建立基于超声的肝硬化食管胃静脉曲张破裂出血无创预测模型

Establishing An Ultrasound-Based Non-Invasive Prediction Model for Esophagogastric Variceal Bleeding in Cirrhosis

目的使用灰阶、多普勒及实时剪切波弹性成像等无创超声建立肝硬化食管胃静脉曲张破裂出血(EVB)预测模型。资料与方法超声测量肝硬化患者门脾静脉主干内径和流速、脾脏长度(SL)、脾脏厚度(ST)、肝脏硬度(LS)、脾脏硬度(SS),随访至出现EVB或截至2020年11月1日,根据随访结局分为出血组和未出血组,对组间有显著差异的指标进行Logistic回归分析,建立回归模型。结果纳入67例肝硬化,出血组15例,未出血组52例。两组ST、SL、LS、SS差异均有统计学意义(t=-2.978、-2.732、-2.423、-2.784,P<0.05)。回归分析发现:ST和SS为独立危险因素(OR=1.063、1.123,P<0.05),建立回归模型(SSST)=0.116×SS+0.061×ST-7.481。LS、SL、SS、ST、SSST预测EVB的ROC曲线下面积分别为0.706、0.733、0.737、0.753、0.867,界值分别为16.5 kPa、132.5 mm、28.4 kPa、45.0 mm、-0.94,敏感度分别为66.7%、100.0%、60.0%、93.3%、73.3%,特异度分别为80.8%、46.2%、82.7%、53.8%、90.4%。结论超声测得SL<132.5 mm、ST<45.0 mm可排除EVB的发生,基于ST和SS的超声无创预测模型SSST具有良好的预测价值和高度特异性。

Purpose Using non-invasive ultrasound technologies such as gray-scale,Doppler and real-time shear wave elastography to develop a prediction model for esophagogastric variceal bleeding(EVB)in liver cirrhosis.Materials and Methods The internal diameter and flow velocity of the portal and splenic vein,spleen length(SL),spleen thickness(ST),liver stiffness(LS)and spleen stiffness(SS)in patients with liver cirrhosis were measured by ultrasound technology.Presenting the EVB or November 1,2020 was defined as the follow-up endpoint.All patients were divided into bleeding group and non-bleeding group based on the follow-up outcomes.Regression analysis was performed on the indicators with statistically significant differences between the groups,establishing a Logistic regression model.Results Sixty-seven patients with liver cirrhosis were finally included,and divided into 15 patients of bleeding group and 52 patients of non-bleeding group.There were significant differences in ST,SL,LS and SS between the two groups(t=-2.978,-2.732,-2.423,-2.784;P<0.05).In regression analysis,ST and SS were found to be the independent risk factors(OR=1.063,1.123;P<0.05),and the regression model SSST=0.116×SS+0.061×ST-7.481 was established.The areas under the receiver operator characteristic curve of LS,SL,SS,ST and SSST for predicting the EVB were 0.706,0.733,0.737,0.753 and 0.867,respectively.The prediction cut-off values were 16.5 kPa,132.5 mm,28.4 kPa,45.0 mm and-0.94,respectively.The sensitivity was 66.7%,100.0%,60.0%,93.3%,73.3%,respectively.The specificity was 80.8%,46.2%,82.7%,53.8%,90.4%,respectively.Conclusion Ultrasonic measurement of SL<132.5 mm,ST<45.0 mm could keep off the occurrence of EVB.The ultrasound non-invasive prediction model SSST based on ST and SS has good predictive value and high specificity.