| Background:Cirrhosis is the terminal stage of CHB(chronic hepatitis B).GEV(gastroesophageal varices)and VH(variceal hemorrhage)are the two life-threatening complications of cirrhosis deriving from CHB.Currently,invasive examinations,including liver biopsy,endoscopy and HVPG(hepatic venous pressure gradient)detection,are unsuitable for the examination and follow-up of patients with CHB.This study aims to construct nomograms based on LF Index(liver fibrosis index)that is determined by the quantitative analysis of tissue diffusion using RTE(real-time elastography imaging).The constructed nomograms are utilized for predicting disease condition and complications in the different phases of hepatitis B cirrhosis,thus providing references for clinical application.Section ⅠObjective:To construct a nomogram combining LF Index based on the quantitative analysis of tissue diffusion using RTE and plasma miRNA-125a for early predicting hepatitis B cirrhosis.Methods:CHB patients(n=94)diagnosed in The First Affiliated Hospital of Wenzhou Medical University from January 2015 to March 2016 were retrospectively analyzed.Liver tissues collected from CHB patients were enrolled in the liver puncture group.Meanwhile,plasma samples from another independent cluster of 100 CHB patients admitted from January 2016 to December 2016 were enrolled in the plasma group.Clinical data and laboratory indexes of each patient were recorded.LF Index was recorded by RTE,while miRNA-125a levels in liver tissues and plasma were detected by qRT-PCR.Correlation between plasma level of miRNA-125a and clinical characteristics&ultrasound parameters in CHB patients was assessed by Spearman correlation and multiple linear regression.Subsequently,univariate and multivariate logistic regression analyses were performed to define the independent factors for diagnosing cirrhosis in puncture group,and these factors were validated in the plasma group.The discrimination of the Nomogram was evaluated with AUC(area under curve),the accuracy with calibration curve,and the applicability with DCA(decision curve analysis).Results:1.Significant differences in GGT level(p=0.001),LF Index(p=0.033)and miRNA125a level(p=0.031)were identified between liver fibrosis group and cirrhosis group.Logistic regression analysis in the liver puncture group uncovered that GGT level(p=0.001),LF Index(p=0.018)and miRNA-125a level(p=0.018)were the independent factors for diagnosing cirrhosis in CHB patients.2.AST level(p=0.007)was independently positively correlated to plasma level of miRNA-125a in CHB patients,whereas HBV-DNA level(p=0.034)displayed a independent negative correlation.3.Logistic regression analysis in the plasma group uncovered that GGT level(p=0.001),LF Index(p=0.004)and miRNA-125a level(p=0.005)were the independent factors for predicting cirrhosis in CHB patients,and thus a nomogram combining LF Index and plasma miRNA-125a was constructed.4.AUC of the nomogram combining LF Index and plasma miRNA-125a for early predicting cirrhosis in CHB patients was 0.861,which was significantly larger than that of miRNA-125a(p=0.001),GGT(p=0.030),LF Index(p=0.032),APRI(p=0.001)and FIB-4(p<0.001).Its sensitivity(84.62%),specificity(74.32%),positive(53.70%)and negative predictive values(93.20%)were obtained(cut-off value>-1.545).5.DCA uncovered that the nomogram combining LF Index and plasma miRNA-125a for early predicting cirrhosis in CHB patients was the most superb,with a satisfactory clinical applicability.Conclusions:A nomogram combining LF Index,plasma miRNA-125a and GGT could early predict cirrhosis in CHB patients,displaying a prospect of clinical application.Section ⅡObjective:To construct two LF Index-based nomograms for non-invasively predicting hepatitis B cirrhosis combined with GEV(gastroesophageal varices),and to verify its model performance.Methods:Patients diagnosed with hepatitis B cirrhosis(n=123)in The First Affiliated Hospital of Wenzhou Medical University from January 2016 to December 2016 were enrolled in the model group.Meanwhile,patients with hepatitis B cirrhosis(n=184)from January 2017 to December 2018 were enrolled in the verification group.Baseline characteristics(e.g.age,gender,etc.)and laboratory indexes of each patient were recorded.Hemodynamic parameters and LF Index were detected by the ultrasound.Endoscopy findings were the gold standard as controls.Correlation between GEV&LF Index and clinical characteristics&ultrasound parameters in patients with hepatitis B cirrhosis was assessed by Spearman correlation and multiple linear regression.Subsequently,independent factors for diagnosing hepatitis B cirrhosis combined GEV were screened by univariate and multivariate logistic regression.Two nomograms for predicting GEV and GEV severity were thereafter constructed.Discrimination and calibration of the constructed nomograms were evaluated by AUC and calibration curve,respectively.Its clinical applicability was assessed via DCA.Internal validation was finally performed.Results:1.Correlation analyses suggested that PVD(portal vein diameter)(p=0.006),SPI(splenoportal index)(p=0.001)and LF Index(p<0.001)were independently positively correlated with GEV severity in patients with hepatitis B cirrhosis.It is identified that albumin level(p=0.001)was independently negatively correlated with LF Index in these patients.Worse condition of GEV was observed in Child-Pugh B/C patients than those in Child-Pugh A patients(p<0.001).2.Significant differences in age(p=0.040),Child-Pugh class(p<0.001),TBil(total bilirubin)(p=0.018),DBil(direct bilirubin)(p=0.008),albumin(p=0.019),ALP(alkaline phosphatase)(p=0.045),INR(international normalized ratio)(p<0.001),PLT(platelets count)(p<0.001),MPV(mean platelet volume)(p=0.048),PVD(p=0.005),MPVV(mean portal vein velocity)(p=0.023),ST(spleen thickness)(p<0.001),SD(spleen diameter)(p=0.002),SPI(p<0.001)and LF Index(p<0.001)were detected between variceal group and non-variceal group.In addition,significant differences in INR(p<0.001),PLT(p<0.001),MPV(p=0.020),PVD(p=0.003),ST(p<0.001),SD(p=0.001),SPI(p=0.001),LF Index(p<0.001)and Child-Pugh class(p<0.001)were identified between low-risk and high-risk GEV groups.3.Logistic regression analysis uncovered that Child-Pugh class(p=0.002),PLT(p=0.044),SPI(p=0.002)and LF Index(p=0.011)were the independent factors for predicting GEV.Moreover,Child-Pugh class(p<0.001),MPV(p=0.028),PVD(p=0.037),SPI(p=0.034)and LF Index(p=0.004)were the independent factors for predicting high-risk GEV4.AUC of the nomogram-GEV in the model group was 0.916,which was significantly higher than that of LF Index(p=0.001),SPI(p=0.002),PSR(platelet/spleen diameter ratio)(p=0.004),King’s score(p<0.001)and Lok index(p=0.006).Similarly,AUC(0.907)of the nomogram-GEV in the verification group was significantly higher than that of LF Index(p<0.001),SPI(p=0.001),PSR(p=0.016),King’s score(p<0.001)and Lok index(p<0.001).In addition,AUC of the nomogram-high risk in the model group was 0.846,which was significantly higher than that of LF Index(p=0.008),SPI(p=0.004),PSR(p=0.009),King’s score(p<0.001)and Lok index(p=0.002).AUC(0.835)of the nomogram-GEV in the verification group was also significantly higher than that of LF Index(p=0.011),SPI(p=0.001),PSR(p=0.032),King’s score(p<0.001)and Lok index(p=0.004).5.The sensitivity(94.51%),specificity(75.00%),positive(91.50%)and negative predictive values(82.80%)of the nomogram-GEV were obtained in the model group(cut-off value>0.076).They were 91.91%,72.92%,90.60%and 76.10%in the verification group,respectively.In addition,sensitivity(89.33%),specificity(66.67%),positive(80.70%)and negative predictive values(80.00%)of the nomogram-high risk were obtained in the model group(cut-off value>0.064),which were 85.45%,71.62%,81.70%and 76.80%in the verification group,respectively.Conclusions:Two LF Index-based nomograms(nomogram-GEV and nomogram-high risk)could predict GEV and high-risk GEV,respectively.Both of them have quite acceptable accuracy and clinical feasibility,and their qualifications are confirmed by internal validation.Our constructed nomograms are conductive to guide screening strategies and preventive treatment for hepatitis B cirrhosis combined GEV.Section ⅢObjective:To construct LF Index-based nomograms for non-invasively predicting hepatitis B cirrhosis combined with VH(variceal hemorrhage)and ReVH(recurrent variceal hemorrhage)via LASSO regression analysis.Methods:VH and ReVH in the verification group described in Section Ⅱ were retrospectively reviewed.Eventually,169 eligible patients were recruited in this section.A total of 24 baseline variables,including laboratory indexes and ultrasonic parameters,were screened via LASSO regression analysis.Subsequently,LASSO models were subjected to multivariate logistic regression analysis alongside their baseline characteristics,which construct two Nomograms for predicting VH and ReVH in patients with hepatitis B cirrhosis,respectively.Discrimination and calibration of the constructed models were evaluated by AUC and calibration curve,respectively.The clinical applicability was assessed via DCA.Results:1.Significant differences in Child-Pugh class(p=0.002),risk stratification of GEV(p<0.001),INR(p<0.001),HAPV(hepatic artery peak velocity)(p=0.027),PVD(p=0.014)ST(p<0.001),SD(p<0.001),SPI(p<0.001)and LF Index(p<0.001)were found between VH and non-VH patients with hepatitis B cirrhosis.Moreover,significant differences in Child-Pugh class(p=0.010),PDW(platelet distribution width)(p=0.003)and LF Index(p=0.002)were detected between first VH and ReVH patients with hepatitis B cirrhosis.2.Using LASSO regression method,three variables(INR,SPI and LF Index)were screened as predictive factors for VH in patients with hepatitis B cirrhosis.Besides,PDW and LF Index were predictive factors for ReVH in patients with hepatitis B cirrhosis.LASSO-VH and LASSO-ReVH models were constructed.Subsequently,the nomogram integrated with the two LASSO models and baseline clinical characteristics via multivariate logistic regression pointed out that LASSO-VH model and risk stratification of GEV were the independent factors for predicting VH in patients with hepatitis B cirrhosis,whereas LASSO-ReVH model and Child-Pugh class were independent factors for predicting ReVH.3.AUC of the nomogram-VH was 0.826,which was significantly higher than that of LASSO-VH model(p=0.005),risk stratification of GEV(p=0.005)and LF Index(p<0.001).AUC of the nomogram-ReVH was 0.821,which was significantly higher than that of Child-Pugh class(p=0.001)and LF Index(p=0.047),rather than the LASSO-ReVH model(p=0.427).4.The sensitivity(90.14%),specificity(64.29%),positive(64.60%)and negative predictive values(90.00%)of the nomogram-VH were obtained(cut-off value>-0.358).They were 94.59%,55.88%,70.00%and 90.50%in the nomogram-ReVH,respectively(cut-off value>-0.718).Conclusions:Two LF Index-based nomograms could predict risks of VH and ReVH in patients with hepatitis B cirrhosis,and their clinical application is expected. |
PDF Full Text Request