Yildirim Abdullah, Coskun Mukremin, Demirtas Abdullah Orhan
Department of Cardiology, University of Health Sciences, Adana City Training and Research Hospital, 01230 Adana, Türkiye.
J Clin Med. 2025 Apr 28;14(9):3035. doi: 10.3390/jcm14093035.
The albumin-bilirubin (ALBI) score, initially a hepatic function marker, may also reflect systemic inflammation and oxidative stress, both linked to the no-reflow phenomenon (NRP). This study investigates the ALBI score's predictive value for the NRP and compares it with conventional risk models. This retrospective, single-center study included 1563 NSTE-ACS patients who underwent PCI between January 2023 and February 2024. Two predictive models were developed: (i) a fitted model with variables selected based on the XGBoost algorithm and SHapley Additive ExPlanations (SHAP) values, and (ii) an ALBI model including the ALBI score. Machine learning via the XGBoost algorithm was used for modeling, with SHAP applied to assess the significance of predictors. The NRP occurred in 14.8% (231/1563) of patients. The ALBI score emerged as an independent predictor (OR = 12.10, 95% CI: 7.75-18.89, < 0.001). The ALBI model demonstrated superior predictive power compared to the fitted model (C-index: 0.860 vs. 0.799), with significant improvements in discrimination (11.1%, < 0.001) and reclassification (14.5%, = 0.002). SHAP analysis ranked the ALBI score (1.025) as the strongest predictor, followed by hs-TnI (0.814), e-GFR (0.582), and pre-dilatation (0.283). The ALBI model exhibited better specificity (AUC: 0.860 vs. 0.798), calibration (Brier score: 0.088 vs. 0.102), and model fit (AIC: 964.7 vs. 1098.3) compared to the fitted model, indicating superior overall performance. The ALBI score significantly enhances the prediction of the NRP in NSTE-ACS patients undergoing PCI, outperforming traditional risk models. Incorporating the ALBI score into predictive frameworks may improve early risk stratification and guide clinical decision-making.
白蛋白-胆红素（ALBI）评分最初是一种肝功能标志物，也可能反映全身炎症和氧化应激，二者均与无复流现象（NRP）相关。本研究调查了ALBI评分对NRP的预测价值，并将其与传统风险模型进行比较。这项回顾性单中心研究纳入了2023年1月至2024年2月期间接受经皮冠状动脉介入治疗（PCI）的1563例非ST段抬高型急性冠状动脉综合征（NSTE-ACS）患者。开发了两种预测模型：（i）一种基于XGBoost算法和SHapley值选择变量的拟合模型，以及（ii）一个包含ALBI评分的ALBI模型。通过XGBoost算法进行机器学习建模，并应用SHapley值评估预测因子的重要性。14.8%（231/1563）的患者发生了NRP。ALBI评分成为独立预测因子（比值比=12.10，95%置信区间：7.75-18.89，P<0.001）。与拟合模型相比，ALBI模型显示出更好的预测能力（C指数：0.860对0.799），在区分度（提高11.1%，P<0.001）和重新分类（提高14.5%，P=0.002）方面有显著改善。SHapley分析将ALBI评分（1.025）列为最强预测因子，其次是高敏肌钙蛋白I（hs-TnI，0.814）、估算肾小球滤过率（e-GFR，0.582）和预扩张（0.283）。与拟合模型相比，ALBI模型表现出更好的特异性（曲线下面积：0.860对0.798）、校准度（Brier评分：0.088对0.102）和模型拟合度（赤池信息准则：964.7对1098.3），表明其总体性能更优。ALBI评分显著提高了接受PCI的NSTE-ACS患者中NRP的预测能力，优于传统风险模型。将ALBI评分纳入预测框架可能会改善早期风险分层并指导临床决策。
