Sasang constitutional types for the risk prediction of metabolic syndrome: a 14-year longitudinal prospective cohort study

Abstract

BackgroundTo examine whether the use of Sasang constitutional (SC) types, such as Tae-yang (TY), Tae-eum (TE), So-yang (SY), and So-eum (SE) types, increases the accuracy of risk prediction for metabolic syndrome.MethodsFrom 2001 to 2014, 3529 individuals aged 40 to 69 years participated in a longitudinal prospective cohort. The Cox proportional hazard model was utilized to predict the risk of developing metabolic syndrome.ResultsDuring the 14 year follow-up, 1591 incident events of metabolic syndrome were observed. Individuals with TE type had higher body mass indexes and waist circumferences than individuals with SY and SE types. The risk of developing metabolic syndrome was the highest among individuals with the TE type, followed by the SY type and the SE type. When the prediction risk models for incident metabolic syndrome were compared, the area under the curve for the model using SC types was significantly increased to 0.8173. Significant predictors for incident metabolic syndrome were different according to the SC types. For individuals with the TE type, the significant predictors were age, sex, body mass index (BMI), education, smoking, drinking, fasting glucose level, high-density lipoprotein (HDL) cholesterol level, systolic and diastolic blood pressure, and triglyceride level. For Individuals with the SE type, the predictors were sex, smoking, fasting glucose, HDL cholesterol level, systolic and diastolic blood pressure, and triglyceride level, while the predictors in individuals with the SY type were age, sex, BMI, smoking, drinking, total cholesterol level, fasting glucose level, HDL cholesterol level, systolic and diastolic blood pressure, and triglyceride level.ConclusionsIn this prospective cohort study among 3529 individuals, we observed that utilizing the SC types significantly increased the accuracy of the risk prediction for the development of metabolic syndrome.

Metabolic Syndrome Prediction Models Using Machine Learning and Sasang Constitution Type

“…In a previous machine learning study, constitution type was a significant predictor of MetS, as were total bilirubin and low-density lipoprotein cholesterol [ 32 ]. Increased performance was also indicated in a previous study in which Lee et al [ 13 ] used statistical analysis to investigate a MetS prediction model and reported that the AUC increased significantly after the incorporation of Sasang constitution type. Sasang constitution type needs to be considered in the development of MetS prediction models in future studies.…”

“…In another study conducted in Taiwan, a risk prediction model for MetS yielded an AUC of 0.83 in subjects aged 35–74 years [ 26 ]. In a Korean study involving a 14-year longitudinal prospective cohort, the AUC of a prediction model was 0.81 without the incorporation of Sasang type and 0.82 with the incorporation of Sasang type [ 13 ]. Studies reporting that the performances of prediction models derived via machine learning methods can be similar to those of conventional statistical methods suggest that machine learning methods are a useful tool for MetS prediction.…”

“…It has previously been reported that MetS risk factors differ by Sasang constitution type [ 13 , 33 ]. One study reported that age and BMI were significant MetS risk factors in TE and SY type individuals, but not in SE type individuals [ 13 ].…”

“…In addition, for body shape, the circumference of the forehead, neck, axilla, chest, ribs, waist, pelvis, and hips was measured. The questionnaire included questions related to general temperature, eating habits, physiological symptoms, sweating, and temperament [ 13 ].…”

“…Specifically, the TE type was associated with a higher risk of MetS than the SY or SE types [ 10 – 12 ]. In addition, several studies suggested that each constitution type is differentially affected by different MetS risk factors [ 13 ].…”