Chen Si, Nie Rui, Luan Haixia, Shen Xiaoran, Wang Yan, Gui Yuan, Zeng Xiaoli, Yuan Hui
Department of Clinical Laboratory, Beijing Anzhen Hospital, Capital Medical University, Anzhen Road No. 2, Chaoyang District, Beijing, 100029, China.
Rheumatol Int. 2025 Apr 23;45(5):118. doi: 10.1007/s00296-025-05822-6.
To investigate the correlation between acute kidney injury (AKI) and 1-year mortality in patients with granulomatosis with polyangiitis (GPA). Clinical data for GPA patients were extracted from the MIMIC-IV (version 3.0) database. Logistic and Cox regression analyses, Kaplan-Meier (KM) survival analysis, and mediation effect analysis were used to assess the association between AKI, renal function indicators, and 1-year mortality in GPA patients. Predictive models were constructed using machine learning algorithms, and tree-based feature selection was applied to evaluate the contributions of AKI and renal function indicators to mortality prediction. A total of 127 GPA patients were included in the analysis. Multivariate logistic regression identified AKI (OR > 1, P < 0.05) as a significant predictor of 1-year mortality. Similarly, multivariate Cox regression analysis revealed AKI (HR > 1, P < 0.05) as an independent risk factor for 1-year mortality. KM survival analysis demonstrated that GPA patients with AKI had significantly lower survival rates than those without AKI (P < 0.0001). Additionally, renal function indicators modestly mediated the relationship between AKI and 1-year mortality in GPA patients. The machine learning analysis indicated that the random forest algorithm performed the best, with an area under the curve of 0.894. Feature selection using tree model analysis highlighted both AKI and renal function indicators as significant contributors to mortality prediction in GPA patients. Our study suggested AKI was an independent risk factor for increased 1-year mortality in GPA patients. Additionally, renal function indicators partially mediated the relationship between AKI and 1-year mortality in these patients.
探讨肉芽肿性多血管炎（GPA）患者急性肾损伤（AKI）与1年死亡率之间的相关性。从MIMIC-IV（版本3.0）数据库中提取GPA患者的临床数据。采用逻辑回归和Cox回归分析、Kaplan-Meier（KM）生存分析以及中介效应分析来评估GPA患者中AKI、肾功能指标与1年死亡率之间的关联。使用机器学习算法构建预测模型，并应用基于树的特征选择来评估AKI和肾功能指标对死亡率预测的贡献。共有127例GPA患者纳入分析。多变量逻辑回归确定AKI（OR>1，P<0.05）是1年死亡率的显著预测因素。同样，多变量Cox回归分析显示AKI（HR>1，P<0.05）是1年死亡率的独立危险因素。KM生存分析表明，发生AKI的GPA患者生存率显著低于未发生AKI的患者（P<0.0001）。此外，肾功能指标在一定程度上介导了GPA患者中AKI与1年死亡率之间的关系。机器学习分析表明，随机森林算法表现最佳，曲线下面积为0.894。使用树模型分析进行特征选择突出显示AKI和肾功能指标均是GPA患者死亡率预测的重要因素。我们的研究表明，AKI是GPA患者1年死亡率增加的独立危险因素。此外，肾功能指标部分介导了这些患者中AKI与1年死亡率之间的关系。
