A Bayesian network-based predictive model for gout onset risk: associations with traditional Chinese medicine constitution in hyperuricemic populations.

OBJECTIVE: To develop a Bayesian network model for predicting gout onset risk in hyperuricemic individuals based on TCM body constitutions, quantify their influence, and provide evidence for early warning and TCM intervention. METHOD: A total of 826 hyperuricemia (HUA) patients were enrolled in this prospective cohort study. All participants underwent standardized TCM constitution identification per "Classification and Identification of Constitution in Traditional Chinese Medicine." Clinical variables including gender, age, BMI, serum uric acid, and triglycerides were collected. A Bayesian network was constructed via the Hill-Climbing algorithm, and its performance was validated by 10-fold cross-validation. RESULT: During follow-up, 217 participants progressed to gout (incidence 26.27%). Phlegm-dampness (24.1%) and damp-heat (21.8%) were the most prevalent TCM constitutions. Their prevalence was significantly higher in the gout group (38.7% and 32.7%) than in the non-gout group (P < 0.001). Bayesian network analysis identified serum uric acid, BMI, phlegm-dampness, and damp-heat constitutions as direct predictors of gout onset. Conditional probabilities: phlegm-dampness 62.3%, damp-heat 54.1%, balanced 8.4%. When phlegm-dampness coexisted with serum uric acid ≥540 μmol/L, gout incidence rose to 81.6%. The model achieved favorable efficacy: accuracy = 0.832, AUC = 0.847, sensitivity = 0.811, specificity = 0.843. CONCLUSION: Phlegm-dampness and damp-heat constitutions are key predisposing TCM patterns for gout in hyperuricemic individuals. The Bayesian network model effectively quantifies probabilistic links between TCM constitutions and gout risk, offering a reliable tool for personalized risk stratification and TCM-guided preventive management.