Incidence and predictors of retreatment in chronic hepatitis B patients after discontinuation of entecavir or tenofovir treatment

Background This study investigated the incidence and predictors of retreatment after discontinuation of either entecavir (ETV) or tenofovir disoproxil fumarate (TDF) treatment in Taiwan. Methods A total of 535 non-cirrhotic chronic hepatitis B (CHB) patients undergoing either ETV (n = 358) or TDF (n = 177) treatment were enrolled. Patients were followed for at least 12 months after stopping ETV or TDF treatment. Most patients (86.3%) fulfilled the retreatment criteria of Taiwan's National Health Plan. Results The 5-year cumulative rates of clinical relapse and retreatment were 52.1% and 47%, respectively, in 160 hepatitis B e antigen (HBeAg)-positive patients, and were 62% and 54.8%, respectively, in 375 HBeAg-negative patients. The median duration from the end of treatment until clinical relapse and retreatment was 40 and 57 weeks, respectively, for all patients. Multivariate Cox regression analysis revealed that discontinuing TDF treatment, old age, male gender, and higher baseline HBsAg levels were independent factors of retreatment in HBeAg-positive patients; old age, HBV genotype B, and higher baseline and end-of-treatment HBsAg levels were independent factors in HBeAg-negative patients. A total of 18.8% of retreated patients satisfied the retreatment criteria of hepatic decompensation according to Taiwan's National Health Plan. Of the 64 patients who had clinical relapse without retreatment, 17 achieved sustained virological remission and 26 did not experience clinical relapse until their last visit after clinical relapse. Four patients developed HBsAg loss. Conclusions The 5-year retreatment rate was about 50% in HBeAg-positive and HBeAg-negative patients. Discontinuing TDF treatment was an independent factor of retreatment in HBeAg-positive patients.


Results
The 5-year cumulative rates of clinical relapse and retreatment were 52.1% and 47%, respectively, in 160 hepatitis B e antigen (HBeAg)-positive patients, and were 62% and 54.8%, respectively, in 375 HBeAg-negative patients. The median duration from the end of treatment until clinical relapse and retreatment was 40 and 57 weeks, respectively, for all patients. Multivariate Cox regression analysis revealed that discontinuing TDF treatment, old age, male gender, and higher baseline HBsAg levels were independent factors of retreatment in HBeAg-positive patients; old age, HBV genotype B, and higher baseline and end-of-treatment HBsAg levels were independent factors in HBeAg-negative patients. A total of 18.8% of retreated patients satisfied the retreatment criteria of hepatic decompensation according to Taiwan's National Health Plan. Of the 64 patients who had clinical relapse without retreatment, 17 achieved sustained virological remission and 26 did not experience clinical relapse until their last visit after clinical relapse. Four patients developed HBsAg loss. PLOS

Introduction
To date, hepatitis B virus (HBV) infection still remains one of the most challenging global public health issues-it may cause chronic hepatitis and lead to cirrhosis and hepatocellular carcinoma [1,2]. Entecavir (ETV) and tenofovir disoproxil fumarate (TDF) are potent nucleos(t)ide analogues (NA) that function as first-line therapies for chronic hepatitis B (CHB) [3][4][5]. However, life-long therapy is usually necessary due to the low rate of hepatitis B surface antigen (HBsAg) loss, and the rate of HBV relapse is typically high after the cessation of NA therapy [6][7][8][9].
In our prior study [10], we reported that the 8-year cumulative rates of virological relapse and clinical relapse after stopping either ETV or lamivudine therapy in hepatitis Be antigen (HBeAg)-positive patients were 55.6% and 47.7%, respectively; the corresponding values in HBeAg-negative patients were 69.3% and 58.9%, respectively [10]. However, the rates of retreatment after discontinuation of NA therapy remain unclear. Recent studies have demonstrated that discontinuation of TDF was associated with earlier relapse than the discontinuation of ETV in CHB patients [7,11,12]. Retreatment rates during off-therapy follow-up between ETV and TDF treatment based on the same criteria of retreatment have been rarely compared. However, consensus for retreatment criteria has yet to be determined according to three international guidelines (Asian Pacific Association for the Study of the Liver (APASL), European Association for the Study of the Liver (EASL), and the American Association for the Study of Liver Diseases (AASLD)) [3][4][5].
A recent study demonstrated that CHB patients with clinical relapse who remained untreated had a 7.34-time higher incidence of HBsAg seroclearance than patients who received retreatment [13]. Additional future studies should focus on clinical outcomes in CHB patients with clinical relapse but without retreatment.
In this study, we investigated the incidence and predictors of retreatment in CHB patients without cirrhosis after the discontinuation of either ETV or TDF treatment. Fig 1 shows the flowchart of patients enrolled in this prospectively retrospective cohort study. We enrolled 358 non-cirrhotic CHB patients who underwent ETV treatment between 2008 and 2015 (112 HBeAg-positive patients and 246 HBeAg-negative patients) and 177 non-cirrhotic CHB patients who underwent TDF treatment between 2011 and 2015 (48 HBeAg-positive patients and 129 HBeAg-negative patients). All of the patients were followed for at least 12 months after the discontinuation of NA therapy. Of the 358 ETV-treated patients, 301 were already included in our prior study to analyze HBsAg loss [10]. Non-cirrhosis was diagnosed based on biopsy (n = 20), a Fibrosis-4 (FIB-4) or liver stiffness measurement at baseline, or combined repeated ultrasound findings and without clinical features such as splenomegaly, thrombocytopenia, ascites, or gastroesophageal varices at baseline. In Taiwan, ETV and TDF has been reimbursed by National Health Plan for HBV treatment since august 2008 and June 2011, respectively. The selection of ETV or TDF was determined via a discussion between the patients and their physicians. All patients were tolerable and did not change drug during treatment.

Patients
Our patients satisfied the NA stopping criteria of the APASL 2012 [14]. During the study period, patients without liver cirrhosis were reimbursed for 3-year NA treatments by Taiwan's  National Health Plan. When determining the treatment plan, the period of NA treatment was defined by the clinical physicians or the patients themselves. In order to avoid extra expenses, most patients (420/535; 78.5%) discontinued ETV or TDF therapy according to the criteria of Taiwan's National Health Plan.
Patients were excluded as described in detail previously [6,10,11]. The study was conducted in accordance with the Declaration of Helsinki of 1975. Written informed consent to participate in the study was obtained from all of the patients. All patients consented to providing serum for the purpose of checking HBV genotype, HBV DNA, and HBsAg levels. The Research Ethics Committees of Chang Gung Memorial Hospital (Kaohsinung, Taiwan) approved this study.

Follow-up scheme
After discontinuation of either ETV or TDF therapy, the patients were followed as described in detail previously [11]. Serum samples from each patient were frozen and stored at -20˚C at baseline, at month 12, and at the end of treatment for further use. Serum HBsAg quantification was investigated using the stored serum retrospectively.

Definition of HBV relapse and criteria of retreatment
The treatment criteria required to initiate NA therapy according to Taiwan's National Health Plan are ALT � 2× upper limit of normal (ULN) (40 U/L) plus HBV DNA > 20000 IU/mL in HBeAg-positive patients, and ALT elevation � 2× ULN in two separate measurements obtained 3 months apart and HBV DNA � 2000 IU/mL in HBeAg-negative patients without cirrhosis. Stopping criteria according to the APASL 2012 [14] include a combination of HBeAg seroconversion and undetectable HBV DNA levels for at least 12 months in HBeAgpositive patients. In HBeAg-negative patients, discontinuation is recommended when the patients have been treated for at least 2 years with undetectable HBV DNA documented on three separate occasions at least 6 months apart.
The definition of clinical relapse was ALT elevation > 2× ULN plus HBV DNA > 2000 IU/ mL after discontinuation of NA therapy [14]. In HBeAg-positive patients, the definition of consolidation duration was the time elapsed from HBeAg seroconversion with undetectable HBV DNA until the end of the treatment regimen. In HBeAg-negative patients, the consolidation duration was calculated from the time of first undetectable HBV DNA until the end of the treatment regimen.
In non-cirrhotic patients, the criteria for retreatment according to Taiwan's National Health Plan are HBeAg reversion with an ALT elevation � 2× ULN with HBV DNA � 20000 IU/mL in HBeAg-positive patients and two measurements of ALT elevation � 2× ULN obtained 3 months apart and HBV DNA � 2000 IU/mL in HBeAg-negative patients. In patients with hepatic decompensation, the criteria for retreatment according to Taiwan's National Health Plan were total bilirubin � 2 mg/dL or prolonged prothrombin time � 3 seconds, regardless of HBV DNA levels.

Serology and HBV genotyping
HBsAg, HBeAg, anti-hepatitis C virus antibodies, anti-hepatitis D virus antibodies, and HBV DNA and HBsAg levels were determined as described in detail previously in the literature [6,10,11].
HBV genotypes were determined using restriction fragment length polymorphism (RFLP) on the surface gene, as described previously in the literature [15]. Direct sequencing of surface genes was performed if the results of HBV genotype by RFLP were ambiguous or indicated a mixed infection. Compared with direct sequencing, the accuracy of HBV genotyping with this RFLP method exceeded 95%.

Statistical analysis
Data are presented as means±standard deviation (SD), proportions, or median (range). Differences between the two of groups of patients were evaluated using the chi-squared test or Fisher's exact test and the Student's t-test for categorical and continuous variables, respectively. The cumulative incidence of clinical relapse and retreatment was estimated using the Kaplan-Meier method and compared using a log-rank test. Univariate and multivariate Cox proportional hazards regression models were used to identify factors associated with clinical relapse and meeting the criteria for retreatment. All variables with P<0.2 in the univariate analysis are required to allow into the multivariate analysis. Cox proportional hazards regression models with the forward method were used to determine independent factors, and variables with P<0.05 were required to stay in the models. All tests were two-sided, and P values less than 0.05 were considered to be statistically significant. Time-dependent receiver-operating characteristic (ROC) curve analysis was used to determine the cut-off values of HBsAg levels for predicting clinical relapse and retreatment within 5 years after discontinuation of either ETV or TDF treatment [16].
The propensity score (PS)-matching method was used to reduce the significant differences in clinical features between the ETV and TDF groups by creating a ratio of 1:1 to 1:2. After subgrouping by HBeAg status, the variables used in the model included age, gender, baseline ALT, HBV genotype, NA-naïve, baseline HBV DNA levels, treatment and consolidation duration, and HBsAg levels at baseline and end of treatment in the ETV and TDF groups. Caliper matching was performed on the propensity score (nearest available matching). Pairs (discontinuing and continuing groups) on the propensity score logit were matched to within a range of 0.2 SD.

Clinical characteristics of the study population
A total of 535 patients were enrolled in the study (160 HBeAg-positive patients and 375 HBeAg-negative patients). The clinical features of the study population are listed in Table 1. A male gender and an HBV genotype B were predominant among the study population. The median duration from the end of treatment until retreatment was 42 weeks in HBeAg-positive patients and 64 weeks in HBeAg-negative patients. Compared with HBeAg-negative patients, HBeAg-positive patients were with younger, more often female and of HBV genotype C, characterized by a shorter consolidation duration, and more likely to exhibit higher levels of ALT, HBV DNA, and HBsAg at entry.
After stopping NA therapy, a total of 286 patients experienced clinical relapse during the follow-up period (78 in the HBeAg-positive group and 208 in the HBeAg-negative group). The median duration from the end of treatment until clinical relapse was 40 weeks (range: 7-411 weeks). In addition, a total of 234 patients received retreatment during off-therapy follow-up. The median duration from the end of treatment until retreatment was 57 weeks (range: 10-411 weeks). Of these patients, 65 were in the HBeAg-positive group and 169 were in the HBeAg-negative group. A total of 202 (86.3%) patients fulfilled the retreatment criteria of Taiwan's National Health Plan. Thirty-two (13.7%) patients received retreatment at their own expense under the patients' willingness.
The clinical characteristics of the patients at baseline and the end of treatment of ETV and TDF treatment are listed in Table 2. The TDF group was generally older, had higher baseline HBV DNA levels, a shorter treatment and consolidation duration, and a lower rate of NAnaïve status compared with patients in the ETV group. There were no significant differences in clinical characteristics at end of treatment between the two groups. To reduce the significant differences in baseline and on-treatment factors, the PS-matching method yielded 277 and 170 patients in the ETV and TDF groups, respectively. The baseline characteristics of both groups according to PS-matching method by HBeAg status are shown in S1 Table.

Optimal cut-off-point of serum HBsAg levels for predicting clinical relapse and retreatment
Because end-of-treatment HBsAg level was an independent predictor of clinical relapse and retreatment in HBeAg-negative patients, we attempted to determine the best cut-off value for predicting clinical relapse and retreatment. Time-dependent ROC analysis revealed that the  ; specificity: 55.2%) within 5 years after discontinuation of either ETV or TDF treatment. We adopted an HBsAg level of 150 IU/mL at the end of treatment as an optimal value for predicting retreatment in HBeAg-negative patients (P < 0.001). The 5-year cumulative retreatment rates were 31.6%, 60.6%, and 64.1% for patients who achieved end-of-treatment HBsAg levels of <150, 150-500, and >500 IU/mL, respectively (Fig 4).

Highest ALT levels and hepatic decompensation upon retreatment
Among the 234 patients who received retreatment, 128, 66, and 12 patients experienced the highest ALT levels (>200, >400, and >1000 U/L upon retreatment, respectively). Total bilirubin � 2 and � 3 mg/dL upon retreatment developed in 44 (18.8%) and 10 (4.3%) patients, respectively, and prolonged PT �3 seconds upon retreatment developed in 12 (5.1%) patients (Table 5). Therefore, 18.8% patients who received retreatment fulfilled the definition of retreatment of hepatic decompensation according to Taiwan's National Health Plan. There was no significant factor at baseline and end of treatment between patients with and without hepatic decompensation upon retreatment (Table 5) In addition, 7 (3%) patients who received retreatment fulfilled the definition of hepatic decompensation according to the 2015 APASL guidelines [3]. Univariate and multivariate analysis revealed that only higher HBV DNA levels at baseline (Odds ratio (OR): 2.16, 95% CI: 1.02-4.58, P = 0.043) were a significant factor of hepatic decompensation according to the  Incidence and predictors of retreatment in CHB patients 2015 APASL guidelines after adjusting for all baseline and end-of-treatment factors. Three patients experienced clinical events of decompensation (2 ascites, 1 hepatic encephalopathy). The time to develop clinical events of decompensation was 15, 40, and 91 weeks after discontinuation of NA therapy. One patient died after timely retreatment.

Clinical outcomes in patients with clinical relapse without retreatment
A total of 64 patients with clinical relapse did not receive retreatment. These patients did not meet the criteria for retreatment according to Taiwan's National Health Plan. Of these patients, 17 (26.6%) achieved sustained viral remission (HBV DNA � 2000 IU/mL) and 26 (40.6%) did not experience clinical relapse (but HBV DNA >2000 IU/mL) from the follow up until the last visit after their clinical relapse. Twenty-eight experienced persistent ALT normalization (� 40 U/L) and 4 developed HBsAg loss after clinical relapse.

Discussion
Our large-scale study revealed that 5-year cumulative retreatment rates were 47% and 54.8%, respectively, in HBeAg-positive and HBeAg-negative patients based on the retreatment criteria of Taiwan's National Health Plan. The retreatment rate represented the rate of persistent hepatitis flare (including ALT and HBV DNA) after discontinuation of NA therapy. In our study, the median duration from the end of treatment until clinical relapse and retreatment was 40 and 57 weeks, respectively. The definitions of retreatment may have a significant impact on the probability of retreatment. According to the retreatment criteria of Taiwan's National Incidence and predictors of retreatment in CHB patients Health Plan, patients could be retreated if two measurements of ALT elevation � 2× ULN obtained 3 months apart and HBV DNA � 2000 IU/mL in HBeAg-negative status after discontinuation of NA therapy. Thus, the criteria might be the main reason for the difference of 17 weeks between clinical relapse and retreatment. One recent study in Taiwan reported that 40 out of 100 patients (40%) (including HBeAgpositive and HBeAg-negative patients at entry) received retreatment for 35 weeks of follow-up after stopping either ETV or TDF therapy [12]. A study in Korea showed that 12 of 45 (26.6%) HBeAg-positive patients and 27 of 68 (39.7%) HBeAg-negative patients received retreatment after stopping either ETV or lamivudine therapy (the criteria of retreatment after HBV relapse were not defined) [17]. A recent study in Greece that enrolled 57 non-cirrhotic, HBeAg-negative CHB patients who discontinued effective � 4-year ETV or TDF therapy showed that the cumulative rate of retreatment was 18% and 26% at 3 and 12 months, respectively [18]. The retreatment criteria of the Greece study were ALT >10× ULN or ALT >5× ULN and total bilirubin >2 mg/dL at the same visit, ALT >3× ULN and HBV DNA >100,000 IU/mL at three sequential visits. A collaborative Greek and Taiwanese study enrolling 130 non-cirrhotic HBeAg-negative patients showed that cumulative retreatment rates were 22% and 40% 12 and 24 months, respectively, after stopping NA therapy according to the retreatment criteria of Greece and Taiwan [19].
The effects of baseline factors in terms of predicting retreatment after stopping NA therapy have not been well established. In our study, multivariate Cox regression analysis revealed that the status of discontinuing TDF therapy, an older age, a male gender, and higher HBsAg levels at baseline were independent factors of retreatment in HBeAg-positive patients; an HBV genotype B, an older age, and higher HBsAg levels at baseline and end of treatment were independent factors of retreatment in HBeAg-negative patients. At the end of treatment, an HBsAg level of 150 IU/mL was an optimal value for predicting retreatment in HBeAg-negative patients.
A collaborative Greek and Taiwanese study demonstrated that the probability of retreatment was significantly associated with lower platelets at NA onset. An older age and the use of TDF as the last NA agent had increasing trends on future retreatment [19]. However, none of the above factors was significantly associated with retreatment in the multivariate regression [19].
Our study found that the predictors of clinical relapse were similar with retreatment. Recent studies have reported that the timing of HBV relapse after discontinuation of TDF was much earlier than that after discontinuation of ETV in CHB patients [7,11,12]. Until now, the mechanism of different relapse patterns between ETV and TDF discontinuation has remained unclear. HBV viral replication can resume from residual covalently closed circular DNA (cccDNA) after discontinuation of NA therapy. Therefore, we speculated that ETV might suppress cccDNA to a lower level and that more time might be required to resume HBV replication after the cessation of therapy. However, this hypothesis requires further study. In our investigation, HBeAg-positive patients who discontinued TDF therapy had significantly higher retreatment rates than HBeAg-positive patients who discontinued ETV therapy; the same situation did not persist for HBeAg-negative patients. This finding may be explained be noting that HBeAg-positive patients in the TDF group had significantly higher clinical relapse rates than HBeAg-positive patients in the ETV group.
Severe acute exacerbation with hepatic decompensation after stopping NA therapy and the effect of ALT elevation upon retreatment are important issues. Our study showed that 54.7%, 28.2%, and 5.1% of patients experienced the highest ALT levels of >200, >400, and >1000 U/L upon retreatment, respectively. A total of 18.8% and 3% patients who received retreatment fulfilled the definitions of hepatic decompensation according to Taiwan's National Health Plan and the 2015 APASL guidelines [3], respectively. Because ALT flare with hepatic decompensation is hard to avoid during off-NA-therapy follow-up, close monitoring after discontinuing NA therapy is necessary for timely retreatment to reduce the risk of hepatic decompensation and mortality. However, the clinical characteristics were similar between patients with and without hepatic decompensation upon retreatment according to criteria of Taiwan's National Health Plan.
A recent study showed that patients with clinical relapse but without retreatment had a higher incidence of HBsAg clearance than patients who received retreatment [13]. In our study, among the 64 patients with clinical relapse but without retreatment, 43 (67.2%) patients did not experience clinical relapse during the follow-up until their last visit after clinical relapse. Among these 43 patients, 17 achieved sustained viral remission (4 developed HBsAg seroclearance). These findings may support the notion that patients with transient ALT elevation and a subsequent decrease in viral load might have better immune control, which may increase HBsAg seroclearance rate. Therefore, close monitoring instead of immediate retreatment is suggested for patients who experience clinical relapse without hepatic decompensation. A recent study suggested that HBsAg levels decrease in patients with hepatitis flaring and that retreatment can be withheld or is not necessary. In contrast, retreatment should be considered for patients with elevated serum HBsAg and ALT levels [20]. Additional studies are necessary to differentiate the effects of ALT flaring on future outcomes to determine the best time for retreatment.
Our study has some limitations. First, this investigation was a retrospective and single-center study. Additional multicenter and prospective studies are necessary to confirm our results.
Second, our study included only Asian participants, among whom HBV genotypes B and C are predominant [15]. Furthermore, the criteria of retreatment were those of Taiwan's National Health Plan. Therefore, the incidence and predictors of retreatment might differ from other HBV genotypes and retreatment criteria. Third, we enrolled a small proportion of patients undergoing retreatment at their own expense, which might have increased the rate of retreatment. Forth, the diagnosis of non-cirrhosis for some patients was based on clinical criteria and ultrasound. Ultrasonography is more likely to result in an underestimation of early stages of cirrhosis.

Conclusions
The 5-year cumulative incidences of retreatment were 47% in HBeAg-positive patients and 54.8% in HBeAg-negative patients after discontinuation of either ETV or TDF therapy according to the criteria of retreatment of Taiwan's National Health Plan. HBeAg-positive patients who discontinued TDF therapy had significantly higher retreatment rates than HBeAg-positive patients who discontinued ETV therapy. Close monitoring for timely retreatment is necessary to reduce the risk of hepatic failure or mortality after discontinuation of NA therapy.
Supporting information S1 Data. Raw data. (XLSX) S1 Table. Baseline characteristics of study population by propensity score matching method. (DOCX)