基于随机森林生存模型的AMI患者PCI术后的不良事件风险分析

Risk factors for adverse events after percutaneous coronary intervention in patients with acute myocardial infarction： an analysis based on a random forest survival model

摘要:

目的 综合分析经皮冠状动脉介入术（percutaneous coronary intervention，PCI）后的急性心肌梗死（acute myocardial infarction，AMI）患者预后影响因素，并构建预测模型和预后评分体系，为临床血管个性化治疗提供参考。方法 本研究回顾性收集从2018年1月至2022年6月所有在江西省南昌大学第二附属医院行PCI术的AMI患者，随访结局是术后首次发生主要心血管不良事件（major adverse cardiovascular events，MACE）。采用十倍交叉验证的Lasso回归确定纳入模型的变量，构建随机生存森林（random survival forest，RSF）模型和Cox比例风险模型，采用受试者工作特征曲线（receiver operating characteristic，ROC）下面积（area under curve，AUC）和校准曲线评估模型性能。根据RSF模型拟合结果绘制风险计算器。结果 研究最终共纳入3 880例AMI患者，其中术后1年内发生主要心血管不良事件473例（12.2%）。Lasso回归筛选出性别、急性心肌梗死类型、高血压等15个变量。多因素Cox回归结果显示，糖尿病、左室射血分数较低（30%~40%）、血管狭窄程度是术后MACE发生的影响因素。验证集中，RSF和Cox模型的AUC分别为0.774（95%CI=0.761~0.787）和0.597（95%CI=0.581~0.613）。绘制的校准曲线提示，该模型在预测1年MACE风险方面具有较高的准确性，构建的RSF评分最佳截断点（Score=133）也能准确区分MACE累计发病风险（P<0.001）。结论 构建的RSF模型及评分综合上述因素，能有效预测术后MACE发病风险并进行风险分层，帮助临床心血管医生制定个性化治疗方案。

Abstract:

Objective To comprehensively analyze the influencing factors for the prognosis of patients with acute myocardial infarction（AMI） after percutaneous coronary intervention（PCI），to construct a prediction model and a prognosis scoring system，and to provide a reference for individualized vascular treatment in clinical practice.Methods A retrospective analysis was performed for all AMI patients who underwent PCI in The Second Affiliated Hospital of Nanchang University from January 2018 to June 2022，with the follow-up outcome of the onset of major adverse cardiovascular events（MACE） for the first time after surgery. The ten-fold cross-validated lasso regression analysis was used to determine the variables to be included in the model，and a random survival forest（RSF） model and a Cox proportional hazards model were constructed. The area under the ROC curve（AUC） and calibration curves were used to evaluate the performance of the model，and a risk calculator was developed according to the fitting results of RSF model.Results A total of 3 880 patients with AMI were finally included in the study，among whom 473（12.2%） experienced MACE within one year after surgery. Lasso regression obtained 15 variables including sex，type of AMI，and hypertension，and the multivariate Cox regression analysis showed that diabetes，low left ventricular ejection fraction（30%~40%），and degree of vascular stenosis were the risk factors for postoperative MACE. In the validation set，the RSF and Cox models had an AUC of 0.774（95%CI=0.761~0.787） and 0.597（95%CI=0.581~0.613），respectively. The calibration curves showed that the model had a relatively high accuracy in predicting the risk of MACE within one year，and RSF score with the optimal cut-off value of 133 could also accurately distinguish the cumulative risk of MACE（P<0.001）.Conclusion The RSF model and the scoring system constructed based on the above factors can effectively predict the risk of postoperative MACE and perform risk stratification，thereby helping cardiovascular physicians to formulate individualized treatment regimens in clinical practice.