A hierarchical Bayesian framework to infer the progression level to diabetes based on deficient clinical data

The increase in lifestyle-related diseases such as heart disease, diabetes, and high blood pressure is a challenging problem that should be resolved. The physiological mechanisms of the human body have long been studied using mathematical models. In particular, to study glucose metabolism, several models that infer insulin sensitivity and β-cell function have been developed. The use of mathematical models to assess progression to diabetes based on clinical data could be effective for preventing the onset of diabetes. However, to assess the progression level, we need clinical data including data from oral glucose tolerance tests, which are not typically performed on patients whose glucose tolerance may be impaired. To address this shortcoming, we developed a hierarchical Bayesian framework to infer the progression of glucose intolerance based on deficient data. We demonstrated how the framework infers the level of progression to diabetes and showed that glucose disposal capacity and insulin-secretory function depend on the fasting glucose and glycated hemoglobin (HbA1c) levels.