Draft guidance consultation

The clinical experts explained that the main aim of treatment for severe haemophilia A and B is to prevent bleeding and long-term damage, especially to joints. This is achieved through prophylaxis to prevent bleeds and on-demand treatment for bleeding episodes when needed. The available treatment options for haemophilia A for long-term prophylaxis are factor VIII replacement therapy, through intravenous injection, to replenish missing clotting factor in the blood, and emicizumab through subcutaneous injection. Emicizumab, a non-factor VIII treatment, is commissioned through an NHS England clinical commissioning policy as prophylaxis for severe congenital haemophilia A in people of all ages without anti-factor antibodies. Emicizumab is a monoclonal antibody administered every 1 to 4 weeks and mimics the activity of factor VIII to restore clotting function. The available treatment option for haemophilia B for long-term prophylaxis is factor IX replacement therapy, through intravenous injection. For both severe haemophilia A and B, standard and extended half-life factor replacement therapies are available and extra on-demand factor replacement therapy can also be used for treating bleeds. The committee noted that NICE's technology appraisal guidance on etranacogene dezaparvovec for treating moderately severe or severe haemophilia B (TA989) recommended it through a managed access programme. But etranacogene dezaparvovec is not a relevant comparator in this evaluation. The committee concluded that treatment for severe haemophilia A includes prophylaxis with factor VIII replacement therapy or emicizumab, and treatment for severe haemophilia B includes prophylaxis with factor IX replacement therapy.

The clinical and patient experts explained that current treatment options do not always prevent bleeding and are associated with administration challenges. Frequent injections for factor replacement therapy can damage veins, resulting in pain on administration and increasing the chance of 'vein collapse'. The frequency of injections is especially challenging in older people and young children who may have poor venous access. The patient experts stated that as they have become older they are limited on which veins they can use because of scar tissue. This can make treatment more painful and difficult to administer. Also, about 5% to 7% of people with haemophilia A develop antibodies (inhibitors) to factor VIII and about 5% of people with haemophilia B develop antibodies to factor IX. Anti-factor antibodies make factor replacement therapies less effective. The committee heard that people with anti-factor antibodies, particularly those with severe haemophilia B, have limited treatment options. The clinical experts explained that the outcomes for these individuals are worse compared with people who do not develop antibodies. They may be wheelchair bound by their second decade because of worsened bleed control and increased bleeds in the joints. The committee noted that because of the anticipated marketing authorisation, people with anti-factor antibodies are outside the scope of this appraisal (see section 2.1). In the NHS, most people with severe haemophilia A are offered emicizumab, an alternative to factor replacement therapy, which is administered subcutaneously. The patient experts explained that there is no single best treatment option for everybody; different people will value aspects of treatment options differently. For example, some people will value a subcutaneous option, while others value an intravenous option. People who participate in sports, for example, may prefer intravenous options so they can have flexibility in adjusting their required dose. The committee acknowledged there are no subcutaneous treatment options available for severe haemophilia B. It concluded that new treatment options for severe haemophilia A and B would be welcomed.

The clinical evidence for marstacimab came from BASIS, an unpublished phase 3 open-label non-randomised one-way non-inferiority cross-over trial. This enrolled people 12 years and over with severe haemophilia A or severe haemophilia B who had no antibodies to factor VIII or IX. The observational phase enrolled 128 people who previously had either:

• factor VIII or factor IX prophylaxis and who had at least 80% adherence with the scheduled prophylaxis regimen during the 6 months before enrolment (n=91), or

• on-demand treatment with 6 or more acute bleeding episodes that required coagulation factor infusion during the 6 months before enrolment and would be willing to continue on-demand treatment during the observational phase (n=37).
  
  The active treatment phase included 83 people who had used factor VIII or IX prophylaxis in the observational phase. They had 300 mg of marstacimab as a loading dose followed by 150 mg once weekly, self-administered by subcutaneous prefilled syringe injection. After 6 months, people could have their dose increased to 300 mg once weekly. This was decided by a clinician based on the dose escalation criteria:

• body weight at least 50 kg

• at least 2 spontaneous bleeds in a 6-month period treated with on-demand factor VIII or IX.
  
  The primary outcome in BASIS was the annualised bleeding rate (ABR) for treated bleeds through the observational phase and active treatment phase. The non-inferiority margin was 2.5 bleeds per year. The committee understood that if non-inferiority of marstacimab compared with factor prophylaxis was shown, then superiority was tested. Key secondary outcomes included annualised joint bleeding rate, spontaneous bleeds, target joint bleeds and total bleeds (treated and untreated) at 12 months after starting marstacimab. The number of people with no treated bleeds was recorded. The company also included data from an open-label extension study of 88 people who had prophylaxis for 6 months before BASIS and continued to have factor prophylaxis during the observational phase.
  
  The committee noted that the BASIS trial design was not aligned with UK practice. The committee concluded that the relevant evidence for marstacimab came from BASIS and its open-label extension study. But it noted the following limitations with the trial design:

• no randomisation within the trial

• being open label, there could be a differential reporting of bleeds

• no UK trial sites (see section 3.6)

• emicizumab was not permitted in the observational phase of the trial, but this is a relevant comparator for people with severe haemophilia A in the NHS

• for people who had routine factor replacement prophylaxis, the proportion who had standard half-life or extended half-life was not reflective of the proportions used in the NHS (see section 3.6).

The company presented results from the 6-month observational phase and 12-month active treatment phase of BASIS, and results for a data cut from the open-label extension study. The EAG advised that dose escalation occurred after 6 months of the active treatment phase and the 12-month results presented by the company for the active treatment phase were censored for dose escalations. The EAG requested additional data at 6 months to view results before any dose escalation happened. The BASIS trial demonstrated an improvement with marstacimab over factor prophylaxis, as measured by the ABR of treated bleeds. The mean model-derived ABR for treated bleeds was 5.08 (95% confidence interval [CI] 3.40 to 6.77) for marstacimab during the 12-month active treatment phase compared with 7.85 (95% CI 5.09 to 10.61) for routine prophylaxis during the 6-month observational phase. This represented a reduction in estimated ABR for treated bleeds of 2.77 (95% CI -5.37 to -0.16). The mean ABR for treated bleeds in haemophilia A was 5.30 for marstacimab during the 12-month active treatment phase compared with 9.16 for routine prophylaxis during the 6-month observational phase. For haemophilia B, the mean ABR for treated bleeds was 4.71 for marstacimab during the 12-month active treatment phase compared with 3.26 for routine prophylaxis during the 6-month observational phase. Six people with haemophilia B increased their dose of marstacimab during the active treatment phase. The mean ABR for treated bleeds for all 18 people with haemophilia B was 3.88 for marstacimab. People having marstacimab had a reduction in ABR for treated joint bleeds from baseline.

There were no trials directly comparing marstacimab with emicizumab, so the company did an indirect treatment comparison (ITC) to establish the relative efficacy. The clinical-effectiveness data for emicizumab came from HAVEN-3. HAVEN-3 was an open-label study including 152 people 12 years and over with severe haemophilia A and no anti-factor antibodies. It had 4 arms including people who:

• had on-demand regimens (instead of factor prophylaxis), randomised to have 1.5 mg/kg emicizumab weekly (arm A), 3 mg/kg every 2 weeks (arm B) or no prophylaxis (arm C)

• had prophylaxis regimens, who had 1.5 mg/kg emicizumab weekly (arm D).

The company developed a 3-state Markov model to determine the cost effectiveness of marstacimab. The health states were 'no bleeds', 'bleeds' and 'death'. Patients could transition between the 'bleeds' and 'no bleeds' health states. All people entering the model were distributed among the 'no bleeds' and 'bleeds' health states. People having factor prophylaxis or marstacimab were distributed based on the BASIS observational and active treatment phase results, respectively, for the proportion of people with zero treated bleeding events. Within the 'bleeds' health state both joint and non-joint bleeds were considered. The company estimated ABR for treated bleeds and annualised joint bleed rate for treated joint bleeds were estimated from a post-hoc analysis of BASIS data for factor prophylaxis and marstacimab, and evidence from the company's ITC for the ABR for emicizumab. In the factor prophylaxis and emicizumab arms, distribution across the 'no bleeds' and 'bleeds' health states was unchanged across the model time horizon. People were assumed to remain on treatment. In the marstacimab arm of the model, the company assumed that the distribution of people in each health state remained constant over time while on treatment, and also assumed that people later redistributed across these health states to model stopping treatment. The company chose a cycle length of 1 year with a half-cycle correction and a time horizon of 64 years. Some people transitioned to death, aligned with general population mortality. That is, no mortality benefit was assumed for marstacimab and people with severe haemophilia A or B were assumed to have the same mortality as the general population. Treated bleeds were associated with a utility decrement per joint and non-joint bleed, and each bleed was associated with a cost. The company modelled a utility decrement per administration for subcutaneous treatments, marstacimab and emicizumab, and for intravenous treatments, factor VIII and factor IX replacement therapies. The committee concluded that the company's model structure was appropriate for decision making.

The company base case used the BASIS data to inform the ABRs of marstacimab and factor prophylaxis treatments and the ITC to inform the bleeding rates of emicizumab compared with marstacimab. The EAG advised that prophylaxis treatments used in BASIS did not reflect NHS standard of care (see section 3.6) and the results of the ITC were highly uncertain (see section 3.7). The EAG decided that the UKHCDO data was the best source to inform the ABRs of factor prophylaxis and emicizumab. It calculated estimates of bleed rates across adult and 12 years and over age groups using weighted averages. The results using the UKHCDO data are confidential and cannot be reported. The UKHCDO uses Haemtrack, a database used by people with haemophilia to record when they experience a bleed. The EAG calculated baseline ABRs for treated bleeds by combining the reported ABR from the UKHCDO report with the balance of standard half-life and extended half-life factor prophylaxis used in BASIS. It then derived a real-world UK-specific baseline ABR estimate for the BASIS population. Bleeding rates for marstacimab were calculated by applying relative effectiveness estimates from BASIS to the real-world UK-specific baseline ABR estimate. This estimate was used when applying the relative effects as estimated from BASIS to derive the efficacy of marstacimab. The committee considered whether the UKHCDO dataset is reliable because it relies on people self-reporting. The clinical experts confirmed that the UKHCDO data is broadly reflective of bleed rates in the NHS. The patient experts stated that recording bleeding episodes on Haemtrack has become easier because it can now be done through an app on a mobile device. So people are more likely to record bleeds compared with previous methods of recording. The committee noted the limitations associated with BASIS and the uncertainty in the company ITC. It accepted the views of the clinical and patient experts that UKHCDO is a reliable source of self-reported bleed rates. The committee concluded that using the UKHCDO baseline ABR for factor prophylaxis and emicizumab is preferable compared to those obtained from BASIS and the ITC. This is because it is more reflective of NHS practice.

The BASIS trial protocol allowed dose escalation after 6 months in the active treatment phase and dose escalation could continue in the open-label extension study. People who dose escalated had their dose of marstacimab increased from 150 mg to 300 mg. A clinician decided whether a person dose escalated once they met the dose escalation criteria defined in the trial; not everybody who met the criteria did dose escalate. The company's base case included dose escalation only in the first year of the model, and this was modelled as 13.25%. The model captured a clinical benefit of dose escalation, but people could not revert to the original dose because ABR for treated bleeds is modelled as constant thereafter. The EAG adjusted the percentage of dose escalation to reflect that nobody who dose escalated stopped treatment, so this figure is higher than the company's estimate. The EAG noted that people were able to dose escalate during the open-label extension study, which is beyond year 1 and therefore not captured in the model. So, the EAG's base case modelled dose escalation for year 2 of the model. The patient experts explained that, in the NHS, dose escalation may take more than a year to implement because people would have a conversation with their clinician and choices around dose escalation are complex. The committee noted that decisions about dose escalation may depend on bleeding rates. It acknowledged that the modelled ABRs were lower using the UKHCDO data (see section 3.6) and that using this data may lead to a lower rate of dose escalation compared with the rate used in the company model. Clinical experts confirmed that some NHS clinical settings have zero tolerance for bleeding episodes, therefore dose escalation could be higher with this practice. If a person had 2 spontaneous bleeds in a 6-month period on a treatment they could either dose escalate, or stop treatment and have factor prophylaxis. The company stated that dose escalation varied between sites in the trial, but this data was not available for the committee. The patient experts explained that in the case of emicizumab, people accept a lower treatment efficacy compared with factor prophylaxis because they value the ease of subcutaneous administration compared with factor VIII administration. So, people may be willing to dose escalate rather than switch treatments to have the benefit of subcutaneous injections. The committee concluded that there is uncertainty about the proportion of people who would dose escalate in the NHS. It noted that the dose escalation rate could be higher or lower than the rate assumed by the company. It recognised that dose escalation may occur beyond year 1, so preferred the EAG's base-case assumption of including dose escalation in the second year of the model to reflect the trial evidence.

In its model, the company assumed that people could stop marstacimab and switch to factor prophylaxis. The company applied a one-off discontinuation rate for marstacimab of 6.02% in the first cycle based on the rates of discontinuation in BASIS. The company stated it did not apply a discontinuation rate for emicizumab because of a lack of data availability. The EAG advised that this is a source of uncertainty and probable bias. The EAG retained this assumption in its base case, but presented a scenario that removed the discontinuation rate for marstacimab so neither treatment was stopped. This increased the incremental cost-effectiveness ratio (ICER) for marstacimab. The clinical expert confirmed that about 10% of people stop using emicizumab. The committee concluded that this is a source of uncertainty in the company's modelling, and that some people on emicizumab would also likely stop treatment. The committee requested further analyses from the company to model a plausible scenario, matching UK practice, that is coherent between the discontinuation of marstacimab and emicizumab. This should include a scenario of applying a treatment discontinuation rate of 10% for emicizumab.

The company did not use the BASIS data to estimate the dosage of factor prophylaxis in its base case. The company took the summaries of product characteristics (SmPCs)-recommended dosage and weighted it by market shares estimated by iQVIA, a global healthcare company providing real-world data. The company also assumed drug wastage and dosing and rounded up these figures. The EAG advised that the dosing figures for factor prophylaxis in the company's base case were 18% higher compared with the mean total prophylaxis dose in the BASIS trial. The EAG did an analysis using the proportion of people having standard half-life and extended half-life factor prophylaxis in BASIS split by haemophilia A and B and applied these weights using iQVIA market share data. This showed a mean total prophylaxis dose 21% higher than the value in BASIS. The EAG suggests this result is either because dosing during BASIS was lower than in the NHS (which may imply suboptimal treatment during the observational phase) or routine prophylaxis dosing in the model is too high. The EAG explored the second option by reducing the total factor prophylaxis dosing to 75% of the company's estimates. The clinical experts confirmed that they use SmPC recommendations for factor dosing because they want to aim for zero bleeds, so dosing is at the higher end of the recommendations. The company stated that their methods of using SmPC dosing aligned with TA989. The EAG noted that the total factor VIII issued per person per year in the model was higher compared with figures in the UKHCDO annual report. The EAG also noted that UKHCDO included factor VIII used for both routine prophylaxis and treating bleeds, but the company's estimates only included factor VIII used for routine prophylaxis. The committee considered both the company's and EAG's approaches. It noted that the company may have overestimated dosing of factor prophylaxis compared with the UKHCDO data, which it considered was representative of NHS practice. It concluded that the EAG's base-case assumption of reducing the factor dosing to 75% of the company base case was preferred.

The company applied no haemophilia-specific ABR estimates or dose escalation for haemophilia A and B. The company stated that the BASIS trial and open-label extension was not powered to detect differences in ABRs of marstacimab between haemophilia A or B (which were subgroups of the trial population). Marstacimab had a smaller treatment effect for severe haemophilia B compared with severe haemophilia A (see section 3.5). The company and EAG both provided scenario analyses using the separate estimates of clinical effectiveness for haemophilia A and B. The clinical experts at the meeting noted that there is likely no biological reason why marstacimab would be less effective for severe haemophilia B. The committee noted that the confidence intervals for haemophilia B crossed the boundary of non-inferiority, which was set by the company as an ABR of 2.5. The committee acknowledged that the haemophilia B population in the trial was small (n=18) and evidence generation in this group is difficult because haemophilia B is less common than haemophilia A. The committee considered whether the model captured different pathway parameters for haemophilia A and B. It noted that the company model may not accurately reflect treatment discontinuation because the same discontinuation rate was applied to both haemophilia A and B. The committee accepted the analyses that included the use of pooled ABR data from people with haemophilia A and B. But it remained concerned that the model had not captured other differences in haemophilia A and B, for example, discontinuation of treatments. The committee requested the company provide further clarity on whether the model captured different treatment pathways and parameters for haemophilia A and B.

In its base case, the company derived differences in quality of life between the 'bleeds' and 'no bleeds' health states through utility decrements applied to acute joint and non-joint bleed events, which only occur in the 'bleeds' state. The company decided it was unfeasible to use EQ-5D scores collected in BASIS because they may not have reflected changes in health-related quality of life related to acute bleed events. The company applied 2 types of utility decrements from the literature dependent upon the type of bleed experienced in the 'bleeds' health state:

• joint bleeds had a utility decrement of 0.28, taken from O'Hara et al. (2018)

• non-joint bleeds had a utility decrement of 0.16, taken from Neufeld et al. (2012).

The company's model applied treatment administration-related utility decrements. It applied a utility decrement of 0.0003 for each intravenous injection for factor prophylaxis, and a utility decrement of 0.0002 for each subcutaneous injection for marstacimab and emicizumab. These values were taken from Johnson et al. (2021). The EAG noted this paper was sponsored by Hoffman-La Roche, the manufacturer of emicizumab which is administered subcutaneously. Johnson et al. (2021) did a time trade off by presenting scenarios for 6 health states. The EAG was concerned that the vignettes did not represent factor prophylaxis administration in the NHS because most people self-administer factor prophylaxis at home, but the vignettes placed more emphasis that factor prophylaxis would be administered in a clinic or hospital. The EAG was also concerned that the disutility estimates for treatment administration were provided at 4 decimal places. So, the actual mean disutility for intravenous administration could be in the range of 0.00025 to 0.00035 and the mean disutility for subcutaneous administration could be in the range of 0.00015 to 0.00025. The EAG noted that because of rounding there may be little to no difference between the 2 central estimates. Also, there were no reported confidence intervals for these values. There is uncertainty whether these confidence intervals would overlap. The EAG's base case provided scenarios for halving utility decrement for treatment administration or applying no utility decrement. The committee was concerned that the time trade off was capturing not only different treatment administration but also different treatment effects, because the bleeding rates were higher in treatments that had to be administered intravenously. The differences may have been exaggerated. The committee noted these 2 issues were not separated out for each health state. The patient experts stated that they could self-administer factor prophylaxis by intravenous injection quickly but noted that as they have become older it is becoming more difficult to find a vein because scar tissue has formed. The patient experts, in their submission, said that there is a high treatment burden associated with frequent intravenous infusions, and that patients and carers value ease of administration and a reduction in treatment burden. The committee concluded that the evidence presented by the company had serious limitations and lacked adequate explanation but acknowledged there would be some benefit for a subcutaneous treatment option. The committee requested that the company provide further evidence and modelling for the effect of administration method on health-related quality of life, including a full explanation and justification for the approaches taken. The committee concluded there is likely to be a difference in utility decrements between treatment administration methods, but this remains a source of uncertainty and the company's approach may have overestimated the benefits of marstacimab. It further concluded that there was insufficient evidence to justify the treatment administration utility decrements used in the company's model.

The committee concluded that the cost-effectiveness estimates were uncertain and further justifications for assumptions are needed (see sections 3.14 and 3.15). It agreed the company's overall model structure was acceptable for decision making (see section 3.8). It concluded that its preferred assumptions were:

• using UKHCDO data to model treatment effectiveness for baseline ABR for factor prophylaxis and emicizumab (see section 3.9)

• including dose escalation in year 2 of the model (see section 3.10)

• reducing the factor VIII and factor IX prophylaxis doses to 75% of the company's base case (see section 3.12)

• using the pooled clinical-effectiveness estimates for haemophilia A and B (see section 3.13)

• using a single utility decrement for bleed events of 0.16 applied for a period of 2.5 days (see section 3.14).

The committee acknowledged there was uncertainty in the evidence about stopping emicizumab and for the treatment-related and bleed rate utility decrements. It would like to see the following analyses and further evidence to enable it to decide on the cost effectiveness of marstacimab:

• analyses to plausibly model discontinuation of treatments including emicizumab, including scenarios in which 10% of people stop emicizumab (see section 3.11)

• exploration and justification for the evidence on bleed rate utility decrements (see section 3.14)

• further evidence and justification to model treatment-related utility decrements (see section 3.15)

• further exploration of the pathway and parameters of haemophilia A and B captured in the model.

Because of confidential commercial arrangements for marstacimab, the comparators and other treatments in the model, the exact cost-effectiveness estimates are confidential and cannot be reported here. Using the committee's preferred assumptions the results showed:

• For haemophilia A, marstacimab was dominated by emicizumab (that is marstacimab is more expensive and less effective than emicizumab) and against a factor VIII basket of weighted comparators the ICER exceeded £1 million per quality-adjusted life-year (QALY) gained.

• For haemophilia B, the ICER for marstacimab against a factor IX basket of weighted comparators exceeded £1 million per QALY gained.
  
  The committee noted that these analyses did not address key uncertainties, including the uncertainty of dose escalation rates, discontinuation rates and utility decrements for treatment administration. The committee concluded that the cost-effectiveness estimates were highly uncertain, and that further evidence was needed to establish plausible ICERs.

The committee noted that a recommendation in severe haemophilia A or B would not be affected by biological sex. Stakeholders advised that some of the treatments for haemophilia A are derived from human blood or human or animal cells. This may not be considered acceptable by people with some religious beliefs. In haemophilia A, the committee noted there are several treatment options. These include emicizumab, which is not derived from human blood products. The committee did not identify this as an equalities issue that would affect its recommendations. Stakeholders and clinical experts explained that some people may have difficulty self-administering an intravenous factor treatment if they have joint damage or a separate disability in addition to haemophilia. The committee noted for haemophilia A there are alternative treatments to intravenous administration, such as emicizumab. The committee concluded that all equalities issues for marstacimab had been considered in its decision making.

The committee considered whether there were any uncaptured benefits of marstacimab. It did not identify additional benefits of marstacimab not captured in the economic modelling. So, the committee concluded that all additional benefits of marstacimab had already been taken into account.

The committee noted the important uncertainties in the modelling, including utility decrements and treatment discontinuation. The committee could not conclude that marstacimab would represent a cost-effective use of NHS resources. So, marstacimab is not recommended.