Givinostat for treating Duchenne muscular dystrophy in people 6 years and over

Duchenne muscular dystrophy is a fatal, rare genetic disorder that begins in childhood and causes systemic loss of muscle function and progressive disability. It is caused by mutations in the dystrophin gene, which is essential for maintaining the structural integrity of muscle fibres. Without dystrophin, muscles gradually weaken and deteriorate, which affects the entire body. Because the dystrophin gene is on the X chromosome, the condition primarily affects boys and young men, although in rare cases girls may also be affected. In the UK, it is estimated that around 100 people are born with Duchenne muscular dystrophy each year and around 1,183 people are currently living with the condition. Symptoms typically begin when a child is between 2 and 5 years old and include delayed walking, frequent falls, and difficulty rising from the floor.

Duchenne muscular dystrophy is a systemic condition, affecting multiple body systems. As it advances, people lose the ability to walk and use their arms. It can cause skin complications due to pressures applied by having limited mobility. Breathing muscles become progressively weaker, making it harder to breathe. The heart's ability to pump blood effectively is also reduced and cardiac problems can lead to unexpected death. Duchenne muscular dystrophy is also associated with neurobehavioural comorbidities, including intellectual disability, attention deficit disorders, and autism spectrum disorder. When walking is no longer possible, people with Duchenne muscular dystrophy must use mobility aids such as wheelchairs. They also need regular healthcare appointments to monitor their spine, heart, and breathing during sleep. The spine can develop scoliosis, which may need surgery. By their teens or early twenties, most people with Duchenne muscular dystrophy need night-time non-invasive ventilation (NIV). This involves wearing a mask or mouthpiece connected to a machine that helps maintain oxygen levels and reduce the strain on their lungs while sleeping.

In the most advanced stages, people need continuous ventilatory support (full-time ventilation). Cough assistance is also often needed to help clear the airways. By adulthood, most people with Duchenne muscular dystrophy need help with all aspects of self-care, including eating, drinking, using the toilet, dressing, washing, and moving. People with Duchenne muscular dystrophy become increasingly dependent on others, including their families and carers, because they need continuous, day-and-night care to support them in their daily lives. Despite improvements in supportive care, Duchenne muscular dystrophy remains a fatal condition, with life expectancy typically under 30 years, most often due to respiratory or cardiac failure. The committee noted the considerable challenges in generating robust evidence for a rare, progressive childhood condition with a long-term disease course such as Duchenne muscular dystrophy. It concluded that it would take these challenges into account in its decision making (see section 3.39).

Some key milestones used to describe the progression of Duchenne muscular dystrophy are:

• loss of ambulation (LOA): no longer being able to walk with or without support or no longer being able to stand with support

• NIV: having to use a ventilator at night to aid breathing

• forced vital capacity (FVC) less than 1 litre: a proxy measure for when someone would need a ventilator all the time.
  
  The clinical experts highlighted that, while these milestones are useful, all aspects of disease progression are important. They explained that focusing on the 3 milestones can be too simplistic. For example, arm and hand function is important for independence. The patient experts agreed with this and highlighted that other outcomes are equally if not more important than the 3 key milestones. A patient expert explained that while LOA was important, losing upper limb function was much more difficult for their son. This was because it meant there were many things he could no longer do, such as feed himself, adjust his glasses or scratch an itch. Another patient expert explained that delays in progression were invaluable to them all through the course of the disease. They explained that they still have some hand function and extending that would allow them to continue to do things they enjoy, such as playing chess, which was very important to their wellbeing. The patient experts also added that any delay from starting full-time ventilation would be incredible. The difference between night-time and full-time ventilation is very big and the impact of a full-time ventilator is enormous. The patient expert explained that the transition to full-time ventilation is extremely difficult, much harder than losing the ability to walk, and the value in delaying it is immeasurable.

Duchenne muscular dystrophy has a profound and far-reaching effect on the lives of people with the condition and their families and carers. Because it is a progressive disease, it leads to increasing physical disability and loss of independence. It also causes significant emotional and psychological strain for people with the condition and the people who care for them. Siblings may experience feelings of worry, guilt, or isolation, and their lives are often shaped by the needs and routines of the family member with Duchenne muscular dystrophy. As the condition advances, people need full-time support, day and night, for daily activities, personal hygiene, mobility, and medical care. This has a considerable and sustained impact on carers. Even in the earlier stages, support is needed to attend medical appointments or respond to emergencies. Patient and carer submissions described the diagnosis of Duchenne muscular dystrophy as devastating. One carer conveyed that witnessing their child progressively lose physical abilities and face a shortened life expectancy, while other children continue to develop normally, had been psychologically overwhelming. They reported severe and lasting emotional impacts, including a loss of identity, career, relationships, and personal wellbeing. They described living with constant grief, ongoing anticipatory loss, and a persistent need to remain vigilant, with no real moments of relief. Even positive experiences were overshadowed by the fear that each milestone could be the last.

Patient organisation submissions emphasised the importance of treatments that can slow the condition's progression and the positive psychological impact that slower progression can have on people with Duchenne muscular dystrophy and their families. Maintaining independent movement, use of their arms and hands, and delaying the need for full-time ventilation is seen as essential for preserving independence, mental wellbeing, and participation in everyday life. Delaying the loss of mobility and maintaining independence during the teenage years is particularly important. This is a time when most young people are becoming more independent, learning to drive, attending university, and spending more time with friends. By contrast, at the same point in life people with Duchenne muscular dystrophy are likely to be experiencing increasing dependence. This can be emotionally challenging and isolating. For some, the loss of mobility is also associated with experiences of bullying or social exclusion, further affecting mental health and self-esteem. Delaying this loss can help preserve a sense of autonomy and self-worth. It may also allow people to continue doing things that are important to them, such as attending university lectures, socialising, and pursuing their interests. These aspects of independence are not only practical but are also deeply connected to emotional wellbeing, identity, and inclusion. Clinical experts noted that people who lose ambulation later tend to manage better overall. They explained that delaying progression was valuable at any stage of Duchenne muscular dystrophy, not just in delaying LOA. The committee understood that Duchenne muscular dystrophy has a substantial and far-reaching impact on people with the condition, as well as on their families and carers. It concluded that there is a clear need for effective treatments for Duchenne muscular dystrophy and that any treatments that delayed progression would be valued by people with Duchenne muscular dystrophy and their families and carers.

There is currently no cure for Duchenne muscular dystrophy. Management focuses on slowing disease progression, preserving muscle function, and improving quality of life through a combination of pharmacological treatments and multidisciplinary care. Established clinical management (ECM) includes using corticosteroids such as prednisolone and deflazacort. These are typically started in early childhood and have been shown to prolong the ability to walk independently and delay the onset of complications. The clinical experts explained that starting steroid treatment as early as possible not only delays LOA but also results in a more pronounced delay in other disease milestones. More recently, vamorolone has been recommended as an alternative corticosteroid (see NICE's technology appraisal guidance on vamorolone for treating Duchenne muscular dystrophy in people 4 years and over, from here TA1031). Supportive care is essential and includes physiotherapy, respiratory and cardiac monitoring, nutritional support, and orthopaedic interventions. As the condition progresses, people may need NIV for periods during the day, at night and then all the time. Other interventions may include spinal surgery for scoliosis and assistive technologies to support mobility and communication. Ataluren is available for people with Duchenne muscular dystrophy caused by a nonsense mutation in the dystrophin gene. Multidisciplinary care is delivered through specialist neuromuscular centres, known as NorthStar Centres. These provide coordinated access to physiotherapists, occupational therapists, respiratory and cardiac specialists, and neuromuscular clinicians.

NICE's scope defined the comparators as ECM without givinostat, and ataluren for people 2 years and over who have a nonsense mutation in the dystrophin gene and are able to walk. In its submission, the company included only ECM without givinostat, consistent with the population and treatment regimens used in the EPIDYS trial (see section 3.9). The company did not include ataluren as a comparator, explaining that it is offered to only a small proportion of people with Duchenne muscular dystrophy (see section 3.4). The company also did not include vamorolone in ECM. It said that this was because the corticosteroid was only recently approved and its long-term efficacy is uncertain. The company said that it would also likely be used in both arms of the analysis, with minimal impact on the relative treatment effect. The EAG agreed that ECM is the most relevant comparator. It also agreed that including vamorolone would likely have little impact on the relative-effectiveness results. But the EAG noted that vamorolone has a different adverse-event profile and cost compared with other steroids, which could affect resource use and health-state utility values. The clinical experts stated that vamorolone is currently prescribed to very few people; one centre reported that only 4 out of 100 people with Duchenne muscular dystrophy were having it. They explained that deflazacort was the most widely used steroid, followed by prednisolone, which aligns with the treatment patterns seen in EPIDYS. The committee concluded that ECM as defined by the company was an appropriate comparator for this evaluation.

The company restricted the population in its submission to people 6 years and over who are ambulant at the start of treatment (although there was no restriction in its submission on people continuing givinostat after LOA). This is consistent with the population in the EPIDYS trial and the EPIDYS open-label extension (OLE) study (see section 3.9 and section 3.10). But givinostat is licensed for the treatment of Duchenne muscular dystrophy in people 6 years and over, regardless of whether they are ambulant at the start of treatment.

The definition of ambulant used in Study 43, EPIDYS and the OLE study was the ability to walk independently. This was assessed by tests such as 4-stair climb test, the time to rise from the floor test or 6-minute walk test (see section 3.9 and section 3.10). In the UK real-world study, LOA was assessed using the 10-metre walk test (see section 3.12). The Cooperative International Neuromuscular Research Group (CINRG) Duchenne Natural History Study (DNHS) used full-time wheelchair use to define loss of ambulation (see section 3.13). The EAG expressed concern about differences between how ambulation is defined in clinical trials and clinical practice, and the challenges of applying a consistent definition across the NHS. Some of the assessments used in the trials, such as the 4-stair climb test, may not always be practical or routinely used in clinical settings. Clinical experts felt that the definitions used in trials and in practice were broadly similar. They noted that some measures, for example the 10-metre walk test, might not give the same results in a clinical trial setting and clinical practice. Both the clinical and patient experts emphasised that the real-world understanding of ambulation is broader and more closely aligned with what is known as the transfer stage. In this stage, people are no longer able to walk or rise from the floor unaided but can still stand with support. This transfer stage is considered an ambulatory state in the economic model (see section 3.19). The clinical and patient experts agreed that the definition of ambulant should include the transfer stage. Clinical experts considered this definition of ambulant to be clinically meaningful and practical to apply consistently in routine care. The committee thought that the definition of ambulant was relevant because of the company's restriction to the ambulant starting population and because it reflected how people transitioned through the economic model, and it could affect how the clinical trial evidence was interpreted. The committee agreed that defining ambulant as being able to walk or stand, with or without support, was appropriate for this evaluation.

The company restricted the population to people who are ambulant at the start of treatment (see section 3.6). This restriction only applied to the starting population, so people who started givinostat when ambulant and later lost ambulation could continue treatment. The clinical experts noted however that people who are non-ambulant at the start of the treatment may still benefit from treatment. So, restricting treatment start based on ambulation status could exclude people who might otherwise experience important benefit from treatment. The patient experts thought that givinostat should be available to all people with Duchenne muscular dystrophy and that it would be unfair to deny it to people who had already lost ambulation. A patient expert explained that using LOA as a cut-off is an arbitrary choice and that they felt it did not make sense to base access on this endpoint. A patient expert organisation submission stated that not extending any recommendation to a non-ambulant starting population would constitute discrimination (see section 3.40). The company explained that at the time of submission, it only had evidence for the ambulant starting population. So, it could only model the cost effectiveness of givinostat in this population. It noted there is a trial in progress called ULYSSES, which is assessing givinostat in a non-ambulant starting population. The committee acknowledged the broader licensed population and the views of clinical and patient experts. It discussed whether excluding the non-ambulant starting population from any recommendation could be considered discrimination, and considered NICE's commitment to eliminating discrimination, advancing equality, and fulfilling its legal and wider equality obligations. The committee felt that excluding a non-ambulant starting population because of a lack of evidence would be a proportionate means of making evidence-based recommendations and ensuring a cost-effective use of NHS resources. It stated that, while it prefers to consider the full population in the marketing authorisation, it could only do so where there is sufficient evidence. It acknowledged the potential benefits in the non-ambulant starting population, but it explained that no evidence in this population was presented. It also explained that it would expect the nature and scale of clinical benefits and the cost effectiveness to be meaningfully different in this group, given the possibility that the ambulant starting population might get a carryover effect from delaying LOA but the non-ambulant starting population would not (see section 3.25). The committee concluded that it was only able to make recommendations for people who start givinostat while ambulant, in line with the evidence that was presented by the company.

EPIDYS was a multicentre, randomised, double-blind, placebo-controlled phase 3 trial that enrolled 179 people aged 6 years and over with genetically confirmed Duchenne muscular dystrophy. All participants were having stable corticosteroid treatment at the start of the trial. The inclusion criteria specified that people had to still be able to walk without help at the time of enrolment. This meant that all recruited people were ambulant at the start of treatment, according to the trial definition of ambulant. Specifically, people had to complete the 4-stair climb test twice, each in 8 seconds or less and within 1 second of variance. They also had to do the time to rise from the floor test in 3 seconds or more but less than 10 seconds. People were stratified into 2 groups based on vastus lateralis fat fraction (VLFF) at baseline:

• group A (n=120), with VLFF over 5% and less than or equal to 30%

• group B (n=59), with VLFF less than or equal to 5% or over 30%.
  
  Participants were randomised in a 2:1 ratio to have either givinostat (n=118) or placebo (n=61) for 18 months. The primary objective was to assess the effect of givinostat on disease progression. Disease progression was measured by the change in the results of the 4-stair climb test between baseline and 18 months. In the full EPIDYS population (groups A and B; n=179), the log-transformed geometric least-squares mean ratio in change from baseline was 0.84 (95% confidence interval [CI]: 0.73 to 0.96; p=0.012). This indicated that givinostat significantly slowed decline in stair-climbing performance compared with placebo. Following completion of EPIDYS, people in the study were eligible to enter the OLE study.

The OLE study (n=207) evaluated the long-term safety, tolerability and efficacy of givinostat. It included people from EPIDYS and Study 43. Study 43 (n=20) is a phase 2 and 3, open-label clinical trial designed to evaluate the safety, tolerability, pharmacokinetics, and early efficacy of givinostat in people with Duchenne muscular dystrophy. As with EPIDYS, people in the study were also having stable corticosteroid treatment at study start. The inclusion criteria specified that people had to be able to walk independently at the time of enrolment. Specifically, people had to complete the 6‑minute walk test twice, achieving at least 250 metres and within 30 metres of variance. People who completed EPIDYS and Study 43 or who were eligible for EPIDYS but were not enrolled could enter the OLE study for continued monitoring of long-term safety and efficacy. In the OLE study, everyone had oral givinostat twice daily, with dosing adjusted based on body weight and blood parameters. Monitoring included regular assessment of platelet counts and other safety markers, with predefined criteria for dose adjustment or treatment discontinuation. Key clinical outcomes included time to LOA (defined as inability to walk 10 metres in 30 seconds or less), changes in motor function (such as stair-climbing ability and NorthStar Ambulatory Assessment scores), and respiratory function (including the need for NIV, or FVC less than 1 litre). At the start of the OLE study, 12 people (6%) were non-ambulant (having lost ambulation during EPIDYS or Study 43). The reported key results from the OLE study for median age at LOA were:

• 16.7 years in people who had givinostat in EPIDYS and Study 43 (the givinostat group; n=119)

• 17.3 years in people who had placebo in EPIDYS (the delayed givinostat group; n=58)

• not calculable in people who were recruited to EPIDYS (all group B) but because of timing were not randomised and were instead entered directly into the OLE study (the givinostat-naive group; n=30).
  
  No results are available for non-ambulatory outcomes such as NIV or FVC less than 1 litre because the data is immature (not enough events occurred). The OLE study is ongoing and is expected to provide further insight into the long-term impact of givinostat on disease progression and safety.

The company did not use relative-effectiveness data from EPIDYS in the economic model because of its short follow up (18 months). Instead, it did unanchored matching-adjusted indirect comparisons (MAICs). The company reweighted givinostat individual patient data from EPIDYS and the OLE study (see section 3.10) to better match the ECM data from both the UK real-word dataset (see section 3.12) and the CINRG DNHS (see section 3.13). The committee acknowledged that the clinical evidence is immature.

In response to the call for additional evidence, the company updated the MAICs using the committee-preferred datasets, the UK real-world dataset (see section 3.12) and the full givinostat population from EPIDYS and the OLE study (see section 3.10). But it still used acceleration factors to estimate age at NIV and FVC less than 1 litre. The unanchored MAIC results suggest that givinostat increases observed age at LOA, estimated age at NIV, and estimated age at FVC less than 1 litre compared with ECM. The results showed:

• for age at LOA:

  • hazard ratio: 0.201 (robust standard error 0.142 to 0.285)

  • givinostat observed median: 17.97 years

  • ECM median: 12.28 years

• for age at NIV:

  • hazard ratio: 0.206 (robust standard error 0.145 to 0.293)

  • givinostat estimated median: 27.00 years

  • ECM median: 18.46 years

• for age at FVC less than 1 litre:

  • hazard ratio: 0.218 (robust standard error 0.149 to 0.321)

  • givinostat estimated median: 34.90 years

  • ECM median: 23.86 years.
    
    The committee noted that the company had updated the MAIC in line with its preferences (see section 3.14 and section 3.15). It thought that the MAIC results were still associated with some uncertainty because of the uncertainties in the approximated givinostat data (see section 3.16) and the ECM data (see section 3.14).

The company base case applied a non-reference 1.5% discount rate to health outcomes only for the full time horizon of the analysis. The committee may consider a discount rate of 1.5% per year to cost and health outcomes for the full time horizon of the analysis if specific criteria are met. The criteria are that the technology:

• is indicated for people who would otherwise die or have very severely impaired life

• is likely to restore people to full or near-full health

• has benefits that are likely to be sustained over a very long period.
  
  The company acknowledged that givinostat does not meet all 3 criteria, but it suggested that the criteria for non-reference discounting excludes Duchenne muscular dystrophy. This is because people with the condition cannot, by definition, be restored to full or near-full health, despite benefits accruing over a long time. The patient experts agreed with the company. They also suggested that this may be a discrimination issue because it implies that curing disability is the ultimate goal and that disabled people do not lead fulfilled lives. They also stated that this criterion may breach NICE's duty to consider equity and long-term societal impact (see section 3.40). The EAG's clinical expert thought that it was possible that in the future, treatments may cure or at least stop the progression of Duchenne muscular dystrophy. Previous evaluations in Duchenne muscular dystrophy have consistently applied the reference-case discount rate of 3.5% across the full time horizon.
  
  The committee recognised that there is no provision in NICE's methods for only applying the 1.5% discounting rate to health outcomes, as the company had done in its base case. It stated that, if it were applied, the reduced rate should also apply to costs. The committee understood the role of the NICE methods and that changing those methods is beyond its remit, but it acknowledged the concerns around the 'restoring to full or near-full health' criterion. It considered how it should interpret and apply this criterion, including in the context of the equality issues raised. The committee noted that 'restoring to full or near-full health' could be interpreted in different ways and did not think that it would necessarily exclude progressive conditions. It recalled NICE's highly specialised technologies guidance on onasemnogene abeparvovec for treating spinal muscular atrophy, which looked at another progressive condition affecting children. In that evaluation, the committee applied the 1.5% discounting rate to the treatment even though many of the children who had it would never be able to walk independently. So, the committee did not think that progressive conditions were excluded from using non-reference-case discounting under the existing criteria. It also thought that in the context of a severe progressive condition such as Duchenne muscular dystrophy, it might reasonably interpret 'restoring to full or near-full health' to mean stopping progression. It noted that while givinostat delays disease progression, it does not stop it. In addition, the committee recalled that longer-term clinical effectiveness of givinostat is uncertain (see section 3.17), so it was unclear if the benefits would be maintained over a very long period. It concluded that givinostat does not meet the second criteria for applying a lower discount rate and might not meet the third criteria, and that a 3.5% discount rate should be applied to both costs and health outcomes.

The company's model structure is based on the HERCULES natural history model. It was developed through Project HERCULES, a collaborative initiative led by Duchenne UK. It is a Markov model that captures the progression of Duchenne muscular dystrophy though 8 clinically defined health states before death (state 9) and was previously used in TA1031. But the clinical experts in TA1031 found that the original natural history model overestimated survival compared with clinical expert opinion. So, the company updated the model. It used the SOLVER function to calibrate transition rates for the first 3 transitions in the model and for mortality transitions from health state 8. It did this to better reflect the UK real-world dataset on 50-year survival and key clinical milestones, such as LOA and initiation of ventilation. The EAG agreed that the UK real-world dataset is an appropriate source to inform the natural history model (see section 3.12). The clinical experts considered the median milestone values in the updated natural history model to be broadly consistent with what is seen in clinical practice. The EAG noted that the natural history model matched the median values from the UK real-world dataset but overestimated the tail of the curve. For example, the median (50th percentile) LOA was 12.2 years in the UK real-world dataset and 12.1 years in the natural history model. But, at the 75th percentile the UK real-world dataset was 13.7 years and the natural history model 15.3 years. The EAG thought that this showed the updated natural history model does not reflect the underlying survival distribution from the UK real-world dataset, and was a source of residual uncertainty. It noted that this may explain why applying the MAIC hazard ratios directly in the model led to implausible results (see section 3.22). The committee noted these residual uncertainties but concluded that the updated natural history model was suitable for decision making.

The HERCULES natural history model was used to estimate health-state occupancy for the ECM arm. There are 8 overall health states. These are implemented as 10 health states in the model, depending on the order in which hand-to-mouth function is lost and night-time ventilation is started (differentiating health states 7a and 7b, and 8a and 8b). In the model there are 9 transitions covering the various living health states. There are 3 ambulatory health states (1 to 3) and 5 non-ambulatory health states (4 to 8):

• health state 1: early ambulatory

• health state 2: late ambulatory

• health state 3: transfer

• health state 4: hand-to-mouth function and no ventilation

• health state 5: no hand-to-mouth function and no ventilation

• health state 6: hand-to-mouth function and night-time ventilation

• health states 7a and 7b: no hand-to-mouth function and night-time ventilation

• health states 8a and 8b: full-time ventilation.

At the first meeting, the company used the unanchored MAIC results with the UK real-world dataset and the post-hoc group from EPIDYS and the OLE study to inform the clinical effectiveness of givinostat. But the company did not use the MAIC hazard ratios directly in the model. It thought that this would triple count the treatment effects on the outcomes. Instead, it used the givinostat MAIC-adjusted medians to estimate model transitions using the SOLVER function (see section 3.23). Each givinostat median was used to estimate hazard ratios for 3 transitions, as follows:

• LOA for informing ambulatory transitions (transitions to health states 2, 3 and 4)

• NIV for informing transitions to health states 5, 6, and 7a (non-ambulatory transitions), and

• FVC less than 1 litre for informing transitions to health states 7b, 8a and 8b (non-ambulatory transitions).
  
  The company said its base case did not model any treatment effect beyond LOA because its SOLVER approach preserved the proportional relationship between the 3 milestones from the ECM in the givinostat arm. The EAG used the unanchored MAIC results with the CINRG DNHS dataset and the post-hoc group from EPIDYS and the OLE study to inform the clinical effectiveness of givinostat. Unlike the company, it applied the MAIC hazard ratio for age at LOA directly to the model. The EAG:

• used the LOA hazard ratio for informing ambulatory transitions (transitions to health states 2, 3 and 4) and

• assumed no difference between givinostat and ECM for all other transitions (hazard ratio of 1 for non-ambulatory transitions).
  
  The EAG explained that their approach relies only on the observed data. It thought that the company's approach did model a treatment effect after LOA. It also thought that if the aim was to not model any treatment effect beyond LOA, the SOLVER approach was needlessly complicated, unlike the EAG's approach. The committee was concerned about the estimates and application of the givinostat treatment effect using the 2 different approaches. It wanted to ensure that the givinostat treatment effect was correctly reflected in the model. So, in its call for additional evidence, it asked for further exploration of modelling the givinostat treatment effect and its application in the model.

At the first meeting, the company used utility values from a recent US study by Audhya et al. (2023), which collected EQ-5D-5L data directly from people with Duchenne muscular dystrophy (n=63). This was because the health-related quality-of-life data in EPIDYS and the OLE study was collected using the Pediatric Outcomes Data Collection Instrument (PODCI) and the Pediatric Quality of Life Inventory (PedsQL). But there was no validated mapping algorithm for PODCI and the company had concerns about the uncertainty introduced by mapping from PedsQL. Audhya et al. used a US population and value set but was mostly based on self-reported data (76%) and used the EQ-5D instrument preferred by NICE (albeit the 5L version rather than the 3L). The company also included a scenario based on Landfeldt et al. (2017). Landfeldt and colleagues wrote a number of publications based on a large, cross-sectional, observational study of people with Duchenne muscular dystrophy from Germany, Italy, the UK and the US (n=770). The EAG preferred to use data from the UK-based burden-of-illness (BOI) study (n=24), done under Project HERCULES. This study used the Duchenne Muscular Dystrophy Quality of Life (DMD-QoL) instrument and was 100% self-reported. The EAG thought that it was more relevant to the NHS setting and was the only study aligned with the health states used in the economic model. The EAG highlighted that the BOI study was the only source to distinguish between night-time ventilation and full-time ventilation (health states 7 and 8), an important clinical distinction.

At the first meeting, the patient expert read out a testimony from a person having full-time ventilation. It strongly challenged the assumption that the 2 ventilation states could be assigned the same utility value. While night-time ventilation allowed for periods of independence and lower care needs, full-time ventilation needed constant supervision, with any interruption posing a life-threatening risk. The patient expert explained that full-time ventilation is associated with a significant increase in anxiety, loss of privacy, and an inability to be left alone. It is also associated with a substantial rise in care costs and a marked decline in quality of life. They highlighted the emotional and practical burdens of managing complex care needs and the impact on daily activities and social interactions. The testimony conveyed a strong desire to return to the night-time ventilation stage, even briefly. The patient experts also raised concerns about the plausibility of the BOI utility values, noting that in some cases the values suggested improvements in quality of life as the disease progressed. For example, the BOI utility set used a lower utility value for health state 2 (late ambulatory, 0.49) than for health state 7 (no hand-to-mouth function, night-time ventilation, 0.52), which completely lacked face validity. The committee was concerned about face validity of the values reported in the studies and how reported utility values were assigned to model health states. It wanted to ensure that the model reflects the increasing impact on people as the condition progresses. So, it asked for patient health-related quality-of-life modelling and assumptions to be further explored during its call for additional evidence.

In response to a call for additional evidence, the company added a scenario using Crossnohere et al. (2021). This study included 263 people (74 from the UK) and used EQ-5D-3L with some self-reporting (23%) but applied a US tariff. But, the company maintained that Audhya et al. (2023) is the most appropriate source because it collected EQ-5D-5L data directly from people with Duchenne muscular dystrophy (76% self-reports). The EAG acknowledged limitations of the BOI study. It highlighted that because the company did not use the trial data, all available options have methodological and plausibility limitations. The EAG felt that the BOI study is still the most appropriate source given it was done as part of the HERCULES project and was used in TA1031. The BOI study, although aligned with the model, had incomplete and limited data for individual disease stages. A patient expert explained that despite being part of the HERCULES project, the BOI study has a number of methodological issues. These include difficulties matching responses with health states and a small sample size. As a result, it is not a reliable source for the patient utilities. The other patient expert had strong concerns about the BOI values. They again highlighted that the values lack face validity because it does not make sense that in a severe progressive disease, quality of life would decline and then go up again.

The patient experts also noted the limitations of using EQ-5D in Duchenne muscular dystrophy. For example, the EQ-5D focuses a lot on 'usual activities' and while many people with Duchenne muscular dystrophy can do their usual activities, they cannot do a range of other things. Also, it is very difficult to estimate quality of life in children. The committee acknowledged the challenges of collecting patient health-related quality-of-life data in rare, progressive diseases, and especially in those that affect children, such as Duchenne muscular dystrophy. It recognised the limitations associated with each data set but noted that quality of life is a fundamental component of cost-effectiveness analysis and that it would have to select a utility set for decision making. The committee heard the clear evidence from the patient experts, and so considered that the BOI study was not a reliable source. It said that trial data would have been preferred but understood that it was not available. From the options available, the committee concluded that the Audhya et al. (2023) utility set, with a multiplier from Landfeldt et al. (2017) to differentiate between state 7 and 8, is suitable for decision making.

At the first meeting, the committee discussed challenges in the modelling in 4 areas:

• Carer utility value data: the company used utility values from a UK vignette study and the EAG used Landfeldt et al. (2016).

• Number of carers modelled: the company modelled 2 carers in all health states (including state 9). The EAG modelled 1 carer per person in the ambulatory health states (health states 1 to 3) and 2 carers per person in the non-ambulatory states (health states 4 to 8), and no carers in state 9.

• Modelling approach: the company used an 'increment relative to midway approach', which calculated an increment relative to health state 4 and applied that increment to the life years gained for givinostat for each carer. The EAG used a 'disutility approach', which calculated decrements by subtracting the utility value for each health state from health state 1.

• Life-extension benefit: the company modelled a decrement of -0.56 (calculated from a utility of 0 for carers in state 9) to the negative incremental life years in state 9 for each carer (multiplying these 2 negatives gave a positive carer quality-adjusted life year [QALY] gain, which the company explained as an 'extension-to-life benefit for carers'). The EAG did not model any carers in state 9, and so there was no direct effect of life extension modelled on carers.
  
  The EAG thought that using the vignette study was not appropriate because direct measurements were available. It explained that Landfeldt et al. (2016) collected EQ-5D-3L, which is preferred by NICE. The committee acknowledged the limitations of all sources of carer utilities. It was concerned about the face validity of the values reported in the studies, how reported utility values were assigned to model health states, and the lack of evidence on number of carers. The EAG's preferred base case still resulted in a carer QALY gain with givinostat, but the gain was much smaller than in the company's base case. The EAG's approach captured an increasing impact of Duchenne muscular dystrophy on carers as the disease progresses. Because extending the time in each health state extends the duration of that impact, it leads to a prediction that extending life in Duchenne muscular dystrophy worsens carer quality of life. Patient experts also raised that the EAG's approach implied that extending the life of people with Duchenne muscular dystrophy worsens carer QALYs. The patient experts said that this was the exact opposite of their experience, which is that extending the life of people with Duchenne muscular dystrophy would improve carer QALYs. One patient expert explained that there is a physical burden of care that would no longer apply after the death of the patient (in state 9). But they said this would be completely outweighed by the emotional anguish of losing a child. They said that it was unreasonable to think that there would be no negative effect of losing a child on health-related quality of life. The committee understood why the EAG's modelling approach created the predictions it did, but it clearly heard the testimony of the patient experts and considered that the EAG's approach was not appropriate for decision making. But it also thought that the company's approach of using of the -0.56 decrement lacked clear justification, evidence and face validity. This approach modelled a QALY gain of 0.56 for each carer for each year of life extension for the patient having givinostat treatment. The committee did not consider this plausible. It concluded that carer health-related quality of life needs to be captured appropriately, but neither the company's nor the EAG's approaches to quantify this had done so. The committee acknowledged the sensitivity and critical importance of this issue for people with Duchenne muscular dystrophy and their families, and the substantial technical challenges associated with it. So, it concluded that further information and modelling was needed and requested further exploration of this in the call for additional evidence.

The company selected Morgan et al. (2024) as the primary source for health-state costs in the economic model. The study was based on the Project HERCULES UK longitudinal study using the Clinical Practice Research Datalink. It included scenario analyses using other sources, including the BOI study and Landfeldt et al. (2014). The EAG initially preferred the BOI study done under Project HERCULES. But, following technical engagement, it agreed with the company that Morgan et al. is the most suitable and robust source for health-state costs. In response to clarification questions, the company included a scenario in which a multiplier was used to increase the cost in state 8. Similarly, it included a scenario using a multiplier and inflating costs to include tertiary-care and medical-aid costs. Nonetheless, the EAG maintained its view that the cost estimates in later stages of the condition, when the burden on NHS and personal social services resources increases, may be underestimated. The committee was concerned that adding tertiary-care and medical-aid costs, and differentiating between health states 7 and 8, were not fully explored. It wanted to ensure that the increasing costs as the condition progresses were reflected in the model. So, it asked for cost modelling and assumptions to be further explored during its call for additional evidence.

In response to the call for additional evidence, the company made no changes to its base case and did not introduce additional scenarios. It highlighted that scenarios differentiating between health states 7 and 8 and costs for tertiary care and medical aid were already explored. The EAG reverted to using the BOI study because it does differentiate between health states 7 and 8 and because it was used in TA1031. It used annual direct medical cost used in TA1031 and applied it in its preferred base case. The committee noted that both the EAG's and the company's base-case costs, although using different sources, are similar. The patient expert explained that despite being part of the HERCULES project, the BOI study has a number of methodological issues and is not a reliable source for the cost data. They further explained that accessing cost information is particularly challenging, especially in the later stages of the condition. The committee acknowledged challenges in collecting reliable cost data for rare, progressive diseases like Duchenne muscular dystrophy and the resulting uncertainty in the available data. It noted that the choice of cost inputs does not significantly impact the overall cost-effectiveness estimates. The committee considered patient expert testimony regarding limitations in the BOI study and concluded that Morgan et al. (2024) should be used to inform decision making.

The committee considered the severity of the condition (the future health lost by people living with the condition and having standard care in the NHS). The committee may apply a greater weight to QALYs, called a severity modifier, if technologies are indicated for conditions with a high degree of severity. The company provided absolute and proportional QALY shortfall estimates in line with NICE's manual on technology appraisal and highly specialised technologies guidance. The company assumed a starting age of 6 years based on the age in the marketing authorisation. The company base case resulted in a severity modifier of 1.7. Applying the committee's preferred assumptions (see section 3.38) also resulted in a severity modifier of 1.7. The committee understood that Duchene muscular dystrophy is a severe, progressive and fatal disease that arises in childhood and causes significant effects on health-related quality of life. So, the committee concluded that the severity weight of 1.7 applied to the QALYs was appropriate.

The committee recalled its preferred assumptions, which were to:

• use ECM based on treatment regimens used in EPIDYS as a comparator (see section 3.5)

• use the UK real-world dataset as a source for ECM data in the unanchored MAIC (see section 3.14)

• use the full givinostat population from EPIDYS and the OLE study as a source of givinostat data for the unanchored MAIC (see section 3.15)

• use the reference-case discount rate of 3.5% for outcomes and costs (see section 3.18)

• use the company's updated natural history model (see section 3.19)

• use SOLVER to apply treatment effect to the model (see section 3.23)

• use the EAG scenario of modelling treatment effect beyond LOA as 100% of treatment effect on LOA for NIV (see section 3.26)

• use Audhya et al. (2023) to model patient utilities (see section 3.28)

• use the 1.7 severity modifier (see section 3.37)

• use Sheffield 3 data as a source for carer health-related quality of life (see section 3.30)

• assume 2 carers per person in the ambulatory health states, and 3 carers per person in the non-ambulatory health states (see section 3.31)

• use the EAG's utility approach to modelling carer health-related quality of life (see section 3.32)

• include the EAG's bereavement disutility scenario (see section 3.33)

• use Morgan et al. (2024) with the company's 1.30-times multiplier differentiating between health states 7 and 8, and the 3.05-times multiplier increasing costs to include tertiary-care and medical-aid costs (see section 3.35 and section 3.36).
  
  When taking into account how severe Duchene muscular dystrophy is and how it affects both quality and length of life, the committee's preferred ICER was marginally higher than £35,000 per QALY gained. Because there is a confidential discount for givinostat, the exact ICER cannot be reported here.

NICE's manual on technology appraisal and highly specialised technologies guidance notes that, above a most plausible ICER of £25,000 per QALY gained, judgements about the acceptability of a technology as an effective use of NHS resources will take into account the degree of certainty around the ICER, aspects that relate to uncaptured benefits and non-health factors, and aspects that relate to health inequalities. The committee will be more cautious about recommending a technology if it is less certain about the ICERs presented. But it will also be mindful that there are certain circumstances in which evidence generation is particularly difficult. The committee understood that above a most plausible ICER of £35,000 per QALY gained, it would need to identify an increasingly stronger case for supporting the technology as an effective use of NHS resources.

The committee noted the high level of uncertainty, specifically the:

• lack of adjustment of potential prognostic factors in the MAIC (see section 3.14)

• overestimation of the tails of survival probabilities by the updated natural history model (see section 3.19)

• inability to use the MAIC hazard ratios directly in the model without generating implausible results (see section 3.22)

• need for SOLVER to estimate model transitions (see section 3.23)

• immaturity of the trial data and requirement to approximate givinostat age at NIV and FVC less than 1 litre data (see section 3.16)

• face validity concerns about why givinostat reduces time in health state 8 despite extending time in all other health states (see section 3.36)

• application of givinostat treatment effect in the model in the context of discontinuation (see section 3.26)

• patient health-related quality-of-life assumptions (see section 3.28)

• carer health-related quality-of-life assumptions (see section 3.30 to 3.33)

• exclusion of local authority or NHS care costs from the model (see section 3.36).
  
  The committee noted the considerable challenges in generating robust evidence for a rare, progressive childhood condition with a long-term disease course such as Duchenne muscular dystrophy. It recognised that long-term outcomes are difficult to capture within feasible study timeframes. Evidence generation is also further constrained by limited sample sizes, heterogeneity in disease progression, and the challenges and complexities of involving children and young people in clinical trials. Together, these factors restrict the availability of high-quality evidence and increase the degree of uncertainty, leading to greater reliance on modelling assumptions. The committee understood that the NICE's manual on technology appraisal and highly specialised technologies guidance specifies that it can make recommendations accepting a higher degree of uncertainty in specific circumstances where evidence generation is particularly difficult. The committee recognised that the difficulties in collecting evidence in Duchenne muscular dystrophy meant that a greater tolerance for uncertainty than usual was appropriate. Moreover, it recognised the significant attempts to mitigate uncertainty in Duchenne muscular dystrophy through Project HERCULES. It considered the other factors that it should take into account in its decision making, and noted in particular the impact of disease progression, and of delaying disease progression, at key stages in young people's lives. Taking all these considerations into account, and considering the full range of evidence and the specific circumstances of this appraisal, it concluded that there was a sufficient case to conclude that givinostat would represent an effective use of NHS resources.

The committee understood the nature of Duchenne muscular dystrophy and its profound impact on people with the condition and their carers and families. It recognised the importance of effective treatments that could delay progression, and noted the clinical effectiveness evidence for givinostat. When taking into account how severe Duchenne muscular dystrophy is, and how it affects both quality and length of life, the most likely cost-effectiveness estimates are within the range that NICE considers an acceptable use of NHS resources (see section 3.39). So, givinostat can be used routinely in the NHS to treat Duchenne muscular dystrophy in people 6 years and over who are ambulant (able to walk or stand, with or without support) at the start of treatment (see section 3.8).