Noninvasive Models To Identify The Need For Antiviral Treatment In Patients With Chronic Hepatitis B

PART I A Noninvasive Model to Identify the Need for Antiviral Treatment in Patients with HBeAg Negative Chronic Hepatitis BBackgroundHepatitis B virus (HBV) infection is prevalent across the world. Approximately one third of the world’s population has serological evidence of past or present infection with HBV, among which an estimated of 350 million people are chronically infected. In patients with chronic hepatitis B (CHB), the covalently closed circular DNA (cccDNA) of HBV is hard to be eradicated from the nucleus of hepatocytes. The progression of HBV infection may evolve to liver cirrhosis, liver failure (LF) as well as hepatocellular carcinoma (HCC). Nearly 15% to 40% HBV carriers will develop serious complications during their lifetime. There are approximately 1 million patients who die of these complications annually.Active HBV replication was found to be the key driver of liver damage ar|d disease progression. Therefore, the primary aim of treatment for chronic HBV infection is to permanently suppress HBV replication, which may decrease pathogenicity and infectivity of the virus. The treatment of chronic hepatitis B (CHB) is mainly based on serum HBV DNA levels, serum alanine aminotransferase (ALT) levels and severity of liver disease. According to the Asian-Pacific Association for the study of the liver (APASL) practice guidelines, HBeAg negative CHB patients with HBV DNA levels >2000IU/ml and ALT≤2 Upper limit of normal (ULN) should be evaluated for liver histological conditions. Only when liver histology shows moderate to severe active necro-mflammation (liver inflammation grade (G)>2) or at least moderate fibrosis (liver fibrosis stage(S)≥2), antiviral treatment should be considered.At present, the evaluation of liver histological condition mainly depends on liver biopsy. However, it is an invasive procedure with inherent risk. The main disadvantages of liver biopsy include poor patient compliance, low reproducibility, high expense, sampling error, limited usefulness for dynamic surveillance and follow-up. After biopsy, nearly 30% of patients feel pain,0.3% have severe complications such as bleeding or pneumothorax, and 0.01%-0.1% even die. Therefore, patients are usually reluctance to undergo liver biopsy and noninvasive predictors for liver histology are urgently needed in clinical work.In recent years, many studies have been carried out to propose noninvasive models for predicting liver fibrosis or cirrhosis. Mainly, they can be divided into two major groups:the elastography imaging and serum markers. These studies imply that constructing noninvasive model with routine serum markers is feasible and has broad application prospect. However, there is no model that is strictly validated and can be used directly for predicting antiviral treatment in patients with HBeAg negative CHB until now.ObjectiveThe aim of this study is to construct a reliable diagnostic model with routine serum markers, which can be used for identifying the need for antiviral treatment in patients with HBeAg negative CHB.MethodsThis study retrospectively enrolled 301 consecutive HBeAg negative CHB patients hospitalized from June 2009 to January 2013. All the patients had HBV DNA ≥2000IU/ml and ALT≤2ULN and had never received antiviral treatment. All the patients underwent percutaneous liver biopsies. Clinical, biochemical, and hematological data were recorded from each patient prior to liver biopsy. According to the enrollment time, all patients were divided into two sets. The training set included 141 patients hospitalized from June 2009 to October 2011. The validation set included 133 patients hospitalized from November 2011 to December 2013.In the training set, univariate analysis was used to identify markers that could differentiate mild liver histological lesion (G<2 and S<2) and moderate to severe liver histological lesion (G≥2 or S≥2). Then, the identified variables entered multivariate stepwise forward logistic regression to construct a predictive model. In the validation set, the model was validated.Statistical analyses of the data were performed with SPSS version 16.0 (Chicago, IL, USA). Non-normal distribution quantitative variables were analyzed hi a logarithmic scale to improve the normality of distribution.χ2 test was used to compare the categorical variables between patients with mild liver histological lesion and those with moderate to severe liver histological lesion. Independent t test were used to compare the quantitative variables between two groups. The variables identified in the univariate analysis entered multivariate stepwise forward logistic regression in order to construct a predictive model. Diagnostic value of this new index was assessed by the area under the receiver operating characteristic curve (AUC). Diagnostic accuracy was assessed by sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV).2-tailed P value less than 0.05 was considered statistically significant.Results1. After univariate analysis, Age, ALT, aspartate aminotransferase (AST), gamma glutamyl transpeptidase (GGT), total bilirubin (TBIL), albumin (ALB), platelet (PLT) and HBV DNA were identified significantly different between patients with mild liver histological lesion and those with moderate to severe liver histological lesion (p<0.05) in the training set.2. In the training set, the variables identified by the univariate analysis were assessed by multivariate stepwise forward logistic regression. After analysis, Age, AST, HBV DNA, PLT were selected as independent predictors and a model was constructed with them. The formula of this model is given below. Mnegative= 0.049*Age+2.858* Ln(AST)-0.352* Log10[HBV DNA]-0.02* PLT+0.2133. The model values of patients with moderate to severe liver histological lesion were significantly higher than those with mild liver histological lesion in the training set (t=-9.387, p<0.01), validation set (t=-7.137, p<0.01) and all patients (t=-11.696, p<0.01).4. When used to predict antiviral treatment in HBeAg negative CHB patients with HBV DNA levels≥2000IU/ml and ALT≤<2ULN, the model had an AUC of 0.858 (standard error [SE] 0.031,95% confidence interval [CI] 0.793-0.908) in the training set. Meanwhile, it had an AUC of 0.843 (SE 0.036,95% CI 0.773-0.898) in the validation set and an AUC of 0.850 (SE 0.023,95% CI 0.805-0.889) in all patients.5.5.16 and 7.26 were chosen as cut-off points for this model in predicting antiviral treatment in HBeAg negative CHB patients with HBV DNA levels ≥2000IU/ml and ALT≤2ULN. Using cut-off point of 5.16, the model showed 94% sensitivity,88% NPV in the training set and 93% sensitivity,85% NPV in the validation set using the low cut-off point of. Using cut-off point of 7.26, it showed 92% specificity,88% positive predictive value in the training set and 94% specificity,92% PPV in the validation set6. We stratified all the patients by age. The model showed an AUC of 0.848 (SE 0.029, 95%CI 0.79-0.90) in the 196 patients with age ≤40yr. Meanwhile, it showed an AUC of 0.821 (SE 0.047,95%CI 0.73-0.89) in the 105 patients with age >40yr.7. Patient with a model value ≤5.16 should not be treated. Patient with a model value >7.26 should be considered for treatment. Liver biopsy is needed only when patient has a model value between the two cut-off points. Therefore, in this study, the model could help 95 of 158 (60%) patients in the training set,84 of 143(59%) patients in the validation set and 179 of 301(59%) in all patients make treatment decision without liver biopsy.Conclusions1. This study constructed a reliable diagnostic model with routine serum markers, which could identify the need for antiviral treatment in patients with HBeAg negative CHB.2. The model showed high diagnostic value and diagnostic accuracy in both training set and validation set.3. The model showed high diagnostic value in patients of different age group.4. The model could be an important complement to liver biopsy and used in patients who were reluctant or unable to accept liver biopsy. In this study, the model could help more than half patients with HBeAg negative CHB make treatment decisions without liver biopsy.PART II A Noninvasive Model to Identify the Need for Antiviral Treatment in Patients with HBeAg Positive Chronic Hepatitis BBackgroundHepatitis B virus (HBV) infection is one of the most serious and prevalent health problems, affecting more than 2 billion people worldwide. HBV infection may cause sustained liver damage. HBV genome may also integrate into the host genome and favor oncogenesis. People with hepatitis B are at an increased risk of developing hepatic decompensation, cirrhosis, and hepatocellular carcinoma (HCC). After diagnosis,8% to 20% of untreated patients with chronic hepatitis B (CHB) developed cirrhosis within five years. Untreated patients with decompensate cirrhosis have a poor prognosis with a 14%-35% probability of survival at 5 years.Antiviral treatment can suppress HBV replication in a sustained manner and improve the histological condition of liver. It can prevent progression of CHB to cirrhosis, HCC, end-stage liver diseases and death. The indications for antiviral treatment are mainly based on the combination of serum HBV DNA levels, serum ALT levels and severity of liver disease. According to the Asian-Pacific Association for the study of the liver (APASL) practice guidelines, HBeAg positive CHB patients with HBV DNA levels≥20,000IU/ml and ALT≤2ULN should be evaluated for liver histological conditions. Antiviral treatment should be performed if liver histology shows moderate to severe active necro-inflammation (liver inflammation grade (G)≥2) or at least moderate fibrosis (liver fibrosis stage(S)≥2).At present, liver biopsy remains the golden standard for assessing liver inflammation and fibrosis. However, many patients are reluctant to undergo liver biopsy. Liver biopsy is invasive, expensive and always associated with some discomfort and complications, In addition, limitations of biopsy include low reproducibility, sampling errors, and modest intra- and inter-observer variation. Therefore, there is a need to establish new noninvasive diagnostic methods for assessing liver histology.In recent years, many studies have been dedicated to developing noninvasive serum markers for predicting liver fibrosis and cirrhosis and several noninvasive markers have been constructed. Direct serum markers such as hyaluronate, laminin, procollagen Ⅲ amino-terrninal peptide (PIIINP), type IV collagen, matrix metalloproteinase (MMP), tissue inhibitor of metalloproteinase-1 (TIMP1) reflect the deposition or removal of extracellular matrix in the liver. Although these markers showed high efficacy in predicting cirrhosis, they were not available in most medical facilities and showed relatively low predicting accuracy for liver fibrosis. Indirect serum markers like Fibrotest, Forn index, aminotransferase to platelet ratio Index (APRI), Goteborg University Cirrhosis Index (GUCI) and FIB-4, have been proposed for predicting liver fibrosis or cirrhosis. With limited expense and widespread availability, they are regarded as potential alternatives to liver biopsy. However, most of these models are proposed in patients with chronic hepatitis C (CHC), which have not yet been validated in patients with CHB. There is no model that is constructed especially for identifying the need for antiviral treatment in patients with HBeAg positive CHB until now.ObjectiveThe aim of our study is to construct a noninvasive model with routine serum markers to indentify the need for of antiviral treatment in patients with HBeAg positive CHB. It may help a large proportion of patients avoid liver biopsy.MethodsFrom June 2009 to December 2013, we retrospectively enrolled 274 patients with HBeAg positive CHB, ALT≤2ULN and HBV DNA≥20,000 IU/ml. All these patients underwent percutaneous liver biopsy and had never received antiviral treatment. Clinical, biochemical, and hematological data were recorded from each patient prior to liver biopsy. According to the enrollment time, they were divided into two sets. The training set included 141 patients hospitalized from June 2009 to October 2011. The validation set included 133 patients hospitalized from November 2011 to December 2013.In the training set, univariate analysis was used to identify markers that were significantly different between patients with mild liver histological lesion (G<2 and S<2) and those with moderate to severe liver histological lesion (G≥2 or S>2). Then, the identified variables entered multivariate stepwise forward logistic regression to construct a predictive model. Finally, the model was validated in the validation set.Statistical analyses of the data were performed with SPSS version 16.0 (Chicago, IL, USA). Non-normal distribution quantitative variables were analyzed in a logarithmic scale to improve the normality of distribution,χ2 test was used to compare the categorical variables between two groups. Independent t test were used to compare the quantitative variables between two groups. Correlation was evaluated by Spearman correlation coefficient as a univariate analysis. The variables identified in the univariate analysis entered multivariate stepwise forward logistic regression in order to construct a predictive model. Diagnostic value of this model was assessed by the area under the receiver operating characteristic curve (AUC). Diagnostic accuracy was assessed by sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV).2-tailed P value less than 0.05 was considered statistically significant.Results1. In the training set, age, hepatitis B surface antigen (HBsAg), albumin (ALB), albumin-globulin ratio (A/G), ALT, aspartate aminotransferase (AST), gamma glutamyl transpeptidase (GGT), total bilirubin (TBIL) and platelet (PLT) were significantly different between patients with mild liver histological lesion and those with moderate to severe liver histological lesion after univariate analysis.2. In the training set, the variables identified by the univariate analysis were then assessed by multivariate stepwise forward logistic regression. After analysis, Age, AST, ALB, PLT were identified as independent predictors and a model was constructed by them. The model formula was presented below. Mpositive=0.065* Age+2.185* Ln(AST)-0.216* ALB-0.011* PLT+6.0643. The model values of patients with moderate to severe liver histological lesion were significantly higher than those with mild liver histological lesion in the training set (t=-10.027, P<0.01), validation set (t=-8.677, P<0.01) and all patients (t=-13.208, P<0.01).4. When used to identify the need of antiviral treatment in HBeAg positive CHB patients with HBV DNA levels ≥20,O00IU/ml and ALT ≤2ULN, the model had an AUC of 0.887 (standard error [SE] 0.029,95% confidence interval [CI] 0.822-0.934) in the training set. Meanwhile, it had an AUC of 0.854 (SE 0.032,95% CI 0.783-0.909) in the validation set and an AUC of 0.871 (SE 0.021,95% CI 0.825-0.908) in all patients.5.3.52 and 5.83 were chosen as cut-off points for this model in predicting antiviral treatment in HBeAg positive CHB patients with HBV DNA levels ≥20,000IU/ml and ALT ≤2ULN. Using low cut-off point of 3.52, it showed 94% sensitivity,90% NPV in the training set and 94% sensitivity,91% NPV in the validation set. Using high cut-off point of 5.83, it showed 92% specificity,88% PPV in training set and 96% specificity, 91% PPV in the validation set.6. We stratified all patients by age and assessed the predictive values of the model in identifying the need for antiviral treatment in different age groups. The model showed an AUC of 0.861 (SE 0.028,95%CI 0.802-0.908) in the 180 (65.69%) patients with age≤40yr. Meanwhile, it showed an AUC of 0.864 (SE 0.039,95%CI 0.778-0.926) in the 94 (34.31%) patients with age>40yr. Then we stratified all patients by ALT level. The model showed an AUC of 0.827 (SE 0.045,95%CI 0.737-0.896) in the 97 (35.40%) patients with ALT<1ULN was presented. Meanwhile, it showed an AUC of 0.885 (SE 0.025,95%CI 0.829-0.928) in the 177 (64.60%) patients with ALT 1-2ULN.7. Patient with a model value≤3.52 should not be treated. Patient with a model value >5.83 should be considered for treatment. Liver biopsy was needed only when patient had a model value between the two cut-off points. With this model,92 of 141 (65%) patients in the training set,80 of 133(60%) patients in the validation set and 172 of all the 274(63%) patients could make treatment decisions without liver biopsy.Conclusions1. This study constructed a noninvasive model with routine serum markers to identify the need for antiviral treatment in patients with HBeAg positive CHB.2. The model showed high diagnostic value and diagnostic accuracy in both training set and validation set.3. The model showed high diagnostic value in patients with different age and ALT levels.4. The model could be an important complement to liver biopsy and help a large proportion of patients avoid liver biopsy. In patients who were reluctant or unable to accept liver biopsy, the model might be especially useful. In this study, the model could help more than half patients with HBeAg positive CHB make treatment decisions without liver biopsy.