process in different people, i.e., subtypes of diabetes. Lastly, it’s worth noting that although C-reactive protein was highest in cluster 5, it had only a small role in predicting inflammatory load. Given the heterogeneity among people with diabetes, understanding the factors underlying its development and risk is welcome.