Draft guidance consultation

The company positioned ruxolitinib as an alternative to the treatments that are currently used in NHS practice for corticosteroid-refractory acute GvHD. NHS England's clinical commissioning policy on treatments for graft versus host disease following haematopoietic stem cell transplantation (PDF only) was issued in 2017. It recommends that people with moderate, severe or very severe acute GvHD (grades 2 to 4) should be treated first with systemic corticosteroids. For acute GvHD that is refractory to corticosteroids, the policy recommends treatment with extracorporeal photopheresis (ECP), a blood-filtering procedure in which white blood cells are collected, treated with a light-activated drug, exposed to ultraviolet light, and returned to the body. The clinical experts noted that ECP is the most common treatment for corticosteroid-refractory acute GvHD in the NHS. But, practice varies, and clinicians use a variety of treatments, many of which are unlicensed for treating acute GvHD. The clinical experts noted that some of this variation is driven by issues with accessing ECP. ECP is available at a few therapeutic apheresis services and a limited number of hospital trusts. The patient experts explained that some people must commit significant time and money to travel for treatment, require invasive venous access, and often have long hospital stays. They also noted that other available treatments have many limitations. For example, immunosuppressants, such as corticosteroids, can cause a range of adverse effects that reduce quality of life, including an increased risk of infections and diabetes. Because people with acute GvHD have suppressed immunity, they are already prone to frequent infections and may require admission to hospital. To prevent this, they may need to isolate, which further impairs their quality of life. The patient experts described how much they would value ruxolitinib because, as an oral treatment, it may be able to be taken at home. This may also reduce infection risk. The committee concluded that the current treatment for acute GvHD has many limitations for clinicians and people with the disease, there is a substantial unmet need for new treatments, and ruxolitinib could address some of these issues.

The clinical-effectiveness evidence for ruxolitinib for acute GvHD came from 2 trials: REACH1 and REACH2. REACH1 was a US-based single-arm phase 2 study of ruxolitinib. REACH2 was a phase 3 randomised, controlled, open-label, superiority trial that compared ruxolitinib with the investigator's choice of standard care. It was done across 22 countries, including the UK. Both REACH1 and REACH2 included people 12 years and over with corticosteroid-refractory acute GvHD following an allogeneic HSCT. In REACH2, 154 people were randomised to ruxolitinib 10 mg twice daily and 155 people were randomised to standard care (see section 3.5 for a discussion on the generalisability of the standard-care treatments). The committee concluded that the randomised trial (REACH2) was important for decision making, but that the uncontrolled trial (REACH1) was not.

The marketing authorisation for ruxolitinib is for 'patients aged 12 years and older with acute graft versus host disease who have inadequate response to corticosteroids'. The EAG noted that only 3% of people in REACH2 were 12 to 17 years, and that its eligibility criteria did not include people with grade 1 acute GvHD. So, REACH2 did not capture these populations. The company referred to advice from clinicians, who suggested there was no difference between adults and young people in acute GvHD manifestation, pathophysiology or treatment options. The clinical experts in the meeting agreed with the company that they would expect little difference between young people and older people in disease manifestations or outcomes, or in the magnitude of effectiveness when using ruxolitinib compared with standard care. The company also noted that the option for treating grade 1 disease was essential for people who are at high risk of developing disease at a severity of grade 2 or higher. The committee concluded that although REACH2 did not capture these populations, it was satisfied that its recommendations were generalisable to young people.

The standard-care arm of REACH2 permitted healthcare professionals and participants to choose the treatment. The EAG questioned whether the standard-care treatments used in REACH2 reflected NHS practice. The clinical experts noted that ECP was the preferred treatment in the NHS and would be used in practice in more people than in REACH2 (where 27% of people had it). They explained that ECP is preferred because of higher perceived efficacy, a lack of good evidence for the other comparator treatments, and the NHS clinical commissioning policy (see section 3.2), which allows ECP to be funded more easily than other treatments. The EAG highlighted that the different standard-care treatments may have different levels of efficacy. So, if ECP is the most effective treatment (as could be inferred by the NHS preference for its use) and was used by a higher proportion of people in NHS practice than in REACH2, then REACH2 may have underestimated standard-care efficacy. The company explained that it did not have the data to conduct subgroup analyses for all standard-care treatments, and that such analyses would break randomisation. The company referred to failure-free and overall survival curves from REACH2 that showed similar outcomes between ECP and the other standard-care treatments. The clinical experts also highlighted that it is difficult to decide that one treatment is more effective than another. This is because of differences between individuals in the trial and variations in treatment protocols in different countries. The committee concluded that it had not been presented with convincing evidence that ECP is more effective than other standard-care treatments. It noted that this uncertainty could not be resolved with the available data and should be accounted for in its decision making. So, although the committee cautioned that a lack of data did not imply a lack of difference between the treatments, it accepted that the results of REACH2 could be generalised to the NHS.

The company used clinical evidence from REACH3 to inform the chronic GvHD health states in the model, acknowledging that some people with acute GvHD go on to develop chronic GvHD. REACH3 was a phase 3 randomised, open-label, multicentre trial. It compared ruxolitinib 10 mg twice daily with the investigator's choice of standard care. It included people who had had an allogeneic HSCT, were 12 years and over, and had moderate or severe corticosteroid-refractory chronic GvHD. It was done across 28 countries, including the UK. The trial allowed people in the standard-care arm to switch to ruxolitinib at or after week 24 if they did not have or maintain a complete or partial response, had adverse effects from a control treatment, or had a flare-up of their chronic GvHD. The committee noted that 38% of people in the standard-care arm switched to ruxolitinib at or after week 24. Because ruxolitinib was effective relative to standard care, the committee concluded that this crossover would probably have had a large impact on the clinical outcomes measured in the trial for standard care of chronic GvHD.

To estimate the cost-effectiveness of ruxolitinib, the company simulated the treatment of NHS patients with acute GvHD with either ruxolitinib or standard care over a lifetime time horizon. The company's model was a state-transition model containing 7 mutually exclusive health states:

• 'Failure-free': the starting health state. People remain in the failure-free health state until they start a new systemic treatment for acute GvHD, have a relapse of their underlying haematological disease, experience non-relapse mortality, or develop chronic GvHD.

• 'Relapse': people have a relapse of their underlying haematological disease.

• 'New systemic treatment': people start a new systemic treatment for acute GvHD.

• 'Chronic GvHD – failure-free': people develop chronic GvHD and remain in this health state until they have a new systemic chronic GvHD therapy or have a relapse of their underlying haematological disease.

• 'Chronic GvHD – relapse': people have chronic GvHD and a relapse of their underlying haematological disease.

• 'Chronic GvHD – new systemic treatment': people start a new systemic treatment for chronic GvHD.

• 'Death': the absorbing state which people can enter from any health state.

The company adjusted the proportion of people having each treatment in the standard-care arm of REACH2 to reflect the likely standard-care costs in the NHS. Clinical advice sought by the company suggested that in the NHS, relative to the trial, more people would have ECP, and that antithymocyte globulin, everolimus and low-dose methotrexate were not used at second line. Table 1 shows the revised treatment proportions used in the company's model.

The company used time-to-event data from REACH2 to estimate the transition probabilities from the 'failure-free' acute GvHD health state to 'new systemic treatment for acute GvHD', 'relapse of underlying haematological disease', 'chronic GvHD' and 'death'. It fitted models to the time-to-event data for failure-free to new systemic treatment, failure-free to relapse, failure-free to chronic GvHD, and failure-free to death. The company chose to fit joint models where there was evidence of proportional hazards, and independent curves where there was not. Joint models apply a single distribution to both treatment arms, whereas independent models fit a separate distribution to each arm. The company chose curves based on statistical goodness-of-fit, clinical plausibility and visual inspection. The EAG disagreed with the company's model fitting. It noted the choice of joint models was inappropriate for failure-free to relapse and failure-free to death because the proportional hazards assumption was not met. It also noted that the curves did not fit well to the underlying Kaplan–Meier data. Because of this uncertainty, the EAG assumed that the only benefit of ruxolitinib was in reducing the risk of moving from the 'failure-free' to 'new systemic treatment' health states. All other curves were based on pooled ruxolitinib and standard-care data, and the transition probabilities were the same for both arms. The committee noted that this counterintuitively decreased the incremental cost-effectiveness ratio (ICER). The EAG explained that this was because fewer people would enter the costly chronic GvHD health state. So, although both the incremental quality-adjusted life years (QALYs) and incremental costs decreased, the incremental costs decreased proportionally more, making ruxolitinib more cost-effective than it would have been otherwise.

In the model, people who enter the health state reflecting a new systemic treatment incur a treatment cost. The company calculated the proportion of people having each subsequent treatment from the pooled ruxolitinib and standard-care arms in REACH2 (table 2).

The company modelled chronic GvHD transition probabilities using data from the standard-care arm of REACH3. The EAG noted that only 10.4% of people in REACH3 had corticosteroid-refractory acute GvHD before they developed chronic GvHD and entered REACH3. But, the company's model implicitly assumed that everyone who entered the chronic GvHD health state previously had corticosteroid-refractory acute GvHD. The EAG questioned whether the clinical characteristics and outcomes would be different between people who did and did not have corticosteroid-refractory acute GvHD before developing chronic GvHD. The clinical experts explained that they would expect little difference in outcomes between people whose acute GvHD resolved before chronic GvHD and people who did not have acute GvHD before chronic GvHD. But, they explained that people who developed chronic GvHD while they still had unresolved acute GvHD may have more severe disease and experience worse outcomes. The company cited data from REACH3 that showed no difference in failure-free survival between people with chronic GvHD who did and did not have previous acute GvHD. The committee concluded that, on balance, the data from REACH3 was a reasonable proxy for modelling the chronic GvHD health states. But, it recognised that this was another source of uncertainty that could not be resolved.

To estimate utility values for the health states, the company fitted a model to pooled EQ-5D data from REACH2 and REACH3. The company noted a substantial increase in utility for people in the acute GvHD 'failure-free' health state after 4 model cycles (112 days). The company added a covariate to its model to account for this. The EAG had issues with the company's utility values. First, the EAG thought that it was inappropriate to pool the utility values from REACH2 and REACH3, given that the populations differed. In response, the company provided models based on separate data. Second, the EAG noted that simulated people who transitioned from the 'failure-free' health state to the 'chronic GvHD – failure-free' health state in the first 4 model cycles experienced a significant utility increase. The EAG thought that this was unlikely, because people who transition in the first 4 cycles are more likely to still be experiencing acute GvHD symptoms alongside developing chronic GvHD. Third, the EAG was concerned that the utility value for people in the 'failure-free' health state after 4 model cycles was significantly higher than the utility value used in the company's previous submissions to other health technology assessment agencies in Canada and Australia, although the EAG noted that these submissions used models with different structures. So, for its base case, the EAG changed the utility value in the 'chronic GvHD – failure-free, first 4 cycles' to be the same as the 'failure-free, first 4 cycles' value. The EAG also did scenarios using the utility models based on separate REACH2 or REACH3 data. The committee concluded that the adjustment to the 'failure-free, first 4 cycles' utility value was appropriate, but noted this change had little effect on the ICER.

The committee had concerns about the face validity of the utility values in the chronic GvHD health states. In particular, it noted that the utility value for the 'chronic GvHD – new systemic treatment' state was similar to the values for the 'failure-free acute GvHD' and 'chronic GvHD – failure-free' health states. It thought that this was implausible, given that people typically experience worse quality of life with each subsequent line of treatment. The company explained that because people with chronic GvHD are managed as outpatients, the utility value could be relatively high. But, the committee concluded that the utility value for the 'chronic GvHD – new systemic treatment' was implausibly high and would probably worsen over time. Because people in the model generate most of their QALYs in the chronic GvHD health state, the committee thought that the model could be sensitive to changes in this utility. So, it requested that the company update its model using plausible and coherent utility values for the chronic GvHD health states that had face validity.

The company calculated treatment duration, and associated costs, using the average proportion of people remaining on treatment at each week of REACH2, and their average dose of ruxolitinib. The committee questioned whether people in REACH2 who developed chronic GvHD continued having ruxolitinib. The company confirmed that some people did continue having ruxolitinib after developing chronic GvHD. The clinical experts noted that this aligned with NHS clinical practice. The committee asked if this was reflected in the model. The company confirmed that the treatment duration in the model was not linked to a health state. But, it added that because the treatment duration was calculated from REACH2, the model implicitly captured those people who continued ruxolitinib after developing chronic GvHD. The committee concluded that the modelled duration of treatment was appropriate and aligned with how ruxolitinib is likely to be used in NHS clinical practice.

The company's model assumed that there would be no wastage of ruxolitinib. The clinical experts noted that acute GvHD can be treated in hospital, with ruxolitinib dispensed by hospital pharmacies. But, there will be some people with acute GvHD who are treated as outpatients. The committee agreed that the risk of wastage is higher among outpatients, and that wastage would increase the cost of ruxolitinib. The committee concluded that some ruxolitinib would be wasted and asked the company to update its model to include wastage of ruxolitinib.

The committee discussed that the risk of acute GvHD is increased when there is a genetic mismatch between donor and recipient. The company highlighted that finding a genetic match is particularly difficult for some ethnic groups, which may lead to an increased incidence of acute GvHD among these groups. The committee also recalled statements from the patient and clinical experts that there are access issues with ECP that may mean significant travel time, and associated costs, for patients and carers. The patient and clinical experts also highlighted equality issues around access to ruxolitinib. First, they noted that ruxolitinib was commissioned in Scotland and Wales. Second, they explained that ruxolitinib was available in England during the COVID-19 pandemic through an NHS England rapid commissioning policy. This policy was withdrawn in 2022, but some people are still able to access ruxolitinib through individual funding requests or local approval from some hospital trusts. The committee acknowledged the concerns raised but concluded that they were not relevant equality considerations that could be addressed in its decision making in a technology appraisal.

The committee discussed whether there were any uncaptured benefits of ruxolitinib. It recalled statements from the patient and clinical experts that many people with acute GvHD would prefer ruxolitinib because it is an oral treatment. They noted that this would be particularly important for immunosuppressed people who could avoid hospital visits for ECP. The committee recognised that if ruxolitinib could permit people to gradually reduce their dose of corticosteroids, the corticosteroid-associated adverse effects would lessen relative to standard care. It also noted statements from carers about the all-encompassing nature of caring for a person with acute GvHD, and that the model does not include estimates of carer disutility. So, the committee concluded that there were uncaptured benefits of ruxolitinib and took these into account in its decision making.

NICE's manual on health technology evaluations notes that, above a most plausible ICER of £20,000 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. The committee will be more cautious about recommending a technology if it is less certain about the ICERs presented. But it will also take into account other aspects, including uncaptured health benefits. The committee noted the high level of uncertainty, specifically:

• The open-label design of the REACH2 and REACH3 trials may have affected the failure-free survival outcome in the standard-care arms of both trials (sections 3.3 and 3.6).

• Increasing the proportion of people having ECP in the standard-care arm increases the costs of standard care but does not change the efficacy. This may underestimate the standard-care treatment effect if different treatment options have different efficacies (sections 3.5 and 3.8).

• The company's model is structured around the failure-free survival outcome. This was a secondary outcome in REACH2. Failure-free survival was primarily driven by switching to new treatment and may have been affected by the open-label design (section 3.7).

• The distribution of follow-on treatments for acute GvHD would probably differ for people who initially had ruxolitinib and for people who initially had standard care (section 3.10).

• Chronic GvHD was modelled by the company using data from REACH3, but only some people in REACH3 had corticosteroid-refractory acute GvHD before chronic GvHD (section 3.11).

• The utility value for the 'chronic GvHD – new systemic treatment' seemed implausibly high (section 3.13).

• Wastage was not included in the company's model (section 3.15).

Because of confidential commercial arrangements for ruxolitinib, the comparators and other treatments in the model, the exact cost-effectiveness estimates are confidential and cannot be reported here. Neither the company's nor the EAG's base-case ICERs included all the committee's preferred assumptions, so the ICERs based on the committee's preferred assumptions are unknown. The following preferred assumptions of the committee aligned with adjustments made by the EAG to form its base case:

• extrapolating survival by assuming treatment benefit of ruxolitinib for time to new systemic treatment only (section 3.9)

• lowering the utility value for the 'chronic GvHD – failure-free, first 4 cycles' health state (section 3.12)

• applying adverse event disutilities so that they are multiplicative.

The committee decided that the cost-effectiveness estimates presented by the company and EAG were uncertain because they did not include all the preferred assumptions. Given the uncertainty, the committee would like to see additional analyses. The committee agreed that it was possible the cost-effectiveness estimates were above the range that NICE considers a cost-effective use of NHS resources. So, the committee concluded that it could not recommend ruxolitinib for treating acute GvHD that has an inadequate response to corticosteroids in people 12 years and over.