基于Cox风险回归预测模型判断HBV-ACLF短期预后的价值

The value of predictive model based on Cox risk regression model analysis in judging the short-term prognosis of HBV-ACLF

目的:研究基于Cox风险回归预测模型在判断乙型肝炎相关性慢加急性肝衰竭(HBV-ACLF)患者短期预后中的临床价值。方法:纳入2017年5月~2020年5月86例HBV-ACLF患者作为研究对象,记录90 d病死率。比较不同预后患者凝血五项指标和TEG指标水平,采用Cox多因素风险回归分析影响HBV-ACLF短期预后的相关因素。建立判断短期预后的预测模型,分析其判断预后的准确性。结果:90 d患者死亡37例,死亡率43.02%。Cox风险回归模型分析结果显示肝性脑病(β=0.446,95%CI=1.031-2.366,P=0.035)、肝肾综合征(β=0.329,95%CI=1.179-2.636,P<0.001)、APTT(β=0.187,95%CI=1.097-1.326,P<0.001)、INR(β=0.359,95%CI=1.163-1.763,P=0.001)及MELD评分(β=0.249,95%CI=1.050-1.568,P=0.015)是HBV-ACLF短期死亡的危险因素。FIB(β=-0.526,95%CI=0.384-0.910,P=0.017)、MA(β=-0.730,95%CI=0.290-0.801,P=0.005)及α角(β=-0.523,95%CI=0.419-0.839,P=0.003)是短期死亡的保护因素。根据Cox风险回归分析结果建立预测模型为Y=0.446X1+0.329X2+0.187X3+0.359X4-0.526X5-0.730X6-0.523X7+0.249X8(肝性脑病=X1,肝肾综合征=X2,APTT=X3,INR=X4,FIB=X5,MA=X6,α角=X7,MELD评分=X8)。预测模型判断90d短期预后的AUC为0.841(SE=0.076,95%CI=0.529-0.943,P<0.001)。结论:HBV-ACLF短期病死率高,基于Cox风险回归模型建立的预测模型对判断HBV-ACLF短期预后准确性高。

Objective:To study the value of predictive model based on the Cox risk regression model analysis in judging the short-term prognosis of hepatitis B virus related acute-on-chronic liver failure(HBV-ACLF).Methods:86 HBV-ACLF patients from May 2017 to May 2020 were as the research objects, the 90 days case fatality rate was recorded.The levels of five coagulation indexes and TEG indexes in patients with different prognosis were compared.The Cox multivariate risk regression analysis was used to analyze the related factors, which affected the short-term prognosis of HBV-ACLF.The predictive model for judging short-term prognosis was established and its accuracy in judging prognosis was analyzed. Results:There were 37 cases died in 90 days, the mortality rate was 43.02%.The Cox risk regression model analysis results showed that the hepatic encephalopathy(β=0.446, 95%CI=1.031-2.366, P=0.035), hepatorenal syndrome(β=0.329,95% CI=1.179-2.636,P=0.000), APTT(β=0.187,95%CI=1.097-1.326,P=0.000), INR(β=0.359,95% CI=1.163-1.763, P=0.001) and MELD score(β=0.249, 95% CI=1.050-1.568, P=0.015) were risk factor for short-term death from HBV-ACLF.FIB(β=-0.526, 95% CI=0.384-0.910, P=0.017), MA(β=-0.730, 95%CI=0.290-0.801, P=0.005) and α angle(β=-0.523, 95%CI=0.419-0.839, P=0.003) were protective factor for short-term death.The prediction model was established as Y=0.446×X1+0.329 X2+0.187 X3+0.359 X4-0.526 X5-0.730 X6-0.523 X7+0.249 X8, which accorded to the results of Cox risk regression analysis(Hepatic encephalopathy=X1, hepatorenal syndrome=X2, APTT=X3, INR=X4, FIB=X5, MA=X6, α angle=X7MELD score = X8). The AUC of the prediction model to judge the short-term prognosis of 90 days was 0. 841(SE= 0. 076, 95%CI= 0. 529-0. 943,P< 0. 001).Conclusion:HBV-ACLF has a high short-term fatality rate. The prediction model based on the Cox risk regression model has a high accuracy in judging the short-term prognosis of HBV-ACLF.