不同终末期肝病模型评估血浆置换治疗慢加急性乙型肝炎肝衰竭患者短期预后

Evaluation on short-term prognosis by using different predictive models for end-stage liver disease in patients with hepatitis B virus-associated acute-on-chronic liver failure treated with plasma exchange

目的应用终末期肝病模型(MELD)、MELD-Na联合公式(MELD-Na)、MESO指数(MESO)、integratedMELD整合公式(iMELD)、MELDNa评分系统(MELDNa)评估慢加急性乙型肝炎肝衰竭患者在血浆置换(PE)治疗后短期预后。方法在122例慢加急性乙型肝炎肝衰竭入院24小时内进行MELD、MELD-Na、MESO、iMELD、MELDNa评分,并随访3个月。应用受试者工作特征曲线(ROC)下面积(AUC)判断五个模型的预测能力。结果在3个月末,PE治疗组患者病死率为43.24%(16/37),与非PE治疗组55.65%(49/85)比,差异无统计学意义(x2=1.466,P=0.143);在PE治疗组,五种模型AUC分别为0.667、0.690、0.670、0.702和0.673,差异无统计学意义,仅iMELD的AUC>0.7,有一定的临床价值,而在非PE治疗组分别为0.901、0.916、0.906、0.876和0.910,差异无统计学意义,PE治疗组与非PE组AUC的比较均具统计学差异,MELD(Z=0.2087,P<0.05)、MELD-Na(Z=0.1971,P<0.05)、MESO(Z=0.2094,P<0.05)、iMELD(Z=0.155,P<0.05)、MELDNa(Z=0.2094<0.05);在PE治疗组,五种模型预测的正确率分别为72.97%、70.30%、75.78%、72.93%和75.68%,差异无统计学意义(x2=0.389,P=0.983),在非PE治疗组分别为83.53%、87.06%、82.35%、78.82%和84.71%,差异无统计学意义(x2=2.266,P=0.687),PE治疗组与非PE治疗组之间相比较,其中MELD、MESO、iMELD、MELDNa均无统计学意义(分别为x2=1.818,P=0.178;x2=0.727,P=0.394;x2=0.499,P=0.480;x2=1.442,P=0.233),而仅MELD-Na具统计学意义(x2=4.916,P=0.027)。结论五种评分模型对非PE治疗的慢加急性乙型肝炎肝衰竭患者短期预后判断均有非常好的预测价值,而在PE治疗的慢加急性乙型肝炎肝衰竭患者,除iMELD对短期预后判断有临床应用价值外,其余四种模型评估患者短期预后的能力较差。

Objective To evaluate the short-term prognosis by different predictive models in patients with hepatitis B-induced acute-on-chronic liver failure （ACLF）treated with plasma exchange （PE） or non-plasma ex-change （non-PE）,including model for end-stage liver disease （MELD）,MELD-Na formula （MELD-Na）,the MESO index （MESO）and integrated IMELD formula （iMELD）. Methods One hundred and twenty-two inpatients with hepatitis B virus-associated ACLF were evaluated by MELD, MELD-Na,MESO,iMELD and MELDNa within 24 hours of administration and were followed up for 3 months. The area under the receiver operating characteristic curve （AUC） was used to evaluate the predictive ability of these five models. Results At the end of 3 months, the mortality in PE group were 43.24%（16/37）,which was not statistically significantly different （x2=1.466,P〉0.05） from that （57.65%,49/85） in non-PE group;In PE group,the AUC of five models were 0.667,0.690,0.670,0.702 and 0.673,respectively,and the differences were not statistically significant;Only the AUC of iMELD was greater than 0.7 and had a clinical value; In non-PE group,the AUC of five models were 0.901,0.916,0.906,0.876 and 0.910, respectively,and these differences were not statistically significant;However,the AUC of each of five models was significantly different between PE group and non-PE group,i.e. MELD（Z=0.2087,P〈0.05）,MELD-Na （Z=0.1971, P〈0.05）,MESO（Z=0.2094,P〈0.05）,iMELD（Z=0.155,P〈0.055） and MELDNa（Z=0.2094〈0.05）;The accuracy rates in PE group were 72.97%,70.30%,75.68%,72.93% and 75.68%,respectively and there were no significant differ-ence （x2=0.389,P=0.983）;The accuracy rates in non-PE group were 83.53%,87.06%,82.35%,78.82%,84.82% and 84.71%,respectively,which were not significantly different（x2=2.266,P=0.687）;The differences in the accuracy rates of the four predictive models （MELD,MESO,iMELD and MELDNa） between PE group and non-PE group were not statistically significantly different （x2=1.818,P=0.178;x2=0.727,P=0.394;x2=0.499,P=0.480;x2=1.442,P=0.233,re-spectively）,only of one（MELD-Na） reached statistical significance（x2=4.916,P=0.027）. Conclusions The models of MELD,MELD-Na,MESO,iMELD and MELDNa have better predictive value on predicting short-term prognosis in non-PE-treated patients with hepatitis B virus-associated ACLF,but in PE-treated patients only iMELD has.