Migalastat for treating Fabry disease

Patient experts and patient groups highlighted the substantial impact of Fabry disease on people with the condition and their families.

• Fabry disease leads to progressive disability from transient ischaemic attacks, strokes, cardiac and renal disease.

• Adults may need dialysis, a kidney transplant or pacemakers and may be physically and mentally disabled.

• Symptoms in adults include hearing impairment, skin rash, gastrointestinal problems and fatigue. For children, symptoms include low energy, fatigue, pain and gastrointestinal problems.

• The effects of the disease can disrupt daily activities and cause absences from work or school.

• Symptoms generally appear in childhood but usually go unrecognised until adulthood, when organ damage has already occurred.

• People with Fabry disease may need a carer relatively early in life; often this responsibility is taken on by family members.

• Many people with Fabry disease have had psychological difficulties coming to terms with a lifelong progressive disorder, particularly before the introduction of enzyme replacement therapy (ERT) in 2001.

• ERT has a number of benefits but it also has limitations. The infusion dosage schedule of every 2 weeks means that people with Fabry disease cannot plan trips away from home. ERT must be kept refrigerated and there are risks of developing an infusion-related infection and antibodies to treatment. There is also a possible need for a homecare nurse or carer to help with administration.

The company submitted evidence from 2 randomised controlled trials (ATTRACT and FACETS) and 2 open-label extension studies. ATTRACT was an 18‑month open-label randomised controlled trial designed to show comparable effectiveness between migalastat and ERT. FACETS was a 6‑month double-blind randomised controlled trial, in which patients who had not had treatment before had either migalastat or placebo.

The final outcomes reported in ATTRACT and FACETS can be grouped into renal function, cardiac function, health-related quality of life and safety outcomes. These outcomes were designed to capture aspects of Fabry disease morbidity that reflect how patients feel or that are used in clinical decision-making. The trials also reported biochemical outcomes of Gb3 and plasma lyso‑Gb3 distributions and activity of the enzyme alpha-galactosidase A (alpha-gal A). These are primarily indicators of migalastat efficacy, but may not directly reflect patients' symptoms and do not themselves have a clear role in clinical decision-making.

Intention-to-treat (ITT) analyses were done based on all randomised patients in each trial. However, the ITT population included some patients who had mutations that were later found not to be amenable to migalastat. This was because the assay used to determine the amenability of mutations was changed to conform to GLP laboratory standards; the updated assay is the one referred to in the marketing authorisation for migalastat. Therefore the company used 'modified ITT' analyses which excluded these patients. In ATTRACT, the modified ITT population excluded patients with other protocol violations as well as non-amenable mutations and was effectively a per-protocol population. The ERG stated that 'modified ITT' is therefore potentially misleading (and has a different meaning in the 2 randomised controlled trials).

The small sample size (n=60) in ATTRACT made a standard non‑inferiority analysis impossible and the company presented its own pre-specified criteria for comparability. Based on these criteria, migalastat would be considered comparable to ERT if the difference between their means for the annualised change in glomerular filtration rate was 2.2 ml/min/1.73 m²/year or less, and the overlap in the 95% confidence intervals for these means was greater than 50%.

In ATTRACT the pre-specified criteria for comparability of migalastat and ERT were met for both the co‑primary outcomes of measured and estimated glomerular filtration rate. In FACETS, the change in glomerular filtration rate was measured at 6 months, although the company stated that this is generally considered too short to show a reliable trend.

In ATTRACT at 18 months, people who switched from ERT to migalastat had a statistically significant decrease from baseline in left ventricular mass index (LVMi; p<0.05), whereas in people who remained on ERT this decrease was not statistically significant. However, there was no statistically significant difference in the change from baseline between the groups. Patients in FACETS who continued into the open-label extension study had LVMi recorded after 18 or 24 months of migalastat; in patients who had migalastat for 24 months, a significant decrease in LVMi from baseline was seen.

ATTRACT included a composite clinical outcome of the rates of pre-specified renal, cardiac and cerebrovascular events and mortality over 18 months. The proportion of patients who had a renal, cardiac or cerebrovascular event was 29% (10/34) of patients who switched from ERT to migalastat compared with 44% (8/18) of patients who remained on ERT. Overall, renal events were the most common, followed by cardiac events. No deaths occurred.

Both ATTRACT and FACETS assessed health-related quality of life using the SF‑36 health questionnaire physical component summary and the Brief Pain Inventory short form. ATTRACT also included the SF‑36 mental component summary, and FACETS used the Gastrointestinal Symptoms Rating Scale. For ATTRACT, the company stated that SF‑36 scores were comparable in the migalastat and ERT groups at baseline and there was little change in these scores over the 18‑month study period. The Brief Pain Inventory pain severity component showed that patients had mild pain at baseline, and this did not change over the 18‑month treatment period. For patients from FACETS continuing in the open-label extension studies, the company reported changes in scores for the same Gastrointestinal Symptoms Rating Scale domains. After 18 or 24 months of migalastat, patients had statistically significant improvements in diarrhoea and indigestion compared with baseline. The company stated that there was a trend for improved reflux and constipation, although symptoms of abdominal pain remained stable. The company reported that SF‑36 results were stable at 24 months. The company also stated that Brief Pain Inventory severity component scores did not change from baseline to month 24. Patients having migalastat reported stabilised cardiac symptoms and kidney function, improved mood swings and freedom from their infusion routine.

The company provided adverse event data from ATTRACT, FACETS and the open-label extension studies. In ATTRACT, between 94% and 95% of patients had a treatment-emergent adverse event, as did 91% of patients in FACETS. Nasopharyngitis and headache were the most common adverse events.

The company submitted a Markov state transition model to estimate the costs and health effects of migalastat compared with ERT in people with Fabry disease. The 10 health states in the model represented the progression of Fabry disease over time. All health states were divided into incident (acute events) and prevalent (long-term). The model took the perspective of the NHS and Personal Social Services. It had a lifetime (52‑year) time horizon, and a cycle length of 1 year. Costs and benefits were discounted at a rate of 3.5% per year.

The model structure and the values for transition probabilities between disease states were based on a Dutch study done in a group with Fabry disease. It was assumed that this was equivalent to a UK Fabry population. A number of structural assumptions were made in the company's model:

• ERTs are equivalent and can be grouped as a 'blended comparator'

• migalastat is clinically equivalent to ERT

• people having migalastat continue treatment until death, whereas some people having ERT stop treatment

• treatment adherence is 100%

• transition probabilities do not vary over time

• people cannot develop 2 complications in 1 model cycle (1 year)

• people with Fabry disease have a similar body weight to the UK general population

• about 50% of people self-administer ERT; for the remainder treatment is given by a nurse at home.

The starting distribution of people in the 5 health states was based on the baseline measurements of the ATTRACT trial population. The company stated that this population is representative of people with Fabry disease in England.

The company also provided details of the agreed patient access scheme, in which migalastat would be provided with a discount. The discount is commercial in confidence and so cannot be reported here. Estimates for costs associated with each health state were provided, including diagnostic, laboratory and imaging tests, primary and secondary care appointments, hospitalisations and treating complications. The costs were derived from NHS reference costs and Personal Social Services Research Unit (PSSRU) data. The frequency of diagnostic, laboratory and imaging tests for all people with Fabry disease was taken from the adult Fabry disease standard operating procedure, with the unit costs taken from the NHS reference costs. The costs for treating adverse events were also considered for each specific adverse event. The costs ranged from £0.06 (headache) to £47.28 (influenza), and were taken from the British national formulary and PSSRU.

The model captured health-related quality of life by assigning utility scores to each health state. The utility scores were taken from the Dutch study and described the health-state utility scores (disutility) for the complication states. Infusion-related utility decrements (disutilities) were based on a discrete choice experiment done by the company with 506 people from the UK general population.

The results of the company's cost–consequence analysis were presented as costs, life years, and quality-adjusted life years (QALYs). Treatment with ERT is associated with 13.36 QALYs and migalastat with 14.33 QALYs, giving an incremental QALY gain of 0.98 for migalastat. The total and incremental costs of migalastat and ERT are confidential and so cannot be reported here. Because equivalent efficacy was assumed between migalastat and ERT, the infusion disutilities were responsible for virtually all (0.97 of 0.98 QALYs) of the differences between migalastat and ERT.

The company explored uncertainty in the economic model through deterministic and probabilistic sensitivity analyses and scenario analyses. The scenario analyses explored assumptions including ERT price discounts, utility scores, effectiveness of ERT and migalastat, patient demographics, perspective of the model, the time horizon, and ERT market share.

The company did a budget impact analysis, in which it estimated that there are 142 people with Fabry disease in the UK for whom migalastat may be considered. This estimate took into account the proportion of people with Fabry disease who have amenable mutations, which was assumed to be 40%. The number of people eligible for migalastat was predicted to increase by 1 person per year. An average body weight of 77.6 kg was used to calculate the ERT doses. The estimated budget impact of migalastat, taking into account the patient access scheme and confidential price discounts for ERT, is commercial in confidence and cannot be reported here.

The ERG stated that the studies providing clinical effectiveness evidence for migalastat are limited and there are concerns about the design of both pivotal randomised controlled trials and the related open-label extension studies. These concerns included:

• small populations and short trial durations

• imbalances in patient baseline characteristics between the trial arms in both randomised controlled trials and

• uncertainty as to how long individual patients had received migalastat because it was not reported how many patients were recruited to the open-label extension study from each arm of FACETS.
  
  One of the ERG's major concerns about the clinical evidence was the uncertainty in the comparability of migalastat and ERT. The pre-specified criteria for non-inferiority allowed a claim of comparability despite very wide confidence intervals for the outcome measures. The ERG was satisfied that the company's adverse event data did not raise any safety concerns over the use of migalastat.

The ERG noted a number of limitations in the company's economic modelling. The Markov model simplified Fabry disease progression. It did not allow people with end‑stage renal disease to have kidney transplants and did not capture different levels of chronic kidney disease, different severities of stroke, or different types of cardiac complications. The ERG also noted that the probability of transition between these disease states remained constant throughout the patient's life; this was considered to be improbable and likely to underestimate the disease state transition probability. The model did not allow for poor adherence or for stopping migalastat at any point. The starting weight of people entering the model was a general population average; the ERG noted uncertainty about whether this was representative of people with Fabry disease. The ERG also noted uncertainty about whether people recruited to ATTRACT were representative of the Fabry population because the trial did not recruit people with severe manifestations of Fabry disease. The mortality rates used by the company led to an overestimation of life expectancy in the model. The ERG noted that the disutility associated with ERT infusion (−0.05) was high and was much greater than the disutility used in the model for developing a new disease complication (−0.018). This infusion disutility was calculated using the results of a discrete choice experiment done in healthy people; the ERG noted uncertainty about the comparability of these values with those of disease complications given the differences in the methods used for estimation.

The ERG did scenario analyses to address flaws and uncertainties in the model. These included:

• changing the price of ERT

• changing the proportions of people starting in each disease state (taken from the Fabry registry)

• increasing the starting age

• including background mortality data from the Office for National Statistics life tables

• reducing patient body weight to reflect the average from ATTRACT

• calibrating transition probabilities to give a life expectancy of 66.5 years

• making discontinuation of migalastat and ERT equal in the model and including discontinuation of migalastat in people with end-stage renal disease

• reducing health-state utilities (taken from alternative sources) and

• reducing the disutility for infusion.
  
  The ERG combined these assumptions into its preferred analysis, which resulted in an incremental QALY gain for migalastat of 0.34 compared with ERT.

The ERG noted that most transition probabilities between the model health states in the company's model did not vary with age, which led to an overestimate of the life expectancy of people with Fabry disease. The ERG stated that its analyses showed the potential effect of these uncertainties, but did not resolve them. The set of assumptions used in the ERG analyses was more conservative because it produced life expectancy estimates that are closer to Fabry registry data and assumed more plausible disutilities for infusions. However, the ERG analyses are based on assumptions that, although informed by some data, represent the ERG's best estimates. The ERG stated that limitations in the evidence remained.

The ERG did sensitivity analyses on the company's budget impact analysis and found that the calculations are most sensitive to the proportion of people who have amenable mutations, the prevalence of Fabry disease, and the proportion of people having treatment.

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.