Heterogeneity in Phenotype and Early Metabolic Response to Lifestyle Interventions in Type 2 Diabetes in China Using a Tree-Like Representation

Deciphering the heterogeneity of type 2 diabetes in prognosis and treatment effect is essential. We used a novel dimensionality reduction approach to describe the type 2 diabetes phenotype continuum and visualize the difference in lifestyle intervention efficacy in Chinese patients. Based on 17,816 participants with newly diagnosed type 2 diabetes (aged ≥40 years) from a nationwide cohort, 12 key phenotypes were residualized for age and sex to construct a two-dimensional tree structure. The tree demonstrated the continuous type 2 diabetes spectrum and region-specific characteristics, with a mixed phenotypic trunk and three extreme phenotypic branches. When mapping data from 325 participants with type 2 diabetes from a randomized controlled trial onto the original tree, lifestyle intervention induced a migration toward the left part of tree, indicating an overall metabolic improvement. Specifically, diet intervention was more effective for glycemic control in the upper part of the tree and featured moderate diabesity and elevated insulin, whereas exercise intervention was more effective for glycemic control in the left side of the tree and featured less adiposity and better overall metabolic status. In summary, this analysis depicted the tree structure representing the underlying pathophysiological features of patients with newly diagnosed type 2 diabetes and identified tree regions with different sensitivity to diet or exercise intervention. The results have the potential to aid lifestyle intervention selection. Article Highlights Deciphering the heterogeneity of diabetes is essential for prognosis prediction and treatment guidance, but current classifications are flawed because they lose continuous phenotypic information. We wanted to determine if the novel data reduction method, the data dimensionality reduction tree (DDRTree), is applicable to visualizing the phenotypic continuum, comorbid conditions, and lifestyle intervention effects in Chinese patients with type 2 diabetes. The DDRTree structure demonstrated the region-specific characteristics of type 2 diabetes. Diet intervention was more effective for glycemic control in the upper part of the tree, featuring moderate diabesity, whereas exercise intervention was more effective in the left side of the tree, featuring less adiposity and better overall metabolic status. The Chinese type 2 diabetes tree structure indicates individualized pathophysiology and guides the selection of lifestyle intervention.