Ombitasvir–paritaprevir–ritonavir with or without dasabuvir for treating chronic hepatitis C

The company submitted 2 completed and 4 ongoing clinical trials as supporting evidence:

Although recommended in the marketing authorisation, 2D plus ribavirin for 24 weeks was not studied for genotype 4 HCV with cirrhosis. The European public assessment report states that data from PEARL I demonstrated that this treatment was efficacious for genotype 1b HCV with cirrhosis. Because the in vitro effects and pharmacodynamics for both components of 2D (paritaprevir and ombitasvir) are similar for genotype 1b and genotype 4 HCV, the report concluded that 24‑week treatment with 2D plus ribavirin for genotype 4 HCV with cirrhosis was likely to be as efficacious as for genotype 1b HCV with cirrhosis.

The primary outcome in all the included trials was sustained virological response at week 12 (SVR12), defined as an HCV RNA level of less than 25 IU per millilitre at 12 weeks after treatment ends. All the completed trials except PEARL I (genotype 4 HCV) planned a comparison with the historical control, telaprevir. Analyses were based on the intention‑to‑treat population (all people who were randomised) or the modified intention‑to‑treat population (all people who were randomised and had at least 1 dose of study treatment).

The results of trials of 3D and 2D, with or without ribavirin, in which treatment matched that specified in the marketing authorisation, and the results of trials included in the company's economic model, are presented in table 4.

The company presented 3 meta‑analyses in which it pooled SVR12 rates from single‑arm trials evaluating 3D for genotype 1 HCV. Analyses were done on the following groups using a random‑effect model:

• all active treatment groups in completed phase 3 clinical trials (SAPPHIRE I, SAPPHIRE II, PEARL II, PEARL III, PEARL IV and TURQUOISE II) plus 1 phase 2 study, M14–103

• all treatment groups in the completed phase 3 trials in line with the marketing authorisation for 3D and

• all active treatment groups in the clinical trial programme for genotype 1 HCV, including from the dose‑finding AVIATOR study, and interim results from 2 ongoing trials, TURQUOISE I and CORAL I.
  
  The pooled SVR12 rate from the meta‑analysis for the 3D treatments recommended in the marketing authorisation was 96.5%.

The company stated that a network meta‑analysis to generate relative estimates of efficacy for 3D and 2D compared with the comparators outlined in the final scope issued by NICE was not feasible.

The completed trials also reported data on health‑related quality of life. This was measured using the SF‑36 physical component score and mental component score; the EQ‑5D‑5L health index score and visual analogue score; and the HCV‑PRO (a patient‑reported outcome tool specific to chronic hepatitis C, which consists of 16 items focusing on physical health, emotional health, productivity, social interactions, intimacy and perception).

Results for health‑related quality of life were reported as the mean change from baseline to the last treatment visit and to 12 weeks after treatment ends. In general, no statistically significant differences in the mean change over either of these periods were seen between treatment groups in most of the trials for most of the patient‑reported outcomes.

The EQ‑5D‑5L health index scores from the trials were used to inform the on‑treatment utility values in the economic model. The EQ‑5D‑5L health index scores were obtained using country‑specific algorithms to map the 5L values to the 3L tariff scores. The US mapping algorithm to convert the 5L values to 3L was used when an individual country‑specific algorithm was not available. The EQ‑5D‑5L scores are academic in confidence and cannot be reported here.

The company presented data on adverse events from the 6 completed trials evaluating 3D and the trial evaluating 2D. The most frequently reported adverse events were fatigue, headache, nausea, pruritus, insomnia, irritability, diarrhoea, anaemia, asthenia, shortness of breath, cough, muscle ache, itching and rash. The proportion of people who had at least 1 adverse event ranged from 67% (for 3D in genotype 1b HCV in PEARL III) to 92% (for 3D plus ribavirin in genotype 1a HCV in PEARL IV). Generally higher rates of adverse events were seen in the groups who had longer treatment and those who had ribavirin. The proportion of people stopping treatment because of adverse events was consistently low across the trials and the highest dropout rate was seen in TURQUOISE II, in people with compensated cirrhosis (2.3% in the 24‑week arm and 1.9% in the 12‑week arm).

The company submitted a Markov state transition model estimating the cost effectiveness of 3D and 2D for people with genotype 1 or 4 HCV. The structure of the model was adapted from the model used in NICE's technology appraisal guidance on peginterferon alfa and ribavirin for the treatment of mild chronic hepatitis C and peginterferon alfa and ribavirin for the treatment of chronic hepatitis C. The model simulated the lifetime disease progression of people with chronic HCV infection. The model adopted a lifetime time horizon (70 years) and a cycle length of 1 year. The model applied half‑year cycle corrections. Costs and health effects were discounted at an annual rate of 3.5%. The perspective of the analysis on costs was that of the NHS and personal social services.

The model had 6 health states simulating progressive liver disease, 3 health states simulating recovery from HCV (for people who had an SVR), and 1 death state.

People entered the model in one of the following health states: mild chronic HCV, moderate chronic HCV, or compensated cirrhosis. They had treatment in the first year of the model. If they had an SVR, people moved to one of the recovery states, which depended on the previous state in which they had treatment. Once in a recovery state, the disease could not progress further. However, reinfection with chronic hepatitis C was possible, with a constant risk across the time horizon. People who did not have an SVR could stay in the same state, or move through the states simulating progressive liver disease (from mild to moderate to compensated cirrhosis, depending on their previous state and the rate of fibrosis progression). From compensated cirrhosis, the disease could progress to decompensated cirrhosis or hepatocellular carcinoma. From decompensated cirrhosis, the person could develop hepatocellular carcinoma, or have a liver transplant. From hepatocellular carcinoma, the person could have a liver transplant. People in the model risked dying at any time, but those with decompensated cirrhosis, hepatocellular carcinoma, and those who had a liver transplant had an additional risk of death from liver disease.

In its original analyses, the company modelled previously untreated and previously treated HCV separately. These groups were further divided by subtype of HCV (genotypes 1a or 1b). In total, the company's original base‑case analyses included 4 different populations.

After the first Appraisal Committee meeting, the company presented revised base‑case analyses separately for previously treated and previously untreated HCV for each treatment regimen as specified in the summary of product characteristics. The results of the revised analyses supersede the original analyses. Therefore only the revised analyses are discussed here.

The baseline characteristics of people in the model, such as age, weight, sex and disease severity, were based on a clinical audit of people with HCV who had treatment at a liver clinic at a London teaching hospital. Overall, 70% of the modelled population were male. The average ages at baseline of people whose HCV was previously untreated and previously treated were 40 and 45 years respectively. The company modelled 3D with or without ribavirin for genotype 1 HCV and 2D with ribavirin for genotype 4 HCV, as per the marketing authorisation. It compared 3D and 2D with peginterferon alfa and ribavirin, sofosbuvir plus peginterferon alfa and ribavirin and simeprevir plus peginterferon alfa and ribavirin. In addition 3D was compared with telaprevir plus peginterferon alfa and ribavirin and boceprevir plus peginterferon alfa and ribavirin. The comparators were modelled in line with their respective marketing authorisations. The company estimated the durations of each modelled treatment from the rates at which people stopped that treatment in the respective clinical trials.

Clinical effectiveness was modelled as the probability of moving to a recovery state, which was based on the SVR12 rates reported in the clinical trials for 3D, 2D and the comparators. The company included estimates of effectiveness from separate trials without any statistical adjustments. When SVRs were available from more than 1 trial, the company pooled the results from the different trials. Because 2D was not studied in people with genotype 4 HCV with cirrhosis the SVR for this group was assumed to be 97%, as reported for the 2D 24‑week treatment in people with genotype 1b HCV in PEARL I.

The company highlighted 2 limitations with the available clinical effectiveness data for simeprevir used in the revised base‑case analyses:

• The marketing authorisation for simeprevir does not allow treatment in people with Q80K positive polymorphism (a genetic mutation) and SVRs for the Q80K negative subgroups were not available.

• The definitions of mild and moderate fibrosis in the simeprevir trials were different from the definitions used in the company's model. The company used pooled SVR from the intention‑to‑treat population in QUEST I and QUEST II and the definitions of fibrosis used in the simeprevir trials in the revised base‑case analyses.

The company assumed in the model that the natural history of genotype 1 and 4 HCV was similar, and so applied the same transition probabilities for both HCV genotypes. Data were sourced from the published literature.

For the health states in the model, the company used utility values obtained from the EQ‑5D scores collected in the UK mild hepatitis C trial and valued using the UK general population tariff (see table 5).

The utility differences associated with treatment were also accounted for in the model. On‑treatment utility decrements or gains were applied during the first year (first cycle) of the model. To estimate the on‑treatment utility difference for 3D and 2D, the company calculated the difference between the EQ‑5D‑3L score at the end of treatment and baseline. EQ‑5D‑3L scores were calculated using a UK mapping algorithm from the EQ‑5D‑5L scores collected in the trials for 3D and 2D. The utility differences associated with the comparator treatments were from other NICE technology appraisal guidance and ranged from a decrement of 0.154 (for telaprevir plus peginterferon alfa and ribavirin in people who had previous treatment) to a utility gain of 0.110 (for boceprevir plus peginterferon alfa and ribavirin in people who had previous treatment).

The company also did 2 scenario analyses around utility values. In scenario 1, the company estimated the utility gain for having an SVR from the difference between the pooled EQ‑5D values collected at baseline and at 12 weeks after treatment in people who had an SVR in the trials (instead of 0.05 used in the base case). In scenario 2, the company explored using alternative values for each health state, estimated from its trials. The company marked the alternative estimate of utility gain used in scenario 1, as well as the utility values for each health state used in scenario 2, as academic in confidence and therefore they cannot be presented here.

The company included 2 categories of resource use in the model that is; health state costs and treatment costs. The health state costs were associated with managing progressive liver disease (in people whose HCV does not respond to treatment) and surveillance after stopping treatment in people who have an SVR. The company's estimate of resource use for health states was based on 2 sources:

• A retrospective chart review of people with chronic hepatitis C that reported resource use according to disease response to treatment (SVR or non‑SVR) done in the East Midlands region of the UK (Backx et al. 2014). The company used these data to estimate costs for all 3 recovery health states and 2 disease states, moderate fibrosis and compensated cirrhosis.

• The cost for the remaining health states, that is mild fibrosis and 3 more advanced disease states, namely decompensated cirrhosis, hepatocellular carcinoma and liver transplant, were based on the models used in NICE's technology appraisal guidance on peginterferon alfa and ribavirin for the treatment of mild chronic hepatitis C and peginterferon alfa and ribavirin for the treatment of chronic hepatitis C. The costs were updated to current values using the Personal and Social Services Research Unit pay and prices inflation index.

Treatment‑related costs included drug acquisition costs and costs associated with on‑treatment monitoring for response and adverse events.

The results of the fully incremental analyses for the treatments recommended in the summary of product characteristics for different groups stratified by treatment history, as requested by the Committee, are in table 6 (using the list price for ombitasvir–paritaprevir–ritonavir with or without dasabuvir). The incremental cost‑effectiveness ratios (ICERs) for the scenario analyses are presented in table 7. Using alternative utility values estimated from the trials increased the ICERs for the 3D (genotype 1 HCV) and 2D (genotype 4 HCV) treatments modestly for most of the populations. The company commented that in the trials, EQ‑5D data at 12 weeks after treatment were collected before people knew their SVR results and therefore, did not capture the psychological and emotional benefit of being cured.

The company stated that in its opinion the marketing authorisation allows for a 12‑week treatment in some people with genotype 1a HCV with cirrhosis. The company presented separate SVRs for genotype 1a and genotype 1b HCV from TURQUOISE II. The results were further stratified for each genotype by treatment history and response to previous treatment. For genotype 1a HCV treated with a 12‑week regimen, the SVR was more than 90% for all subgroups except for people whose HCV did not respond to previous peginterferon alfa and ribavirin treatment (SVR 80%). The company stated that in its regulatory submission it proposed a 24‑week treatment only for this subgroup of people with genotype 1a HCV with cirrhosis. Results presented in the summary of product characteristics showed higher relapse rates in genotype 1a HCV treated with a 12‑week regimen than with a 24‑week regimen. In TURQUOISE II there were 13 incidences of relapse and 11 of these were in people with genotype 1a HCV treated with a 12‑week regimen.

The company did a post‑hoc analysis to identify the predictors of relapse in genotype 1a HCV treated with a 12‑week regimen and found that for people with 3 favourable baseline laboratory values (alpha fetoprotein [AFP] less than 20 ng/ml, platelets 90×10⁹/litre or more and albumin 35 g/litre or more), relapse rates were similar with the 12 and 24‑week treatments. The company noted that the summary of product characteristics acknowledged this post‑hoc analysis. On that basis, the company considered a 12‑week regimen for people with genotype 1a HCV with cirrhosis and these favourable baseline laboratory values to be within the marketing authorisation. However, the company did not provide SVR data or any economic analyses exclusively for this group.

The company also explored scenarios assuming that some people with genotype 1a HCV with cirrhosis would have treatment for 12 weeks. It assumed that everyone except those whose HCV did not respond to previous peginterferon alfa and ribavirin treatment would have a 12‑week treatment. All people with genotype 1a HCV with cirrhosis had a 24‑week treatment in the base case. To inform these analyses the company used corresponding SVRs for each population from the subgroup analyses of TURQUOISE II. The resulting ICERs for 3D plus ribavirin compared with peginterferon alfa plus ribavirin were £5,985 per quality‑adjusted life year (QALY) gained for the previously untreated HCV group and £8,812 per QALY gained for the previously treated HCV group.

The company presented probabilistic sensitivity analyses for 32 different populations. These also included the 12 populations for whom the Committee requested revised base‑case analyses. The company presented the results graphically in the form of cost‑effectiveness acceptability curves. The results showed that for a maximum acceptable ICER of £30,000 per QALY gained, 3D or 2D were the optimal treatment strategies for most of the revised base‑case population except for people with genotype 1a HCV with cirrhosis and genotype 4 HCV with cirrhosis. In these 2 populations sofosbuvir plus peginterferon alfa and ribavirin was the optimal treatment strategy.