Prediction of sustained response to peginterferon alfa-2b for hepatitis B e antigen–positive chronic hepatitis B using on-treatment hepatitis B surface antigen decline

Authors

  • Milan J. Sonneveld,

    1. Gastroenterology and Hepatology, Erasmus MC University Medical Center, Rotterdam, The Netherlands
    Search for more papers by this author
  • Vincent Rijckborst,

    1. Gastroenterology and Hepatology, Erasmus MC University Medical Center, Rotterdam, The Netherlands
    Search for more papers by this author
  • Charles A. B. Boucher,

    1. Virology, Erasmus MC University Medical Center, Rotterdam, TheNetherlands
    Search for more papers by this author
  • Bettina E. Hansen,

    1. Gastroenterology and Hepatology, Erasmus MC University Medical Center, Rotterdam, The Netherlands
    2. Biostatistics, Erasmus MC University Medical Center, Rotterdam, TheNetherlands
    Search for more papers by this author
  • Harry L. A. Janssen

    Corresponding author
    1. Gastroenterology and Hepatology, Erasmus MC University Medical Center, Rotterdam, The Netherlands
    • Department of Gastroenterology and Hepatology, Erasmus MC University Medical Center Rotterdam, ’s Gravendijkwal 230, room Ha 204, 3015 CE Rotterdam, The Netherlands
    Search for more papers by this author
    • fax: +31 (0) 10 436 5916


  • Dr. Janssen is a consultant for, and received grants from Roche, Bristol-Myers Squibb, Novartis, Gilead, and Schering-Plough.

Abstract

Serum hepatitis B surface antigen (HBsAg) levels may reflect the immunomodulatory efficacy of pegylated interferon (PEG-IFN). We investigated within a large randomized trial whether quantitative HBsAg levels predict response to PEG-IFN in patients with hepatitis B e antigen (HBeAg)-positive chronic hepatitis B. Serum HBsAg was measured in samples taken at baseline and weeks 4, 8, 12, 24, 52, and 78 of 221 patients treated with PEG-IFN alfa-2b with or without lamivudine for 52 weeks. HBsAg decline was compared between treatment arms and between responders and nonresponders. Response was defined as HBeAg loss with HBV DNA < 10,000 copies/mL at 26 weeks after treatment (week 78); 43 of 221 (19%) patients achieved a response. One year of PEG-IFN with or without lamivudine resulted in a significant decline in serum HBsAg, which was sustained after treatment (decline 0.9 log IU/mL at week 78, P < 0.001). Patients treated with combination therapy experienced a more pronounced on-treatment decline, but relapsed subsequently. Responders experienced a significantly more pronounced decline in serum HBsAg compared to nonresponders (decline at week 52: 3.3 versus 0.7 log IU/mL, P < 0.001). Patients who achieved no decline at week 12 had a 97% probability of nonresponse through posttreatment follow-up and no chance of HBsAg loss. In a representative subset of 149 patients similar results were found for prediction through long-term (mean 3.0 years) follow-up. Conclusion: PEG-IFN induces a significant decline in serum HBsAg in HBeAg-positive patients. Patients who experience no decline from baseline at week 12 have little chance of achieving a sustained response and no chance of HBsAg loss and should be advised to discontinue therapy with PEG-IFN. (HEPATOLOGY 2010)

Chronic hepatitis B (CHB) is a major health problem, affecting more than 350 million people worldwide. Prolonged infection with the hepatitis B virus (HBV) may ultimately result in severe liver-related morbidity and mortality, and treatment of CHB is therefore indicated in patients with persistent liver inflammation.1-4 The ideal outcome of treatment of CHB would be complete eradication of HBV, but this is only scarcely, if ever, achieved, for HBV covalently closed circular DNA (cccDNA) persists in host hepatocytes.5 Therefore, the main goal of therapy is to halt the progression of liver inflammation to fibrosis, cirrhosis or hepatocellular carcinoma.6, 7

Current treatment options for CHB consist of nucleo(s)tide analogues and (pegylated) interferons (PEG-IFN). Antiviral treatment with nucleo(s)tide analogues aims at inhibiting viral polymerase activity,8 and the most recently approved nucleo(s)tide analogues can effectively maintain suppression of HBV DNA levels for prolonged periods of time in the vast majority of patients.9-11 Nevertheless, PEG-IFN remains an important first-line treatment option for CHB, especially in hepatitis B e antigen (HBeAg)-positive disease, because a long-term off-treatment sustained response can be achieved in about 25% of patients after a finite treatment course.12-14 Response to IFN-based therapy in these patients is accompanied by high rates of hepatitis B surface antigen (HBsAg) seroconversion,15 a reduced incidence of hepatocellular carcinoma, and prolonged survival.16, 17

The development of a durable off-treatment response is attributed to the immunomodulatory effect of PEG-IFN,18 which results in a decrease in intrahepatic cccDNA.19 The cccDNA levels at the end of therapy are indeed predictive of a sustained off-treatment response,20 but the clinical utility is limited because these can only be assessed invasively. Recent studies report an excellent correlation between decline in intrahepatic cccDNA and serum HBsAg levels in HBeAg-positive patients.5, 21 A decline in serum HBsAg levels may therefore reflect the efficacy of PEG-IFN in decreasing intrahepatic cccDNA and consequently predict a sustained off-treatment response.

The aims of our study were to investigate the effects of 1 year of PEG-IFN with or without lamivudine (PEG-IFN ± LAM) therapy on serum HBsAg levels in patients with HBeAg-positive CHB, and to describe the relationship between on-treatment HBsAg decline and a sustained off-treatment response.

Abbreviations

ALT, alanine aminotransferase; AUC, area under the receiver-operating characteristic curve; cccDNA, covalently closed circular DNA; CHB, chronic hepatitis B; HBV, hepatitis B virus; HBeAg, hepatitis B e antigen; HBsAg, hepatitis B surface antigen; LAM, lamivudine; LTFU, long-term follow-up; PEG-IFN, pegylated interferon; ULN, upper limit of normal.

Patients and Methods

Patients.

In this study, serum HBsAg levels were assessed in HBeAg-positive CHB patients who were previously enrolled in an investigator-initiated multicenter randomized controlled trial and a subsequent long-term follow-up (LTFU) study.12, 13 Patients were eligible for the initial study if they had been HBsAg-positive for at least 6 months prior to randomization, were HBeAg-positive on two occasions within 8 weeks prior to randomization, had elevated serum alanine aminotransferase (ALT) levels of 2-10 times the upper limit of normal (ULN), and had a serum HBV DNA concentration above 1.0 × 105 copies/mL. Key exclusion criteria were: antiviral therapy within 6 months prior to randomization, presence of viral coinfections, preexisting cytopenia, or decompensated liver disease. Treatment comprised of PEG-IFN alfa-2b 100 μg weekly (PegIntron; Schering-Plough, Kenilworth, NJ) in combination with placebo or LAM (Zeffix; GlaxoSmithKline, Greenford, UK) 100 mg daily for 52 weeks. To limit the probability of early treatment discontinuation, the dose of PEG-IFN was reduced to 50 μg per week after 32 weeks of treatment. Patients attended the outpatient clinic at least every 4 weeks for routine examinations and laboratory assessments during both the treatment and the posttreatment follow-up phase of the initial study. For the LTFU study, patients were reevaluated at one additional visit at the local participating center. The mean duration of follow-up was 3 years.12

Inclusion criteria for the present analysis were completion of the 26-week follow-up phase of the main study and availability of a baseline serum sample for HBsAg quantification. Of the 266 patients in the initial study, 221 fulfilled these criteria. Of these patients, 149 participated in the associated LTFU study.12 The study was conducted in accordance with the guidelines of the Declaration of Helsinki and the principles of Good Clinical Practice. All patients gave written informed consent according to standards of the local ethics committees.

Laboratory Measurements.

Serum HBsAg was quantified in samples taken at baseline, during the treatment period (weeks 4, 8, 12, 24, and 52) and during follow-up (week 78) using the ARCHITECT HBsAg assay (Abbott Laboratories; range 0.05-250 IU/mL).22 HBV DNA quantification for the initial study was performed with 4-week intervals using an in-house-developed TaqMan polymerase chain reaction assay (lower limit of quantification = 400 copies/mL) based on the EuroHep standard.23 For the LTFU study, HBV DNA was measured with the Cobas TaqMan HBV assay (Roche Molecular Systems, Branchburg, NJ), with a dynamic range of quantification of 174-6.4 × 108 copies/mL (30-1.1 × 108 IU/mL). It has previously been demonstrated that there is an excellent correlation between the two assays.12 HBeAg was assessed using enzyme immunoassay (AxSYM; Abbott Laboratories, Abbott Park, IL) or enzyme-linked immunosorbent assay (DiaSorin SpA, Saluggia, Italy). ALT was measured locally in accordance with standard procedures and is presented as multiples of the ULN. HBV genotype was assessed using the INNO-LiPA assay (Innogenetics).

Statistical Analysis.

For the current study, a composite endpoint of HBeAg loss and HBV DNA level <10,000 copies/mL was chosen for definition of response.24 Patients who were retreated after the initial study were considered nonresponders at LTFU. Associations between variables were tested using Student t test, chi-squared test, Pearson correlation, or their nonparametric equivalents when appropriate. The differences in HBsAg decline between treatment arms and (non)responders were analyzed using repeated measurement models with an unstructured covariance allowing heterogeneity across compared groups. Discrimination, or the ability of HBsAg concentration and decline at various time points to distinguish patients who will develop a response from those who will not, was quantified by the area under the receiver-operating characteristic curve (AUC). Our aim was to use on-treatment HBsAg levels to identify a stopping rule that would enable a clinician to discontinue patients who had a very low chance of response as early as possible, while maintaining >90% of responders on treatment. The optimal cutoff in HBsAg decline was identified using a grid-search of possible cutoff points at weeks 4, 8, 12, and 24. For each cutoff point, the chi-squared test was calculated together with the sensitivity and the negative predictive value (NPV). The highest chi-squared test identified the optimal cutoff point.25 SPSS, version 15.0 (SPSS Inc., Chicago, IL) and the SAS 9.2 program (SAS Institute Inc., Cary, NC) were used to perform statistical analyses. All statistical tests were two-sided and were evaluated at the 0.05 level of significance.

Results

Patient Characteristics.

The characteristics of the 221 patients are shown in Table 1 according to assigned treatment regimen. Patients were comparable across both groups with regard to age, race, HBV genotype distribution, baseline prevalence of cirrhosis, and ALT and HBV DNA levels. Overall, 43 (19%) patients had a response at week 78, and these patients were distributed equally across the two study arms. Baseline mean serum HBsAg was 4.4 log IU/mL in both treatment groups. Serum HBsAg was positively correlated with HBV DNA (r = 0.66, P < 0.01) and inversely correlated with age (r = −0.16, P = 0.02) but did not correlate with ALT. Variation was observed in pretreatment HBsAg levels between genotypes, with the highest baseline levels in genotypes A and D (mean = 4.5 log IU/mL for both) and lower levels in genotypes B (mean = 4.3 log IU/mL) and C (mean = 3.8 IU/mL) (P < 0.001 for genotype C versus other genotypes with Bonferroni correction).

Table 1. Patient Characteristics According to Treatment Regimen
CharacteristicsPEG-IFN and Placebo (n = 111)PEG-IFN and Lamivudine (n = 110)P Value
  • *

    Multiples of upper limit of the normal range.

  • †HBeAg loss and HBV DNA <10,000 copies/mL.

Demography
 Mean (SD) age, in years34 (13)33 (12)0.52
 Male90 (81%)83 (75%)0.33
 Body mass index24.2 (3.4)25.1 (4.5)0.10
Race  0.31
 Caucasian80 (72%)80 (73%) 
 Asian25 (23%)19 (17%) 
 Other6 (5%)11 (10%) 
Laboratory results
 Mean (SD) ALT*4.3 (3.0)4.2 (3.0)0.75
 Mean (SD) HBV DNA, log copies/mL9.1 (0.8)9.1 (1.0)0.87
 Mean (SD) HBsAg, log IU/mL4.4 (0.6)4.4 (0.7)0.74
HBV Genotype  0.52
 A40 (36%)34 (31%) 
 B10 (9.0%)10 (9.1%) 
 C18 (16%)14 (13%) 
 D41 (37%)46 (42%) 
 Other/mixed2 (1.8%)6 (5.5%) 
Response at week 78
 Response#20 (18 %)23 (21%)0.61
 HBV DNA < 400 copies/mL9 (8.1%)13 (12%)0.38
 HBeAg loss43 (39%)41 (37%)0.89
 HBsAg loss8 (7.3%)11 (10%)0.48

On-Treatment HBsAg Decline According to Treatment Regimen.

Overall, HBsAg levels decreased significantly through 52 weeks of therapy (mean decline = 1.2 log IU/mL, P < 0.001), and the decrease was sustained after 26 weeks of follow-up (mean decline compared to baseline = 0.9 IU/mL, P < 0.001). Patterns of HBsAg decline for both treatment groups are depicted in Fig. 1. Declines were similar in both treatment arms at weeks 4, 8, and 12, but slightly more pronounced in the combination (PEG-IFN + LAM) compared to the monotherapy group (PEG-IFN + placebo) at week 24 (mean decline = 1.0 log IU/mL versus 0.6 log IU/mL, P = 0.04) and at week 52 (mean decline = 1.46 and 0.87 log IU/mL for combination therapy and monotherapy, respectively, P = 0.04). This difference was not sustained through posttreatment follow-up (mean decline of 0.98 and 0.86 log IU/mL for combination and monotherapy at week 78, respectively, P = 0.63). Considering the equal response rates and HBsAg levels at week 78 in the two treatment groups, we analyzed the relationship between HBsAg decline and treatment response in all 221 patients.

Figure 1.

Mean change in serum HBsAg from baseline in patients treated with monotherapy (PEG-IFN + placebo) and combination therapy (PEG-IFN + LAM).

HBsAg Decline According to Treatment Response at Week 78.

Baseline mean HBsAg levels were comparable in the 43 patients who achieved a response at week 78 and those who did not; 4.4 versus 4.3 log IU/mL in nonresponders and responders, respectively (P = 0.19). Mean HBsAg declines from baseline for responders and nonresponders at week 78 are shown in Fig. 2. Nonresponders showed a modest decline through 52 weeks of therapy (0.69 log IU/mL, P < 0.001), and relapsed during follow-up (decline from baseline at week 78 was 0.35 log IU/mL, P < 0.001 compared to week 52). Mean decline from baseline in responders was 3.3 log IU/mL at week 52 and 3.4 at week 78 (P < 0.001 for both when compared to baseline). Responders thus showed a more vigorous decline in HBsAg starting at week 4, and this difference increased through 52 weeks of therapy and was sustained during posttreatment follow-up (P < 0.005 for week 4 and P ≤ 0.001 for all other time points compared to nonresponders).

Figure 2.

Mean change in serum HBsAg from baseline in patients who achieved a response (HBeAg loss and HBV DNA < 10,000 copies/mL) at week 78 and those who did not.

Prediction of Response.

Because HBsAg decline patterns differed depending on treatment response, we investigated the discriminatory capabilities of HBsAg decline at weeks 4, 8, 12, and 24 for predicting response. Using receiver operating characteristic curve analysis, AUCs were 0.70, 0.76, 0.75, and 0.78 for decline at week 4, 8, 12, and 24, respectively, for predicting response at week 78. We also investigated the discriminatory values of absolute HBsAg levels (in log IU/mL) and HBV DNA decline, but these proved inferior to HBsAg declines.

Next, we proceeded to investigate the optimal cutoff point, according to our preset criteria, in HBsAg decline at week 4, 8, 12, and 24 for prediction of response. A cutoff of any decline in serum HBsAg level from baseline (i.e., the HBsAg level on-treatment was lower than the level measured at baseline: log(HBsAgon-treatment) − log(HBsAgbaseline) < 0) proved superior. Subsequently, prediction of response at weeks 12 and 24 was superior to weeks 4 and 8, because it allowed for more patients to be stopped, while maintaining >90% of responders on-treatment (Fig. 3). In addition, week 12 was superior to week 24 because it allowed for earlier discontinuation of therapy, while maintaining high predictive values for both response and HBsAg loss (Table 2).

Figure 3.

Individual HBsAg declines at week 12, stratified by response at week 78. All but two patients who achieved a response at week 78 experienced a decline in HBsAg from baseline to week 12. Lines represent median decline.

Table 2. Positive and Negative Predictive Values for Any HBsAg Decline at Week 12 and 24 for Prediction of Response and HBsAg Loss at Week 78
 Response Week 78*HBsAg Loss Week 78
 NoYesPPVNPVNoYesPPVNPV
  • *

    Response is defined as HBeAg loss and HBV DNA < 10,000 copies/mL.

Any decline, week 12 Yes1043525%1221712%
No61297%630100%
Any decline, week 24 Yes1223623%1401811%
No47492%510100%

At week 12, 69% of patients achieved a decline in HBsAg when compared to baseline. Of the 31% who did not, only 3% achieved a response at week 78. Consequently, the NPV of the presence of any decline in HBsAg at week 12 is 97% for prediction of response at week 78. Comparable NPVs were found for prediction of response at week 24 (Table 2, Fig. 4). Of those patients who developed a decline at week 12, 25% achieved a response at week 78, and 12% achieved HBsAg loss.

Figure 4.

Flowcharts for any decline in serum HBsAg levels from baseline at (A) week 12 or (B) week 24 in relation to sustained off-treatment response at week 78.

Prediction of Response Through LTFU.

Of the 149 patients with LTFU data available, 36 (24%) had a response at LTFU. Similar decline patterns were observed for responders and nonresponders at LTFU when compared to (non)responders at week 78; responders showed a steeper on-treatment decline. Declines were 0.53 log IU/mL versus 2.76 log IU/mL at week 52, for (non)responders, respectively (P = 0.007 for weeks 4 and 8, P ≤ 0.002 for all other time points), and the difference was sustained after treatment. Furthermore, of the patients who did not achieve a decline through 12 weeks of therapy, only 5% achieved a sustained response through LTFU and none lost HBsAg (Table 3).

Table 3. Positive and Negative Predictive Values for any HBsAg Decline at Week 12 and 24 for Prediction of Response and HBsAg Loss at LTFU
 Response LTFU *HBsAg Loss LTFU
 NoYesPPVNPVNoYesPPVNPV
  • *

    Response is defined as HBeAg loss and HBV DNA < 10,000 copies/mL.

Any decline, week 12 Yes652931%801415%
No42295%440100%
Any decline, week 24 Yes763028%901615%
No32391%350100%

Discussion

We report the first large study on serum HBsAg decline during PEG-IFN treatment for HBeAg-positive CHB in relation to a sustained off-treatment response. One year of therapy with PEG-IFN significantly reduced serum HBsAg levels, and the decrease was sustained through post-treatment follow-up. HBsAg decline was significantly more pronounced in patients who achieved a response (HBeAg loss and HBV DNA < 10,000 copies/mL). Furthermore, we found that reliable prediction of nonresponse to PEG-IFN is possible as early as week 12 of therapy, based on the absence of a decline in serum HBsAg. Patients who do not experience a decline in serum HBsAg from baseline to week 12, comprising 31% of our study population, have a minimal chance of achieving a sustained off-treatment response. Our results can help clinicians in their decision of whether to continue PEG-IFN therapy based on an individual patient's probability of nonresponse.

PEG-IFN can induce an off-treatment sustained response in a substantial proportion of patients with HBeAg-positive CHB,12–15 but its clinical use is compromised by the frequent occurrence of side-effects26 and the uncertainty as to whether a patient will actually benefit from this therapy. Reliable prediction of nonresponse at baseline or during the first weeks of therapy is therefore essential to optimal utilization of this agent. Recently, a baseline prediction model has been published, based on data from the two largest studies involving PEG-IFN in HBeAg-positive CHB.24 The model enables the clinician to predict response (HBeAg loss and HBV DNA < 2000 IU/mL [∼10,000 copies/mL]) of HBeAg-positive patients to PEG-IFN, based on readily available data, such as HBV genotype, HBV DNA and ALT levels, age, and sex. Although the model provides considerable support when considering a patient for PEG-IFN therapy, substantial uncertainty remains as to whether an individual patient will respond to a 1-year course of PEG-IFN. On-treatment monitoring of viral replication using HBV DNA, HBeAg and HBsAg levels may aid decision-making and frequent HBV DNA monitoring is therefore recommended in treatment guidelines.3 However, modeling of HBV DNA kinetics during PEG-IFN therapy has shown only limited clinical utility,27, 28 and reliable prediction of nonresponse is only possible at week 24 of therapy (NPV = 86%).29

Recent technical advances have allowed for the quantitative assessment of HBsAg in serum. HBsAg is secreted from the hepatocyte during viral replication as part of the HBV nucleocapsid, or as part of noninfectious viral particles.30 Several studies have reported that serum HBsAg levels correlate with intrahepatic cccDNA levels in HBeAg-positive patients.21, 31 On-treatment HBsAg decline may therefore reflect the efficacy of PEG-IFN in decreasing intrahepatic cccDNA and consequently predict a sustained response.21, 31 This hypothesis was first tested in patients who are HBeAg-negative, and it was found that patients with low HBsAg levels at the end of treatment had the highest probability of achieving a sustained off-treatment response.32 Furthermore, another study showed that patients who did not achieve a 0.5 log decline in serum HBsAg from baseline to week 12 of therapy had only 10% probability of achieving a response (NPV = 90%).33

Our observations in HBeAg-positive patients corroborate these results on the excellent predictive capabilities of on-treatment HBsAg decline. In our study population, patients who did not achieve a decline in serum HBsAg concentration from baseline to week 12 of therapy had only 3% chance of achieving a sustained off-treatment response. The resulting NPV of 97% is superior to that achieved using HBV DNA and comparable to HBeAg monitoring.29 Furthermore, our findings indicate that prediction of nonresponse to PEG-IFN is possible as early as week 12, as opposed to week 24 when using serum HBV DNA or HBeAg levels29 and that prediction of nonresponse using HBsAg decline can accurately indentify those patients with a low probability of sustained response through 3 years of post-treatment follow-up. Furthermore, if our on-treatment stopping rule was applied combined with the baseline prediction model,24 the AUC increased from 0.75 for the stopping rule alone to 0.79 for the combination, showing that application of both models to guide therapy decisions may be beneficial.

Other studies have reported that HBsAg levels of <1500 IU/mL at week 12 or week 24 of therapy were highly predictive of sustained HBeAg seroconversion 6 months post-treatment.34 We found comparable positive predictive values (PPVs) for HBsAg levels <1500 IU/mL at week 12 for response at LTFU (PPV = 55%) and for loss of HBsAg at LTFU (PPV = 35%). Prediction did not improve at week 24, with PPVs of 53% for response at LTFU, and 41% for HBsAg loss at LTFU. Anyhow, these results have limited clinical significance, because even patients with HBsAg levels >1500 IU/mL at either of these time points have a considerable probability of response. If one were to discontinue therapy in all patients with HBsAg >1500 IU/mL at week 24, one would miss out on 48% of patients with a response at LTFU in our study population.

A possible caveat of our study is that we pooled data from the two treatment arms for the formulation of our stopping rule. Patients who received combination therapy experienced a somewhat larger decline from week 24 to week 52. To account for this, we validated our stopping rule in both treatment groups, and found that it performed equally well in both populations. Sensitivity analysis confirmed that a cutoff of any decline was superior in both groups. Additionally, our LTFU population comprised only a subgroup of the total study group (149 of 221). However, it was previously shown that the LTFU group was representative of the entire study cohort,12 and we confirmed these findings (data not shown). Also, the cutoff of any decline performed well in both groups (Tables 2 and 3). Furthermore, one could argue that we should have chosen a different definition of response. In this study, we defined response as off-treatment sustained HBeAg loss combined with HBV DNA < 10,000 copies/mL (∼2000 IU/mL), because HBeAg loss 6 months after treatment has been reported to be highly durable12 and because patients with low HBV DNA levels are less likely to develop HBV related liver complications or require antiviral therapy according to recent guidelines.3, 35-37 Moreover, this endpoint is in line with other recently published articles on response to PEG-IFN in HBeAg-positive CHB,24 and the high negative predictive values were maintained if HBeAg seroconversion combined with HBV DNA <10,000 copies/mL was applied as an endpoint.

In conclusion, a 1-year course of PEG-IFN results in a significant decline in serum HBsAg in patients with HBeAg-positive CHB. The decline is considerably more pronounced in patients who achieve a response (HBeAg loss and HBV DNA <10,000 copies/mL) when compared to nonresponders. Patients who do not experience a decline in HBsAg levels through 12 weeks of therapy have a low chance of achieving a sustained off-treatment response (<5%) and no chance of HBsAg loss, and should therefore be considered for treatment discontinuation.

Ancillary