Aliment Pharmacol Ther 2010; 32: 1323–1331
Background On-treatment predictors of response to peginterferon can guide individualization of therapy in chronic hepatitis B virus infection.
Aim To investigate the use of serum hepatitis B surface antigen quantification to predict sustained response.
Methods Hepatitis B e antigen-positive chronic hepatitis B patients who received peginterferon for 32–48 weeks with or without lamivudine combination were studied. Sustained response was defined as hepatitis B e antigen seroconversion and chronic hepatitis B virus DNA <10 000 copies/mL until 12 months post-treatment.
Results Twenty-one of 92 (23%) patients achieved sustained response. At month 6, the area under receiver operating characteristics curve for hepatitis B surface antigen to predict sustained response was 0.77 (95% confidence interval 0.65–0.89, P < 0.001). An hepatitis B surface antigen cutoff at 300 IU/mL at month 6 could give the maximum combination of sensitivity (62%) and specificity (89%) to predict sustained response. Nine of 21 (43%) sustained responders vs. 9 of 71 (13%) nonsustained responders had >1 log hepatitis B surface antigen reduction at month 6 (P < 0.001). Combined hepatitis B surface antigen ≤300 IU/mL and >1 log reduction at month 6 had sensitivity, specificity, positive and negative predictive values of 43%, 96%, 75% and 85% to predict sustained response, respectively.
Conclusion On-treatment serum hepatitis B surface antigen can predict response to peginterferon therapy in chronic hepatitis B.
Chronic hepatitis B virus (HBV) infection is the most common cause of hepatocellular carcinoma in most parts of Asia.1 In a recent meta-analysis, treatment with interferon could reduce the risk of hepatocellular carcinoma development by 33%.2 Although there are limited data suggesting that peginterferon is more effective than conventional interferon, peginterferon is a more convenient treatment option with once weekly dosing.3 Responders to peginterferon therapy rarely have hepatitis relapse after stopping treatment in long-term follow-up studies.4–7 Unfortunately, only approximately one-third of patients will respond to peginterferon treatment, and combination with lamivudine cannot improve the post-treatment virological response.8
Baseline factors including HBV genotype, HBV DNA and alanine aminotransferase (ALT) levels can provide some prediction on the treatment response.9 However, in Asia where genotype B and C HBV are most prevalent, these baseline factors have limited ability to discriminate responders from nonresponders.9, 10 On-treatment response predictors that can identify the nonresponders for early treatment termination can avoid unnecessary cost and side effects of peginterferon.11 There are some preliminary data suggesting that extended peginterferon beyond 48 weeks might improve the response rate in suboptimal responders.12 Definition of on-treatment response predictors will therefore be an important step towards individualization of peginterferon therapy in chronic hepatitis B.
Hepatitis B surface antigen (HBsAg) quantification has been shown to correlate with the concentration of covalently closed circular DNA in the liver.13, 14 As clearance of covalently closed circular DNA is believed to be an immune-mediated process, monitoring of serum HBsAg becomes a logical tool to predict the response to peginterferon therapy. Studies among hepatitis B e antigen (HBeAg)-negative chronic hepatitis B patients suggested that reduction in serum HBsAg level could predict post-treatment virological suppression and loss of HBsAg.15, 16 In this study, we aimed to investigate the role of serum HBsAg level to predict treatment response among HBeAg-positive patients who have received peginterferon therapy in our centre. On the basis of these data, we would define the best timing and cutoff of HBsAg concentration that could predict treatment response in our patients.
Chronic hepatitis B patients who had positive HBeAg and completed per protocol peginterferon-alfa treatment in previous clinical trials were studied.17–19 Among patients who received combination peginterferon and lamivudine treatment, only those who started peginterferon either before or together with lamivudine were included in the analysis. Co-infection with hepatitis C virus was excluded. Sustained response (SR) was defined as HBeAg seroconversion and HBV DNA ≤10 000 copies/mL at the end of the entire course of peginterferon and/or lamivudine treatment sustained until 12 months post-treatment. All study protocols have been approved by local ethics committee. All patients have signed written informed consent for the studies including the usage of stored residual serum samples for virologic studies.
Residual serum samples were stored at −80 °C freezer. Serum HBV DNA and HBsAg were measured at baseline, month 3 and month 6 to predict SR. For patients among whom month 3 serum samples were not available due to the protocol design, month 4 serum samples would be studied instead. The serum HBV DNA and HBsAg levels at end of peginterferon treatment and 12 months post-treatment were also measured. HBV genotyping was determined in the residual serum sample at the baseline visit.
Hepatitis B surface antigen was quantified by Architect HBsAg QT (Abbott Diagnostic, Wiesbaden, Germany) according to the manufacturer’s instruction.20 The sensitivity of Architect assay ranged from 0.05 to 250 IU/mL. Samples with HBsAg titre higher than 250 IU/mL were diluted to 1:500 to 1:1000 to bring the reading within the range of the calibration curve. All HBsAg quantifications were run in duplicate with a coefficient of variance less than 10%. HBV DNA was quantified by TaqMan real-time polymerase chain reaction assay as described previously.21 The range of HBV DNA detection was from 102 to 109 copies/mL. HBV genotyping was determined by restriction fragment length polymorphism and confirmed by direct sequencing in case of doubt as described previously.22
Statistical analysis was performed by spss (version 15.0, Chicago, IL, USA). Serum HBV DNA was logarithmically transformed for analysis. Continuous variables were expressed as mean ± standard deviation or median (range), and compared by Student’s t-test or Mann–Whitney U-test as appropriate. Categorical variables were compared by Pearson’s chi-square test. Area under receiver operating characteristics (ROC) curves were used to assess the performance of serum HBV DNA and HBsAg at different time points during treatment to predict SR, and the differences between the areas under the ROC curves were compared by DeLong test. The HBsAg level cutoffs that could predict SR with >90% sensitivity (high sensitivity), >90% specificity (high specificity) and maximum sum of sensitivity and specificity (maximum sum) were derived from the coordinates of the ROC curves. Sensitivity, specificity, positive (PPV) and negative (NPV) predictive values, and positive and negative likelihood ratios for different HBsAg cut-offs were computed. All statistical tests were two-sided. A P value of < 0.05 was considered statistically significant.
Clinical characteristics of patients
Overall, ninety-nine HBeAg-positive patients were recruited in the previous three studies started with peginterferon therapy.17–19 Seven patients did not complete the per protocol treatment and prematurely withdrew from the trials. The remaining 92 patients were eligible for analysis. Sixty-three patients received 32-week peginterferon alfa-2b treatment, among whom 45 had combination with 1-year lamivudine and 18 had combination with 2-year lamivudine treatment.17, 18 Twenty-nine patients received 48-week peginterferon alfa-2a monotherapy19 (Figure 1).
Twenty-one (23%) patients achieved SR (Table 1). Patients who achieved SR tend to have lower baseline serum HBV DNA than those who failed to achieve SR (7.47 ± 0.95 log copies/mL vs. 8.11 ± 1.0 log copies/mL, P = 0.012). Similarly, the median baseline serum HBsAg levels also tend to be lower among sustained responders than nonsustained responders (3301 IU/mL vs. 6232 IU/mL, P = 0.012). Otherwise, there was no difference in the age, gender ratio, body mass index, ALT levels and HBV genotypes between sustained responders and nonsustained responders. On long-term follow-up for 5.2 ± 1.8 years post-treatment, 20 of the 21 (95%) patients with SR remained in HBeAg seroconversion.
|SR N = 21||Non-SR N = 71||All N = 92|
|Age (year)||35 ± 12||34 ± 10||34 ± 10|
|Male gender||13 (62%)||49 (69%)||62 (67%)|
|Body mass index (kg/m2)||23.7 ± 4.2||22.8 ± 3.2||23.0 ± 3.5|
|Alanine aminotransferase (IU/L)||129 (52–339)||135 (48–2260)||134 (48–2260)|
|Log HBV DNA (copies/mL)||7.47 ± 0.95*||8.11 ± 1.0*||7.96 ± 1.02|
|HBsAg (IU/mL)||3301 (10–39 488)**||6232 (41–168 200)**||4893 (10–168 200)|
|B||6 (29%)||15 (21%)||21 (23%)|
|C||14 (67%)||51 (72%)||65 (71%)|
|Mixed B/C||1 (5%)||4 (6%)||5 (5%)|
|D||0 (0%)||1 (1%)||1 (1%)|
On-treatment predictors of SR
Serum HBV DNA and serum HBsAg tend to decrease during treatment. Patients with SR had a more significant reduction in serum HBsAg and HBV DNA than those without SR (Figure 2a and 2b). The same trend was observed among patients on different treatment regimes (Table 2). Both on-treatment serum HBV DNA and HBsAg levels at month 3 and month 6 could predict SR (Table 3). The areas under ROC curve at month 6 tend to be higher than those in month 3 for both HBV DNA (P = 0.06) and HBsAg (P = 0.19). An HBsAg cutoff at 300 IU/mL at month 6 could give the maximum sum of sensitivity (62%) and specificity (89%) to predict SR among HBsAg cutoffs at any time point (Table 4). Among 21 (21% of the 92) patients who had HBsAg ≤300 IU/mL at month 6, 13 (62%) of them could achieve SR. On the contrary, among the 71 patients who had HBsAg greater than 300 IU/mL at month 6, only 8 (11%) could achieve SR (P < 0.001).
|Number of patients||Baseline||Month 3||Month 6||End of peginterferon||1-year post-treatment|
|Regime 1||HBV DNA||SR||8||7.40||3.08||2.27||2.77||3.04|
|Regime 2||HBV DNA||SR||7||7.74||3.34||3.03||2.82||3.09|
|Regime 3||HBV DNA||SR||6||7.26||5.28||3.29||2.14||2.25|
|Month||SR||Non-SR||Area under ROC curve||95% confidence interval||P|
|Log HBV DNA (copies/mL)|
|0||7.47 ± 0.95||8.11 ± 1.0||0.67||0.55–0.79||0.019|
|3||3.80 ± 1.47||4.90 ± 1.70||0.69||0.57–0.81||0.008|
|6||2.80 ± 1.01||4.42 ± 1.74||0.78||0.68–0.88||<0.001|
|0||3301 (10–39 489)||6232 (41–168 200)||0.64||0.51–0.78||0.047|
|3||1192 (4–19 685)||5594 (2–125 250)||0.73||0.60–0.85||0.002|
|6||180 (0.3–12 965)||3449 (0.6–77 112)||0.77||0.65–0.89||<0.001|
|Reduction in log HBsAg (IU/mL)|
|3||0.37 (−0.09 to 2.63)||0.09 (−0.97 to 1.34)||0.67||0.55–0.79||0.017|
|6||0.85 (−0.16 to 4.00)||0.31 (−1.40 to 5.16)||0.70||0.58–0.83||0.005|
|Selection criteria||Log HBV DNA (copies/mL)||Sn (%)||Sp (%)||PPV (%)||NPV (%)||LR (+)||LR (−)||DA (%)|
|Maximum Sn + Sp||8.00||76||51||31||88||1.6||0.5||49|
|Month 3||High Sn||5.50||91||44||32||94||1.6||0.2||50|
|Maximum Sn + Sp||5.50||91||44||32||94||1.6||0.2||50|
|Month 6||High Sn||4.50||95||44||33||97||1.7||0.1||48|
|Maximum Sn + Sp||3.50||76||70||43||91||2.5||0.3||62|
|Selection criteria||HBsAg (IU/mL)||Sn (%)||Sp (%)||PPV (%)||NPV (%)||LR (+)||LR (−)||DA (%)|
|Maximum Sn + Sp||4000||57||68||34||84||1.8||0.6||57|
|Month 3||High Sn||7000||91||44||32||94||1.6||0.2||49|
|Maximum Sn + Sp||1500||57||80||46||86||2.9||0.5||65|
|Month 6||High Sn||4500||91||41||31||94||1.5||0.2||45|
|Maximum Sn + Sp||300||62||89||62||89||5.6||0.4||72|
Reduction in serum HBsAg at months 3 and 6 from baseline could also predict SR (Table 2, Figure 2c). Again, the area under ROC for log HBsAg reduction at month 6 tend to be higher than that at month 3, but the difference fell short of statistical significance (P = 0.25). Eighteen patients had HBsAg reduction greater than 1 log at month 6, and 9 (50%) of them could achieve SR. All these nine sustained responders also had HBsAg ≤300 IU/mL at month 6. On the contrary, among the 74 patients who had HBsAg reduction equal to or less than 1 log at month 6, only 12 (16%) could achieve SR (P = 0.002).
Combined HBsAg algorithm
At month 6, twelve (13%) of 92 patients had both HBsAg reduction by greater than 1 log and serum HBsAg ≤300 IU/mL. Nine (75%) patients with this combined HBsAg response developed SR. On the other hand, twelve (15%) of the 80 patients who did not have this combined response had SR (P < 0.001). The sensitivity, specificity, PPV, NPV, positive likelihood ratio and negative likelihood ratio of this combined HBsAg response were 43%, 96%, 75%, 85%, 10.8 and 0.6 respectively.
Twelve of 71 (11%) patients who did not achieve either an HBsAg of ≤300 IU/mL or a 1 log reduction of HBsAg, i.e. no HBsAg response, at month 6 had SR. Four of 9 (44%) patients who had partial HBsAg response (HBsAg ≤300 IU/mL but HBsAg reduction by ≤1 log) at month 6 could achieve SR. Based on the on-treatment HBsAg response at month 6, an algorithm could be developed to predict SR in HBeAg-positive patients who received peginterferon therapy (Figure 3).
Two patients had clearance of HBsAg. One patient who was infected by genotype B HBV received 32-week of peginterferon alfa-2b combined with 52-week of lamivudine. He had HBsAg loss at the end of peginterferon therapy. His serum HBsAg was 79481 IU/mL at baseline, 11394 IU/mL at month 3 and 0.55 IU/mL at month 6 (reduction of 5.2 log from baseline). Another patient who was infected by genotype B HBV received 48-week of peginterferon alfa-2a monotherapy. He had loss of HBsAg at 1-year post-treatment. His serum HBsAg was 2979 IU/mL at baseline, 7.02 IU/mL at month 3, and 0.3 IU/mL at month 6 (reduction of 4.0 log from baseline).
In this study, we have demonstrated that serum HBsAg quantification can be used as an on-treatment predictor for peginterferon therapy in HBeAg-positive chronic hepatitis B patients. An absolute HBsAg level of ≤300 IU/mL and a reduction of HBsAg by greater than 1 log at month 6 offered the best prediction to SR at 1 year after stopping treatment. According to our previous experience and the long-term follow-up data, most patients who had SR at 1-year posttreatment tend to remain in remission thereafter.4
On the basis of recommendation of the regional guidelines, we have adopted a stringent criterion for SR, which was defined as HBeAg seroconversion and HBV DNA lower than 10 000 copies/mL at 1 year posttreatment.23–25 Overall, 23% of patients developed SR. Sustained responders tend to have lower baseline HBsAg and HBV DNA than the nonsustained responders. However, the area under ROC curve of these two parameters were lower than 70% and could hardly be useful to guide patient selection for peginterferon treatment in clinical practice. There was no difference in the SR between patients infected by genotype B and genotype C HBV.
Hepatitis B virus DNA has been used for on-treatment monitoring in the roadmap models of nucleos(t)ide analogues.11 In one of our previous reports among HBeAg-positive patients treated by peginterferon-alfa-2b and lamivudine combination, HBV DNA at week 8 could offer an early prediction of SR.26 However, in another study using peginterferon-alfa-2a monotherapy, HBV DNA at month 6 could offer a better prediction of SR than the earlier time points.19 Nonetheless, the accuracy of prediction by on-treatment HBV DNA monitoring was only modest in the larger scaled, multi-centre studies using either peginterferon therapy.27, 28 In this study, we have pooled together patients in different clinical trials to evaluate on-treatment predictors of SR. As the designs of these studies were heterogeneous and some patients had combination of peginterferon and lamivudine, our study could not evaluate the performance of serum HBV DNA. We would expect a more dramatic reduction in HBV DNA among patients on combination treatment than in those who were on peginterferon monotherapy. Although we have demonstrated a good prediction by HBV DNA at month 6 for SR, we could not draw any conclusion on the use of HBV DNA monitoring. However, combination of peginterferon and lamivudine seemed to have little effect on the serum HBsAg reduction,16, 29 and the pooling of patients from different clinical trials could increase the patient number and event rate of SR for the analysis.
Our results were consistent with that of the pivotal peginterferon-alfa-2a trials, which demonstrated a more dramatic reduction in serum HBsAg among patients who had sustained off-treatment response as compared with those who failed to respond.16, 30 Thirty-two (35%) patients in our cohort had HBsAg <1500 IU/mL at month 6, which was very comparable to that reported (32%) in the global peginterferon-alfa-2a trial.30 The differences in HBsAg levels were significant between sustained responders and nonsustained responder at month 3 and 6 of treatment. At month 3, HBsAg cutoff of 1500 IU/mL could provide the best prediction for SR. Comparing the area under the ROC curves for the absolute HBsAg level and HBsAg reduction between month 3 and month 6, it seemed that month 6 was a more accurate time point to predict SR than month 3. Instead of an HBsAg cutoff of 1500 IU/mL, we found that a lower HBsAg level of 300 IU/mL at month 6 could provide the best prediction for SR.30 Sixty-two percent of patients could have response accurately predicted by an HBsAg cutoff of 300 IU/mL, which has approximately 90% specificity and negative predictive value for SR.
In a French study among HBeAg-negative patients, eleven of the twelve patients who had HBsAg reduction by equal to or greater than 1 log at month 6 could achieve undetectable HBV DNA 6 months after stopping peginterferon therapy.15 In the multi-centre peginterferon alfa-2a study for HBeAg-negative patients, 30% of patients who had HBsAg reduction by greater than 1 log at month 12 had HBsAg clearance 3 years after treatment.16 These findings were consistent with a recent longitudinal study among 117 untreated chronic hepatitis B patients followed up for >8 years, in which HBsAg reduction reflected a good immune control among HBeAg-negative patients.31 However, among HBeAg-positive patients, it seemed an absolute level of HBsAg rather than the magnitude of HBsAg reduction at month 6 could provide a better prediction of sustained HBeAg seroconversion.30 In contrast to the excellent prediction by HBsAg in HBeAg-negative patient, both HBsAg level and HBsAg reduction could only achieve modest prediction for SR in our cohort of HBeAg-positive patients. Therefore, we proposed to combine the absolute HBsAg level of ≤300 IU/mL and HBsAg reduction by greater than 1 log at month 6 to generate an algorithm for prediction of SR to peginterferon therapy. Among the minority (12 patients; 13%) who could achieve a combined HBsAg response at month 6, 75% would develop SR. On the other hand, among those (71 patients; 77%) who could not achieve an HBsAg of 300 IU/mL, regardless of the HBsAg reduction, at month 6, only 11% could achieve SR.
Our study has a few limitations. First, some patients had received continuation of lamivudine treatment for 24–72 weeks after the cessation of peginterferon. We could not assess virological relapse immediately after stopping peginterferon due to the effect of lamivudine. As combination with lamivudine was unlikely to improve the SR of peginterferon,32, 33 we believed that assessment at 1 year after stopping lamivudine could accurately evaluate the SR of peginterferon treatment. Second, some patients received only 32-week peginterferon treatment, which is slightly shorter than the standard 48-week treatment as recommended by the regional guidelines.23–25 So far, there was no evidence showing superiority of 48-week peginterferon treatment over 32-week peginterferon treatment. Nonetheless, the SR of our patients (22%) was comparable to that among patients infected by genotypes B and C HBV reported in the combined analysis of two pivotal studies using similar criteria for SR.9 Third, our study was a post hoc analysis of three previously conducted studies. We did not find any difference on the baseline ALT level between patients with and without SR, which was not in agreement with previous large scaled studies with peginterferon.9, 33 This was probably due to our small sample size and the wide range of baseline ALT levels. Future prospectively designed, larger scaled studies on peginterferon treatment would be warranted to define how HBV DNA and HBsAg could be used together to predict SR. As illustrated in a recently published study in HBeAg-negative patients, combination of month 3 HBV DNA and HBsAg changes could predict poor responders for early cessation of peginterferon therapy.34 Last, this was a single-centred experience and the results needed to be confirmed by multi-centred studies or studies in other geographical regions among different ethnic patient groups.
In conclusion, we found that serum HBsAg at month 6 can predict SR to peginterferon treatment in HBeAg-positive chronic hepatitis B. Patients with favourable on-treatment HBsAg response should be reassured and encouraged to complete the treatment course. Whether a 6-month course of peginterferon is sufficient will require further investigations. It is controversial on the best strategy for patients with no HBsAg response. One possibility is switching to nucleos(t)ide analogue therapy to avoid the unnecessary inconvenience and adverse effects of peginterferon. At present, there is no evidence suggesting that combination of lamivudine with peginterferon can improve the efficacy of peginterferon therapy, but further study is needed to assess the benefit of newer nucleos(t)ide analogues.18, 32, 33 Future studies should explore the benefit of extended peginterferon beyond 48 weeks, particularly among patients who have partial HBsAg response at month 6.12 Nonetheless, the predictive value of HBsAg was not yet optimal, as only a minority (13%) of patients have favourable combined HBsAg response, and the false positive rate (11%) for a lack of HBsAg response was still high. Future investigation on the use of HBsAg monitoring, possibly in combination with HBV DNA and quantitative HBeAg, will be needed to transform peginterferon therapy for chronic hepatitis B into an individualized, response-dependent treatment.
Declaration of personal interests: Henry L-Y. Chan has served as a speaker, a consultant and an advisory board member for F. Hoffmann-La Roche, Abbott Diagnostic, Novartis Pharmaceutical, Bristol-Myers Squibb and Merck. Vincent W-S Wong has served as a speaker for Abbott Diagnostic and Novartis Pharmaceutical. Other authors have no personal conflict of interest. Declaration of funding interests: This study was funded in full by Research Fund for the Control of Infectious Diseases (RFCID) grant (08070242) by Food and Health Bureau, Hong Kong to Henry L-Y. Chan. No writing support was received.