4 Evidence submissions
The evaluation committee (section 8) considered evidence submitted by the company for eliglustat, a review of this submission by the evidence review group (ERG) and evidence submitted by clinical experts, patient experts and NHS England.
Nature of the condition
4.1
Patient experts described how:
-
Type 1 Gaucher disease can have a profound impact on health-related quality of life.
-
Symptoms of Gaucher disease are not easily recognised and diagnosis can take a long time.
-
The disease is rare, so there is little information about it, which can lead to frustration and anxiety for people who have it.
-
The disease has an immediate impact on family life, social interactions and work.
-
There is social stigma associated with Gaucher disease because of a lack of understanding about it, and an unmet need for mental health and psychosocial support.
-
Haematological, bone and visceral symptoms are key factors affecting the health-related quality of life of people with type 1 Gaucher disease. As the disease progresses, it can cause anaemia and thrombocytopenia, which lead to fatigue, joint pain and reduced mobility. Severe disease is associated with bone damage, with an increased incidence of fragility fractures, pain and loss of self-reliance.
4.2
The patient experts reported that people with Gaucher disease face the challenge that they usually have no visible disability, except for a few older people who use a wheelchair or walking aids. This can make it difficult for them to access the care, support and services they need, such as benefits and employment support (for example, rest breaks, reduced working hours, time off for appointments and treatment).
4.3
The main treatment option is enzyme replacement therapy (ERT – imiglucerase or velaglucerase). This is given by regular intravenous infusion, which is time consuming and burdensome for patients and caregivers. Miglustat is an oral therapy, which provides an alternative for people for whom ERT is not suitable. Supportive therapy may include blood products, bisphosphonates or analgesics. NHS England and clinical experts stated that current clinical practice in England is to titrate the dose of ERT and use the lowest effective dose. The company stated that miglustat is used in a very small number of people. The clinical and patient experts noted that people with type 1 Gaucher disease choose ERT whenever possible because miglustat is associated with tolerability and safety issues, and modest efficacy. The company submission outlined that the management of Gaucher disease needs an individualised approach to treatment that takes into consideration disease manifestations, disease burden and quality-of-life needs.
Clinical evidence
4.4
The company conducted a systematic literature review and identified the following key phase 3 randomised controlled trials of eliglustat for type 1 Gaucher disease:
-
ENCORE was an open-label trial comparing eliglustat (n=106) with imiglucerase (n=54) in patients whose disease was stable with ERT. Patients had 50 mg, 100 mg or 150 mg eliglustat twice daily titrated according to trough plasma concentration, or 30–130 U/kg/month of imiglucerase. The statistical design of the ENCORE trial was to test non-inferiority in the primary composite outcome, that is, the percentage of patients who remained stable for 52 weeks in the following parameters: haemoglobin levels decreased by 1.5 g/dL or less from baseline, platelet counts decreased 25% or less from baseline, spleen volume increased 25% or less from baseline and liver volume increased 20% or less from baseline. The non-inferiority margin was 25%. The analysis was stratified by ERT dose (see table 1 for further details).
-
ENGAGE was a double-blind placebo-controlled trial comparing eliglustat (n=20) with placebo (n=20). The company submission referred to the population as being treatment naive. However, inclusion criteria allowed for patients who had previously had treatment with ERT as long as they had not had it within 9 months of recruitment to the trial. Patients in the eliglustat arm were given 50 mg on day 1; 50 mg twice daily from day 2 to week 4; and 50 mg or 100 mg twice daily from week 4 to week 39.
4.5
The company submission also included supportive information from a phase 3 trial (EDGE) and a phase 2 trial (NCT00358150)
-
EDGE was a double-blind trial that compared once daily (100 mg or 200 mg) eliglustat with twice daily (50 mg or 100 mg) eliglustat in 170 patients with type 1 Gaucher disease. The trial started with a lead-in of up to 18 months, during which time patients had eliglustat 50 mg or 100 mg twice daily for at least 4 months, until therapeutic goals were achieved. Data were only provided for the open-label lead-in phase.
-
The phase 2 trial (NCT00358150) included 26 patients who had not had ERT in the 12‑months before the study. Eliglustat was administered at 50 mg twice daily from day 1 to day 20, after which the dosage could be increased to 100 mg twice daily if trough plasma concentrations were less than 5 ng/ml. The primary outcome measure was a composite requiring improvement from baseline to week 52 in at least 2 of the 3 main efficacy parameters, which were spleen volume, haemoglobin level and platelet count.
Clinical results – ENCORE
4.6
The ENCORE study showed that 84.8% of patients on eliglustat and 93.6% on imiglucerase met the primary composite endpoint of stability at 52 weeks. Stability was maintained for 104 weeks in 87.8% of patients (n=95) having eliglustat. Further details of the primary outcome results are presented in table 1. In both treatment groups, more than 92.0% of patients had stable disease in each component of the composite endpoint.
Table 1 ENCORE study results (per protocol set*)
4.7
Of the secondary outcomes (absolute and percentage changes in haemoglobin, platelet count and organ volumes at week 52 and week 104), the difference was statistically significant between treatment groups only for absolute and percentage changes in haemoglobin levels, for which there was a larger reduction for eliglustat (−0.28, 95% CI −0.52 to −0.03, p=0.03). The company stated that this difference was not clinically meaningful because it remained within the normal range. There were small or no differences in bone-related outcomes: spine bone mineral density (0.06), lumbar spine T‑score (0.01) and Z‑score (0.0), total femur bone mineral density (0.19), and total femur T‑score (0.03) and Z‑score (0.02). Data on the Gaucher Disease Type 1 Severity Scoring System (GD‑DS3) were collected. This is the main measure used to score the severity of type 1 Gaucher disease in clinical practice in England. The range of GD‑DS3 scoring is from 0 to 19 (0 to 3 indicates borderline to mild disease; 3 to 6, moderate disease; 6 to 9, marked disease; above 9, severe disease). Scores were all below 3 indicating mild disease, and they showed no clinically important improvements, with little change from baseline to week 52.
4.8
The company also presented a post-hoc subgroup analysis according to pre-treatment with either velaglucerase alfa or imiglucerase. The company stated that the results showed that:
-
eliglustat had similar efficacy, both post-imiglucerase and post-velaglucerase alfa, with continued stability
-
haemoglobin levels showed a similar change from baseline to week 52 in both groups
-
spleen and liver volume outcomes also showed no statistically significant change from baseline in both groups.
Clinical results – ENGAGE
4.9
Eliglustat was associated with a 27.77% reduction in spleen volume from baseline, which translated to a statistically significant mean difference of 30.03% in spleen volume (the primary outcome measure) compared with the placebo group (p<0.001). This reduction in spleen volume continued through to week 78, with a mean reduction of 44.60% in the eliglustat group. Additionally, by week 78, disease in patients who started eliglustat at week 39 showed a similar response to that at week 39 in patients randomised to eliglustat at week 0.
4.10
The company submission stated that eliglustat showed efficacy compared with placebo on all secondary endpoints. At 39 weeks, there were statistically significant differences in liver volume (−6.64%, 95% CI −11.37 to −1.91; p=0.0072), haemoglobin levels (1.22 g/dL, 95% CI 0.57 to 1.88; p=0.0006) and platelet count (41.06%, 95% CI 23.95 to 58.17; p<0.0001). These results were maintained at week 78.
4.11
The GD‑DS3 scores showed no clinically important improvements at 39 weeks. The company reported that there was a clinically significant decrease in bone marrow burden scores for 5 patients in the trial, with 3 shifting from marked/severe to moderate bone marrow infiltration.
Clinical results – EDGE
4.12
The company submission presented the interim analysis for the 18‑month lead-in period only. The primary composite outcome was the proportion of patients in whom therapeutic goals were maintained or reached. It was based on measures of bone crisis, haemoglobin levels, platelet counts, and spleen and liver volumes. All 5 therapeutic goals were reached in 137 (83%) patients. The company stated that the analysis of the randomised part of the study had not been completed at the time of submission.
Clinical results – NCT00358150
4.13
For the composite primary outcome, statistically significant improvements in haemoglobin, platelet counts, and liver and spleen volumes were maintained throughout 4 years of treatment, showing long-term change from baseline with eliglustat.
Adverse events
4.14
The company presented a safety analysis that pooled data from 393 patients with type 1 Gaucher disease who had eliglustat in the clinical trial programme. The overall results of the pooled safety analysis showed that eliglustat was generally well tolerated, with few patients (3%) stopping treatment because of adverse events. Adverse events were mostly mild (78%) or moderate (44%), and were not thought to be related to eliglustat in 79% of patients. The most common events were headache (17%), joint pain (14%), nasopharyngitis (13%), upper respiratory tract infection (11%), diarrhoea (10%) and dizziness (10%).
Economic evidence
4.28
The company developed a cost–consequence analysis using a 10 health state semi-Markov model (that is, the transition probabilities used in the model depended on a patient's initial health state). The model, comparing eliglustat with imiglucerase and with velaglucerase alfa, included 2 patient groups: those who were treatment naive and those who were taking ERT and whose disease was considered clinically stable. Within each of these populations, the model also considered subgroups based on metaboliser status. The company did not present a comparison with miglustat, stating that it is used in less than 2% of patients, and is associated with issues around tolerability and efficacy. The company also stated that eliglustat is not expected to be used in place of miglustat in this small population.
4.29
The starting age of people in the treatment-naive population was assumed to be 32 years based on the mean age in the ENGAGE trial. The starting age of people in the population whose disease was stable with ERT who switched to eliglustat was assumed to be 38 years. Health states were defined by a patients' scores on the GD‑DS3 severity scoring system. In the model, people were grouped by: mild (GD‑DS3 score 0 to 3.5), moderate (3.5 to 6.5), marked (6.5 to 9.5), and severe (more than 9.5) disease. People could move between any of the living states in each cycle, remain in their current state, or move to the absorbing death state. All people with moderate, marked and severe disease were assumed to have at least 1 instance of bone or joint pain or bone crisis, based on the contribution of this domain to the overall GD‑DS3 score.
4.30
For people whose disease was stable with ERT, transition probabilities in the first year were based on the ENCORE trial and thereafter based on data from the GD‑DS3 score study, a registry validating the GD‑DS3 scoring system. The model assumed differential clinical effectiveness in the first year and then equal effectiveness in subsequent years. For the treatment-naive population, treatment effectiveness was assumed equal and based on the eliglustat arm of the ENGAGE study.
4.31
The model used a time horizon of 70 years and a cycle length of 1 year. The company stated that this was appropriate given the limited data available. The analysis was conducted from the perspective of the NHS and Personal Social Services, and costs and benefits were discounted at a rate of 3.5% per year.
4.32
Some of the assumptions used in the company's model were:
-
The treatment efficacy of eliglustat and the comparators is equal in the treatment-naive population.
-
After the trial period, the state transitions derived from GD‑DS3 Score Study data are the same for eliglustat and all the comparators.
-
After their disease becomes stable on the selected treatment, people might stop treatment for up to 3 years and switch to a different therapy. (A stopping rate of 1.9% was applied for the treatment-naive population for both eliglustat and ERT. For the population whose disease was stable with ERT, a 1.9% stopping rate was applied for eliglustat but it was assumed that patients having ERT would not stop treatment.)
-
Mortality is the same for eliglustat and ERT across all health states, and mortality rate does not increase with disease severity.
The outcomes at 39 weeks from the ENGAGE trial were used for people at 1 year in the model.
4.33
Quality-of-life data were derived from the GD‑DS3 score study, which also collected SF‑36 data. The SF‑36 scores were mapped to EQ‑5D utilities using a published algorithm. Utility decrements were applied to patients having treatment to reflect the impact of adverse events. The ERG agreed that the GD‑DS3 score study provided the most complete set of utility values. The model also incorporated preference for oral therapy over infusion therapy in the base-case analysis via a utility increment of 0.12, which was applied in every cycle. This value was taken from a vignette study that was commissioned by the company.
4.34
Costs for drug acquisition, administration, and monitoring and management were included in the model. Differential monitoring and management costs were applied to each health state, broadly increasing with severity of disease. No costs associated with adverse events were included in the model, and the company assumed that additional training of healthcare staff was not needed for administration of eliglustat or the comparators. No administration costs were included in the model for eliglustat. Table 3 presents the costs included in the model. Additionally, direct medical and social service costs were included, ranging from £2,583.05 per year for the mild health state with no clinical symptoms of bone disease to £6,411.63 for the severe health state with severe skeletal complications.
4.35
Confidential discounts were available for eliglustat, imiglucerase and velaglucerase alfa, and results incorporating the confidential prices were explored by the ERG for all analyses in a confidential appendix.
Table 3 Costs per treatment per patient per year based on the list prices
Model results
4.36
The company estimated that the lifetime benefit associated with using eliglustat in place of ERTs (driven almost entirely by the quality-of-life improvement associated with mode of administration) was 2.44 QALYs for people who had not had treatment before and 2.28 QALYs for people whose disease was stable with ERT.
4.37
The results of the incremental costs for eliglustat compared with imiglucerase and velaglucerase alfa in people whose disease was stable with ERT and those who were not having treatment at time of starting eliglustat are presented in table 4. The results are based on list prices; confidential discounts are available for eliglustat and ERT.
Table 4 Summary of incremental costs in company's base-case cost-effectiveness model
4.38
The company presented one‑way sensitivity analyses to explore uncertainty. Incremental costs were most heavily influenced by patient weight because this determined the dosing and costs of the ERT comparators. Other influential parameters were those used to model overall survival of patients, the number of doses of ERT patients were assumed to have per month and the duration over which patients could stop eliglustat. Varying the utility increment assigned to eliglustat for its more favourable administration method was the biggest driver of the difference in QALYs.
Budget impact analysis
4.39
The company presented a 5‑year budget impact model to estimate the costs of eliglustat to the NHS. It was based on estimates of total costs generated by the cost–consequence model. Some other key assumptions made by the company were:
-
Newly diagnosed patients were assumed to start treatment on eliglustat rather than imiglucerase/velaglucerase alfa.
-
Costs were based on the licensed dose of eliglustat and the dosing of ERTs used in the ENCORE clinical trial.
-
Effects of mortality and stopping treatment were included in the estimated total costs.
-
Model results for intermediate or extensive metabolisers were used (most patients in the trials).
4.40
The company stated that there was uncertainty over uptake rates, which would be driven both by clinician and patient preference, and by NHS purchasing decisions.
4.41
The budget impact calculations estimated the difference in costs over 5 years if eliglustat were to be introduced as a treatment option. The company estimated that using eliglustat would result in additional costs of £84,559 in year 1 after launch, leading to a total cost of £571,487 in year 5 (a cumulative total of £1,623,219). These results are based on the list prices for eliglustat and ERT.
Evidence review group review
4.42
The ERG highlighted 2 main concerns about the structure of the model developed by the company: the use of long-term transitions and the use of the GD‑DS3 score system to define health states. The ERG considered the company's approach to generating long-term transition probabilities to be complicated, stating that it reduced the transparency of the model, so making validation difficult. The ERG stated that, because the same transition probabilities were applied to both treatment and comparator groups, it was unclear why a simpler approach was not used. Additionally, the ERG stated that the GD‑DS3 score appeared to be insensitive to changes in disease status, so did not reflect differences between the treatments seen in the ENCORE trial. This meant that differences between the treatment and comparators were not accounted for in the model. This resulted in a bias towards equivalence in clinical benefits, so underestimating the differences between eliglustat and imiglucerase seen in the ENCORE study.
4.43
The ERG stated that assuming long-term equivalence of eliglustat and ERT underpinned the calculation of long-term benefits, and had the potential to impact on estimated incremental QALYs. The ERG considered that this assumption had not been adequately justified in the company's submission. It stated that short-term non-inferiority results in the ENCORE trial did not imply non-inferiority in the long term.
4.44
The ERG questioned whether the inclusion of a large number of health states was necessary. The ERG acknowledged that more health states can improve the accuracy of a model. However, the advantage of this approach is offset when the model has a greater complexity and reduced transparency as a result. The ERG commented that this was particularly important because data for type 1 Gaucher disease are limited.
4.45
The ERG questioned the company's assumption that eliglustat and ERT were equivalent in people who had not had previous treatment. It considered that the evidence from the ENCORE trial should have been incorporated instead.
4.46
The ERG considered that the company's assumptions about stopping treatment were reasonable given the lack of data available.
4.47
The ERG stated that mortality risk would increase with severity of disease, so disagreed with the company's assumption on mortality. The ERG explored this assumption in its analyses.
4.48
The ERG considered that the dose of eliglustat in the model was in line with practice. However, the ERG noted that the efficacy data were taken from ENCORE, in which 48% of patients had a higher dosage of eliglustat (150 mg twice daily) for most of the trial.
4.49
The ERG disagreed that there will be no administration costs associated with eliglustat because it is an oral therapy, and explored incorporating a minimum pharmacy dispensary cost. Additionally, the ERG considered that the company overestimated the administrative costs for ERT delivered at home because it was implausible that it would be higher than the cost of hospital administration.
4.50
The ERG was concerned with the costs for ERT in the model, noting that the company did not include any vial wastage. The ERG reiterated that there was considerable evidence to suggest that substantially lower doses of ERT are used in practice (see section 4.26), so the higher dose of ERT treatment assumed in the model overestimated the ERT acquisition cost. The ERG also noted that patients who had not had previous treatment in the model were assumed to have had the same dose of ERT as patients whose disease was stable. However, the clinical adviser to the ERG suggested that newly diagnosed patients are typically less severely affected than patients who start treatment in childhood and so do not need such intensive dosing.
4.51
The ERG stated that the budget impact model was linked directly to the cost–consequence model, so its concerns around the company's model were also applicable to the company's budget impact analysis. The ERG noted a number of issues with the budget impact analysis beyond those identified in the cost–consequence model. These related to:
-
The costs incorporated in the budget impact model, which were taken from the cost–consequence model, represented the average lifetime costs when allowing for mortality rather than the costs of treating the disease in 1 patient for 5 years. The ERG stated that the latter was relevant to the budget impact analysis and the company's approach underestimated total costs. With regard to stopping treatment, the ERG stated that the effects of switching were double counted because both the cost–consequence model and the budget impact analysis accounted for switching.
-
The choice of treatment for the incident population in the absence of eliglustat: the ERG suggested that it is plausible that all patients are offered velaglucerase alfa rather than some patients having imiglucerase.
-
The composition of the Gaucher population (the budget impact model excluded poor metabolisers): the ERG stated that this may have overestimated the costs of treatment with eliglustat.
ERG exploratory analyses
4.52
The ERG conducted exploratory analyses to address the uncertainties it had identified in the company's cost–consequence model. It presented its own base-case analysis with its preferred assumptions, including:
-
additional administration costs for eliglustat (£14.40 monthly dispensary cost)
-
revised administration costs for ERT treatments (home therapy cost equal to hospital cost)
-
revised estimate of the QALY benefits of oral therapy (estimate of 0.05)
-
revised modelling of mortality to allow for increased mortality risk for people with marked and severe disease
-
reduction in dose of ERT to bring it in line with UK practice (25 U/kg)
-
using ENCORE effectiveness data in the treatment-naive population during the first cycle.
4.53
The impact of the ERG's analyses, based on list prices for ERT treatments, was to reverse the company's results for intermediate and extensive metabolisers for the comparison with velaglucerase: eliglustat was no longer cost saving (see table 5). For the comparison with imiglucerase, the incremental costs estimated by the ERG were substantially higher than those estimated by the company (see table 6). The cost savings with eliglustat for poor metabolisers, based on the ERG's analyses, were substantially lower compared with imiglucerase and velaglucerase. The key driver of the change in results was the dose of ERT treatment used.
4.54
The ERG also highlighted that the QALY benefits of eliglustat compared with imiglucerase and velaglucerase alfa were reduced to around 1.05, driven by alternative assumptions about the size of the incremental benefit for oral therapy.
Table 5 ERG base-case analysis – incremental QALYs and costs (eliglustat versus velaglucerase alfa) – based on list prices of eliglustat and ERT
Table 6 ERG base-case analysis – incremental QALYs and costs (eliglustat versus imiglucerase) – based on list prices of eliglustat and ERT
4.55
The ERG presented an exploratory analysis for the budget impact analysis. This included the assumptions in section 4.54 but also assumed zero mortality and no treatment stopping. Based on these revised cost assumptions, and using list prices, the budget impact of eliglustat was estimated by the ERG at an additional cost of £11,677,472 in year 5 and £36,428,402 over 5 years. Additionally, the ERG explored the impact of assuming that 4% of eliglustat patients would be poor metabolisers, based on the proportion in the ENGAGE trial. This reduced the budget impact to £11,123,765 in year 5 and £34,701,740 over 5 years.
Additional evidence
4.56
Following consultation on the evaluation consultation document, the company provided 4‑year data from ENCORE. Data for mean haemoglobin concentration, platelet count, and spleen and liver volumes remained stable for up to 4 years. Year to year, all 4 measures remained collectively stable in 85% or more of patients, as well as individually in 92% or more. Mean bone mineral density Z‑scores (lumbar spine and femur) remained stable and were in the healthy reference range.
4.57
The company provided results from Ibrahim et al. (2016), based on an indirect comparison of eliglustat with imiglucerase. Data for eliglustat was from a phase 2 study and ENGAGE. Data for imiglucerase was from a cohort of patients who had not had treatment, with comparable baseline haematological and visceral parameters from the International Collaborative Gaucher Group Registry. The company stated that results for spleen and liver volumes, haemoglobin and platelet count showed that no clinically meaningful differences in efficacy were found.
4.58
The company agreed that it was appropriate to use real world dose data for ERT but it stated that real world weight data must also be used when estimating the total administered dose. The company stated that using the ENCORE weight of 67.5 kg resulted in a mean total dose of 4 vials of ERT, but using real world weights of between 71.8 kg and 75.0 kg resulted in a mean total dose of 5 vials.
4.59
The ERG stated that the dose of ERT in the ERG base case came from English prescribing data that reported average units per month independent of weight, so the average weight in the model was not relevant. However, the ERG presented exploratory analyses using estimates of ERT use based on real world weight and the weight-dependent ERT dosing rate. This had the impact of increasing the costs of ERT in the model.
4.60
Full details of all the evidence are in the submissions received for this evaluation, and in the ERG report, which are all available in the committee papers.