HBV相关慢加急性肝衰竭患者短期预后危险因素分析及诺模图预测模型的建立

A Nomogram Model for Risk Factor Analysis of Short-Term Prognosis in HBV Patients Associated Acute-on-Chronic Liver Failure

目的本研究旨在建立诺模图模型来预测乙型肝炎病毒相关慢加急性肝衰竭患者(HBV-ACLF)短期(90 d)生存率的方法。方法收集2017年6月至2020年9月收治的240例HBV-ACLF患者,根据随访90 d时的转归分为死亡组(n=53)和存活组(n=187),我们使用ROC曲线来评估每个连续性数值变量预测HBV-ACLF预后的能力,并计算它们的最佳截断值(cutoff值),使用卡方检验和多因素COX回归分析来确定与预后显著相关的危险因素。利用这些预测因子建立诺模图模型,利用受试者工作特征曲线下面积(AUC)比较各种模型对HBV-ACLF患者90 d预后的预测价值。结果HBV-ACLF患者90d总死亡率为22.08%(53/240)。通过卡方检验及多因素COX回归分析,我们发现年龄(≥48.5岁)(HR=2.585,95%CI=1.428~4.677,P=0.002)、NEU/LYM(≥4.389)(HR=2.379,95%CI=1.279~4.425,P=0.006)、PA(<40.8 mg/L)(HR=0.490,95%CI=0.274~0.879,P=0.017)、PT(≥23.65 s)(HR=2.150,95%CI=1.106~4.180,P=0.024)、MELD(≥20.929)(HR=3.087,95%CI=1.195~7.969,P=0.020)、CTP(≥10.5)(HR=3.460,95%CI=1.726~6.938,P<0.001)是慢加急性肝衰竭患者短期预后的独立危险因素,与预后显著相关。以50%的死亡率作为分类的分界点,对应点为324来验证诺模图模型。该诺模图模型预测HBV-ACLF患者短期预后的准确率为90.9%。诺模图的敏感度和特异度分别为94.34%和72.19%。诺模图模型、CTP、MELD、ICGR15的AUC分别为0.909、0.889、0.778、0.674,该诺模图模型的AUC均大于其他3个模型(Z值分别为17.852、9.193、4.384,均P<0.001)。结论该诺模图模型对HBV-ACLF患者90 d的短期预后的预测价值优于CTP、MELD、ICGR15。

Objective To establish a nomogram model to predict the short-term(90 days)survival rate of patients with hepatitis B virus(HBV)associated acute-on-chronic liver failure(HBV-ACLF).Methods From June 2017 to September 2020,240 patients with HBV-ACLF were collected.They were assigned to death group(n=53)and survival group(n=187)according to the outcome of 90-days follow-up.The receiver operating characteristic(ROC)curve was used to evaluate each continuous variable to predict prognosis of HBV-ACLF,and their optimal cutoff values were calculated.Chi-square test and multivariate Cox regression analysis were used to identify risk factors significantly associated with prognosis.Then,nomogram model was established using these predictors,and area under ROC curve(AUC)was used to compare predictive value of various models for 90-days prognosis of HBV-ACLF patients.Results 90-days mortality rate of HBV-ACLF patients was 22.08%(53/240).By chi-square test and multivariate Cox regression analysis,we found that age(≥48.5)(HR=2.585,95%CI=1.428~4.677,P=0.002),NEU/LYM(≥4.389)(HR=2.379,95%CI=1.279~4.425,P=0.006),PA(<40.8 mg/L)(HR=0.490,95%CI=0.274~0.879,P=0.017),PT(≥23.65 s)(HR=2.150,95%CI=1.106~4.180,P=0.024),MELD(≥20.929)(HR=3.087,95%CI=1.195~7.969,P=0.020),CTP(≥10.5)(HR=3.460,95%CI=1.726~6.938,P<0.001)were independent risk factors for short-term prognosis of patients with HBV-ACLF,which were significantly correlated with prognosis.50%mortality rate was used as cut-off point of classification,and corresponding point 324 was used to verify nomogram model.Accuracy of the nomogram model in predicting short-term prognosis of HBV-ACLF patients was 90.9%.Sensitivity and specificity of nomogram were 94.34%and 72.19%,respectively.AUC of nomogram model,CTP,MELD,and ICGR15 were 0.909,0.889,0.778,and 0.674,respectively.AUC of nomogram model was higher than that of other three models(Z=17.852,9.193,and 4.384,respectively,all P<0.001).Conclusion Predictive value of nomogram model for 90-days short-term prognosis of HBV-ACLF patients would be better than that of CTP,MELD,and ICGR15.