Antiviral treatment for treatment-naïve chronic hepatitis B: systematic review and network meta-analysis of randomized controlled trials

The growing burden of chronic hepatitis B (CHB) infection poses a significant public health concern worldwide [1]. The main goal of CHB therapy is to improve survival and quality of life by preventing disease progression to cirrhosis and liver failure, and consequently hepatocellular carcinoma development. Additional goals of antiviral therapy are to prevent mother-to-child transmission, hepatitis B reactivation, and the development of hepatitis B virus-associated extrahepatic manifestations [2].

Interferon-alfa (IFN) was the first drug approved by the Food and Drug Administration (FDA) for the treatment of CHB in 1992. In 1998, lamivudine (LAM) was available as the first oral therapy for CHB [3]. However, LAM’s effectiveness has been limited because of the development of antiviral resistance [4]. Over the years, other drugs have become available including pegylated interferon-alfa (PEG-IFN), adefovir (ADV), entecavir (ETV), telbivudine (TBV), tenofovir disoproxil fumarate (TDF), and recently tenofovir alafenamide (TAF) [3, 5]. Among these drugs, in particular, IFN and PEG-IFN can produce long-term immune control without the need for long-term treatment in a small proportion of patients. For other treatments, lifelong administration may be required. Initiating therapy with these medications involves consideration of drug-specific trade-offs such as high and potentially lifelong medication costs, limited adherence, potential side effects, and the risk of antiviral resistance.

ETV, TDF, and TAF were approved for the treatment of CHB by FDA in 2005, 2008, and 2016 respectively [6]. They appeared to have a higher genetic barrier to resistance than the other nucleos(t)ides [4, 7]. Patients have also expressed the need for treatments with higher cure rates, higher barriers to resistance, better side effect profiles, and better coverage by drug plans [8].

In 2010, we conducted a systematic review and Bayesian meta-analysis to evaluate the relative efficacy of the first 12 months of CHB antiviral treatments [9]. We concluded that TDF and ETV were the most potent oral antiviral agents for hepatitis B e antigen (HBeAg)-positive patients, and TDF the most effective for HBeAg-negative patients. Nearly 10 years later, there are more data on long-term treatment and follow-up. Virological breakthrough has been observed with long-term therapy which may or may not be related to the emergence of genotypic resistance [10, 11]. New adverse events have been reported notably drug effects on renal function and bone metabolism [12]. There were also new clinical trial data for the recently approved drug, TAF, and for new drug combinations.

The objective of this review was to provide a comprehensive and contemporary look at the effectiveness of CHB antiviral treatments. Our focus was on treatment-naïve adult patients diagnosed with HBeAg-positive or HBeAg-negative CHB infection without co-infections and without decompensated cirrhosis, hepatocellular carcinoma, and liver transplantation. Patients were not further stratified by their baseline hepatitis B virus (HBV) DNA level, serum alanine aminotransferase (ALT) level, or HBV genotype even though these are recognized predictors of response to certain therapies. This was partly due to sample size considerations and partly because the review was intended to be a comprehensive overview.

We conducted a thorough exhaustive search of available literature and identified the evidence from randomized controlled trials (RCTs) investigating the comparative effectiveness among the CHB treatments (PEG-IFN, ADV, LAM, ETV, TBV, TDF, TAF as monotherapy or combination therapy) on treatment-naïve adult populations through a systematic review. We synthesized both direct and indirect evidence on efficacy for all possible comparisons using network meta-analysis (NMA).

The protocol for this systematic review was an extension of our earlier systematic review carried out under the sponsorship of the Ontario Drug Policy Research Network (ODPRN) and reported in 2015 [13].

We used network meta-analysis to integrate all available randomized trial evidence for the treatment of CHB by conducting direct and indirect comparisons among multiple treatment strategies. For each treatment within the Bayesian framework, we estimated the posterior probability of achieving a specific outcome and presented the results in the form of rankogram. However, we have to interpret this probability ranking with caution. The choice of the most desirable treatment should not be based solely on the ranking, but should take the relative effect into account [64].

Treatment recommendations depend not only on effectiveness but also on many other factors such as adverse events, potential for viral resistance, cost, patients’ preference and values, and availability in the care setting. In addition, the choice has to be made across five relevant outcomes for HBeAg-positive patients and two outcomes for HBeAg-negative patients. A treatment that is best in one outcome may not fare well in another outcome.

Across all outcomes and in both HBeAg-positive and HBeAg-negative populations, TAF emerged as the treatment with the most consistent performance (Fig. 5a, b). TAF was approved by FDA in 2016 for the treatment of CHB. It is a prodrug of tenofovir and has a better safety profile with respect to renal function and bone mineral density compared to TDF [65‐68]. It is rapidly converted to the active metabolite intracellularly. With reduced systemic exposure to the active metabolite, the off-target kidney and bone exposure are thereby reduced. It has the same resistance profile as TDF [69]. However, TAF is costly (estimated US$15,570.90 per patient annually at US average wholesale price of $42.66/25 mg tablet) [70], though one published study has suggested that TAF may nonetheless be cost effective [71].

Prior to 2017, TDF and ETV were the recommended oral drugs for CHB in the international guidelines because they have high effectiveness and a high barrier to resistance [72, 73]. With TDF, renal complications (nephropathy and Fanconi syndrome) and osteomalacia have been reported [74], but no resistance has been detected to date [7]. ETV has been associated with lactic acidosis and a low but slowly emerging resistance pattern [74, 75].

We found significant inconsistencies when studies with PEG-INF were included in the NMA with oral nucleos(t)ides. Among these studies, PEG-IFN was used in different dosages; the duration of treatment was shorter than the oral nucleos(t)ides, and PEG-INF was combined in different order with oral nucleos(t)ides in different combination therapies. This means that there were effect modifiers in the identified studies that suggested the evidence should not be combined. For the purpose of this NMA, the studies identified were too heterogeneous to combine, which manifested as inconsistency in the network. On the basis of this clinical heterogeneity, we decided to further exclude studies that included the PEG-IFN treatment.

Pegylated interferon-α (PEG-IFN) is a synthetic cytokine and is believed to act on the cell-mediated immunity [76]. PEG-IFN treatment has the benefit of finite treatment duration, a higher rate of HBeAg and HBsAg seroconversion, and no drug resistance [77]. However, it is associated with more adverse events (flu-like symptoms, neutropenia, anemia, thrombocytopenia, depression, neuropathy, and dermatological side effects) than any of the oral drugs [78]. The need for parenteral administration also adds to the low preference and compliance. For all these reasons, PEG-IFN is hardly used in clinical practice nowadays. It has also been shown that extended nucleos(t)ide therapy could result in HBsAg loss (functional cure) rates similar to those reported with PEG-IFN therapy [79]. In recent years, using regression analysis, clinicians are able to identify and select specific patients with positive predictors for sustained response to PEG-IFN and reduce unnecessary exposure for patients who are not likely to respond [80, 81]. Certain combination strategies between a nucleos(t)ide and PEG-IFN have been shown to exhibit synergistic therapeutic effect resulting in greater viral suppression and higher rates of HBeAg loss and HBsAg loss [82, 83].

In this review, we included all eligible RCTs published prior to the end of 2017. Our results are consonant with current clinical guidelines [2, 72, 84] which recommend TAF, TDF, and ETV to be the first-line treatment for CHB, and PEG-IFN for selective patient groups. These recommendations are based on evidence reviews, consensus of expert panels, and ratings on the Grading of Recommendations, Assessment, Development and Evaluation system. Our NMA provides comprehensive comparative evidence for these treatment recommendations.

Our analysis is also broadly consonant with other evidence reviews. In comparison with the NMAs published in 2010 and 2015 [9, 85], this analysis has included a new drug, TAF, and many new combination strategies. There are some differences in ranking. All three NMAs ranked TDF first for virologic suppression for HBeAg-positive patients. For ALT normalization, the two previous NMAs ranked TDF first for HBeAg-positive patients and TBV first for HBeAg-negative patients. We ranked TAF first for HBeAg-positive patients and TBV first for HBeAg-negative patients.

Our study is in agreement with the findings of another two NMAs in that TBV was superior to most other nucleos(t)ides in terms of HBeAg loss and seroconversion [86, 87]. This was until the arrival of TAF which superseded TBV.

Since our previous review in 2010, drug combinations have been investigated increasingly to delay the emergence of resistance, to treat patients with previous treatment failure [88, 89], and most importantly to increase immune control over the virus, as indicated by HBeAg loss and HBsAg loss [38, 44]. However, there are relatively few studies at present, and sample sizes are small.

Our study has a few limitations that merit discussion. Firstly, we included studies that spanned 19 years. The laboratory methods used for quantification of HBV DNA have evolved tremendously in the last decade with detection threshold as low as < 10 IU/ml (approximately 50 copies/ml). Consequently, the definition for virologic response varied widely between older studies and recent studies and ranged from < 20,000 IU/ml (100,000 copies/ml) to < 15 IU/ml (75 copies/ml). Even though we only included in our analysis those outcome data with detection limits of ≤ 200 IU/ml (1000 copies/ml), this is still an important source of variability. Secondly, in order to include as many interventions as possible to inform the network, we did not specify participants’ baseline characteristics in our eligibility criteria. This could be an important source of heterogeneity. However, we tried to reduce baseline variability by excluding studies with decompensated cirrhosis, hepatocellular carcinoma, and liver transplant cohorts. Additional file 4 showed the variation in age, gender, baseline HBV DNA level, and baseline ALT level among the included studies. Studies with more favorable patient selection criteria are expected to produce more favorable results. Viral suppression is highly dependent on the baseline HBVDNA levels and treatment duration. It has been shown that baseline HBV DNA and HBsAg levels are strong predictors of virologic response to different CHB treatments [90‐92]. ALT normalization is highly dependent on the inclusion criteria of ALT level and the presence of concomitant disease such as nonalcoholic steatohepatitis. Other baseline variables such as HBV genotypes also have correlation with treatment outcome [93, 94]. This variation in baseline clinical characteristics is a recognized source of heterogeneity in NMA [95, 96]. Thirdly, the small number of studies and patients limited the ability to conduct pairwise comparisons. This is particularly true for combination therapies. These small sample sizes produced wide credible intervals and reduced the number of closed loops in the network, especially for the HBeAg-negative population. This is also true for outcomes with rare events such as HBsAg loss, where the observed wide credible intervals indicate large uncertainty in differences between treatments. Fourthly, for some studies where information comes from fewer studies and more rare events, the estimate of between-study heterogeneity is more heavily dependent on the choice of the (vague) prior. As a consequence, the estimate of uncertainty for some of the relative effectiveness parameters is inflated. The mean estimates or relative effects, however, have remained unchanged. Lastly, given that our review was limited to studies written in English, useful studies in other languages may have been missed.

The currently available treatment options are either nucleos(t)ide-based or interferon-based or their combinations. For both hepatitis B e antigen-positive and -negative populations, tenofovir disoproxil fumarate and tenofovir alafenamide are the most effective agent for virologic suppression. Our NMA confirmed the low probabilities and overall difficulty of achieving such endpoints as HBeAg loss/seroconversion or HBsAg loss. Even when virologic suppression is achieved, relapse rate is high and long-term treatment is necessary. Research for new treatment strategies and targets is necessary to achieve a functional cure. Currently, many new potential drugs targeting directly at the genomic viral reservoirs (cccDNA and integrated viral DNA) and agents boosting innate and HBV-specific immunity are in the pipelines [97‐100]. Until these new therapies arrive, best practice for care of HBV patients should be based on current evidence, as reported here.