Ledipasvir–sofosbuvir for treating chronic hepatitis C

Clinical effectiveness

3.1 The company did a systematic literature review to identify studies evaluating the clinical effectiveness of ledipasvir–sofosbuvir for treating chronic hepatitis C. It presented 10 studies of ledipasvir–sofosbuvir with and without ribavirin in adults whose chronic hepatitis C was either previously untreated or previously treated (described as treatment naive and treatment experienced in the Committee papers). The company focused its clinical effectiveness submission on 3 phase III non‑randomised controlled trials in people with HCV genotype 1 (ION‑1 [previously untreated], ION‑2 [previously treated] and ION‑3 [previously untreated]). The other 7 studies submitted by the company were phase II studies and were included as 'supportive evidence'. Of these 7 studies:

Cost effectiveness

3.19 The company submitted a Markov state‑transition model that reflected the natural history of chronic hepatitis C. It compared ledipasvir–sofosbuvir (with or without ribavirin) with the comparators defined in the final scope of the appraisal. The company's economic model had 9 states, according to disease stage and treatment response. The same model structure was used for all people irrespective of HCV genotype or treatment experience. The company used a monthly cycle length for the first 18 cycles, then 3‑monthly until year 2 and yearly thereafter. The company did the economic analysis from an NHS and personal social services perspective and chose a lifetime time horizon (from age 40 [for previously untreated HCV] or 45 [for previously treated HCV] until people reached 100 years). Costs and health effects were discounted at an annual rate of 3.5% and a half‑cycle correction was applied from year 3.

3.20 The cost effectiveness of ledipasvir–sofosbuvir was assessed in populations defined by HCV genotype, which included those with cirrhosis:

3.21 The company used patient characteristics from the HCV Research UK database to inform the mean age, the proportion with cirrhosis, and weight of the population entering the model. People entered the model in either the non‑cirrhotic or compensated cirrhosis stages of disease. People who started treatment in the non‑cirrhotic state and were cured would not become symptomatic again. However, people with cirrhosis whose HCV was cured were still at risk of progression to the decompensated cirrhosis and hepatocellular carcinoma state. Those who did not clear the virus after treatment remained in their respective health states, or progressed to more severe stages of chronic HCV. All people in the decompensated cirrhosis health state were assumed to be candidates for liver transplant. The company chose transition probabilities for disease progression from several publications used in recent NICE technology appraisals of treatments for HCV (Cardoso et al. 2010; Fattovich et al. 1997; Grishchenko et al. 2009; Hartwell et al. 2011; Shepherd et al. 2007; Siebert et al. 2005; Thompson et al. 2008). Transition probabilities from the non‑cirrhotic to cirrhotic health state varied according to age and genotype according to published literature. Age- and sex‑specific general population mortality was also applied to each health state in the company's model.

3.22 Treatment-effect data for ledipasvir–sofosbuvir were based on SVR12 from the relevant ION studies and ELECTRON‑2 (except for the previously treated genotype 3 HCV population for whom only SVR4 data were available). Treatment‑effect data for the comparators were taken from publications or the summary of product characteristics. Because there was no mixed treatment comparison (see section 3.18), the estimates of the relative effectiveness of ledipasvir–sofosbuvir with the comparators were based on naive indirect comparisons. In its base‑case analysis the company used the ledipasvir–sofosbuvir data from the genotype 1 HCV population for the analysis of the genotype 4 HCV population because:

3.23 Resource use and costs in the company's economic model included those for treatment (drug and administration), monitoring during treatment, adverse events and for each health state (that is, monitoring of people after treatment has stopped). Drug costs were based on the list prices in the 'British national formulary' (BNF; August 2014). Monitoring costs were based on NHS reference costs, published literature or the company's clinical expert opinion (if a published source was unavailable). An additional cost for an 'initial patient evaluation' was included for people with previously untreated HCV, but the monitoring requirements for people having interferon‑containing treatments or interferon‑free treatments did not differ. Treatment durations were also used to estimate drug and monitoring costs, and the proportion of people on a given ledipasvir–sofosbuvir treatment duration was generally based on the company's clinical expert opinion. Costs for each of the health states in the company's economic model were taken from the published literature (Grishchenko et al. 2009; Longworth et al. 2014; Wright et al. 2006) and inflated to 2012–13 prices. Adverse event costs were taken from the BNF and NHS reference costs. The company assumed that the cost of each adverse event also depends on whether the event is actively treated in an outpatient setting, by a hospital registrar or specialist.

3.24 To estimate the health‑related quality of life, the company used EQ‑5D utility values from Wright et al. (2006) that were based on a UK trial of mild chronic hepatitis C. For people who had an SVR, the company's economic model included a utility benefit of 0.04 taken from Vera‑Llonch et al. (2013). The company's economic model also captured the health‑related quality of life of people while on treatment (independent of whether they had cirrhosis or not). The company assumed that the health‑related quality of life of people treated with ledipasvir–sofosbuvir without ribavirin did not change while on treatment, but reduced in people having treatments that included ribavirin or interferon. The company stated that these 'on treatment' decrements were assumed to include any effect on health‑related quality of life from treatment‑related adverse events.

3.25 The company's deterministic cost‑effectiveness results for ledipasvir–sofosbuvir compared with the comparators for each population and ordered by cost are given in tables 4–8.

ERG comments on the clinical effectiveness

3.32 The ERG considered that it was unlikely that any studies of ledipasvir–sofosbuvir relevant to this appraisal were missed. However, the ERG noted that there were gaps in the reporting of the searches done by the company, and the company's searches for comparator evidence and adverse events were not systematic.

3.33 The ERG stated that although the 3 phase III studies were open‑label, they were generally at low risk of bias. It commented that the phase II studies had small sample sizes but provided data consistent with the phase III trials. The ERG stated, however, that subjective health‑related quality‑of‑life outcomes were subject to bias.

3.34 The ERG noted that historical controls were used because there was no control arm. It commented that there are limitations with using historical controls, particularly when there are changes in the definition of, or diagnostic methods used to detect, the condition under consideration. However, the ERG stated this was unlikely to be an issue for hepatitis C, and that their clinical expert advised that the use of historical controls in this context was considered to be reasonable.

3.35 The ERG's clinical experts suggested that the diagnostic criteria used for the disease and the SVR outcomes used in the studies were representative of clinical practice in England. The ERG commented that using SVR12 was appropriate because there is a high correlation between SVR12 and SVR24. However, it noted that SVR4 is not a suitable surrogate end point for cure because there is a chance of relapse between 4 and 12 weeks.

3.36 The ERG highlighted that the results of the company's subgroup analyses for factors not stratified at randomisation were potentially biased because the respective subgroups may not be well‑balanced across treatment arms.

3.37 The ERG commented that ION‑1 (people with cirrhosis) and ION‑3 could create a network of evidence, although without any links to the comparator treatments. The ERG considered it would have been useful for the company to:

3.38 The ERG commented that the efficacy of ledipasvir–sofosbuvir does not appear to depend on the patient characteristics pre‑specified in the company's subgroup analyses. However, given that several patient characteristics were pre‑specified, it seems reasonable to assume that these characteristics may affect the efficacy of some comparators. The ERG concluded that in a given study, SVR rates for comparator treatments are much more likely to vary compared with SVR rates for ledipasvir–sofosbuvir (that is, using SVR rates from a single comparator study introduces more uncertainty than using SVR rates for ledipasvir–sofosbuvir from a single study).

3.39 The ERG's clinical experts stated that HCV genotype 1 subtype 1a, baseline viral load and IL28B CC genotype had less effect on response to treatment with ledipasvir–sofosbuvir compared with current treatment options. The ERG's clinical experts advised that there was unlikely to be any meaningful differences in baseline characteristics between the populations of the ledipasvir–sofosbuvir and comparator studies that would significantly affect outcomes. The ERG concluded that although baseline characteristics appear similar between intervention and comparator trials, the possibility that other factors differed across trials cannot be ruled out (see section 3.47).

ERG comments on the cost effectiveness

3.40 The ERG considered that the company's model structure was broadly appropriate and in line with previous economic analyses of treatments for hepatitis C.

3.41 The ERG's clinical experts noted that people with genotype 3 HCV and cirrhosis have active treatment because of disease severity. Therefore, it considered that the company's exclusion of 'no treatment' as a comparator for these populations was appropriate.

3.42 The ERG stated that boceprevir and telaprevir do not have a UK marketing authorisation for treating genotype 4 HCV and should not be considered as comparators for this population.

3.43 The ERG commented that it was unclear whether the baseline proportion of people with cirrhosis used in the company's model reflects the HCV population in England.

3.44 The ERG commented that the details about how transition probabilities from the non‑cirrhotic to the compensated cirrhosis health states had been estimated were insufficient for the ERG to critique the robustness of the approach. The ERG highlighted that the transition probabilities from compensated or decompensated cirrhosis to hepatocellular carcinoma were considerably higher for this appraisal (0.0631; Cardoso et al. 2010) than those used by the same company in its economic model for NICE's technology appraisal guidance on sofosbuvir for treating chronic hepatitis C (0.014; Fattovich et al. 1997).

3.45 The ERG commented that the company's assumption that people cannot die or have disease progression until 12–24 weeks after completing treatment lacks credibility. However, the ERG acknowledged that the size of bias was likely to be small and would favour treatment options given over longer treatment durations.

3.46 The ERG highlighted that using SVR4 data for the genotype 3 HCV population is likely to overestimate the effectiveness of ledipasvir–sofosbuvir (see sections 3.22 and 3.35).

3.47 The ERG noted that the selection criteria for comparator SVR rates in the company's economic model were not clear, and that the company's submission did not indicate the range of SVR estimates possible for the comparators. The ERG stated that because the SVR rates for comparators were estimated from single studies, rather than a meta‑analysis of all relevant studies, it was not clear whether they were conservative or optimistic rates. The ERG commented that using naive indirect comparisons breaks randomisation and fails to fully reflect uncertainty around the SVR rates. The ERG concluded that the cost‑effectiveness results may be biased by the selection of individual studies and confounded by the effect of other factors such as differences in study design, patient characteristics and trial protocols.

3.48 The ERG stated that the results should be interpreted with caution when using the company's weighted‑average approach (that is, the company included more than 1 treatment duration of ledipasvir–sofosbuvir within its base‑case analysis for some populations listed in section 3.20 [for example, previously untreated genotype 1 HCV] and therefore the efficacy and cost inputs used in the model depended on the assumed proportion of people receiving each treatment duration). The ERG also noted that the company's submission did not explicitly state the treatment durations estimated from the company's weighted‑average approach. The ERG concluded that using cost‑effectiveness results dependent on the company's weighted‑average approach may result in the recommendation of some options that represent an inefficient use of NHS resources, particularly when taking into account that there may be clear clinical reasons why specified treatment durations of ledipasvir–sofosbuvir should be considered for specific subgroups of people.

3.49 The ERG highlighted that for people with previously untreated genotype 1 HCV without cirrhosis, the company used a 79% to 21% split between the 8‑week and 12‑week ledipasvir–sofosbuvir treatments. This was because data from the HCV Research UK database showed that 79% of this population had a pre‑treatment viral load less than 6 million IU/ml (based on the company's post‑hoc subgroup analysis, the company stated this group is likely to have 8 weeks' treatment; see section 3.6). However, the ERG considered that this criterion is not consistent with the recommendations for this treatment indication in the ledipasvir–sofosbuvir marketing authorisation and is based on a post‑hoc analysis of the ION‑3 study.

3.50 The ERG considered that the company did not sufficiently explain how choices were made in the selection of costs and utility values used in its economic model, nor did the company specify the source used for resource use estimates.

3.51 The ERG commented that the publication (Vera‑Llonch et al. 2013) used by the company to reflect the utility gain associated with achieving SVR was derived using a US EQ‑5D tariff of 0.04. It suggested that the utility gain associated with an SVR taken from Wright et al. (2006), which reflects the preferences of the general public in England (that is, using a UK EQ‑5D tariff of 0.05) would be more appropriate. The ERG stated that the 'on treatment' decrements used by the company were applied to the entire cycle in the 'on treatment' health state rather than for the duration of treatment, but the ERG noted that the effect of this bias was likely to be small.

3.52 The ERG highlighted that health effects on people with HCV (that is, potential for reinfection) and health effects between people (that is, transmission of HCV) were excluded from the company's model. The ERG explained that:

3.53 The ERG highlighted that the company's base‑case analysis uses point estimates of parameters (that is, a deterministic approach) rather than the expectation of the mean (that is, a probabilistic approach). However, the ERG considered that the results from the deterministic analyses and probabilistic analyses were similar in the company's economic model.

3.54 The ERG considered that the results of the company's probabilistic analyses were limited because of:

3.55 The ERG presented ICERs for several exploratory analyses, using deterministic analyses because of the computation time and complexity associated with running probabilistic analyses. The ERG did the following additional analyses:

Equality

3.56 The company stated that people with genotype 4 HCV have a particularly high unmet need, and that minority ethnic groups have a higher proportion of people who have genotype 4 HCV than people who have genotypes 1 or 3 HCV in the UK. The company provided NICE with evidence from an HCV genotype surveillance report commissioned by the company to be produced by Public Health England. This shows the proportion of people of white or white British family origin with genotype 1 or 3 HCV as 81% and 72% respectively, whereas minority ethnic groups represent 8% and 18% respectively. The proportion of people with genotype 4 HCV of white or white British family origin is 44%, whereas minority ethnic groups represent 39% (see table 13). Additionally, the company presented commercial‑in‑confidence evidence that a disproportionate number of people with HIV co-infection have genotype 4 HCV compared with people without HIV co‑infection.

Company's additional evidence

3.57 The company provided additional evidence in response to consultation. The company focused its response on:

3.58 The company stated that the marketing authorisation allows some people with previously treated genotype 1 or 4 HCV with cirrhosis to have 12 weeks (rather than 24 weeks) of treatment with ledipasvir–sofosbuvir. It explained that these people are those 'deemed at low risk for clinical disease progression and who have subsequent retreatment options'. The company considered that the European Medicines Agency included this criterion so 'individual clinical judgement could determine those patients whose HCV would benefit from 12 weeks' treatment with ledipasvir–sofosbuvir'. However, the company sought advice from its clinical advisors to define 'low risk of clinical disease progression' and 'subsequent retreatment options' that could be used to identify these people in clinical practice. The company stated that 12 weeks' treatment with ledipasvir–sofosbuvir should be considered if all the following criteria are met:

3.59 The company presented pooled SVR12 data for people with previously treated genotype 1 HCV with cirrhosis and a platelet count of 75,000/mm³ or more from the ION studies. The company noted that people with a history of HCV‑related decompensation episodes or who were previously treated with an NS5A inhibitor were excluded from the ION studies, and information on people with features of portal hypertension was not collected. The SVR12 rate for 12 weeks' ledipasvir–sofosbuvir was 96% (197 out of 206 people). The company presented an analysis for people with previously treated genotype 1 or 4 HCV with cirrhosis for 12 weeks' ledipasvir–sofosbuvir treatment 'deemed at low risk of clinical disease progression and who have subsequent retreatment options'. This analysis incorporated the approach taken in the ERG's exploratory analyses as preferred by the Committee (see section 4.11), and used the pooled SVR12 data from the ION studies (see table 14).

ERG comments on company's additional evidence

3.62 The ERG's clinical advisers suggested that there was no clear definition that could be used to identify people with previously treated genotype 1 or 4 HCV with cirrhosis deemed 'at low risk for clinical disease progression and who have subsequent retreatment options' in clinical practice. Therefore, the ERG urged caution in interpreting the company's ICERs for this subgroup.

3.63 The ERG was concerned about how people with previously untreated genotype 3 HCV could be considered as 'intolerant' to or 'ineligible' for interferon therapy if they had no prior treatment.

3.64 The ERG was concerned that the company had not included 'no treatment' as an option in the previously treated genotype 3 HCV analyses. The ERG commented that when including 'no treatment' in the analysis of previously treated genotype 3 HCV that was unsuitable for interferon therapy in people with cirrhosis, sofosbuvir plus ribavirin was extendedly dominated (that is, a combination of 2 of its comparators provides equal health at a reduced cost). The ERG's ICER for ledipasvir–sofosbuvir in a fully incremental analysis ranged from £19,668 per QALY gained (using Cardoso) to £33,148 per QALY gained (using Fattovich) compared with 'no treatment'.

3.65 Full details of all the evidence are available.