Bayesian Multivariate Longitudinal Modeling of Metabolic Syndrome in Blood Donors

Metabolic syndrome is a cluster of risk factors that elevate the risk of insulin resistance-related diseases, such as heart disease, stroke, and type 2 diabetes. It affects approximately 24.3% of the European population. Diagnosis is based on exceeding established thresholds in at least three of five physiological markers: blood glucose, triglycerides, systolic blood pressure, waist circumference, and HDL cholesterol. This study introduces a Bayesian model for the analysis of metabolic syndrome in blood donors using longitudinal data from Associazione Volontari Italiani del Sangue, which consist of lifestyle habits and blood exams. We analyze the five target variables: HDL cholesterol, waist circumference, glucose, systolic blood pressure, and triglycerides, modelling their dependencies and the influence of covariates. The joint likelihood of these target variables is designed as the product of the conditional univariate distributions, using the R package BDgraph. Bayesian inference is obtained via the software Stan.