Factors That Predict Response of Patients With Hepatitis B

Factors That Predict Response of Patients With Hepatitis B e Antigen–Positive Chronic Hepatitis B to Peginterferon-Alfa


Gastroenterology Dec 2009

Erik H.C.J. Buster, Bettina E. Hansen‡, George K.K. Lau§, Teerha Piratvisuth∥, Stefan Zeuzem¶, Ewout W. Steyerberg#, Harry L.A. Janssen

Department of Gastroenterology and Hepatology, Erasmus MC University Medical Center Rotterdam, Rotterdam, The Netherlands

‡ Department of Epidemiology and Biostatistics, Erasmus MC University Medical Center Rotterdam, Rotterdam, The Netherlands

# Department of Public Health, Erasmus MC University Medical Center Rotterdam, Rotterdam, The Netherlands

§ Department of Medicine, Queen Mary Hospital, University of Hong Kong, Hong Kong SAR, China

∥ Department of Medicine, Songklanakarin Hospital, Songkla, Thailand

¶ Medizinische Klinik I, Klinikum der Johann Wolfgang Goethe-Universität, Frankfurt am Main, Germany


Received 25 March 2009; accepted 26 August 2009. published online 07 September 2009.

ABSTRACT

Background & Aims


Therapy with pegylated interferon (PEG-IFN)–alfa results in sustained response in a minority of patients with chronic hepatitis B virus (HBV) infection and has considerable side effects. We analyzed data from the 2 largest global trials of hepatitis B e antigen (HBeAg)-positive patients with chronic hepatitis B to determine which are most likely to respond to PEG-IFN–alfa therapy.

Methods


The study included 542 patients treated with PEG-IFN–alfa-2a (180 μg/wk, 48 wk) and 266 patients treated with PEG-IFN–alfa-2b (100 μg/wk, 52 wk). Eighty-seven patients were excluded, leaving 721 patients for analysis. A sustained response was defined as HBeAg loss and HBV-DNA level less than 2.0 × 103 IU/mL 6 months after treatment. Logistic regression analysis was used to identify predictors of sustained response and a multivariable model was constructed.

Results


HBV genotype, high levels of alanine aminotransferase (ALT; ≥2 × upper limit of normal), low levels of HBV DNA (<2.0 × 108 IU/mL), female sex, older age, and absence of previous IFN therapy predicted a sustained response. Genotype A patients with high ALT and/or low HBV-DNA levels had a high predicted probability (>30%) of a sustained response. The strongest predictors of response were a high level of ALT in genotype B patients and a low level of HBV DNA in genotype C patients. Genotype D patients had a low chance of sustained response, irrespective of ALT or HBV-DNA levels.

Conclusions


The best candidates for a sustained response to PEG-IFN–alfa are genotype A patients with high levels of ALT or low levels of HBV DNA, and genotypes B and C patients who have both high levels of ALT and low HBV DNA. Genotype D patients have a low chance of sustained response.

Funding The Peginterferon Alfa-2a HBeAg-Positive Chronic Hepatitis B Study was supported by a research grant from Roche (Basel, Switzerland). The HBV99-01 study was organized and sponsored by the Foundation for Liver Research (SLO), Rotterdam, The Netherlands. Financial support and study medication for the HBV99-01 study was provided by Schering-Plough International (Kenilworth, NJ) and GlaxoSmithKline (Greenford, UK).


Hepatitis B is a major global health problem. The World Health Organization reported that there are more than 400 million carriers in the world, approximately 75% of whom reside in Asia and the Western Pacific.1 In this part of the world, hepatitis B virus (HBV) infection usually is acquired perinatally or in early childhood. Most patients from these areas typically have hepatitis B e antigen (HBeAg)-positive chronic hepatitis B with high HBV-DNA levels, and they develop moderate to severe hepatic inflammation with increased alanine aminotransferase (ALT) levels after 10–30 years of infection.2 In contrast, patients infected in late childhood, adolescence, or adulthood present with increased aminotransferase levels after a shorter duration of infection. Although spontaneous HBeAg seroconversion occurs in the majority of HBeAg-positive patients, the duration of hepatic inflammation can be prolonged and severe, and may result in cirrhosis. Therefore, antiviral treatment is indicated in patients who remain HBeAg positive with high HBV-DNA levels after a 3- to 6-month period of increased ALT levels.3


Successful treatment of chronic HBV infection with loss of HBeAg, decline in serum HBV-DNA level, and normalization of ALT level is associated with favorable long-term outcome, independent of the antiviral drug used.4, 5 In HBeAg-positive patients, sustained clearance of HBeAg from serum is associated with a higher likelihood of losing hepatitis B surface antigen (HBsAg), reduced incidence of cirrhosis and hepatocellular carcinoma (HCC), and improved survival.5, 6, 7, 8 Of currently available drugs for the treatment of chronic hepatitis B, pegylated interferon (PEG-IFN) still results in the highest rate of off-treatment sustained response after a 1-year course of therapy.9, 10, 11, 12 Furthermore, responders to IFN-based therapy have a considerable chance of losing HBsAg, which has been observed in 12%–65% of patients within 5 years after HBeAg loss.7, 8, 13, 14, 15, 16, 17, 18 Treatment with PEG-IFN, however, often is complicated by the occurrence of side effects such as flu-like symptoms, cytopenia, and depression.19 Nucleos(t)ide analogues such as lamivudine, adefovir, entecavir, and tenofovir on the other hand are well tolerated, but because of the modest seroconversion rate and the high risk of posttreatment relapse, prolonged therapy usually is required.20 Nowadays, maintenance of virologic response over prolonged periods is feasible,21 but still may pose a considerable risk for resistance in the long term.22, 23, 24


Because treatment with both PEG-IFN and nucleos(t)ide analogues has proven effective, but also have their advantages and limitations, the question arises of what treatment regimen should be used as first-line therapy in which patients. Both the chance of achieving response and specific patient characteristics play a role in the decision on what type of antiviral therapy should be started. Recently performed studies with 1 year of PEG-IFN in HBeAg-positive patients identified high baseline ALT levels, low baseline HBV-DNA levels, absence of previous IFN therapy, low baseline HBeAg levels, and HBV genotype A or B as predictors of response.10, 25, 26 Current guidelines do not provide specific recommendations as to which patients should be treated with peginterferon27; the earlier-mentioned studies were considered to provide insufficient evidence for such recommendations. The aim of this study therefore was to develop a solid model for the prediction of sustained response to PEG-IFN in individual HBeAg-positive patients, which would allow physicians throughout the world to choose the optimal candidates for treatment with this drug.

Discussion


We combined the data of the 2 largest studies investigating PEG-IFN in HBeAg-positive chronic hepatitis B to develop a model for the prediction of response to PEG-IFN in all HBV genotypes. Although the model is based on the data of patients enrolled in randomized clinical trials with predefined inclusion and exclusion criteria, generalizability of our results probably is good because of the large sample size and wide geographic distribution of the patients. We provided nomograms that can be used to calculate the predicted probability of response in individual patients. A rapid estimate can be obtained from the provided flowcharts.


We recommend starting PEG-IFN therapy in patients with the highest chance of achieving sustained response (Table 2). We arbitrarily chose those with a predicted probability of sustained response of at least 30% to be good candidates for PEG-IFN therapy. About 25% of patients included in this study had a predicted probability of sustained response above this level. This includes all HBV genotype A–infected patients, except for those with low ALT and high HBV-DNA levels. In addition, genotypes B- and C–infected patients with high ALT and low HBV-DNA levels have a high likelihood of response to PEG-IFN. All remaining patients are moderate candidates for PEG-IFN except for those with genotype D, who have a rather low chance of achieving sustained response and are in our view generally not candidates for treatment with PEG-IFN. It should be noted that Table 2 provides recommendations for patient groups, not individual patients. In selected patients, the given response rate in Table 2 thus may differ slightly from the more accurate predicted probability of response obtained from the nomograms.

With the licensing of PEG-IFN and an additional 5 nucleos(t)ide analogues for the treatment of chronic hepatitis B in past years, choice of antiviral therapy has become more important and more complex at the same time. Because both treatment with IFN-based therapy and nucleos(t)ide analogue therapy have proven effective and can improve long-term outcome, the pros and cons of these drugs as well as patient-specific characteristics should be taken into consideration. All of the major practice guidelines have advocated IFN-based therapy as potential first-line therapy for both HBeAg-positive and HBeAg-negative patients,3, 27, 33, 34 particularly because sustained response and HBsAg loss seem to occur more often with IFN and PEG-IFN than with the direct antiviral agents.20 To reduce the risk of relapse, nucleos(t)ide analogue therapy can be extended for several months after HBeAg seroconversion because this reduces relapse rates.35, 36 HBeAg seroconversion was sustained in 86% of patients treated with telbivudine or lamivudine who discontinued therapy after at least 6 months of maintenance therapy.37, 38 Whether these responses also can be sustained in the long term, however, still is unknown.


However, the use of PEG-IFN currently accounts for no more than 10% of all prescriptions for the treatment of chronic hepatitis B.39 The relatively low use of PEG-IFN may be explained by its significant side effects and need for administration by injection. Furthermore, recommendations on the use of PEG-IFN in specific subsets of patients who are most likely to have a sustained response and HBsAg seroconversion were lacking. When we are able to identify patients with a high likelihood of response to PEG-IFN, the proportion of patients achieving sustained response after treatment with this drug probably can be increased.


Most studies investigating IFN-based therapy in HBeAg-positive chronic hepatitis B found that high baseline ALT level, low baseline HBV-DNA level, and HBV genotype A or B were associated with response.10, 25, 26, 40 In addition to these factors, we identified sex and age as predictors of response to PEG-IFN. It should be mentioned that in both studies more men than women were included. We found that the influence of sex, age, HBV DNA, and previous IFN therapy was significantly different across HBV genotypes. HBV genotype thus has great influence on the outcome of PEG-IFN therapy. Therefore, contrary to a statement on this topic in the newest guidelines from the European Association for the Study of the Liver,27 we believe that determination of HBV genotype is essential in all patients in whom sustained off-treatment response is pursued. Other potential approaches to tailor PEG-IFN therapy in chronic hepatitis B include quantification of serum HBeAg and HBsAg.41 These approaches still are being validated and are not routinely available to most physicians. Because of limited availability in clinical practice, we also chose not to include liver histology.


Previously we presented a model based on 266 HBeAg-positive patients participating in a single randomized trial.42 However, a vast majority of these patients were infected with HBV genotype A or D; only a small proportion of patients harbored HBV genotype B or C. To gain a good prediction model for all HBeAg-positive patients, we now combined the data of the 2 largest randomized trials investigating PEG-IFN in HBeAg-positive chronic hepatitis B. We showed that a model based on readily available baseline factors can provide an adequate prediction of sustained response. Ideally, a large confirmatory group would have been used for external validation. Unfortunately, such a group is not available. Clinical trials that currently still are ongoing may allow for further validation of the model in the near future.


Because substantial viral replication may persist despite HBeAg loss in some patients, a combined end point of HBeAg clearance from serum and low HBV-DNA level is crucial in HBeAg-positive chronic hepatitis B. Particularly, patients with HBV genotype non-A infection can develop mutations in the precore or core promoter region and still may be at risk for progressive liver disease despite HBeAg loss.6, 43 Both clearance of HBeAg and suppression of HBV replication are key events in the natural course and during antiviral therapy in HBeAg-positive chronic hepatitis B. HBeAg loss after IFN-based therapy was associated with reduced progression to cirrhosis and HCC, and improved survival.5, 44 In addition, large population studies have established a clear link between HBV viremia and the risk for HBV-related complications.45, 46 Serum HBV DNA was the strongest predictor of progression to cirrhosis and HCC, with a significantly higher risk for patients with an HBV-DNA level greater than 10,000 copies/mL (2000 IU/mL) as compared with those with a serum HBV-DNA level less than 300 copies/mL (relative risk, 2.5; 95% CI, 1.6–3.8; and relative risk, 2.3; 95% CI, 1.1–4.9 for developing cirrhosis and HCC, respectively). Although the proportion of patients with undetectable HBV DNA was relatively low shortly after PEG-IFN therapy,10 it further increased with prolonged duration of follow-up evaluation.6 Because the presence of anti-HBe at 6 months posttreatment was not associated with long-term sustainability of response to PEG-IFN,6 the combined end point of HBeAg loss and low HBV DNA seems optimal.


The parties involved in this study agreed not to perform a direct comparison between the 2 formulations of PEG-IFN. However, the previously reported results of 2 studies included in this retrospective analysis were very similar.10, 11 Unfortunately, we cannot provide recommendations for HBeAg-negative patients because we only had data of HBeAg-positive patients treated with PEG-IFN. Because of the low sustained response rate in HBeAg-negative patients,47 PEG-IFN is given relatively less often to HBeAg-negative as compared with HBeAg-positive patients. Prediction of response to PEG-IFN therefore seems of greater importance in HBeAg-positive than in HBeAg-negative chronic hepatitis B.


In conclusion, we provide a practical tool to calculate the probability of sustained response to PEG-IFN in individual HBeAg-positive patients, which easily can be used in clinical practice and thus can allow for the selection of the optimal candidates for PEG-IFN therapy. Unfortunately, we were not able to perform external validation because such a database is not available. Clearly, this should be performed when an appropriate patient group is available. We recommend consideration of PEG-IFN therapy in all genotype A patients with either high ALT or low HBV-DNA levels, and in genotypes B– and C–infected patients with both high ALT and low HBV-DNA levels. HBeAg-positive genotype D–infected patients are generally not good candidates for treatment with PEG-IFN.

Results


Of 808 patients eligible for participation in this study, 87 were excluded because of missing values (n = 76) or infection with HBV genotype other than A–D (n = 11), leaving 721 patients for analysis. HBeAg loss, HBeAg seroconversion, and HBV-DNA level less than 10,000 copies/mL (<2.0 × 103 IU/mL) were observed in 254 (35.2%), 232 (32.2%), and 174 (24.1%) of 721 patients, respectively. Sustained response, defined as HBeAg loss and HBV-DNA level less than 10,000 copies/mL (<2.0 × 103 IU/mL) at 6 months posttreatment, was observed in 158 of 721 patients (21.9%). Sustained response was observed in 22.4% of patients treated with PEG-IFN alone and in 21.4% of those treated with PEG-IFN and lamivudine combination therapy (P = .73). Sustained response was observed in 37% of patients with genotype A, 25% with genotype B, 20% with genotype C, and 8% with genotype D.


There were no differences in baseline characteristics between patients enrolled and those excluded from participation in this study, except for a lower rate of previous IFN therapy among the participants than the excluded patients (14% vs 24%; P = .01). Baseline characteristics of patients with and without sustained response are given in Table 1. Patients with sustained response were older, more often were female, had higher baseline ALT and lower HBV-DNA levels, and were more likely to have genotype A but less likely to have genotype D infection compared with those without sustained response. The proportion of patients who previously were treated with IFN or lamivudine therapy did not differ between patients with sustained response and those without.

Predictors of Sustained Response


Factors associated with an increased likelihood of sustained response included HBV genotype A infection, high baseline ALT level, low baseline HBV-DNA level, female sex, and older age (Table 1). There was no association between sustained response and previous treatment with IFN or lamivudine on univariate analysis. By using multivariate analysis, high baseline ALT was found to be an independent predictor of sustained response (OR, 1.57 per 1 log10 × ULN increase; 95% CI, 1.19–2.09; P = .002). In addition, HBV genotype was associated with sustained response, with higher rates of sustained response in patients with genotype A (OR, 1; reference) than B (ORB vs ORA, 0.46; 95% CI, 0.21–0.99; P = .05), C (ORC vs ORA, 0.30; 95% CI, 0.16–0.59; P < .001), or D (ORD vs ORA, 0.08; 95% CI, 0.02–0.31; P < .001). The influence of sex, age, HBV DNA, and previous IFN therapy was significantly different across HBV genotypes (P < .02 for the interaction between HBV genotype and each of these factors). These variables therefore also were included in the model. We here describe the most important predictive factors. Genotype C– and D–infected females had a significantly higher chance of sustained response compared with males (OR, 2.78; 95% CI, 0.51–5.11; and OR, 7.69; 95% CI, 1.48–39.90; P < .02). Older age was associated with a significantly higher chance of sustained response in genotype A–infected patients (OR, 1.04 per year increase in age; 95% CI, 1.01–1.08; P = .01). High baseline HBV-DNA level was associated with a lower likelihood of sustained response in patients with genotypes A (OR, 0.57; 95% CI, 0.40–0.82; P = .003) and C (OR, 0.77; 95% CI, 0.65–0.91; P = .002). Previous IFN therapy resulted in a significantly lower chance of sustained response in patients with genotype A or D (OR, 0.21; 95% CI, 0.07–0.58; P = .003). We found no differences in the predictors of response for the 2 treatment groups.

Performance of the Model


The distribution of the predicted probabilities of sustained response in genotypes A–D is shown in Figure 1. The agreement between the predicted probabilities and the observed frequency of sustained response was good (P = .27 by the Hosmer–Lemeshow goodness-of-fit test). A multivariable model including the variables of age, sex, ALT level, HBV-DNA level, HBV genotype, and previous IFN therapy had adequate discriminative ability as shown by an AUC of 0.72 (95% CI, 0.67–0.77). The AUC was 0.75 (95% CI, 0.65–0.85), 0.65 (95% CI, 0.55–0.75), 0.68 (95% CI, 0.61–0.75), and 0.78 (95% CI, 0.65–0.92) for genotypes A–D, respectively. After bootstrap validation, the AUC was 0.69 (95% CI, 0.60–0.77). Because the influence of the predictors was significantly different across genotypes, a validated formula for the prediction of sustained response was generated for each HBV genotype separately. The PEG-IFN HBV Treatment Index is based on these formulas (Figure 2). An automated calculator can be found at www.liver-gi.nl/peg-ifn.

Application of the Model in Clinical Practice


To allow for application of the model in clinical practice, a nomogram for IFN-naive patients was generated from the validated formula for each of the HBV genotypes separately (Figure 2). These nomograms can be used for calculating the probability of sustained response in individual HBeAg-positive patients based on their sex, age, and ALT and HBV-DNA levels. Average response rates based on the presence of low (<2 × ULN) or high ALT levels (≥2 × ULN), and low (<9 log10 copies/mL [<2.0 × 108 IU/mL]) or high HBV-DNA levels (≥9 log10 copies/mL [≥2.0 × 108 IU/mL]) are shown in Figure 3.

Materials and Methods


Patients and Study Design


Individual data of 542 patients treated with PEG-IFN–alfa-2a 180 μg/wk for 48 weeks (271 patients with added lamivudine 100 mg/day) and 266 patients treated with PEG-IFN–alfa-2b 100 μg/wk for 52 weeks (130 patients with added lamivudine 100 mg/day) were analyzed.10, 11 Posttreatment follow-up evaluation lasted 6 months. Addition of lamivudine did not influence response rates at the end of the follow-up period (6 months posttreatment) in any way. For the current study, sustained response was defined as clearance of HBeAg from serum and HBV DNA less than 10,000 copies/mL (2000 IU/mL) at 6 months after treatment.


The inclusion and exclusion criteria were reported in detail previously and were similar for the 2 studies.10, 11 In short, patients were eligible if they had been HBsAg positive for at least 6 months, were HBeAg positive, had increased serum ALT levels between 1 and 10 times the upper limit of normal (ULN), had serum HBV-DNA levels greater than 1.0 × 105 copies/mL (2.0 × 104 IU/mL), and had findings on a liver biopsy within the preceding 12 months that were consistent with the presence of chronic hepatitis B. Exclusion criteria included decompensated liver disease, antiviral therapy within 6 months before randomization, viral co-infections (hepatitis C virus, hepatitis delta virus, or human immunodeficiency virus), or pre-existent neutropenia or thrombocytopenia.

Laboratory Testing


During therapy and posttreatment follow-up evaluation, all patients were monitored monthly by routine physical examination, as well as biochemical and hematologic assessments. ALT level was assessed locally and therefore expressed as times ULN. HBV DNA was assessed monthly using an in-house–developed Taqman (Applied Biosystems Inc, Foster City, CA) polymerase chain reaction assay based on the Eurohep standard (lower limit of detection, 373 copies/mL) or the Cobas Amplicor HBV Monitor Test (Roche Diagnostics, Basel, Switzerland).28 We compared the 2 HBV-DNA quantification assays and found an excellent correlation between the 2 assays (r = 0.930, P < .001). Plotting the difference against the average of the assays showed no significant correlation (Bland–Altman test; r = 0.12, P = .49), strengthening the conclusion that both assays are comparable in the dynamic range.6 HBeAg, anti-HBe, HBsAg, and anti-HBs were measured with the use of the AxSYM test (Abbott, Abbott Park, IL). HBV genotype analysis was performed by INNO-LiPA Assay (Innogenetics, Gent, Belgium).

Statistical Analysis


Statistical analysis was performed using the SPSS 15.0 program (SPSS, Inc, Chicago, IL) and the R 2.3.1 Project for Statistical Computing (Harrell's Design, HMisc and Foreign libraries).29 A P value less than .05 was considered statistically significant (all 2-tailed). We performed univariate logistic regression analysis to identify predictors of sustained response among the variables of age, sex, HBV genotype (A–D), serum HBV DNA (log10 copies/mL), ALT (ln ALT × ULN), treatment allocation (PEG-IFN monotherapy or combination therapy of PEG-IFN and lamivudine), and previous treatment with IFN or lamivudine. We used multivariable logistic regression analysis with backward stepwise selection, using a P value greater than .05 for removal of variables, to construct a multivariable linear model that provides a natural logarithm transformed prediction of sustained response. We used restricted cubic spline functions to assess the linearity of the effect of continuous variables. Interactions between variables were explored. Odds ratios (ORs) were calculated with 95% confidence intervals (95% CI). Because there were interactions between HBV genotype and other factors, ORs for HBV genotype were calculated for 33-year-old (mean age), IFN-naive men with ALT and HBV-DNA levels fixed at 2 × ULN and 1.0 × 109 copies/mL (2.0 × 108 IU/mL), respectively.


Discrimination, which is the ability to distinguish patients who will achieve sustained response from those who will not, was quantified by the area under the receiver-operating characteristic curve (AUC). An AUC of 0.5 indicates no discriminative ability at all, whereas an AUC of 1.0 indicates perfect discrimination. Internal validity was assessed with bootstrap sampling.30, 31 Two hundred bootstrap samples were drawn with replacement and with the same size as the original sample. The final prediction model was constructed by applying the penalized maximum likelihood estimation acquired from the bootstrap samples.32 Nomograms for IFN-naive patients were constructed based on the logistic regression formulas. A nomogram allows for the approximate graphic computation of sustained response with points allocated to each variable based on the logistic regression formula. To develop a simple rule for each of the genotypes independent of sex and age, the predicted probability of sustained response in different patient subgroups was calculated with the logistic regression formulas. Because sex and age also predicted sustained response, but were not included in the flowchart, age was fixed at the respective mean value of each subgroup and the mean predicted probability of sustained response for males and females was calculated. For serum ALT, a low level (<2 × ULN) was considered to be between 1 × ULN and 2 × ULN, and a high level (≥2 × ULN) was considered to be between 2 × ULN and 10 × ULN. For serum HBV DNA, a low level (<1.0 × 109 copies/mL [<2.0 × 108 IU/mL]) was considered to be between 1.0 × 107 copies/mL (2.0 × 106 IU/mL) and 1.0 × 109 copies/mL (2.0 × 108 IU/mL), and a high level (≥1.0 × 109 copies/mL [≥2.0 × 108 IU/mL]) was considered to be between 1.0 × 109 copies/mL (2.0 × 108 IU/mL) and 1.0 × 1011 copies/mL (2.0 × 1010 IU/mL). These cut-off levels were chosen because the majority of patients had ALT (95%) and HBV-DNA (80%) levels between these values. In addition, these cut-off levels generally are used in clinical practice and are recommended by international guidelines for the treatment of chronic hepatitis B.3, 27, 33, 34