3 The manufacturer's submission
The Appraisal Committee (appendix A) considered evidence submitted by the manufacturer of entecavir and a review of this submission by the Evidence Review Group (ERG; appendix B).
3.1 The manufacturer approached the initial decision problem by comparing entecavir monotherapy with interferon alfa2a and 2b, peginterferon alfa2a, lamivudine, adefovir dipivoxil and telbivudine. The population under consideration was adults with compensated liver disease and active chronic hepatitis B (that is, evidence of viral replication and active liver inflammation). The primary outcome measures outlined in the decision problem were virological response (HBV DNA), histological improvement (inflammation and fibrosis), biochemical response (for example, ALT levels), development of viral resistance and HBeAg/hepatitis B surface antigen (HBsAg) seroconversion rate. secondary outcome measures were survival and adverse affects of treatment.
3.2 The manufacturer's submission presented evidence on the clinical effectiveness of entecavir from five randomised controlled trials (RCTs) that compared entecavir with lamivudine. Three of the studies were carried out in people who had not previously received nucleoside analogue treatment. One trial compared entecavir with lamivudine in people with HBeAgpositive hepatitis B, another included only people with HBeAgnegative disease and another included a mixed group with either HBeAgpositive or HBeAgnegative chronic hepatitis B. The remaining two studies were in people with lamivudinerefractory disease; one included only people with HBeAgpositive disease and the other included people with either HBeAgpositive or HBeAgnegative chronic hepatitis B.
3.3 The results of the five RCTs (n = 2438) showed that entecavir was statistically superior to lamivudine in terms of the number of people with HBV DNA suppression, ALT normalisation and histological improvement after one year of treatment. There was no statistically significant difference between the treatments in the number of people with HBeAgpositive chronic hepatitis B achieving HBeAg seroconversion. The number of people with any adverse events or serious adverse events was similar for entecavir and lamivudine. The number of people who withdrew during the first year because of adverse events was similar for entecavir and lamivudine, except in one trial where significantly more people in the lamivudine group withdrew from the study due to adverse events. The number of deaths during treatment was low (< 1% in all groups).
3.4 There were no trials that compared all treatment options in any one population; the manufacturer therefore conducted a series of network metaanalyses but only for the nucleosidenaive populations. The models used entecavir as the baseline treatment as it was common to all analyses, and all the models assumed fixedtreatment effects. The ERG identified as a strength of the mixed treatment comparison (MTC) that it was supported by a reasonably sound systematic review process; albeit noting caveats around the ambiguity about the number of trials that were included. The ERG considered the following as weaknesses of the MTC.

Relatively few studies in some of the networks; for example, only two peginterferon alfa2a RCTs were included, one in HBeAgpositive patients and one in HBeAgnegative patients.

There was a paucity of outcome for year two treatment; the entecavir year two data were unpublished and would not have been subjected to external journal peer review that the data from the other trials included in the MTC would have undergone, and peginterferon alfa2a was omitted entirely from the network as no year two data were identified.

No definition of the criteria by which entecavir was judged to be 'significantly better' or 'equivalent' to other drugs.

No assessment, or at least discussion or reflection on the results of the MTC, and the methodology used to construct it in general.

No discussion on how the results of the MTC compared to the results of the manufacturer's systematic review of entecavir (that is, how mixed direct and indirect evidence compared with direct evidence).

No discussion or rationale was presented for use of a fixed over a randomeffects model.
3.5 The results of the metaanalyses showed that for HBeAgpositive chronic hepatitis B, entecavir had a significantly higher predicted probability of HBV DNA response than all comparators and an equivalent predicted probability of seroconversion to all comparators at 1 and 2 years. Entecavir also had a significantly higher predicted probability of ALT normalisation than lamivudine (at both 1 and 2 years) and peginterferon alfa2a (at 1 year), and was reported to be equivalent to telbivudine (at both 1 and 2 years). Entecavir had a significantly higher predicted probability of histological improvement compared to lamivudine at 1 year, and was reported to be equivalent to telbivudine. For HBeAgnegative disease, the network metaanalysis found that entecavir had a significantly higher predicted probability of HBV DNA response at 1 and 2 years compared with lamivudine and peginterferon alfa2a, and was reported to be equivalent to telbivudine at both 1 and 2 years. Entecavir had a significantly higher predicted probability of ALT normalisation compared with all comparators at 1 year, but appeared similar to comparators at 2 years. Entecavir had a significantly higher predicted probability of histological improvement compared with lamivudine at 1 year, and was reported to be equivalent to telbivudine.
3.6 The available RCTs in people with HBeAgpositive, lamivudineresistant disease were smaller so the manufacturer stated that the likelihood of no events occurring in one of the arms was much higher. The manufacturer therefore presented a 'simple' indirect comparison using lamivudine as the common reference. The results showed that entecavir and lamivudine/adefovir had similar rates of seroconversion and viral load reduction.
3.7 Headtohead studies evaluating the relative rates of genotypic resistance were not available. Similarly a formal network metaanalysis of resistance rates was deemed by the manufacturer not to be possible because the data would come from nonRCT evidence and the patient populations were too heterogeneous to be combined in such an analysis. Instead, the manufacturer presented a descriptive analysis of the rates of genotypic resistance across available nucleoside analogues taken from the literature. This showed that entecavir had a lower rate of genotypic resistance than lamivudine, telbivudine and adefovir dipivoxil at 2, 3 and 4 years, and only a slightly higher rate than adefovir dipivoxil at 1 year (adefovir dipivoxil 0%, entecavir 0.2%).
3.8 The manufacturer's submission presented an economic analysis comprising two Markov models (one for HBeAgpositive disease and one for HBeAgnegative disease). The HBeAgpositive disease model consisted of 14 health states that were defined as untreated chronic hepatitis B, spontaneous HBeAg seroconversion, HBsAg loss, resistance, flare, compensated/active cirrhosis, inactive cirrhosis, decompensated cirrhosis, hepatocellular carcinoma, liver transplantation, postliver transplantation, treated chronic hepatitis B, treatmentinduced HBeAg seroconversion and death. The HBeAgnegative disease model also differentiated between response to initial treatment and response to salvage therapy, resulting in 15 health states. The models were designed to compare entecavir with lamivudine, peginterferon alfa2a and telbivudine, and both had a lifetime horizon. The estimated treatment duration for entecavir was 2 years in the HBeAgpositive model and 5 years in the HBeAgnegative model. The estimates of efficacy used in the economic model were based on the indirect comparison.
3.9 The basecase analysis for people with HBeAgpositive disease resulted in an incremental costeffectiveness ratio (ICER) of £14,329 per additional qualityadjusted life year (QALY) gained for entecavir compared with lamivudine. A comparison of entecavir with peginterferon alfa2a resulted in an ICER of £8403 per additional QALY gained. A comparison of entecavir with telbivudine resulted in telbivudine dominating entecavir.
3.10 The basecase analysis for people with HBeAgnegative disease resulted in an ICER of £13,208 per QALY gained for entecavir compared with lamivudine. A comparison of entecavir with peginterferon alfa2a resulted in an ICER of £7511 per QALY gained and a comparison of entecavir with telbivudine resulted in an ICER of £6907 per QALY gained.
3.11 The basecase analysis for people with lamivudinerefractory disease, comparing entecavir with adefovir dipivoxil plus lamivudine, resulted in entecavir dominating.
3.12 The ERG questioned the clinical validity of some of the assumptions in the manufacturer's model, in particular the basecase treatment duration assumptions of 2 years for people with HBeAgpositive disease and 5 years for people with HBeAgnegative disease. Comparing entecavir with lamivudine, the ERG's exploratory scenario analyses found that increasing the treatment duration from 2 to 5 years for people with HBeAgpositive disease increased the ICER from £14,329 in the manufacturer's base case to £22,107 per QALY gained. Even longer treatment durations gave higher ICERs – £27,120 per QALY gained for 10 years' treatment and £30,334 per QALY gained for 20 years' treatment. The ERG noted the scenario analysis used by the manufacturer in which the assumption of a lifetime treatment duration for people with HBeAgnegative disease was used. This resulted in an ICER of £16,850 and £11,100 per QALY gained when compared with lamivudine and peginterferon alfa2a respectively, with entecavir dominating telbivudine.
3.13 The ERG also conducted exploratory scenario analyses of the HBeAgnegative model, assuming a lifetime treatment duration. In this scenario people who progressed to compensated cirrhosis continued receiving treatment unless (or until) they developed decompensated cirrhosis. The same rate of progression to decompensated cirrhosis was assumed for all alternative treatments (1.8% per year based on the estimate used for lamivudine in 'Adefovir dipivoxil and peginterferon alfa2a for the treatment of chronic hepatitis B' [NICE technology appraisal 96] – see section 6). This resulted in an ICER of £27,124 per QALY gained, when comparing entecavir with lamivudine.
3.14 The assumption that all people present for treatment in the precirrhotic state of the disease was not supported by the ERG clinical specialists. The ERG scenario analyses for people with HBeAgnegative disease assumed that 90% of people start treatment with chronic hepatitis B without cirrhosis and 10% of people start treatment with compensated cirrhosis. This produced an ICER of £34,006 per QALY gained when comparing entecavir with lamivudine. When the proportion of people presenting with cirrhosis at the start of treatment is set to 20%, the ICER increases to £42,608 per additional QALY gained.
3.15 During the consultation period for this appraisal, the manufacturer submitted revised costeffectiveness estimates for the HBeAgnegative population at the request of the Committee. This revised model considered lifetime treatment duration and assumed that treatment with entecavir continued when people progressed to compensated cirrhosis. (In the original model a 5year treatment duration was used and it was assumed that treatment was discontinued when cirrhosis developed.) A 1.8% rate of progression from compensated to decompensated cirrhosis was used. The cost of adefovir dipivoxil treatment following the development of resistance in people who had not yet developed cirrhosis was also included and this treatment was assumed to be continued when the disease progressed to active cirrhosis. This revised base case gave an ICER for entecavir versus lamivudine of £20,463 per QALY gained. A further scenario was modelled in which people who developed resistance to lamivudine after developing compensated cirrhosis were also assumed to switch to adefovir dipivoxil. This resulted in an ICER of £15,531 per QALY gained.
3.16 The Committee also requested that the assumption that all people with HBeAgnegative chronic hepatitis B are in the precirrhotic state when they start treatment should be changed to reflect NHS practice. The manufacturer presented a modified analysis for different proportions of people starting treatment after developing cirrhosis, from none to 20%. Assuming that 10% of people start treatment with compensated cirrhosis, the ICER was £24,335 per QALY gained (assuming adefovir dipivoxil treatment costs for people with resistance but only in those without cirrhosis) and £17,083 per QALY gained (with adefovir dipivoxil treatment costs for all people with resistance). Assuming that 20% of people start treatment with compensated cirrhosis, the ICER was £29,176 per QALY gained (assuming adefovir dipivoxil treatment costs for people with resistance only in those without cirrhosis) and £19,023 per QALY gained (assuming adefovir dipivoxil treatment costs for all people with resistance).
3.17 For the purposes of clarification, the Committee requested that the manufacturer provide disaggregated outcomes from their simulation analysis, including the number of events of cirrhosis and hepatocellular carcinoma in each treatment group. For comparison, observational data on the relationship between HBV DNA and the incidences of cirrhosis and hepatocellular carcinoma in people with chronic hepatitis B were also provided by the manufacturer. In general, the incidences of cirrhosis and hepatocellular carcinoma were higher in the observational studies (mixed cohorts) than in the models for both HBeAgpositive and HBeAgnegative populations. However, the manufacturer suggested that this was not unexpected, given that the populations in the studies were untreated.
3.18 Full details of all the evidence are in the manufacturer's submission and the ERG report.