borderline significanceP = .054
Tertile-based analysis revealed a nonlinear dose–response pattern: compared with T1 (2.0–19.8 mmol/L), T3 (24.0–51.0 mmol/L) exhibited persistent protective effects in logistic regression (Model 3 OR: 0.77, 95% CI: 0.64–0.93, P = .006) and Cox (Model 3 HR: 0.78, 95% CI: 0.66–0.92, P = .003) models, while T2 (20.0–23.9 mmol/L) exhibited borderline significance in logistic regression (OR: 0.85, P = .054) but remained protective in time-to-event analysis (HR: 0.79, P = .002; Table 4 ). 4.