Incidence and Predictors of Relapse After Stopping Antiviral Therapy in Pediatric Chronic Hepatitis B

Background: There are no definite end-points for stopping therapy in pediatric chronic hepatitis B (CHB). The study objective was to evaluate the incidence of relapse after stopping antiviral therapy and to identify its predictors. Methods: All hepatitis B surface antigen (HBsAg) positive children presenting to our hospital, who had been on antivirals for at least 2 years with undetectable hepatitis B virus-deoxyribonucleic acid (HBV-DNA) and normal alanine aminotransferase (ALT) on 3 consecutive occasions over last 12 months were included. Antivirals were stopped if liver biopsy showed histological activity index <5 and fibrosis (Ishak) <3. Virological relapse was defined as the elevation of HBV-DNA (>2000 IU/mL) and biochemical relapse as a rise in ALT levels to >2 times the upper limit of normal. Those having biochemical relapse were started on pegylated interferon alpha-2b-based sequential therapy. Results: Of the 114 children with CHB screened, 31 HBsAg-positive children fulfilled inclusion criteria and antivirals were stopped in them. Virological and biochemical relapse was seen in 12 (38.7%) and 5 (16.1%) children within 12 months of stopping antiviral treatment. On Cox regression, hepatitis B e antigen (HBeAg) positive status at the time of stopping antiviral therapy (HR: 6.208, 95% CI: 1.630–23.638) and longer time taken for HBV-DNA to become undetectable while on antivirals (HR: 1.027, 95% CI: 1.000–1.055) were the independent predictors of relapse. Conclusion: Discontinuation of antiviral treatment in children with CHB resulted in relapse in one-third of the patients. Relapse was frequent in those who were HBeAg-positive at the time of stopping therapy and in those who required longer therapy for HBV-DNA to become undetectable.

T he goal of antiviral therapy in patients with chronic hepatitis B (CHB) is to improve survival by preventing the progression of liver damage to cirrhosis, end-stage liver disease, hepatocellular carcinoma (HCC) and death. 1,2 Most children are in the stage of hepatitis B e antigen (HBeAg)-positive CHB infection (previously known as the immune-tolerant stage) and do not require therapy. Spontaneous HBeAg seroconversion is reported to occur in less than 2% per year among those younger than 3 years and 4-5% per year in older children. 3 Current guidelines recommend treatment for children with HBeAg-positive chronic hepatitis, characterized by persistently elevated alanine aminotransferase (ALT). 4 Tenofovir is the recommended antiviral in children older than 12 years and entecavir in children 2-12 years old. 4 There are clear guidelines for stopping antiviral therapy in adults. [5][6][7][8] In absence of hepatitis B surface antigen (HBsAg) loss, most guidelines recommend stopping antivirals only after at least 12 months of HBeAg seroconversion; and the presence of undetectable hepatitis B virus-deoxyribonucleic acid (HBV-DNA) with persistently normal ALT levels. [5][6][7][8] After discontinuation of antivirals in adults, virological relapse (VR) has been reported in 42%-68%, whereas biochemical relapse (BR) has been reported in 21.7%-63% of patients on follow-up. 9,10 There is no ambiguity about the decision to stop antiviral therapy in children who achieve a functional cure, defined as loss of detectable serum HBsAg with or without seroconversion to hepatitis B surface antibody. 4 However, only 1.7%-8.3% of children treated with antivirals achieve this target of functional cure. [11][12][13][14] There is no consensus about the duration of antiviral therapy in the vast majority who continue to remain HBsAg-positive. Initiation of antivirals at a very young age in children, coupled with a lack of definite endpoint for stopping therapy, often implies an extremely long cumulative dose and duration of exposure to antivirals, thereby increasing the risk of adverse events and drug resistance. [11][12][13][14] An extensive literature search failed to show studies designed to evaluate incidence and predictors of relapse after stopping antivirals in children. Therefore, we planned this pilot research to determine the incidence and identify risk factors of early relapse on stopping antiviral therapy in children with CHB.

METHODS
This was a prospective interventional study conducted at a tertiary care pediatric hepatology center in New Delhi over a period of 2 years (from January 2019 to December 2020). Approval was obtained from the institutional ethical committee vide letter no IEC/2019/67/MA-06. The study was conducted in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki Declaration and its later amendments. Informed, written consent was obtained from the parents/ guardians of the children. The criteria for stopping antivirals in the study were: (1) children between the ages of 2 and 18 years with CHB; (2) having received oral antivirals for at least 24 months; (3) with persistently normal ALT levels over the last 12 months (at least 3 consecutive ALT <1.5 times the upper limit of normal (ULN); (4) undetectable HBV-DNA on 3 consecutive occasions in the last 12 months; and (5) histological activity index (HAI) <5, Ishak fibrosis stage <3 on liver histology. Normal ALT levels were defined as per the cutoffs obtained from the 'SAFETY study', which has defined separate cutoffs in healthy boys (25.8 IU/L) and girls (22.1 IU/L), free of viral hepatitis or fatty liver disease. 15 Children with a family history of cirrhosis, HCC or co-infection with hepatitis C virus and HIV were excluded. Liver Biopsy was performed to evaluate the HAI and fibrosis as per the modified Ishak system. 16 The details of the HAI score and fibrosis staging have been described in Text, Supplemental Digital Content 1, http://links.lww.com/INF/E752. Children who had become HBsAg-negative did not undergo repeat liver biopsy at the time of stopping therapy. Ultrasound-guided percutaneous liver biopsy was performed through an intercostal approach in a mid-axillary line using a Tru-Cut liver biopsy needle (BARD Biopsy Systems, Peripheral vascular Inc. Mexico) (18 G size, 10 cm needle length). The need for short general anesthesia was decided on a case-to-case basis in younger children. Additionally, quantitative estimation of covalently closed circular DNA (ccc-DNA) was also performed in the liver tissue. ccc-DNA levels were measured by real-time polymerase chain reaction (PCR) based in-house assay on lightcycler 480 (LC480) (Roche Diagnostics, Almere, Netherlands) instrument using following primers: Forward primer: 5′-CTCCCCGTCTGTGCCTTCT-3′; Reverse Primer: 5′-GCCCCAAAGCCACCCAAG-3′. This is a qualitative real-time PCR-based assay using sybergreen probes for the target detection. 17 Every run included a negative plasma sample, water and positive control with a known concentration of ccc-DNA (10 3 copies/mg). The results were then analyzed with LightCycler 480 software v1.5.0.39 (Roche Diagnostics). Calculations for the amount of ccc-DNA were carried out based on the standard curve plotted with the dilutions made by using positive control. The results were expressed in copies/mg of the liver biopsy tissue received after its weight measurement. HBV-DNA was measured in the plasma samples by using an automated real-time PCR kit, COBAS AmpliPrep/ COBAS TaqMan, v2.0 (Roche Diagnostics, GmbH, Mannheim, Germany). This is an automated hepatitis B viral load quantitative assay needing 650 µL of serum or plasma, with the linear range of quantification being 1.3-8.23 log 10 IU/ml in all of them. HBsAg quantification was performed by the chemiluminescent immunoassay (CLIA) method (Abbott Laboratories, Chicago, IL). 18 HBeAg was detected using a semi-quantitative chemiluminescence-based assay which gives results in a signal to cutoff (S/Co) ratio, the value of which is directly proportional to the amount of analyte in the sample. The higher the S/Co value higher is the amount of analyte in the sample. A S/Co ratio greater than 1 is reported as HBeAg detected by the assay. Similarly, anti-HBe was analyzed and reported using a semi-quantitative assay with S/Co ratio less than 0.9 reported as the presence of anti-HBe in serum.
After stopping antivirals, the patients were followed up at 1 month, 3 months and every 3 months thereafter (Fig. 1). Workup at each follow-up included: ALT, quantitative HBV-DNA (qHBV-DNA), quantitative HBsAg (qHBsAg), HBeAg, alfa-fetoprotein (AFP) and ultrasonography of the abdomen. Virological relapse was defined as an elevation of qHBV-DNA >2000 IU/mL. Biochemical relapse was defined as an elevation of ALT >2 times ULN. The primary outcome measure of the study was to determine the incidence of relapse after stopping antiviral therapy. A secondary exploratory objective was to identify the predictors of relapse after stopping antiviral therapy. The predictors studied included clinical parameters, antivirals used, HBeAg seroconversion status, time taken for HBV-DNA to become undetectable on therapy, HBV genotype, ccc-DNA levels, duration of normal ALT before stopping therapy and qHBsAg at the time of stopping antivirals. Those with only VR were continued on regular follow-up whereas those with VR as well as BR were treated with combination therapy: 8 weeks of antivirals (entecavir in those younger than 12 years; tenofovir in those older than 12 years), followed by the combination of peg-IFN alpha 2b (60 mcg/m 2 /week) for 48 weeks. Figure 1 depicts the study design and follow-up protocol.

STATISTICAL ANALYSIS
All analyses were performed with SPSS 22.0 (SPSS, Chicago, IL). As this was a pilot study, all the children fulfilling the inclusion criteria during the study period were enrolled. Continuous variables were expressed as median and inter-quartile range (IQR), whereas categorical variables were expressed as proportions. Both categorical and continuous variables were analyzed using Cox analysis. For identifying predictors of VR, univariate analysis was followed by Cox regression analysis. Variables which were significant predictors of VR on univariate analysis and had clinical relevance were entered in the Cox regression multivariate analyses for predictors of VR. Several models were evaluated with different combinations of variables based on their significance in univariate analysis, confounding factors, their relevance based on existing literature and preventing multicollinearity at the same time. A maximum of 1 variable was included in the models for every 5 events. Cox regression was performed using the forward (likelihood ratio) method. The variables remaining in the equation at the last step of regression analysis were deemed independent predictors of relapse. The strength of association was represented as hazard ratio (HR) with its 95% confidence intervals (CI). Receiver operating characteristic curves were made for independent continuous variables and the area under the receiver operating characteristic curve (AUROC) was calculated. Youden's index was used to determine the best cutoff of the particular variable and sensitivity as well as specificity were determined at that cutoff.

RESULTS
A total of 114 CHB children were screened during the study period, of which 39 (34.2%) children satisfied the biochemical and virological criteria for inclusion in the study. Eight of them were HBsAg-negative; whose treatment was stopped without a repeat liver biopsy. Parents of all remaining 31 HBsAg-positive children gave consent for inclusion in the study, liver biopsy and consideration of stopping antiviral treatment. All 31 children had HAI <5 and fibrosis <3 in liver biopsy performed at the time of enrollment and all their parents agreed for close follow-up. Table 1

Relapse After Stopping Antivirals
All 31 HBsAg-positive children enrolled in the study were followed up for a minimum of 12 months. Twelve of these 31 children (38.7%) developed VR. Five children (16.1%) additionally had a BR. Most (58.3%) of the VR occurred within a month of antiviral stoppage (Fig. 2). The BR was more common in those developing VR at 3 months or beyond after stopping antivirals. In 4 children, VR, as well as BR, were detected at the same time-point during follow-up (Fig. 2). Of the 7 children who developed only VR, 6 (85.7%) developed it within 1st month after stopping antiviral treatment. None of the 8 children with HBsAg-negative status developed VR or BR until 12 months of follow-up.

Risk Factors for Development of Relapse
On univariate analysis, HBeAg-positive status at the time of stopping therapy (HR: 7.058, 95% CI: 1.825-27.29, P = 0.005), semi-quantitative HBeAg (HR: 1.002, 95% CI: 1.001-1.003, P < 0.001), semi-quantitative anti-HBe (HR: 1.025, 95% CI: 1.003-1.049, P = 0.028) and longer time taken for qHBV-DNA to become undetectable on antiviral treatment (HR: 1.024, 95% CI: 1.003-1.046, P = 0.028) were the significant predictors of VR after stopping antiviral therapy (Table 2). Various models were evaluated on Cox multivariate regression analysis using several combinations of variables which were either significant on univariate analysis (HBeAg-positive at the time of stopping therapy, semi-quantitative HBeAg, semi-quantitative anti-HBe and time taken for qHBV-DNA to become undetectable) or deemed clinically important based on literature review (quantitative HBsAg and ccc-DNA on liver tissue); while at the same time avoiding multicollinearity (Table 3). HBeAg-positive status and semi-quantitative HBeAg were not entered together in any model to prevent confounding. HBeAg-positive status at the time of stopping therapy was a strong independent   A S/Co ratio of >16 on semi-quantitative HBeAg was a predictor of VR with 75% sensitivity and 89.5% specificity (AUROC 81.1%). ccc-DNA in liver tissue although higher in those with relapse (4.19 log 10 copies/ml (IQR: 3.5-6.99) than in those without relapse 3.2 log 10 copies/ml (IQR: 2.5-5.1), was not statistically different between the 2 groups.

Treatment of Children with Relapse
Four (80%) of the 5 children with BR were treated with sequential therapy. The fifth child only had a transient BR; hence, he was treated with antivirals alone. Of the 4 children who were treated with sequential therapy, 3 were HBeAg-positive whereas 1 was HBeAg-negative. None of the 3 HBeAg-positive children, showed HBeAg seroconversion at 1 year of therapy. Two of them achieved virological clearance (qHBV-DNA undetectable). All 4 achieved normalization of their ALT levels at 1 year of therapy. The rest of the patients with only VR are being continued on strict follow-up every 3 months for timely identification of BR and possible re-treatment. Most of the guidelines in adult hepatitis B recommend stopping antivirals after 12-36 months of HBe seroconversion with undetectable HBV-DNA and persistently normal ALT levels. [6][7][8] There are no studies specifically designed to evaluate the timing and criteria for stopping antiviral therapy in children. Very low HBsAg seroconversion (1.7-8.3%) and HBeAg seroconversion (21-31.3%) rates observed in children on antivirals imply longer, often indefinite antiviral therapy. [11][12][13][14] Longer duration of therapy increases the cumulative risk of adverse events as well as the risk of development of antiviral resistance. 4 The present study provides some insights into patient selection for stopping antivirals in children who fail to achieve HBsAg seroconversion. VR rate of 38.7% and BR rate of 16.1% at 12 months observed in children in our study is consistent with relapse rates observed in adults after antiviral discontinuation. The VR rates reported have varied from 42% at 3 months to 68% at 6 years of follow-up, whereas BR rates have varied from 21.7% at 3 months to 63% at 6 years of follow-up after antiviral discontinuation. 9,10 HBeAg-positive status at the time of stopping therapy and the longer time taken for qHBV-DNA to become undetectable were the most important predictors of relapse in the present study. The optimal endpoint for stopping therapy would be the loss of HBsAg, often termed 'functional cure' but it is only rarely achievable in children. 7,13,14 Short of it, HBeAg seroconversion, coupled with HBV-DNA loss and ALT normalization characterizes a state of good immune control with a low replicative phase of the disease. 7 The high rates of early relapse (69.2%) seen in HBeAg-positive children in the present study suggest that treatment withdrawal should not be attempted in the absence of HBeAg seroconversion. The important predictors of relapse after stopping antivirals in adults include (1) higher qHBsAg level at the time of stopping the therapy 19-21 ; (2) shorter duration of consolidation therapy 20,22-24 ; (3) slower HBV-DNA decline while on therapy 25 ; (4) baseline pretreatment HBV-DNA 26,27 ; (5) hepatitis B core-related antigen (HBcrAg) 28,29 ; and (6) hepatitis B virus-ribonucleic acid (HBV-RNA)-. 29 The maximum  19 qHBsAg levels were not significantly higher in the patients who relapsed in the present study either on the univariate or in the multivariate model. HBcrAg and HBV-RNA are among the newer serological markers predicting relapse either singly or in combination. 28,29 We evaluated the role of liver tissue ccc-DNA which although higher in relapse was not statistically different. This could be due to the smaller sample size and would need evaluation in larger multicentric studies. Recently, Lai et al 30 investigated the effect of antiviral withdrawal in patients with undetectable ccc-DNA in the hepatocytes in adults. 31 All 13 patients with undetectable ccc-DNA randomized to stop therapy had VR. Although either of HBeAg-positive status or a quantitative HBeAg could be used to predict relapse, it may be more prudent to use qualitative HBeAg as many centers may not have the facility for doing a quantitative assay.

DISCUSSION
The main strength of our study is the pilot effort in determining the relapse rate and its risk factors in a carefully curated cohort of children with CHB. Even though our study period coincided in part with the COVID-19 pandemic which could have limited the enrollment, we managed to ensure timely follow-up of each enrolled patient using online consultation wherever necessary. Our study is limited by a small sample size being a single-center study. Although limited by a short follow-up of 12 months, we continue to follow-up these patients for timely identification and treatment of new relapses. We also could not identify the predictors of BR due to small number of patients (n = 5) developing BR. The predominance of boys in the study is a limiting factor and unfortunately reflects the health-seeking behavior of the population in this part of the world. The predominance of genotype D in our cohort is representative of the population. To conclude, virological and biochemical relapse rates were 38.7% and 16.1%, respectively within 12 months of stopping antivirals in HBsAg-positive children with most relapses occurring within a month of stopping therapy. HBeAg-positive status at the time of stopping therapy and the longer time required for HBV-DNA to become undetectable while on therapy are important predictors of relapse after stopping antivirals.