心脏移植早期死亡的影响因素及预测模型构建

Influencing factors of early mortality after heart transplantation and constructing a prediction model

目的:探索影响心脏移植受者早期死亡的危险因素及列线图预测模型构建。方法:回顾性纳入2018年4月至2022年6月在郑州市第七人民医院行心脏移植的163例受者的临床资料,根据受者早期(术后90 d)存活情况,分为存活组(132例)和死亡组(31例),分析比较两组的临床资料,选出危险因素变量,绘制列线图对早期死亡概率进行预测,利用ROC曲线(AUC)下面积、Harrell的C指数和校准曲线用于评估和内部验证该模型的性能。应用决策曲线分析来评估其临床效用。结果:存活组和死亡组在机械通气、神经系统病变、体外膜氧合使用、红细胞计数≤3.52×1012/L、肺动脉平均压力>27 mmHg、肺血管阻力>4.01 Wood Unit、白蛋白≤33 g/L、天冬氨酸转氨酶>50 U/L、血红蛋白≤108 g/L、血小板计数≤109×109/L、总胆红素>57μmol/L方面差异存在统计学意义(均P<0.05),是心脏移植受者早期死亡的危险因素。同时根据实际情况和差异变量,围绕血红蛋白≤108 g/L、白蛋白≤33 g/L、血小板计数≤109×109/L、总胆红素>57μmol/L、天冬氨酸转氨酶>50 U/L、神经系统病变和肺动脉平均压>27 mmHg 7个变量,构造的列线图用来预测心脏移植早期死亡概率,其模型评估AUC曲线下面积0.917,C指数0.910,校准曲线良好,决策曲线分析显示该列线图可使心脏移植受者受益。结论:根据心脏移植受者实际情况和差异变量构造的列线图用来预测心脏移植受者早期死亡概率可信度高,对心脏移植受者的选择和个体化治疗具有重要的意义。

Objective To explore the risk factors for early mortality in heart transplant(HT)recipients and construct a nomogram prediction model.Methods From 2018 to 2022,preoperative clinical data were retrospectively reviewed for 163 consecutive HT recipients.Risk factor variables were shortlisted by univariate correlation analysis based upon early(90-day)postoperative patient survival.Lasso regression was then employed for screening all variables and common variables were combined.A nomogram was constructed for predicting the probability of early mortality after considering actual circumstance.Receiver operating characteristic(ROC)curve,area under the ROC curve(AUC),Harrell's C-index and calibration curves were employed for evaluating and internally validate the performance of the model.Decision curve analysis was performed for assessing clinical utility of the model.Results In survival and mortality groups,mechanical ventilation,nervous system lesions,use of extracorporeal membrane oxygenation,red blood cell count≤3.52×1012/L,mean pulmonary arterial pressure>27 mmHg,pulmonary vascular resistance>4.01 Wood Unit,albumin≤33 g/L,aspartate aminotransferase>50 U/L,hemoglobin≤108 g/L,platelet count≤109×109/L and total bilirubin>57μmol/L demonstrated statistically significant differences(P<0.05).At the same time,according to actual situations and different variables,hemoglobin≤108 g/L,albumin≤33 g/L,platelet count≤109×109/L,total bilirubin>57μmol/L,aspartate aminotransferase>50 U/L,nervous system lesions and average pulmonary arterial pressure>27 mmHg were seven variables.And a nomogram with relatively high reliability was constructed for predicting the probability of early mortality post-HT(nomogram model evaluation,AUC 0.917,C index 0.910 and good calibration curve).Decision curve analysis indicated that the nomogram could benefit HT recipients.Conclusions Risk factors have been identified for early mortality in HT recipients.And the nomogram prediction model offers a simple and reliable tool for predicting early mortality post-HT.It has important implications for individualized treatment of HT candidates.