Digestive and Liver Disease 54 (2022) 1044–1051 Contents lists available at ScienceDirect Digestive and Liver Disease journal homepage: www.elsevier.com/locate/dld Liver, Pancreas and Biliary Tract On-therapy HBsAg kinetics can predict HBsAg loss after nucleos(t)ide analogues interruption in HBeAg-negative patients. The cup is half full and half empty Teresa Broquetas a,b,c, Juan-Jose Hernandez d, Montserrat Garcia-Retortillo b,c, Lidia Canillas b,c, Marc Puigvehí b,c, Nuria Cañete b,c, Susanna Coll b,c, Ana Viu b,c, Esther Garrido b,c, Miquel Mico d, Xavier Bessa b,c, José A. Carrión a,b,∗ a Departament de Medicina, Universitat Autònoma de Barcelona, Barcelona, Spain b Liver Section, Gastroenterology Department, Hospital del Mar, 25-29 Passeig Marítim, Barcelona 08003, Spain c IMIM (Hospital del Mar Medical Research Institute), Barcelona, Spain d Laboratori de Referència de Catalunya, El Prat de Llobregat, Spain a r t i c l e i n f o a b s t r a c t Article history: Background: Nucleos(t)ide analogues withdrawal may improve HBsAg loss rates. However, conditions to Received 24 August 2021 select patients are not well established. Accepted 29 December 2021 Aims: to evaluate the impact of HBsAg kinetics before treatment interruption on post-treatment response. Available online 19 January 2022 Methods: Longitudinal, ambispective study in non-cirrhotic chronic hepatitis B HBeAg-negative patients, Keywords: analysing on-treatment and post-treatment HBsAg kinetics. On-treatment HBsAg kinetics diagnostic ac- Antiviral curacy (AUROC) to identify HBsAg loss was evaluated. HBsAg Results: 52 HBeAg-negative patients stopped treatment after 8.2 years, and 6 (11.5%) achieved HBsAg loss Hepatitis B one year after withdrawal. Multivariate analysis showed that on-treatment HBsAg kinetics was related to Kinetics HBsAg loss (OR=0.10; 95%CI=0.016–0.632; p = 0.014) with a high diagnostic accuracy (AUROC=0.935). A significant HBsAg decline ≥1 log10 IU/mL showed a positive and negative predictive value of 50% and of 97.6%, respectively. After treatment interruption, HBsAg decline speed (log10 IU/mL/year) accelerated in patients treated >6 years (from -0.06 to -0.20, p<0.05) and remained stable in treated <6 years (from -0.12 to -0.12 p=ns). Conclusions: On-treatment HBsAg kinetics can predict post-treatment HBsAg loss rate. Half of patients with a significant HBsAg decline can eliminate HBsAg the first year after withdrawal. Post-treatment HB- sAg decline is faster not only in patients who lost the HBsAg but also in those who remain HBsAg- positive. © 2022 The Authors. Published by Elsevier Ltd on behalf of Editrice Gastroenterologica Italiana S.r.l. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) 1. Introduction The hepatitis B virus (HBV) infection remains a global public health problem. In chronic hepatitis B (CHB), long-term adminis- HBsAg, hepatitis B surface antigen; HBeAg, hepatitis B e antigen; NA, nu- tration of nucleos(t)ide analogues (NA) with high barrier to resis- cleos(t)ides analogues; HBV, hepatitis B virus; OR, odds ratio; CI, confidence inter- val; CHB, chronic hepatitis B; ETV, entecavir; TDF, tenofovir disoproxil fumarate; tance, i.e., entecavir (ETV) or tenofovir disoproxil fumarate (TDF), EoT, end of treatment; TE, transient elastography; VR, virological relapse; CR, clini- is the treatment of choice [1]. The optimal therapeutic endpoint is cal relapse; ALT, alanine aminotransferase; ULN, upper limit of normality; IL, inter- the hepatitis B surface antigen (HBsAg) loss, which indicates sup- leukin; IQR, interquartile range; Peg-IFN, pegylated interferon; AUROC, area under pression of HBV replication and viral protein expression. However, receiver operating characteristic; S, sensitivity; Sp, specificity; PPV, positive predic- in CHB e antigen (HBeAg)-negative patients the decline of HBsAg tive value; NPV, negative predictive value; +LR, positive likelihood ratio; -LR, nega- tive likelihood ratio. during NA therapy is very slow (−0.1 log10 IU/mL/year) and HBsAg ∗ Corresponding author. loss very infrequent (0.6–4.6%) [2–5]. Therefore, European Associ- E-mail address: 95565@parcdesalutmar.cat (J.A. Carrión). ation for the study of the Liver (EASL) and American Association https://doi.org/10.1016/j.dld.2021.12.017 1590-8658/© 2022 The Authors. Published by Elsevier Ltd on behalf of Editrice Gastroenterologica Italiana S.r.l. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) T. Broquetas, J.-J. Hernandez, M. Garcia-Retortillo et al. Digestive and Liver Disease 54 (2022) 1044–1051 for the Study of Liver Diseases (AASLD) guidelines recommended the IMIM (Hospital del Mar Medical Research Institute) and were during years to maintain NA until HBsAg loss [6,7] and to monitor extracted in fasting conditions and centrifuged at 30 0 0 rpm before on-treatment HBsAg kinetics [2,5,8]. preservation at −30 °C [5]. The Asian-Pacific consensus suggested in 2008 to discon- The on-treatment HBsAg kinetics was evaluated at different tinue NA therapy in CHB HBeAg-negative patients with unde- time points, calculated as delta of HBsAg levels from NA initia- tectable HBV-DNA in three consecutive determinations separated tion to year 1 (Delta_1), year 3 (Delta_3) and to EoT (Delta_EoT) by 6 months [9]. Important studies, not only in Asian population and the off-treatment HBsAg kinetics as delta of HBsAg from EoT [10] but also in Caucasian patients [11,12] have shown that NA to one year after interruption. Virological relapse (VR) was defined withdrawal after years of viral suppression, can improve HBsAg as positive HBV-DNA at any time point. Significant virological re- loss rates. Thus, EASL guidelines accepted in 2017 that NA could be lapse (SVR) as HBV-DNA above 20 0 0 IU/mL. Clinical relapse (CR) discontinued in HBeAg-negative patients after 3 years of DNA sup- as an elevation of alanine aminotransferase (ALT) above 2 times pression [1]. Recent studies evaluating NA withdrawal have shown the upper limit of normality (ULN) and HBV-DNA >20 0 0IU/mL at that low HBsAg levels, at baseline and at the end of treatment any time point [18]. Sustained off-treatment response was defined (EoT), are related to HBsAg loss [10,13,14]. However, the optimal as persistent ALT<2xULN and HBV-DNA<20 0 0IU/mL and patients duration of therapy before discontinuation remains unclear and in “grey-zone” as ALT>2xULN or DNA>20 0 0IU/mL at week 48 after there are not well-established conditions to select these patients withdrawal. Retreatment with NA was indicated if patient fulfilled [15]. any of the following criteria: severe flare (ALT>10xULN in two con- The hypothesis of our study was that HBsAg kinetics dur- secutive blood test for 2 weeks), moderate flare (ALT 5–10xULN in ing NA therapy could affect the HBsAg kinetics after NA with- two consecutive blood test for 4 weeks) or mild persistent flare drawal. Thus, the primary aim was to evaluate HBsAg decline be- (ALT 2–5xULN and DNA>2.0 0 0 IU/ml persisting for more than 6 fore and after treatment withdrawal in non-cirrhotic CHB HBeAg- months). negative patients. Secondary aim was to evaluate the influence of on-treatment HBsAg kinetics on post-treatment responses. 2.3. Statistical analysis 2. Materials and methods Quantitative variables were expressed as medians and in- terquartile ranges (IQR, Q1-Q3). Categorical variables were ex- 2.1. Patients and study design pressed as proportions. Continuous variables were compared by the Mann–Whitney U test, Wilcoxon or Kruskall-Wallis when ap- This is a single centre, longitudinal, ambispective study propriate and categorical by the Pearson chi-square test or Fisher analysing HBsAg levels in non-cirrhotic CHB HBeAg-negative pa- test. Differences between patients who achieved HBsAg loss and tients during NA therapy and after withdrawal. Patients were eligi- those who did not, were analysed by univariate analysis. Variables ble if they had received a stable NA dose during a minimum of 3 showing a P value<0.05 were included in a multivariate forward years and achieved virological response (HBV-DNA below the limit stepwise logistic regression analysis to determine independent pre- of quantification <13 IU/mL). dictors of HBsAg loss and expressed as odds ratio (OR) and 95% Recruitment period was from December 2017 to October 2019. confidence interval (95%CI). The diagnostic accuracy of HBsAg de- Exclusion criteria were: CHB HBeAg-positive patients, human im- cline to identify patients at risk of losing HBsAg was assessed us- munodeficiency virus or hepatitis D virus coinfection, immuno- ing the area under the receiver operator characteristic (AUROC) suppressive therapy, history of hepatocellular carcinoma, transient curve (95%CI). The optimal HBsAg decline cut-off value to iden- elastography (TE) >9.4 kPa [16], absence of HBsAg determina- tify HBsAg loss was selected on the basis of sensitivity (S), speci- tion before NA treatment, or inability to perform a close follow- ficity (Sp), positive predictive value (PPV), negative predictive value up. HBsAg levels were determined before NA treatment, at year (NPV), positive likelihood ratio (+LR) and negative likelihood ra- 1 and 3 after initiation and 1 year before withdrawal. Proto- tio (-LR). Bootstrapping was used to perform an internal validation col visits were at EoT, and at weeks 4, 12, 24 and 48 after of the HBsAg kinetic diagnostic accuracy by generating 10 0 0 re- interruption. sampling sets with random replacement. The results of the inter- The study protocol was approved by the Ethical Committee of nal bootstrap validation gave estimation for the AUROC with the our Institution “Comitè Ètic d’Investigació Clínica - Parc de Salut median (Percentile 5-Percentile 95). The correlation between treat- Mar”, study reference 2018/7939/I, in accordance with the ethical ment duration (years) and HBsAg decline (log 10 IU/mL) was eval- guidelines of the 1975 Declaration of Helsinki. uated by Pearson’s coefficient (r). The cumulative HBsAg loss rate was evaluated by the Kaplan-Meier method (Breslow and Log-rank 2.2. Variables and clinical definitions tests). All statistical tests were two-sided and a P value<0.05 was The HBV-genotype was collected from electronic data. Demo- considered significant. Analyses were performed with the SPSS® graphic data and TE were assessed at EoT. After NA cessation, liver 25.0 (SPSS Inc., Chicago, IL, USA) and the AUROC of HBsAg decline function, HBV-DNA, HBsAg levels, HBeAg, anti-HBe and anti-HBs and Bootstrapping were calculated with MedCalc® v19.1.3 (Med- were assessed at every protocol visit. Calc Software). HBV-DNA was measured by polymerase chain reaction with a limit of quantification of 13 IU/mL (Versant HBV DNA 1.0®, 3. Results Siemens Medical Solutions Diagnostics, New York, USA). Serum HB- sAg quantification was introduced in our laboratory in July 2014 3.1. Study population and baseline characteristics and was evaluated by Electro-chemiluminescence immunoassay Elecsys® HBsAgII (Roche Diagnostic, Rotkreuz, Switzerland). The From January 1999 to December 2017, 148 CHB HBeAg-negative assay ranged from 0.05 to 52,0 0 0 IU/mL. In highly concentrated patiens started NA treatment in our hospital. Twenty-seven (18.2%) samples above the upper limit, the value of manual dilution was were lost during follow-up and 9 (6.2%) lost the HBsAg during multiplied by the dilution factor [17]. In patients who started NA therapy (6 under NA treatment [5] and 3 under pegylated inter- treatment before July 2014, the HBsAg was analysed in cryopre- feron [peg-IFN] add-on strategy [17]). Therefore, 112 patients were served serum samples part of the private collection (C.0 0 0 0956) of evaluated, and sixty were excluded: 20 (17.8%) with cirrhosis, 6 1045 T. Broquetas, J.-J. Hernandez, M. Garcia-Retortillo et al. Digestive and Liver Disease 54 (2022) 1044–1051 Table 1 HBsAg kinetics and NA treatment duration. N = 52 NA treatment duration NA treatment duration NA treatment duration P value 3–6 years (n = 11, 21.2%) 6–9 years (n = 22, >9 years (n = 19, 42.3%) 36.5%) Before antiviral treatment Males, n (%) 39 (75) 7 (63.6) 17 (77.3) 15 (78.9) ns Caucasian, n (%) 30 (57.7) 3 (27.3) 10 (45.5) 17 (89.5) <0.001 HBV Genotype, n (%) ns A 13 (25) 3 (27.3) 6 (27.3) 4 (21.1) B 2 (3.8) 0 (0) 1 (4.5) 1 (5.3) C 2 (3.8) 0 (0) 1 (4.5) 1 (5.3) D 32 (61.5) 8 (72.7) 12 (54.5) 12 (63.2) E 3 (5.8) 0 (0) 2 (9.1) 1 (5.3) HBsAg (IU/mL) 3821 (1587–7144) 5493 (1698–7005) 3206 (1558–8092) 4084 (1700–7007) ns HBsAg (IU/mL), n (%) ns <1000 10 (19.2) 2 (18.2) 4 (18.2) 4 (21.1) 1000–10,000 34 (65) 8 (72.7) 15 (68.2) 11 (57.9) >10,000 8 (15.4) 1 (9.1) 3 (13.6) 4 (21.1) During NA treatment NA treatment, n (%) ns Tenofovir 34 (65) 8 (72.7) 15 (68.2) 11 (57.9) Entecavir 17 (33) 3 (27.3) 7 (31.8) 7 (36.8) Lamivudine 1 (2) 0 (0) 0 (0) 1 (5.3) Treatment duration (years) 8.17 (6.5–10.3) 3.98 (3.5–5.1) 7.95 (6.8–8.4) 11.25 (10.2–13.6) <0.001 Delta_1 HBsAg (log10 IU/mL) −0.01 (0.03-(−0.09)) −0.03 (−0.19-(−0.02)) 0.01 (−0.05–0.07) −0.02 (−0.17–0.01) ns Delta_3 HBsAg (log10 IU/mL) −0.12 (−0.09(−0.24)) −0.15 (−0.44-(−0.07)) −0.07 (−0.17–0.03) −0.15 (−0.28-(−0.04)) ns Delta_1pre-EoT HBsAg (log10/mL) −0.09 (−0.15-(−0.01)) −0.12 (−0.17-(−0.01)) −0.09 (−0.14-(−0.07)) −0.07 (−0.16-(−0.02)) ns Delta of HBsAg per year (log10IU/mL/year) −0.06 (−0.11-(−0.03)) −0.06 (−0.14- (−0.02)) −0.06 (−0.11-(−0.02)) −0.06 (−0.10 -(−0.04)) ns End of treatment (EoT) Age (years) 52 (43–59) 44 (42–53) 53 (43–61) 55 (50–59) ns HBsAg (IU/mL) 817 (197–2486) 1350 (332–2639) 817 (221–3721) 558 (57–1628) ns HBsAg (IU/mL) 0.053 <100 8 (15.4) 0 (0) 2 (9.1) 6 (31.6) 100–1000 19 (36.5) 4 (36.4) 10 (45.5) 5 (26.3) >1000 25 (48.1) 7 (63.6) 10 (45.5) 8 (42.1) Delta_EoT HBsAg (log 10 IU/mL) −0.51 (−0.93–0.21) −0.29 (−0.88-(−0.07)) −0.45 (−0.74-(−0.15)) −0.82 (−1.3-(−0.51)) 0.017 Delta_EoT HBsAg, n (%) 0.105 <−1 log10 (IU/mL) 42 (80.8) 10 (90.9) 19 (86.4) 13 (68.4) ≥−1 log10 (IU/mL) 10 (19.2) 1 (9.1) 3 (13.6) 6 (31.6) 48 Weeks after NA interruption HBsAg (IU/mL) 364 (30–1973) 1022 (339–2394) 590 (104–2584) 130 (0–602) 0.097 HBsAg (IU/mL), n (%) 0.017 <100 13 (26.5) 1 (10) 4 (19) 8 (44.4) 100–1000 18 (36.7) 3 (30) 9 (42.9) 6 (33.3) >1000 18 (36.7) 6 (60) 8 (38.1) 4 (22.2) Delta_1post-EoT HBsAg (log10 UI/mL) −0.19 (−0.57-(−0.08)) −0.12 (−0.18-(−0.02)) −0.19 (−0.41-(−0.09)) −0.41 (−1.30-(−0.06)) ns NA reintroduction, n (%) 3 (5.8) 1 (9.1) 1 (4.5) 1 (5.3) ns Sustained off-treatment response, n (%) 21 (40.4) 4 (36.4) 13 (49.1) 4 (21.1) ns HBsAg loss, n (%) 6 (11.5) 0 (0) 0 (0) 6 (31.6) 0.003 Quantitative variables are expressed as median (IQR). Qualitative variables are expressed as n (%). HBsAg: hepatitis B surface antigen; HBV: hepatitis B virus; NA: nu- cleos(t)id analogue; EoT: end of treatment; Sustained off-treatment response (ALT<2xULN and HBV-DNA<20 0 0IU/mL). (4.1%) without baseline serum sample, and 34 (30.4%) declined −0.82 in those treated >9 years (n = 19)(p = 0.017) (Table 1 and their participation. Fifty-two patients were included, 75% were Fig. 1). However, the speed of HBsAg decline was the same in males, the median (IQR) age was 52 (43–59), 65% received TDF, the three treatment periods −0.06 vs. −0.06 and −0.06 log10 and treatment duration was 8.17 (6.51–10.29) years. HBV-genotype IU/mL/year; p=ns). was evaluated in all patients and 32 (61.5%) were infected by geno- The decline of HBsAg one year after NA interruption was −0.19 type D. Main characteristics of included patients are depicted in log10 IU/mL. Therefore, the speed of HBsAg decline after stopping Table 1. treatment was faster than during therapy (−0.19 vs. −0.06 log10 IU/mL/year; p<0.001). The HBsAg decline after NA withdrawal 3.2. HBsAg kinetics during therapy and after withdrawal compared to one year before, was similar in patients treated from 3 to 6 years (−0.12 vs. −0.12; p=ns) but accelerated in those The HBsAg level (IU/mL) was 3821 (1587–7144) before antivi- treated from 6 to 9 (−0.19 vs. −0.09; p = 0.015), or >9 years ral treatment, 817 (197–2486) at EoT, and 364 (30–1973) 48 weeks (−0.41 vs. −0.07; p = 0.01)(Table 1 and Fig. 1). After NA interrup- after withdrawal. The HBsAg decline during NA therapy (log10 tion, the speed of HBsAg decline was faster not only in patients IU/mL) was −0.01 (0.03-(−0.09)) at year 1, −0.12 (−0.09(−0.24)) with HBsAg loss (−1.33 vs. −0.14 log10 IU/mL/year; p = 0.046) at year 3 and −0.51 (−0.93-(−0.21)) at EoT. The speed of HBsAg but also in those with persistence of HBsAg (−0.18 vs. −0.05 log10 decline during NA therapy (log10 IU/mL/year) was −0.06 (−0.11- IU/mL/year; p<0.001). (−0.03)). We observed a correlation between treatment duration (years) and HBsAg decline during treatment (log 10 IU/mL)(r= 3.3. Predictors of HBsAg loss −0.51; p<0.001). Therefore, the Delta_EoT (log 10 IU/mL) was higher as longer the treatment was: −0.29 in patients treated from Differences between patients who achieved HBsAg loss (n = 6) 3 to 6 years (n = 11), −0.45 in treated from 6 to 9 (n = 22), and and those who did not (n = 46) were evaluated by univariate 1046 T. Broquetas, J.-J. Hernandez, M. Garcia-Retortillo et al. Digestive and Liver Disease 54 (2022) 1044–1051 Fig. 1. HBsAg kinetics and treatment duration. The kinetics of HBsAg (Delta_EoT HBsAg)(log 10 IU/mL) are depicted according to treatment duration in Fig. 1a (from 3 to 6 years; n = 11), Fig. 1b (from 6 to 9; n = 22) and Fig. 1c (longer than 9 years; n = 19). Each line represents one patient, and distribution of HBsAg values in each time point is represented by Box and whisker plots. The decline of HBsAg was higher as longer the treatment was, but the speed of HBsAg decline was the same independently of treatment duration. After stopping treatment, the speed of HBsAg decline was faster than during therapy, especially in those receiving NA longer than 6 years. analysis (Table 2). HBsAg level before NA therapy was similar be- not (12.8 vs. 7.9 years; p = 0.001). All patients with HBsAg loss tween groups (3062 and 3822 IU/mL; p=ns) as the Delta HBsAg (100%) were treated longer than 9 years, compared with 28.3% of at first and third year of treatment (−0.01 vs. −0.01 and −0.12 vs. those without HBsAg loss (p = 0.004). Moreover, the speed of HB- −0.19 log 10 IU/mL, respectively, p=ns). However, treatment du- sAg decline was faster in patients with HBsAg loss compared with ration was longer in patients with HBsAg loss compared who did those who did not (0.14 vs. 0.05 log10 IU/mL/year; p = 0.006). As 1047 T. Broquetas, J.-J. Hernandez, M. Garcia-Retortillo et al. Digestive and Liver Disease 54 (2022) 1044–1051 Table 2 HBsAg kinetics and variables associated with HBsAg loss. HBsAg+ (n = 46, N = 52 88.5%) HBsAg- (n = 6, 11.5%) P value OR (95% CI), p value Before antiviral treatment Males, n (%) 39 (75) 34 (73,9) 5 (83.3) ns Caucasian ethnicity, n (%) 30 (57.7) 24 (52.2) 6 (100) 0.026 HBV Genotype D, n (%) 32 (61.5) 28 (60.9) 4 (66) ns HBsAg (IU/mL) 3821 (1587–7144) 3822 (1578–7159) 3062 (1822–4495) ns HBsAg (IU/mL) ns <1000 10 (19.2) 9 (19.6) 1 (16.7) 1000–10,000 34 (65) 30 (65.2) 4 (66.7) >10,000 8 (15.4) 7 (15.2) 1 (16.7) During NA treatment Antiviral treatment, n (%) ns Tenofovir 34 (65) 30 (65.2) 4 (66.7) Entecavir 17 (33) 15 (32.6) 2 (33.3) Lamivudine 1 (2) 1 (2.2) 0 (0) Treatment duration (years) 8.17 (6.5–10.3) 7.9 (5.9–10.1) 12.8 (10.3–16.2) 0.001 Treatment duration, n (%) 0.004 3–6 years 11 (21.1) 11 (23.9) 0 (0) 6–9 years 22 (42.3) 22 (47.8) 0 (0) >9 years 19 (36.6) 13 (28.3) 6 (100) Add-on Peg-IFN, n (%) 19 (36.6) 18 (39) 1 (16.7) ns Delta_1 HBsAg (log 10 IU/mL) −0.01 (0.03-(−0.09)) −0.01 (−0.08–0.03) −0.01 (−0.19–0.03) ns Delta_3 HBsAg (log 10 IU/mL) −0.12 (−0.09(−0.24)) −0.12 (−0.20- (−0.01)) −0.19 (−0.33-(−0.04)) ns Delta_1pre-EoT HBsAg (log10 IU/mL) −0.09 (−0.15-(−0.01)) −0.09 (−0.14-(−0.02)) −0.16 (−0.29-(−0.01)) ns Delta_EoT HBsAg/year −0.06 (−0.11-(−0.03)) −0.05 (−0.10-(−0.02)) −0.14 (−0.16-(−0.11)) 0.006 (log10IU/mL/year) End of treatment (EoT) Age (years) 52 (43–59) 51 (43–60) 54 (53–56) ns HBsAg_EoT (IU/mL) 817 (197–2486) 1176 (258–2957) 21.4 (14.3–401) 0.002 HBsAg (IU/mL) 0.001 <100 8 (15.4) 4 (8.7) 4 (66.7) 100–1000 19 (36.5) 17 (37.0) 2 (33.3) >1000 25 (48.1) 25 (54.3) 0 (0) Delta_EoT HBsAg (log 10 IU/mL) −0.51 (−0.93–0.21) −0.47 (−0.87- (−0.15)) −1.75 (−2.11- (−1.45)) <0.001∗ 0.10 (0–016–0.632), p = 0.014 Delta_EoT HBsAg, n (%) 0.001 <−1 log10 (IU/mL) 42 (80.8) 41 (89.1) 1 (16.7) ≥−1 log10 (IU/mL) 10 (19.2) 5 (10.9) 5 (83.3) 48 Weeks after NA interruption HBsAg (IU/mL) 364 (30–1973) 590 (135–2489) <0.05 (BLD) 0.001 Delta_1post-EoT HBsAg/year −0.19 (−0.57-(−0.08)) −0.18 (−0.40-(−0.07)) −1.33 (−2.60-(−1.16)) 0.02 (log10IU/mL/year) Quantitative variables are expressed as median (IQR). Qualitative variables are expressed as n (%). HBsAg: hepatitis B surface antigen; HBV: hepatitis B virus; NA: nu- cleos(t)id analogue; Peg-IFN: pegylated interferon; EoT: end of treatment; BLD: below limit of detection. consequence, patients with HBsAg loss showed greater Delta_EoT tients at risk of losing HBsAg are depicted in the Supplementary HBsAg (−1.75 vs. −0.47 log10 IU/mL; p <0.001) and lower HBsAg Tables 1 and 2, respectively. levels before interruption (21.4 vs. 1176 IU/mL; p = 0.002). Multivariate analysis including HBsAg_EoT, treatment duration 3.5. Probability of HBsAg loss according to HBsAg kinetics during and Delta_EoT, showed that only the Delta_EoT was associated with antiviral treatment HBsAg loss after NA interruption (OR=0.10; 95%CI=0.016–0.632; p = 0.014). A Delta_EoT ≥ −1 log10 IU/mL was observed in 10 (19.2%) pa- tients (Supplementary Table 2). Patients with a Delta_EoT ≥ −1 3.4. Diagnostic accuracy of HBsAg decline to predict HBsAg loss log10 IU/mL were usually Caucasian (90% vs. 50%, p = 0.02) but HBsAg levels before NA were similar between groups (3330 vs. The diagnostic accuracy of HBsAg kinetics during antiviral treat- 3891 IU/mL; p=ns). Patients with a Delta_EoT ≥ −1 log10 IU/mL ment was excellent to identify patients at risk of losing HBsAg after had a longer treatment duration (10.2 vs. 7.9 years; p = 0.02) and treatment withdrawal. The AUROC (95% CI) of HBsAg decline was faster HBsAg decline (−0.15 vs. −0.05 IU/mL/year; p = 0.001). Thus, 0.935 (0.83–0.98) (Fig. 2). The bootstrap method showed a median HBsAg at EoT was lower in patients with a Delta_EoT ≥ −1 log10 AUROC (Percentile 5-Percentile 95) for the HBsAg kinetics of 0.75 - IU/mL (114 vs. 1277 IU/mL; p = 0.01). Therefore, 5 (50%) out of 10 0.99 to identify HBsAg loss. patients with a Delta_EoT ≥ −1 log10 IU/mL showed HBsAg <100 The optimal Delta_EoT cut-off to identify patients at risk of IU/mL before interruption and 5 achieved HBsAg loss during the losing HBsAg was > −1.4 log10 IU/mL (S = 83.3%, Sp= 95.7%, first year after withdrawal (p<0.001 in both cases). PPV=71.4% and NPV=97.8%). Other cut-offs were evaluated for an The cumulative rate of HBsAg loss at weeks 4, 12, 24, 36 and easier applicability in real clinical practice (Fig. 2). The Delta_EoT 48 after withdrawal was 1.9%, 1.9%, 5.8%, 9.6% and 11.5%. The prob- cut-off ≥ −1 log10 IU/mL showed good accuracy to identify HBsAg ability of HBsAg loss one year after NA interruption was 50% in pa- loss (S = 83.3%, Sp = 89.1%, PPV = 50% and NPV = 97.6%). Char- tients with a Delta_EoT ≥ −1 log10 IU/mL and 2.4% in those with acteristics of the included patients according to the optimal (−1.4 Delta_EoT < −1log10 IU/mL (log-rank p<0.001; Breslow p<0.001) log10 IU/mL) and useful (−1 log10 IU/mL) cut-offs to identify pa- (Fig. 3). Moreover, patients with a Delta_EoT ≥ −1log10 IU/mL 1048 T. Broquetas, J.-J. Hernandez, M. Garcia-Retortillo et al. Digestive and Liver Disease 54 (2022) 1044–1051 Fig. 2. Diagnostic accuracy of HBsAg decline to identify patients at risk of losing HBsAg. Diagnostic accuracy (AUROC, 95% CI) of HBsAg decline to identify patients at risk of losing HBsAg after treatment interruption. Evaluation of different cut-offs according to sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio, and negative likelihood ratio. AUROC: area under receiver operating characteristic; CI: confidence interval; S: sensitivity; Sp: specificity; PPV: positive predictive value; NPV: negative predictive value; +LR: positive likelihood ratio; -LR: negative likelihood ratio. Fig. 3. Cumulative probability of HBsAg loss after NA interruption according to HBsAg kinetics during NA therapy. Patients with a Delta_EoT HBsAg ≥ −1log10 IU/mL showed an HBsAg loss cumulative probability of 50% one year after NA interruption compared to 2.5% in those with a Delta_EoT HBsAg < −1log10 IU/mL (log-rank= p<0.001; Breslow p=<0.001). that one year after treatment interruption persisted with HBsAg- was identified in 51 (98.1%), SVR in 24 (46.2%) and CR in 5 (9.6%) positive showed lower HBsAg levels compared to those with a patients. Severe flare leading to NA reintroduction was observed Delta_EoT < −1log10 IU/mL (103 vs. 766 IU/mL; p = 0.002). in 3 (5.8%) but neither acute liver failure nor hepatic decompen- sation occurred. Patients with NA reintroduction were Caucasian 3.6. Response after NA withdrawal (n = 1), Asian (n = 1) and Hispanic (n = 1), males (n = 3), 2 in- fected by genotype A and 1 genotype D, all of them treated with The HBsAg loss was observed in 6 (11.5%) patients and sero- TDF during 4.7, 6 and 15 years. The three patients showed an HB- conversion (positive anti-HBs) in 4 (66.7%) of them. There were sAg decline < 1 log10 IU/mL and HBsAg levels before interruption no differences in baseline characteristics or in on-treatment HB- of 1781, 1632 and 807 IU/mL, respectively. Patients were retreated sAg kinetics between patients with and without seroconversion. with the same NA and rapidly achieved virological response. Dif- However, the 4 patients who developed positive anti-HBs had been ferences on response rates according the HBsAg kinetics during NA treated with TDF and the two others with ETV. Virological relapse therapy are depicted in Supplementary Table 2 and Supplementary 1049 T. Broquetas, J.-J. Hernandez, M. Garcia-Retortillo et al. Digestive and Liver Disease 54 (2022) 1044–1051 Figure 1. Considering the HBsAg-positive patients without NA re- the HBsAg during the first year after interruption. A recent meta- treatment at week 48 (n = 43), 21 (48.8%) remained in sustained analysis has shown that antiviral therapy duration can be crucial off-treatment response and 22 (51.2%) in grey zone. Among pa- to achieve a persistent viral remission after treatment interruption tients in grey zone, 20 (91%) showed a SVR and only 2 (9%) a in HBeAg-negative patients [20]. Therefore, we consider that treat- CR. Four out of 5 (80%) patients with an HBsAg decline ≥ 1 log10 ment longer than 3 years could be beneficial before interruption IU/mL that one year after NA withdrawal persisted with HBsAg- [1,9,15]. positive remained in sustained off-treatment response (persistent Patients with HBsAg loss, showed a greater on-treatment HB- ALT<2xULN and HBV-DNA<20 0 0IU/mL), compared to 48% (18 out sAg decline and lower HBsAg values before interruption. The on- of 38) of patients with HBsAg decline < 1 log10 IU/mL (p = 0.17). treatment HBsAg decline was independently associated with HB- sAg loss and was a good predictor of HBsAg loss (AUROC=0.935). 3.7. HBsAg kinetics according to HBV genotype and NA therapy The optimal HBsAg decline cut-off was > 1.4 log10 IU/mL. How- ever, to make the use of HBsAg kinetics easier in real clinical prac- The HBsAg kinetics was evaluated before and after treatment tice, we also evaluated the cut-off ≥1 log10 IU/mL that showed withdrawal according to HBV genotype D vs. other genotypes (Sup- good PPV (50%) and excellent NPV (97.6%). Therefore, half of pa- plementary Table 3). Patients infected by genotype D (n = 32) tients with an HBsAg decline ≥1log10 IU/mL achieved the HBsAg showed lower HBsAg levels (IU/mL) before treatment initiation loss during the first year after NA withdrawal. Moreover 40% of compared to those infected by other genotypes (2497 vs 5708; them remained in sustained off-treatment response with low HB- p = 0.03). However, there were no differences in the on-treatment sAg levels (103 IU/mL) and no patients needed to be retreated. On HBsAg kinetics at year 1, at year 3, at EoT or in the speed of HBsAg the other hand, despite only 2.4% of patients with an HBsAg de- decline. After treatment interruption, there were no differences be- cline < 1 log10 IU/mL achieved the HBsAg loss, the 40% of them tween patients according to HBV genotype. persisted in sustained off-treatment response. Regarding the type of NA therapy, we did not find differences t is important to note that after NA interruption the speed between patients treated with TDF (n = 34) or ETV (n = 17) of HBsAg decline accelerated in patients treated longer than 6 on treatment duration (8.0 vs. 8.5 years; p=ns), levels of HB- years. It has been recently postulated that long-term HBV-DNA sAg_EoT (697 vs. 1143 IU/mL; p = 0.31), Delta-EoT (−0.62 vs. suppression can reinvigorate exhausted CD8+ T cells and restore −0.48 log10 UI/mL; p=ns), Delta_1post-EoT HBsAg (−0.18 vs. −0.19 the immune control against infected hepatocytes after withdrawal log10 UI/mL; p=ns), rates of HBsAg loss (11.8 vs. 11.8%; p=ns), VR [14,22,23]. In consonance, our study has clearly shown that the (97.1 vs. 100%; p=ns), SVR (65 vs. 35.3%; p=ns), CR (11.8 vs. 5.9%; speed of HBsAg decline after treatment interruption is 3 times p=ns), or retreatment (8.8 vs. 0%; p=ns). However, patients treated faster than during therapy, not only in patients who lost the HB- with TDF compared to those treated with ETV showed earlier VR sAg but also in HBsAg-positive patients, being another argument in (4 vs. 12 weeks; p<0.001), SVR (12 vs. 30 weeks; p<0.001) and favour to stop treatment in these patients. CR (10 vs. 48 weeks; p = 0.14). In patients who had received Peg- In terms of safety, only three patients developed a severe flare IFN (n = 19) [17] the add-on therapy was finished a median of during the first 12 weeks after withdrawal and were retreated with 2.3 years before EoT and no differences in HBsAg loss rate were the same NA showing an excellent response. In contrast, more found (5.3 vs. 15.2%; p=ns). However, patients with add-on Peg- than 90% of HBsAg-positive patients remained without antiviral IFN showed faster (−0.10 vs. −0.05 log10 IU/mL/year; p = 0.02) and treatment. Patients receiving ETV showed later virological relapse. greater HBsAg decline (−0.74 vs. −0.46 log10IU/mL; p = 0.056). Similarly, recent Asian studies [24,25] have reported that patients treated with ETV can develop the clinical or virological relapse 4. Discussion later than those treated with TDF [26]. Thus, we consider that NA type should be considered for the monitoring after interruption. Our longitudinal study in non-cirrhotic HBeAg-negative patients Another interesting point of our cohort was that 37% added-on with NA treatment withdrawal has shown that on-treatment HB- Peg-IFN two years before EoT [17]. Patients with add-on Peg-IFN sAg kinetics can predict theHBsAg decline after treatment interrup- showed faster HBsAg decline despite no differences in HBsAg loss tion and the frequency of HBsAg loss . rate were found. Hence, an add-on strategy with Peg-IFN could be Our study has shown a high incidence of HBsAg loss during useful to accelerate HBsAg kinetics and to short NA duration before the first year after treatment interruption (11.5%). Studies on Asian interruption [17,27]. population have shown lower HBsAg loss rates (1.8%) [10,19] com- Our study has some limitations. The limited number of included pared to European cohorts (from 9% to 22%) [11–14]. A recent sys- patients compared to Asian cohorts [10,19], a short time of follow- tematic review of 25 studies showed a 2% HBsAg loss rate. How- up, and the lack of an external validation. Nevertheless, it is im- ever, only two studies included Caucasian patients, and most were portant to note that these limitations have been compensated per- Asian patients infected by HBV genotypes B or C [20]. In contrast, forming an internal validation of the HBsAg kinetics diagnostic ac- Kuhnhenn et al. have recently described low levels of HBsAg in curacy and evaluating all included patients without any loss during HBeAg-negative patients infected by genotype B or D [21]. In our follow-up. Moreover, the number of our included patients are sim- cohort, the 61.5% of patients were infected by genotype D. We have ilar to the previous European studies [11–14]. On the other hand, confirmed that patients with genotype D had lower levels of HBsAg our study has several strengths. All included patients were HBeAg- at the initiation of NA treatment. However, we could not demon- negative at the NA initiation, who are the patients with less ev- strate differences between patients on treatment HBsAg kinetics or idence on NA interruption. The long NA therapy has allowed to in the rate of HBsAg loss according to HBV genotypes. identify a significant correlation between treatment duration and Our study has shown a significant correlation between treat- HBsAg decline and to analyse the HBsAg kinetics before and af- ment duration and HBsAg decline. Therefore, the HBsAg decline ter NA interruption in three different treatment time periodstime. was higher as longer the treatment was. The annual decline Moreover, the evaluation of HBsAg kinetics before and after ther- of HBsAg during NA treatment was very stable and slow (0.06 apy has demonstrated an accelerated effect after withdrawal. log10IU/mL/year) as we have previously reported [5]. The HBsAg In conclusion, on-treatment HBsAg kinetics can predict the decline during first 3 years of treatment was very low and HB- post-treatment HBsAg decline and the frequency of HBsAg with sAg kinetics was not associated with HBsAg loss. However, patients high accuracy. Half of patients with a significant HBsAg decline with longer treatment duration had more probabilities of clearing (≥ 1 log10 IU/mL) can eliminate HBsAg during the first year 1050 T. Broquetas, J.-J. Hernandez, M. Garcia-Retortillo et al. Digestive and Liver Disease 54 (2022) 1044–1051 after withdrawal compared to only few patient who did not show [4] Marcellin P, Wong DK, Sievert W, et al. Ten-year efficacy and safety of teno- this kinetics. Importantly, after NA interruption HBsAg decline is fovir disoproxil fumarate treatment for chronic hepatitis B virus infection. Liver Int 2019;39:1868–75. faster not only in patients who lost the HBsAg but also in those [5] Broquetas T, Garcia-Retortillo M, Hernandez JJ, et al. Quantification of HBsAg who remain HBsAg-positive. to predict low levels and seroclearance in HBeAg-negative patients receiving nucleos(t)ide analogues. PloS one 2017;12:e0188303. [6] Lok AS, McMahon BJ. Chronic hepatitis B: update 2009. Hepatology Conflict of interest 2009;50:661–2. [7] European association for the study of the L. EASL clinical practice guidelines: None declared. management of chronic hepatitis B virus infection. J Hepatol 2012;57:167–85. [8] Zoutendijk R, Hansen BE, van Vuuren AJ, et al. Serum HBsAg decline during long-term potent nucleos(t)ide analogue therapy for chronic hepatitis B and Financial support statement prediction of HBsAg loss. J Infect Dis 2011;204:415–18. [9] Liaw YF, Leung N, Kao JH, et al. Asian-Pacific consensus statement on the man- agement of chronic hepatitis B: a 2008 update. Hepatol Int 2008;2:263–83. JAC has received a grant from Instituto de Salud Carlos III, Min- [10] Jeng WJ, Chen YC, Chien RN, et al. Incidence and predictors of hepatitis B sur- isterio de Economía y Competitividad (PI14/00540), co-funded by face antigen seroclearance after cessation of nucleos(t)ide analogue therapy in Fondo Europeo de Desarrollo Regional (FEDER), Unión Europea, hepatitis B e antigen-negative chronic hepatitis B. Hepatology 2018;68:425–34. Una manera de hacer Europa. The funders had no role in study [11] Hadziyannis SJ, Sevastianos V, Rapti I, et al. Sustained responses and loss of HBsAg in HBeAg-negative patients with chronic hepatitis B who stop design, data collection and analysis, decision to publish, or prepa- long-term treatment with adefovir. Gastroenterology 2012;143:629–36 e1. ration of the manuscript. There was no additional external funding [12] Berg T, Simon KG, Mauss S, et al. Long-term response after stopping tenofovir received for this study. disoproxil fumarate in non-cirrhotic HBeAg-negative patients - FINITE study. J Hepatol 2017;67:918–24. [13] Papatheodoridis GV, Rigopoulou EI, Papatheodoridi M, et al. DARING-B: dis- Ethics approval statement continuation of effective entecavir or tenofovir disoproxil fumarate long-term therapy before HBsAg loss in non-cirrhotic HBeAg-negative chronic hepatitis B. Antivir Ther 2018;23:677–85. The study protocol was approved by the Ethical Committee of [14] Garcia-Lopez M, Lens S, Pallett LJ, et al. Viral and immune factors associated our Institution “Comitè Ètic d’Investigació Clínica - Parc de Salut with successful treatment withdrawal in HBeAg-negative chronic hepatitis B Mar”, study reference 2018/7939/I. patients. J Hepatol 2020. [15] Mak LY, Seto WK, Fung J, et al. Use of HBsAg quantification in the natural history and treatment of chronic hepatitis B. Hepatol Int 2020;14:35–46. Patients consent statement [16] Vigano M, Paggi S, Lampertico P, et al. Dual cut-off transient elastography to assess liver fibrosis in chronic hepatitis B: a cohort study with internal valida- tion. Aliment Pharmacol Ther 2011;34:353–62. All patients provided a written informed consent. [17] Broquetas T, Garcia-Retortillo M, Mico M, et al. Hepatitis B surface antigen and hepatitis B core-related antigen kinetics after adding pegylated-interferon Guarantor of the article and final version of the manuscript to nucleos(t)ids analogues in hepatitis B e antigen-negative patients. World J Hepatol 2020;12:1076–88. [18] Liaw YF, Kao JH, Piratvisuth T, et al. Asian-Pacific consensus statement on the Jose A. Carrión Rodríguez certifies that all the authors have ap- management of chronic hepatitis B: a 2012 update. Hepatol Int 2012;6:531–61. proved the final version of the manuscript. [19] Liem KS, Fung S, Wong DK, et al. Limited sustained response after stopping nucleos(t)ide analogues in patients with chronic hepatitis B: results from a randomised controlled trial (Toronto STOP study). Gut 2019;68:2206–13. Acknowledgements [20] Papatheodoridis G, Vlachogiannakos I, Cholongitas E, et al. Discontinuation of oral antivirals in chronic hepatitis B: a systematic review. Hepatology This study was funded in part by a grant from Instituto de 2016;63:1481–92. [21] Kuhnhenn L, Jiang B, Kubesch A, et al. Impact of HBV genotype and mutations Salud Carlos III, Ministerio de Economía y Competitividad (Grant on HBV DNA and qHBsAg levels in patients with HBeAg-negative chronic HBV No. PI14/00540), co-funded by Fondo Europeo de Desarrollo Re- infection. Aliment Pharmacol Ther 2018;47:1523–35. gional (FEDER), Unión Europea, Una manera de hacer Europa. [22] Bertoletti A, Ferrari C. Adaptive immunity in HBV infection. J Hepatol 2016;64:S71–83. [23] Honer Zu Siederdissen C, Rinker F, Maasoumy B, et al. Viral and host responses Supplementary materials after stopping long-term nucleos(t)ide analogue therapy in HBeAg-negative chronic hepatitis B. J Infect Dis 2016;214:1492–7. [24] Kuo MT, Hu TH, Hung CH, et al. Hepatitis B virus relapse rates in chronic hep- Supplementary material associated with this article can be atitis B patients who discontinue either entecavir or tenofovir. Aliment Phar- found, in the online version, at doi:10.1016/j.dld.2021.12.017. macol Ther 2019;49:218–28. [25] Chiu SM, Kuo YH, Wang JH, et al. Associations of HBV genotype B vs C infec- References tion with relapse after cessation of entecavir or tenofovir therapy. Clin Gas- troenterol Hepatol 2020;18:2989–97 e3. [1] European Association for the Study of the LiverElectronic address eee, Euro- [26] Kuo YH, Wang JH, Hung CH, et al. Combining end-of-treatment HBsAg and pean Association for the study of the L. EASL 2017 clinical practice guidelines baseline hepatitis B core-related antigen reduce HBV relapse rate after teno- on the management of hepatitis B virus infection. J Hepatol 2017;67:370–98. fovir cessation. Hepatol Int 2021. [2] Seto WK, Wong DK, Fung J, et al. Reduction of hepatitis B surface antigen levels [27] Brakenhoff SM, de Man RA, Boonstra A, et al. Hepatitis B virus RNA decline and hepatitis B surface antigen seroclearance in chronic hepatitis B patients re- without concomitant viral antigen decrease is associated with a low probabil- ceiving 10 years of nucleoside analogue therapy. Hepatology 2013;58:923–31. ity of sustained response and hepatitis B surface antigen loss. Aliment Phar- [3] Kim GA, Lim YS, An J, et al. HBsAg seroclearance after nucleoside analogue macol Ther 2021;53:314–20. therapy in patients with chronic hepatitis B: clinical outcomes and durability. Gut 2014;63:1325–32. 1051