| BackgroundHepatic sinusoidal obstruction syndrome(HSOS)is a vascular disease of the liver caused by toxic damage to the endothelial cells of the hepatic sinusoidal.The main cause of HSOS in China is the consumption of herbal medicines containing pyrrolizidine(PAs),with the herb Gynura segetum a member of the Asteraceae family,being the most common.In recent years,with the widespread use of herbal medicines worldwide,the incidence of pyrrolizidine alkaloids-related hepatic sinusoidal obstruction syndrome(PA-HSOS)has been increasing year by year,with a mortality rate of over 80%in severely ill patients,and has become a clinical problem that cannot be ignored.As patients with PA-HSOS do not have characteristic clinical manifestations,the early diagnosis and identification of high-risk patients is currently a challenge in clinical management.ObjectiveWe retrospectively analyzed the clinical characteristics and prognosis of patients with PA-HSOS,explored the factors associated with poor prognosis,and constructed a mortality risk prediction model.By comparing the predictive efficacy of the new model with common liver disease prognostic models applied alone and in combination,we aimed to select the best prognostic scoring model to guide clinical early identification of high-risk patients.MethodsThe clinical data of 57 patients with PA-HSOS treated at the Provincial Hospital of Shandong First Medical University from December 2013 to May 2022 were collected.A follow-up endpoint of 180 days was used to divide the group into death and survival groups.Single-factor logistic regression and Lasso regression were used to screen prognostic risk factors,and the screened variables were included in multivariate logistic regression analysis to determine prognosis-related risk factors,and a prediction model for the risk of death in PA-HSOS was constructed,and the prediction model was tested for goodness-of-fit using the Hosmer-Lemeshow test.Subject working characteristic(ROC)curves were plotted for the new model and different scoring models predicted individually and jointly,and the predictive efficacy of different scoring models for prognosis was evaluated by comparing the area under the curve(AUC),sensitivity,and specificity of the ROC curves.Results(1)A total of 57 patients with PA-HSOS were included in this study,35(61.4%)were male and 22(38.6%)were female,male:female=1.59:1.the mean age was 62.89±11.64 years.The main clinical manifestations of PA-HSOS patients were abnormal liver function and coagulation dysfunction.17 cases died at 180 days of follow-up,with a morbidity and mortality rate of 29.8%.Abdominal distension(94.7%),ascites(100%),hepatomegaly(73.7%),bilateral lower limb edema(35.1%),dysphagia(33.3%),and abdominal pain(29.8%)were the most common symptoms and signs.(2)There was no statistical difference between the patients in the death and survival groups in terms of age,sex,number of days in hospital,symptoms,signs,diabetes mellitus,hypertension,coronary artery disease,presence of hepatic encephalopathy,presence of spontaneous bacterial peritonitis,whether glucocorticoid treatment was given,and whether TIPS treatment was performed(P>0.05).Compared with the surviving group,TBIL,DBIL,IBIL,CR,BUN,PT,INR,WBC count,NEUT,NLR,and the presence of hepatorenal syndrome were significantly higher and PTA was significantly lower in the death group,and the differences between the two groups were statistically significant(P<0.05).Patients who received low-molecular heparin anticoagulation therapy had a better prognosis(P<0.05).(3)Univariate and multivariate logistic regression analysis revealed that creatinine(CR)and prothrombin activity(PTA)were 180-day risk of death in patients with PA-HSOS with a statistically significant difference(P<0.05).A new prediction model was constructed by logistic regression equation:logit(P)=0.036×CR-0.073×PTA+0.280,and the area under the ROC curve(AUC)of the new model(PTA-CR)was 0.829(95%CI:0.694-0.965),with a sensitivity of 58.8%,specificity of 100%,and The cut-off value was 0.58,and the model had good discriminatory ability.The goodness-of-fit test for the new model was performed with P=0.141,and the model worked well.(4)Comparing the ROC curves of the PTA-CR score model with the CTP score,MELD score,MELD-Na score,and MELD 3.0 score model alone for predicting the risk of death at 180 days in patients with PA-HSOS,the results showed that the PTA-CR score had the largest AUC and highest specificity of 100%for predicting the risk of death;the CTP score had the highest sensitivity of 88.2%for predicting the risk of death.A comparison of the predictive accuracy of the PTA-CR model with the other four models by Delong’s test showed that the PTA-CR score was better than the MELD-Na score in determining prognosis,with a statistically significant difference(P=0.043).No significant differences were seen between the comparison with the remaining scoring models.(5)Comparing the ROC curves of the PTA-CR score model with the CTP score,MELD score,MELD-Na score and MELD 3.0 score in tandem and parallel to predict the 180-day risk of death in patients with PA-HSOS respectively,the CTP score in parallel to the PTA-CR score had the highest sensitivity of 100%,which was higher than the sensitivity when the model was applied alone,but the specificity was only 22.5%,reducing the accuracy of the prediction;the MELD 3.0 in tandem PTA-CR score had the highest specificity of 100%.Conclusions(1)Serum creatinine and PTA were independent predictors of 180-day mortality risk in PA-HSOS patients.(2)The death risk prediction model(PTA-CR)established based on Logistics regression analysis has good differentiation and goodness of fit,with specificity of 100%,sensitivity of 58.8%and cut-off value of 0.58.(3)By comparing with different scoring models,the PTA-CR model had the largest area under the ROC curve,the highest specificity and better sensitivity,which may be helpful for early clinical assessment of patient prognosis. |
PDF Full Text Request