个体化预测肝硬化食管胃底静脉曲张破裂出血患者不良预后列线图模型的建立与验证

Establishment and verification of a nomogram model for individualized prediction of poor prognosis in patients with cirrhosis of esophagogastric variceal bleeding

目的建立个体化预测肝硬化食管胃底静脉曲张破裂出血(EGVB)患者不良预后的列线图模型并验证其效能,为EGVB的防治提供科学依据。方法回顾性分析2010年1月至2018年12月浙江省医疗健康集团杭州医院收治的389例肝硬化EGVB患者的临床资料。患者均随访3年,其中预后不良232例(预后不良组),预后良好157例(预后良好组)。比较两组患者入院时的一般临床资料和实验室检查结果。采用受试者工作特征(ROC)曲线分析影响肝硬化EGVB患者不良预后因素的最佳截断值,多因素Logistic回归分析影响EGVB患者不良预后的独立危险因素,R语言软件4.0"rms"包建立预测肝硬化EGVB患者不良预后的列线图模型,采用校正曲线对列线图模型进行内部验证,决策曲线评估列线图模型的预测效能。结果预后不良组年龄、腹水率、肝表面粗糙率、终末期肝病模型评分(MELD评分)、Child-Turcotte-Pugh评分(CTP评分)、丙氨酸氨基转移酶(ALT)、天冬氨酸氨基转移酶(AST)、国际标准化比值(INR)和总胆红素(TBIL)明显高于预后良好组[(62.48±6.21)岁比(58.71±5.93)岁、51.29%(119/232)比35.03%(55/157)、60.78%(141/232)比42.03%(66/157)、(13.89±1.93)分比(11.32±1.69)分、(8.93±0.77)分比(7.46±0.63)分、(37.73±5.21)U/L比(32.13±5.03)U/L、(64.19±11.31)U/L比(57.36±10.29)U/L、1.73±0.41比1.61±0.39和(24.31±2.63)μmol/L比(19.86±2.17)μmol/L],白蛋白、血红蛋白和血钠明显低于预后良好组[(36.21±4.51)g/L比(39.12±4.96)g/L、(86.31±8.27)g/L比(92.28±9.67)g/L和(136.58±18.24)mmol/L比(141.21±19.26)mmol/L],差异有统计学意义(P<0.01或<0.05)。ROC曲线分析结果显示,年龄、MELD评分、CTP评分、白蛋白、ALT、AST、血红蛋白、INR、TBIL和血钠预测肝硬化EGVB患者不良预后的最佳截断值分别为55岁、14.20分、9.30分、35 g/L、38 U/L、67 U/L、88 g/L、1.90、25μmol/L和135 mmol/L。多因素Logistic回归分析结果显示,年龄≥55岁、腹水、MELD评分≥14.20分、CTP评分≥9.30分、白蛋白<35 g/L以及INR≥1.90是肝硬化EGVB患者不良预后的独立危险因素(HR=1.528、1.439、1.637、1.795、1.521和1.596,95%CI 1.165~1.891、1.088~1.790、1.308~1.966、1.385~2.205、1.262~1.780和1.259~1.933,P<0.05或<0.01)。整合影响肝硬化EGVB患者不良预后的独立危险因素,构建列线图模型,其预测效能较好(C-index为0.839,95%CI 0.781~0.948)。列线图模型预测肝硬化EGVB患者不良预后的校正曲线结果显示,实际曲线趋近于理想曲线;当高风险阈值>0.02时,列线图模型预测肝硬化EGVB患者不良预后时可提供明显的额外临床净收益,且其提供的临床净收益高于单个独立危险因素。结论基于年龄、腹水、MELD评分、CTP评分、白蛋白和INR等影响肝硬化EGVB患者不良预后的独立危险因素所构建的列线图模型在筛选与识别肝硬化EGVB患者不良预后方面具有较大的临床价值。

Objective To establish a nomogram model for individualized prediction of poor prognosis in patients with cirrhosis of esophagogastric variceal bleeding(EGVB),and verify its efficacy,so as to provide a scientific basis for the prevention and treatment of EGVB.Methods The clinical data of 389 patients with cirrhosis of EGVB from January 2010 to December 2018 in Hangzhou Hospital of Zhejiang Medical and Health Group were retrospectively analyzed.All patients were followed up for 3 years,including 232 cases with poor prognosis(poor prognosis group)and 157 cases with good prognosis(good prognosis group).The general clinical data and laboratory results were compared between 2 groups.Receiver operating characteristic(ROC)curve was used to analyze the optimal cut-off value of poor prognosis factors in patients with cirrhosis of EGVB;multivariate Logistic regression analysis was used to analyze independent risk factors of poor prognosis in patients with cirrhosis of EGVB.A nomogram model to predict poor prognosis in patients with cirrhosis of EGVB was established with R language software 4.0"rms"package.Internal validation of the nomogram model was performed using correction curves,and the prediction efficiency of the nomogram model was evaluated using decision curves.Results The age,ascites rate,liver surface roughness rate,end-stage liver disease model score(MELD score),Child-Turcotte-Pugh score(CTP score),alanine aminotransferase(ALT),aspartate transaminase(AST),international standard ratio(INR)and total bilirubin(TBIL)in poor prognosis group were significant higher than those in good prognosis group:(62.48±6.21)years old vs.(58.71±5.93)years old,51.29%(119/232)vs.35.03%(55/157),60.78%(141/232)vs.42.03%(66/157),(13.89±1.93)scores vs.(11.32±1.69)scores,(8.93±0.77)scores vs.(7.46±0.63)scores,(37.73±5.21)U/L vs.(32.13±5.03)U/L,(64.19±11.31)U/L vs.(57.36±10.29)U/L,1.73±0.41 vs.1.61±0.39 and(24.31±2.63)μmol/L vs.(19.86±2.17)μmol/L,the albumin,hemoglobin and serum sodium were significantly lower than those in good prognosis group:(36.21±4.51)g/L vs.(39.12±4.96)g/L,(86.31±8.27)g/L vs.(92.28±9.67)g/L and(136.58±18.24)mmol/L vs.(141.21±19.26)mmol/L,and there were statistical differences(P<0.01 or<0.05).ROC curve analysis results show that the optimal cut-off values of age,MELD score,CTP score,albumin,ALT,AST,hemoglobin,INR,TBIL and serum sodium for predicting poor prognosis in patients with cirrhosis of EGVB were 55 years old,14.20 scores,9.30 scores,35 g/L,38 U/L,67 U/L,88 g/L,1.90 scores,25μmol/L and 135 mmol/L,respectively.Multivariate Logistic regression analysis results showed that age≥55 years old,ascites,MELD score≥14.20 scores,CTP score≥9.30 scores,albumin<35 g/L and INR≥1.90 were independent risk factors for poor prognosis in patients with cirrhosis of EGVB(HR=1.528,1.439,1.637,1.795,1.521 and 1.596;95%CI 1.165 to 1.891,1.088 to 1.790,1.308 to 1.966,1.385 to 2.205,1.262 to 1.780 and 1.259 to 1.933;P<0.05 or<0.01).To construct a nomogram model that integrates independent risk factors for poor prognosis in patients with cirrhosis of EGVB,the predictive power of the model was good(C-index 0.839,95%CI 0.781 to 0.948).The corrected curve of nomogram model to predict poor prognosis in patients with cirrhosis of EGVB was close to the ideal curve;when the high risk threshold>0.02,nomogram model provided a significant additional clinical net benefit to predict poor outcome in patients with cirrhosis of EGVB,which was higher than the individual risk factors.Conclusions The nomogram model based on age,ascites,MELD score,CTP score,albumin,INR and other independent risk factors that affect the high risk of poor prognosis in patients with cirrhosis of EDVB has great clinical value in screening and identifying high risk of poor prognosis in patients with cirrhosis of EDVB.