Energy landscape analysis of health checkup data clarified multiple pathways to diabetes development in obese and non-obese subjects

Abstract

Aims: To clarify the pathways from a healthy state to the diabetes onset via pre-disease states, we applied energy landscape analysis (ELA) to Specific Health Checkup data in Japan.

Methods: This retrospective and observational cohort study analyzed data from 4,928 males aged 56.0 ± 3.2 years, including 242 individuals with diabetes, over a period of 5.26 ± 3.21 years. A total of 22,326 records were examined using six features: hemoglobin A1c, plasma glucose, high-density lipoprotein-cholesterol, body mass index (BMI), uric acid, and alanine aminotransferase. ELA was also applied to subdata from the 242 individuals with diabetes.

Results: ELA revealed three stable states: healthy, intermediate, and unhealthy (pre-diabetes) states. The intermediate state was characterized by obesity. Obese individuals with BMI ≥ 25 kg/m² (n = 1,460) preferred a pathway via the intermediate state, whereas non-obese individuals with BMI < 25 kg/m² (n = 3,468) preferred to transit directly to the unhealthy state. There was a significant difference between the preferences of the two groups (p = 0.0085, chi-squared test). Two distinct pathways were also observed for obese and non-obese individuals with diabetes.

Conclusions: We demonstrated that ELA could indicate different pathways of diabetes development in obese and non-obese individuals in a data-driven manner. These insights could inform more targeted diabetes prevention measures, such as reducing visceral fat in obese individuals and protecting beta-cells in non-obese individuals.

Keywords: diabetes; energy landscape analysis; multiple pathways; obesity; pre-disease state; specific health checkup data.