Abstract
Introduction: Total bile acids (TBAs) are emerging as potential prognostic biomarkers in acute ischemic stroke (AIS). This study aimed to evaluate the association between TBA levels and long-term outcomes in AIS patients receiving intravenous thrombolysis (IVT). Methods: A total of 231 AIS patients treated with IVT were prospectively enrolled. TBA levels were measured on admission. The primary outcome was the 3-month modified Rankin Scale (mRS) score. Logistic regression, restricted cubic splines (RCS), and decision curve analysis (DCA) were used to assess associations. Machine learning (ML) models were employed to validate predictive performance. Results: High TBA (> 5 μmol/L) patients showed significantly better functional outcomes (91.2% vs 60.2%, P < 0.001). Multivariate analysis confirmed higher TBA independently predicted favorable outcomes (adjusted OR = 0.74, 95%CI:0.59-0.93), with TBA-integrated models showing superior discrimination (AUC = 0.970 vs ≤ 0.64 for NIHSS/TOAST). Restricted cubic spline analysis revealed a J-shaped non-linear relationship between TBA levels and outcome probability. Critically, a predictive model combining TBA with clinical factors demonstrated superior discriminative ability (AUC = 0.970), significantly outperforming traditional scores (NIHSS AUC = 0.64; TOAST AUC = 0.55). Decision curve analysis confirmed the model's clinical utility. Machine learning validation, particularly using Random Forest (accuracy: 93.8%, AUC: 93.14%, Brier score: 0.072), further substantiated TBA's predictive value. Feature importance analysis identified TBA (25.85) and hemoglobin (24.34) as the primary predictors, substantially exceeding others (e.g., NT-proBNP:3.60; admission NIHSS: 3.41; eGFR-EPI: 3.28). Conclusion: TBA is independently associated with functional outcomes after IVT and may serve as a novel prognostic biomarker in AIS.