Pegunigalsidase alfa for treating Fabry disease

There is no cure for Fabry disease, but treatments are available that relieve the symptoms and slow progression of damage to the kidneys and heart. In the UK, people with Fabry disease typically start treatment when one of the criteria outlined in the British Inherited Metabolic Disease Group guidelines are met. These include evidence of Fabry-related general symptoms (such as uncontrolled pain) and kidney and cardiac disease. The clinical experts explained that treatment options include infusion with agalsidase alfa or agalsidase beta, which are enzyme replacement therapies (ERTs) that replace the non-functioning enzyme, or migalastat. NICE has not evaluated ERTs but does recommend migalastat (taken orally) as a treatment option for Fabry disease in people over 16 with an amenable mutation (see NICE's highly specialised technology guidance on migalastat for treating Fabry disease, from here HST4). The committee noted that the company considered that people with amenable mutations would be offered migalastat first if it is suitable. For this reason, the company did not consider migalastat to be a relevant comparator for this appraisal. The clinical experts stated that if a person had an amenable mutation either migalastat or an ERT can be offered first, and the choice is based on the person's preference. The clinical experts also noted that people who have amenable mutations tend to have milder Fabry disease. They highlighted that many people with an amenable mutation may choose migalastat because it is taken orally. But, they also noted that some people may not choose to have migalastat because of the need to fast for 2 hours before and 2 hours after having it. It is also taken every 2 days, so some people may have difficulty remembering to take the treatment consistently. The clinical experts shared that it is possible for people to switch treatment from ERT to migalastat and vice versa. They stated that the current treatments slow disease progression but there is an unmet need for further treatment options for Fabry disease. The committee concluded that pegunigalsidase alfa would be an additional ERT for people with and without an amenable mutation, and migalastat, agalsidase alfa and agalsidase beta were relevant comparators.

The company's key clinical evidence is from BALANCE, an international, randomised, double-blind, phase 3 clinical trial with 24‑month follow-up. The trial was done in adults (aged 18 to 60 years) with Fabry disease and impaired kidney function, who had previously had agalsidase beta. The trial was designed to test whether pegunigalsidase alfa was statistically non-inferior in clinical effectiveness to agalsidase beta. The EAG noted that not everyone with Fabry disease has kidney impairment or would have already had treatment with agalsidase beta. So, the results may not be generalisable to people who have not had a previous treatment or do not have kidney impairment. In addition, the EAG noted that kidney impairment is more common in people with classic Fabry disease than those with non-classic Fabry disease. The clinical experts stated that deterioration in kidney function tends to be late in the disease progression, so people with kidney impairment have worse treatment outcomes. The clinical experts expected that the assessment of non-inferiority of pegunigalsidase alfa compared with agalsidase beta in BALANCE would be generalisable to the whole population who would have an ERT. The EAG also noted that there was a higher proportion of men (72% compared with 56%) and people with classic Fabry disease (56% compared with 52%) in the agalsidase beta arm compared with the pegunigalsidase alfa arm. Also, more people in the poorer kidney function group (that is, estimated glomerular filtration rate [eGFR] slope of less than −5 ml/minute/1.73 m²/year compared with a slope greater than −5 ml/minute/1.73 m²/year) had agalsidase beta. The clinical experts agreed that there were some imbalances but considered that this would likely have less of an impact on the analysis because kidney impairment was a predefined inclusion criterion and people in both arms of the trial would have similar kidney function. The clinical experts did not consider the difference in the proportions of people in the kidney function subgroups to be significant or to impact the results. The committee acknowledged the imbalances in the baseline characteristics but concluded that for the purpose of its decision making, data from BALANCE could be considered generalisable to the whole population who would have pegunigalsidase alfa.

The primary outcome in BALANCE was annual rate of change (slope) in eGFR, which is a measure of declining kidney function over time. The trial was intended to measure if pegunigalsidase alfa is non-inferior to agalsidase beta. The committee noted the EAG's concerns that the trial's statistical analysis plan changed over the course of the clinical trial. It heard from the company that this related to its FDA regulatory submission and the protocol amendment happened after the last patient entered the trial and before the trial database was locked. The median eGFR slope difference between pegunigalsidase alfa and agalsidase beta after 24 months was −0.359 ml/minute/1.73 m²/year (95% confidence interval −2.444 to 1.726). The company's prespecified criteria for non-inferiority was that the lower limit of the 95% confidence interval had to be greater than −3.0 ml/minute/1.73 m²/year. Based on this, it considered the non-inferiority criteria met. The mean value at 12 months did not conclusively meet the criteria but the later data cut (24 months) did. Subgroup analysis did not show any difference in eGFR slope based on gender or form of Fabry disease (see section 3.3), although the confidence intervals were wide due to the small population in the clinical trial. The committee concluded that based on the available evidence it considered pegunigalsidase alfa non-inferior to agalsidase beta.

There was no direct comparison of pegunigalsidase alfa with agalsidase alfa in a randomised clinical trial. In its submission, the company assumed all 3 ERTs were clinically equivalent. The company's assumption was based on 2 randomised controlled trials (Sirrs et al. 2014 and Vedder et al. 2007) that showed no statistical difference between agalsidase alfa and agalsidase beta, and on the BALANCE trial, which showed that pegunigalsidase alfa was non-inferior to agalsidase beta. The company noted that an indirect comparison for the treatments was not feasible, and any analysis would be uncertain because of the heterogenous evidence base. The EAG raised concerns regarding the company's assumption of clinical equivalence. It noted that Sirrs et al. included only around one-third (94 out of 294) of the people needed to observe the prespecified difference in outcome. Also, in Vedder et al. people were treated with only one-fifth (0.2 mg/kg compared with 1 mg/kg) of the dose of agalsidase beta used in BALANCE. So, these trials do not provide supportive evidence for clinical equivalence of pegunigalsidase alfa compared with agalsidase alfa and agalsidase beta. The clinical experts noted that clinical trials for rare diseases can sometimes be underpowered because of the difficulty with recruiting people from a small population. The clinical experts also stated that there is no strong evidence of pharmacological difference between agalsidase alfa and agalsidase beta, and in clinical practice they are broadly considered to be similarly effective. The clinical experts noted that because pegunigalsidase alfa contains the polyethylene glycol molecule, it would be expected to be broken down less quickly in the body. This would potentially lead to longer exposure time, which would be beneficial to people having the treatment. But, they noted that this potential benefit in exposure time had not translated into better outcomes in BALANCE. The committee noted that the company had not presented a comparison of pegunigalsidase alfa with migalastat and had stated in response to technical engagement that an indirect comparison was unfeasible. The committee was aware that HST4 concluded that it was reasonable to assume clinical equivalence between migalastat and ERTs, although the data used to determine clinical equivalence in HST4 also had limitations. The committee heard from clinical experts that, in practice, whether a person has ERT or migalastat would depend on that person's choice of administration method rather than any difference in clinical effectiveness (see section 3.2). The committee concluded that, overall, the limitations of the data meant it was not possible to conclude that pegunigalsidase alfa and its comparators were clinically equivalent. But, it was reasonable to assume for the purpose of this appraisal that pegunigalsidase alfa, agalsidase alfa, and agalsidase beta were similarly clinically effective.

In BALANCE, a similar proportion of people who had pegunigalsidase alfa and agalsidase beta experienced an adverse event. But the company noted that the overall number of events (rate per 100 exposure years) was lower for people who had pegunigalsidase alfa. The committee asked the clinical experts if pegunigalsidase alfa would be expected to provide fewer adverse events than other ERTs. The clinical experts responded that early data suggests that people in the pegunigalsidase alfa group had fewer anti-drug antibodies, which reduce treatment efficacy. They speculated that this may result in benefits for pegunigalsidase alfa in the longer term. The clinical experts noted that the adverse event of interest in BALANCE is infusion-related reaction and the rate of this event was lower (by about 3 cases per 100 infusions) in the pegunigalsidase alfa arm than the agalsidase beta arm. The EAG noted that in BALANCE, slightly more people in the agalsidase beta arm had treatment for 24 months than in the pegunigalsidase alfa arm. This may have affected the observed rates of adverse events. The committee further noted that people in BALANCE had all previously had agalsidase beta and may have tolerance to it. This might have biased the adverse event outcomes against pegunigalsidase alfa. The committee concluded that pegunigalsidase alfa was a similarly tolerable treatment to agalsidase beta.

The company made a case for a cost-minimisation analysis (an approach that assumes equivalent outcomes and compares costs only) rather than a cost-utility analysis to assess the cost effectiveness of pegunigalsidase alfa compared with agalsidase alfa and agalsidase beta. This was because the company stated that there was no difference in clinical effectiveness or quality of life of the 3 treatments, so it was appropriate to compare only the costs. The EAG also presented an exploratory cost-utility assessment between pegunigalsidase alfa and migalastat. The EAG's assessment assumed clinical equivalence between both treatments, with a decrease in quality of life associated with taking an intravenous rather than an oral treatment. The company confirmed at technical engagement that if migalastat were considered a comparator, it agreed with the EAG's approach to measure cost effectiveness. The committee acknowledged the rarity of Fabry disease and the potential difficulty of gathering robust utility evidence in rare conditions (see section 3.11). It concluded that cost minimisation was appropriate for the comparison with the ERTs. This was because it was satisfied that the clinical effectiveness and quality of life would be similar, and it would use the EAG's analysis for the comparison with migalastat in its decision making.

The company used a Markov state transition model with 10 health states capturing symptoms and complications related to Fabry disease. These included pain, end-stage kidney disease and cardiac complications. The modelled cohort had a mean age of 40, had symptomatic Fabry disease and included the same proportion of men and women. The cohort were modelled for 60 years. The model structure was similar to that used in HST4. This was in turn based on a model from a Dutch study (Rombach et al. 2013) that evaluated the cost effectiveness of ERTs and standard care in a Dutch Fabry disease cohort. In the model, people progressed to worse health states or died. The distribution of the modelled cohort at entry across the health states was based on the global Fabry Registry. But, the model excluded people with end-stage kidney disease because they were not considered appropriate to start a new treatment. The EAG noted that at baseline, people could only have a single symptom health state. The EAG stated that this did not align with its clinical expert's opinion that at 40 years, multiple complications would likely have already developed. The committee heard from clinical experts that it is uncommon for people to have a single health state complication (see section 3.3). But overall, the model structure was reflective of Fabry disease in terms of kidney and cardiac symptoms occurring later as the disease progressed. The committee concluded that, although the health states in the model may not reflect the combinations of symptoms people may have, the model structure was reasonable for decision making.

The company's submission included transition probabilities (the chance of moving between health states) from Rombach et al., the same dataset used for HST4. The model had different transition probabilities for men and women. The company noted that it was not feasible to derive transition probabilities from the pegunigalsidase alfa trials because they did not follow a large enough population over a long enough period to generate robust transition probabilities. The company added that robust estimates of transition probabilities are difficult to achieve because Fabry disease is a rare condition and disease progression through health states occurs over a lifetime (about 60 years). So, it considered Rombach et al., which included 72 people having ERT, to be the most robust dataset. The EAG considered that the transition probabilities did not reflect the rate or extent of disease progression described by its clinical expert or by the company. It noted that in the model, around half of the population die in their baseline health state and less than 1% of people are estimated to have more than 1 symptom (for example, end-stage kidney disease and cardiac complication). The clinical experts during the committee meeting agreed that the transition probabilities did not reflect the progression seen in clinical practice. In particular, the clinical experts highlighted that more people would be expected to have cardiac and kidney complications than were modelled. The committee was aware that newer data may be available from the Fabry disease registry (Clinical Practice Research Datalink), which was identified in the company submission. The company noted that Rombach et al. had 11 years of data from a single patient centre. The data is therefore less heterogenous and may also be less subject to bias than the Fabry disease registry, in which people were enrolled without clear exclusion criteria. The company stated that this meant that the characteristics of people enrolled in the registry may vary from centre to centre. The clinical experts did not have concerns about selective enrolment, noting that most people with Fabry disease in the UK are enrolled in a registry. One clinical expert noted that people under their clinical care are enrolled in a registry containing around 450 people. They considered this registry appropriately captures the Fabry disease population across all health states. The EAG considered the uncertainty around the model's transition probabilities unresolved. But it noted that the impact of using different transition probabilities on the incremental costs would likely be minimal. This is because the treatment arms would be affected equally because of the company's assumption of clinical equivalence between the modelled treatment arms (see section 3.5 and section 3.7). But, the EAG noted that external validity would be important for future Fabry disease appraisals where measured differences in outcomes are used in the model. The committee concluded that the transition probabilities likely lacked external validity. But it noted that because of the clinical equivalence assumption (see section 3.5), using a different source of transition probabilities might have limited impact on the incremental cost estimates.

For HST4, the committee concluded that the mortality probability data used in the model led to an unexpectedly high life expectancy (83.4 years) for people with Fabry disease. For the current pegunigalsidase alfa submission, the company used Fabry Registry data (Waldek et al. 2009), which estimates the life expectancy for men and women to be 58.2 years and 74.7 years respectively. The EAG could not validate this adjustment in the company's base-case results but applied the mortality adjustment in its own base-case. The committee concluded that the Fabry Registry is an appropriate data source for estimating mortality.

Although the company presented a cost-minimisation analysis as its base case, which did not include utility values, it also presented a cost-utility scenario analysis. This scenario assumed equal clinical effectiveness and quality of life between each treatment arm and no disutility for adverse events between treatment arms. It therefore produced the same results as the company base case. The cost-utility model used utility values for each health state from Arends et al. (2018) and adjusted by the mean baseline utility value in BALANCE (0.762). The committee was aware that the company collected EQ-5D-5L data from BALANCE, which was mapped to EQ-5D-3L using crosswalk regression method described by Hernández Alava et al. (2017). However, the company reported that it could not estimate robust utilities for every health state. This was because of the low number of Fabry clinical events, and utility data from BALANCE could only be derived for 2 of the 10 health states: pain and other symptoms. The EAG preferred the company's original base-case approach (that is, Arends et al. adjusted with the BALANCE baseline utility value for all health states), if the complete utility values for all health states could not be obtained from BALANCE. In the EAG's exploratory analyses comparing pegunigalsidase alfa with migalastat, the EAG assumed a disutility of 0.025 associated with having an intravenous treatment rather than an oral treatment. This was based on HST4 in which the committee had determined a disutility of 0.5 was too high and 0.025 was plausible. The patient experts stated that it was not possible to say whether the exact assumed value was plausible, but taking an intravenous rather than an oral treatment was expected to have a minimal effect on quality of life. The committee noted that it can be difficult to generate utility values for rare conditions such as Fabry disease. It concluded that it would have preferred robust utility data from BALANCE but given the assumption of clinical equivalence the company's base-case dataset is sufficient for decision making. It further concluded that the disutility assumed in the EAG exploratory analysis was reasonable.

The company sourced its administration costs from NHS Reference Costs 2020/2021, and Personal Social Services Research Unit 2021. For the maintenance treatments, the company assumed 50% of people would need nurse-assisted administration while the other 50% would self-administer and only have 1 nurse visit per year. The EAG's clinical expert reported that around 90% of people would likely need nurse-assisted administration because they are not fully independent to deliver their own infusion. The EAG applied this assumption in its exploratory base case. The clinical experts at the committee meeting stated that not everyone who self-administers their treatment does so independently; some people may still need nurse assistance for preparation and other support. They estimated that around 10% to 20% of people with Fabry disease would be fully independent and require no nurse-assisted administration. The company also included the costs of other healthcare professionals such as GPs, physiotherapists, psychiatrists and social workers, as part of the follow-up costs. The EAG's clinical expert noted that the cost of social worker visits would not be funded by the NHS. So, in its base case the EAG excluded the cost of social worker visits. The committee concluded that social worker costs should be captured in the model as part of personal social service costs that are in the NICE reference case, but it preferred the EAG's estimates of the cost of administration.

The company obtained costs for acute complications from NHS Reference Costs 2020/2021. The EAG noted that the company used simple averages rather than weighted average for cost categories that included multiple codes representing varying severity. The EAG also noted that the company assumed all tests (including anti-ERT antibody tests) in the general management of Fabry disease are paid for by the NHS. Whereas, the EAG's clinical expert and the experts at the committee meeting stated that some tests are currently paid for by manufacturers of current standard care. The company stated that it would also cover these costs for pegunigalsidase alfa. The EAG's expert further noted that the annual number (frequency) of some of the routine management tests was different to the company's estimates. The EAG conducted a scenario using its clinical expert's estimates, which had a minor impact on the incremental costs. The committee concluded that the EAG's frequency estimates are appropriate. The committee also noted the company's statement that it will cover the cost of similar tests currently paid for by manufacturers of current standard care.

The company's base-case deterministic incremental cost estimates suggest that pegunigalsidase alfa is cost saving compared with agalsidase alfa (saving £476,243) and agalsidase beta (saving £470,950). The results represent costs over the lifetime of 1 person with Fabry disease. The probabilistic estimates were also cost saving, but the results are considered confidential by the company and cannot be reported here. The EAG applied its preferred assumptions in its base case by:

• Increasing the number of people needing nurse-assisted treatment infusion to 90% (see section 3.12).

• Using weighted average for estimating the cost of acute complications (see section 3.13).

• Removing costs of social care visits (see section 3.12).

• Adjusting mortality to reflect average life expectancy in people with Fabry disease (see section 3.10).

• Using the EAG's clinical expert's general management resource use estimates (see section 3.13).
  
  With the EAG's preferred assumptions, the deterministic base-case incremental cost estimates suggest that pegunigalsidase alfa is cost saving compared with agalsidase alfa (saving £386,796) and agalsidase beta (saving £396,288). The EAG's additional scenario analysis, which compared pegunigalsidase alfa with migalastat, assumed:

• non-inferiority between the treatments

• no difference in disutilities related to adverse events

• an annual disutility of 0.025 (based on HST4) applied for people having pegunigalsidase alfa, which is an intravenous infusion, and

• no administration cost for migalastat because it is an oral treatment.
  
  The EAG considered the analysis illustrative only and noted that a full analysis by the company would be preferable. Migalastat has a confidential discount, so the results cannot be reported here. However, the results showed that pegunigalsidase alfa provides fewer quality adjusted life years (QALYs) but is less costly than migalastat.

The company's model applied a single treatment discontinuation rate of 0.5% across treatment arms, which was also used in HST4. The committee noted that the number of discontinuations in BALANCE for pegunigalsidase alfa and agalsidase beta (5 people compared with 1 person) was low. But it raised concerns that subsequent treatment costs for people who switch treatment from pegunigalsidase alfa were not included in the model. The committee was concerned that if people switched to more expensive treatments after pegunigalsidase alfa these costs could affect any potential cost savings in the modelled pegunigalsidase alfa treatment arm. The committee noted that omitting subsequent treatments from the model and using an assumption of treatment discontinuation rates meant that the exact cost savings associated with pegunigalsidase alfa were uncertain but it remained reasonable to conclude that pegunigalsidase alfa is cost saving compared with its comparators. In addition, the committee shared the EAG's concerns that the cost effectiveness of agalsidase alfa and agalsidase beta have not been evaluated by NICE and it could not exclude the possibility that pegunigalsidase alfa was being compared with cost-ineffective treatments. However, it recognised that agalsidase alfa, agalsidase beta, and migalastat represent established treatments for Fabry disease and that pegunigalsidase alfa was cost saving compared with these treatments.

The committee concluded that pegunigalsidase alfa was cost saving compared with agalsidase alfa, agalsidase beta, and migalastat. Although there were fewer modelled QALYs with pegunigalsidase alfa compared with migalastat, this was driven by the different ways the treatments are administered rather than a difference in clinical effectiveness or adverse events. Overall, the committee concluded that pegunigalsidase alfa was similarly clinically effective, as tolerable as the treatments used in the NHS, and costs less. Therefore, the committee recommended pegunigalsidase alfa for treating Fabry disease.