3 Committee discussion
The appraisal committee (section 5) considered evidence submitted by Astellas Pharma, a review of this submission by the evidence review group (ERG), the technical report, and responses from stakeholders. See the committee papers for full details of the evidence.
The appraisal committee discussed the following issues (issues 1, 2, 3, 4, 5, 7 and 8), which were outstanding after the technical engagement stage.
3.1 Acute myeloid leukaemia (AML) is a rapidly progressing form of leukaemia, often diagnosed after an emergency admission to hospital. The FLT3 mutation is associated with poorer outcomes, such as a higher risk of relapse. Current treatment for relapsed or refractory AML is limited. The condition is managed with salvage chemotherapy, which is administered as an inpatient treatment and is associated with side effects and debilitating complications. Gilteritinib is an oral tablet that is self-managed and can be taken at home. Patient experts explained that it would improve their quality of life if they could avoid the disruption and loss of autonomy associated with inpatient treatment. They explained that the potential for improved quality of life is important to them, as well as the potential for improved survival. The committee concluded that people with relapsed or refractory AML would welcome a new treatment that improves survival and quality of life, particularly one that is taken orally at home.
The ADMIRAL trial provides the main clinical evidence for gilteritinib compared with salvage chemotherapy
3.2 The clinical evidence came from ADMIRAL, an open-label, randomised trial, which compared gilteritinib with the investigator's choice of salvage chemotherapy. The comparator arm included:
low-dose cytarabine (LoDAC)
mitoxantrone, etoposide and cytarabine (MEC)
fludarabine, idarubicin, granulocyte-colony stimulating factor and high-dose cytarabine (FLAG-IDA).
The primary outcome measure in ADMIRAL was overall survival. In response to consultation the company provided updated data from the ADMIRAL study (September 2019 data cut). Treatment with gilteritinib increased median overall survival compared with salvage chemotherapy from 5.6 months to 9.3 months (hazard ratio 0.68; 95% confidence interval 0.53 to 0.88, p=0.0013). The committee concluded that salvage chemotherapy was an appropriate comparator.
Best supportive care is a relevant comparator but the evidence presented to support its relative efficacy is not reliable
3.3 Best supportive care was not included as a comparator in ADMIRAL. The clinical experts noted that, in clinical practice, most people would have salvage chemotherapy. But they added that best supportive care is a relevant option in a small proportion of patients who choose not to have salvage chemotherapy because of toxicity and lack of fitness for treatment. Stakeholders at technical engagement considered that best supportive care could be a relevant option for 10% to 20% of patients in this population. The company included a blended comparator of salvage chemotherapy based on ADMIRAL in its economic model results. It did not include best supportive care as a comparator in its original base-case results. However, it did include it as a separate comparator in a scenario analysis by applying a hazard ratio of 2.86 to gilteritinib overall survival, informed by a naive indirect comparison. This was because there was no direct evidence comparing gilteritinib with best supportive care. The ERG had concerns about the methods, assumptions and sources used to inform the company's indirect comparison for best supportive care, including:
the indirect comparison assumes that LoDAC is equivalent to salvage chemotherapy
the source of the values used to calculate the hazard ratio between gilteritinib and best supportive care was unclear
proportional hazards are assumed, which may not be appropriate because it is not clear whether the assumption was assessed.
The committee noted the ERG's concerns about the methods of including best supportive care and did not consider that the indirect comparison was reliable. After technical engagement, the company updated its analysis to include best supportive care in the blended comparator. This reduced the cost-effectiveness estimates. The ERG noted that the company's analysis assumed the characteristics of people receiving best supportive care are the same as for people receiving salvage chemotherapy, for example the stem cell transplant rate, which it considered was implausible. The company's updated base-case model, provided after consultation, also included best supportive care in the weighted comparator arm. It suggested that 20% of people have best supportive care and half of them could have gilteritinib. The company suggested that these people would have the same outcomes as the gilteritinib arm in ADMIRAL. It assumed in its model that people having best supportive care would not have stem cell transplant so the probability of receiving transplant in the weighted comparator group is reduced by 10%. The gilteritinib group stem cell transplant rate was unchanged. The ERG highlighted that this assumption meant that people who choose to have best supportive care would have the same likelihood of receiving stem cell transplant if they had had gilteritinib, which was unrealistic. The issues with the indirect treatment comparison and how it was applied in the model discussed at the first committee meeting remained. The committee concluded that best supportive care was a relevant comparator as well as salvage chemotherapy. But it agreed that the company's method of including best supportive care in the blended comparator was not appropriate and therefore accepted analyses comparing gilteritinib with salvage chemotherapy.
The proportion of people who would have received midostaurin in clinical practice in the NHS may be higher than the proportion in ADMIRAL
3.4 NICE technology appraisal guidance on midostaurin (another FLT3 inhibitor) recommends it for use in the NHS for newly diagnosed acute FLT3‑mutation-positive AML. In ADMIRAL, 13% of the gilteritinib group and 11.3% of the salvage chemotherapy groups had received prior FLT3 inhibitors. If, in clinical practice in the NHS, the proportion of people who have had prior midostaurin is higher than in ADMIRAL, the efficacy of gilteritinib may be different to that seen in the trial. The company presented a subgroup analysis of people in ADMIRAL who had had prior FLT3 inhibitors, such as midostaurin. The results showed that, for patients with no prior FLT3 inhibitor (n=325), gilteritinib statistically significantly improved overall survival (hazard ratio 0.620; 95% confidence interval 0.470 to 0.818). For the 46 patients with prior use of an FLT3 inhibitor, the treatment difference was not statistically significant (hazard ratio 0.705; 95% confidence interval 0.346 to 1.438). However, this subgroup analysis only included a small number of patients and may be unreliable. The clinical experts confirmed that they would give gilteritinib after midostaurin in clinical practice. They stated that gilteritinib is a more potent FLT3 inhibitor and they did not believe that prior exposure to midostaurin would affect response to gilteritinib, although this is uncertain. The clinical expert estimated that there were about 600 people a year in the NHS who have relapsed or refractory FLT3‑positive AML. Comments from technical engagement suggested that around 50% to 60% of patients with newly diagnosed FLT3‑positive AML may have midostaurin. The committee concluded that currently the proportion of people with relapsed or refractory disease who may have received prior midostaurin in clinical practice in the NHS is higher than the proportion in ADMIRAL.
3.5 In its original model, the company assumed that all patients who were alive at 3 years were 'cured', regardless of whether their disease had progressed or they had had a stem cell transplant. After 3 years, survival was modelled using an uplifted general population mortality rate (standardised mortality ratio of 2.0). The 3‑year cure assumption was based on NICE technology appraisal guidance on midostaurin, published literature, and clinical advice given to the company. The company did not present any evidence from ADMIRAL to support the cure assumption. The clinical expert suggested that most relapses would be within 12 months. The ERG noted that the Kaplan–Meier curves from ADMIRAL did not show a plateau, which would have suggested a cure. At technical engagement, stakeholders agreed that it was clinically plausible to assume that patients alive after 3 years were cured. However, after technical engagement the company updated its model to include a 2‑year cure point, instead of 3 years. It did not give evidence or a clear rationale for why it had changed the cure point. After consultation the company provided additional evidence to support the 2‑year cure point assumption in the model. It provided data from different studies to show the flattening of curves between 18 months and 24 months. However, some of these studies were not in the same population as the ADMIRAL trial. The clinical experts explained that a 2‑year cure point is clinically plausible. They explained that FLT3-positive AML is a highly aggressive form of AML and relapses occur early in this population, ranging from 6 months to 2 years. The experts also pointed out that mortality after 2 years is likely to be a late consequence of stem cell transplant. However, the committee noted that using a 2-year cure point appears to overestimate the long-term overall survival for the gilteritinib arm in the observed period of the trial. It also noted that there were 3 deaths after 2 years in the gilteritinib arm of the trial. The committee had concerns about applying a 2-year cure point, because the population in the evidence used to support the 2‑year cure point was different to the ADMIRAL trial, and because of the lack of good visual fit of the extrapolated 2-year cure to the Kaplan–Meier data. However, taking into account clinical expert advice, it concluded that a cure point between 2 years and 3 years was plausible, and it was more likely to be closer to 2 years.
3.6 In the company's model, post-stem cell transplant overall survival was based on a Gompertz curve fitted to data from a study by Evers et al. (2018). The company did not use ADMIRAL data for this group of patients from the company submission and the model because there was limited follow up and a small sample size. However, patients in the Evers study did not all have FLT3 mutations so were not directly comparable to the population who would be eligible for gilteritinib in clinical practice. The company also highlighted data from a study by Ustun et al., which it used in a scenario analysis. This study included people with FLT3‑positive AML but most people in the study did not have relapsed or refractory disease. The ERG highlighted the company's model's predictions and the proportion of patients alive at the end of the final data cut off from ADMIRAL. It said that, because of these, to meet the 3‑year cure rate from the company's original model, the majority of surviving (censored) patients in the ADMIRAL gilteritinib-treated stem cell transplant group would need to be considered 'cured'. The ERG considered that the ADMIRAL trial was the most relevant data source, and did an analysis using ADMIRAL data to inform overall survival for people who had a stem cell transplant, which it included in its base case. The ERG pooled both treatment groups from ADMIRAL and fitted a log-normal parametric curve to the data until the 3‑year cure point. At technical engagement, stakeholders agreed that the ADMIRAL data should be considered. The committee considered that the ADMIRAL trial was the most appropriate because it included the population relevant to this appraisal.
3.7 To model overall survival for the post-stem cell transplant group, the company applied a hazard ratio to the Gompertz model (see section 3.6) to reflect an additional survival benefit associated with gilteritinib maintenance therapy after stem cell transplant. The company derived the hazard ratio from an indirect comparison using data from Evers 2018. The company acknowledged that the results from ADMIRAL (September 2018 data cut) do not show a favourable effect of gilteritinib after stem cell transplant. However, it noted that there were small patient numbers and high levels of censoring. The company believed that, if the patients with salvage chemotherapy were followed up for longer, a benefit of gilteritinib maintenance therapy would be seen. The ERG considered that the company's approach was inconsistent. The company did not use ADMIRAL data to model post-stem cell transplant overall survival but it did use it, with the data from Evers, to calculate the hazard ratio for the additional benefit of gilteritinib. The ERG did an analysis using a hazard ratio of 1 to indicate no additional benefit of maintenance therapy, which it included in its base case. The clinical experts and other stakeholders at technical engagement confirmed that gilteritinib would be used as maintenance therapy after stem cell transplant in clinical practice, although there is little evidence to support this practice. In response to consultation the company provided updated data from the ADMIRAL study that did not suggest a benefit for gilteritinib over chemotherapy for overall survival after stem cell transplant (hazard ratio 1.108; 95% confidence interval 0.53 to 2.29, p=0.7836). The company did not update the indirect comparison with the updated 2019 data. The committee had already concluded that ADMIRAL data should be used to model post-stem cell transplant overall survival, so agreed that the additional benefit of maintenance therapy included by the company was not relevant.
3.8 In its response to consultation the company reintroduced the gilteritinib maintenance therapy hazard ratio (0.69) for overall survival based on the naive indirect comparison using data from Evers et al. (see section 3.6). The company provided new evidence from the ADMIRAL trial comparing the overall survival of people who restarted gilteritinib after stem cell transplant and those who did not (hazard ratio 0.46). However, in ADMIRAL people could only restart gilteritinib in certain conditions, such as being in complete remission after stem cell transplant. This might lead to selection bias because those patients may be fitter than those who would receive gilteritinib maintenance treatment in clinical practice. The clinical experts confirmed that, in clinical practice maintenance therapy is the preferred treatment strategy, but people may not have to be in complete remission to restart gilteritinib. Therefore, more people would be eligible for treatment than in the trial. The committee also noted that in this analysis, patients who had chemotherapy before stem cell transplant had better overall survival than those who had gilteritinib before stem cell transplant. The company acknowledged that the true hazard ratio was likely to be somewhere between 0.46 and 1. The committee also recalled the overall survival after stem cell transplant data from the trial, which did not show gilteritinib to be effective (see section 3.7). The committee was also aware that including a maintenance therapy hazard ratio leads to a survival projection that is more favourable than the observed gilteritinib data from the trial. It acknowledged that there is interaction in the model between the cure point and the hazard ratio associated with maintenance therapy. Using the company's maintenance hazard ratio (0.69) would mean that a later cure point (later than 2 years) would be required for the extrapolations to fit the observed data. Therefore, when combining these assumptions – as in the company's updated base case – model predictions extremely overestimate the overall survival for the gilteritinib arm. Although the committee understood there might be a clinical benefit to gilteritinib maintenance treatment after stem cell transplant, it did not see robust evidence supporting this benefit. It agreed that no change to its previous conclusion was needed and therefore concluded that an additional benefit of maintenance therapy and associated costs should not be included in the model. The committee also concluded that treatment with gilteritinib should not restart as maintenance therapy after stem cell transplant.
3.9 In its original model, the company did not include wastage for gilteritinib. The ERG considered that tablets could be wasted in clinical practice, for example, if patients died or their disease progressed while they were on treatment. The ERG did an exploratory analysis to include 14 days' supply of wastage for all patients who died before the 3‑year cure point. After technical engagement, the company updated its model to include wastage for 7 days' supply of gilteritinib. The clinical expert explained that normally a 28‑day pack would be given to each patient at a time. Therefore, the committee considered that it was reasonable to assume 14 days' supply of gilteritinib may be wasted if someone died before the 3‑year cure point. In its response to consultation the company explained that most people would stop treatment in a managed way after consulting clinicians, therefore in the company's updated base case 7 days of wastage was used. Clinical experts confirmed that treatment is closely monitored and tests are done before dispensing a pack of gilteritinib. Because the disease is closely monitored, it is highly unlikely that someone's condition would deteriorate in the first 2 weeks after starting a new pack of gilteritinib. The experts also confirmed that compliance is good and that treatment with gilteritinib would usually stop after completing a course of therapy. The committee concluded that wastage of 7 days' supply of gilteritinib should be accounted for in the model.
3.10 The company included the costs of gilteritinib and chemotherapy as one-off costs in the first cycle of the model. The ERG noted that this was an unconventional approach that meant:
discounting could not be applied properly
gilteritinib treatment duration was underestimated because some patients were still having gilteritinib at data cut off and this was not accounted for
treatment duration was not linked to progression.
The ERG stated that, if the drug costs had been applied in each cycle, the incremental cost-effectiveness ratio (ICER) would likely increase, although it did not know by how much. The committee agreed that drug costs should have been applied in each cycle.
The benefit of taking an oral tablet at home compared with having chemotherapy in hospital should be captured in the model
3.11 At technical engagement, the clinical expert highlighted that a potential benefit of gilteritinib is that it is an oral treatment that does not need to be administered in hospital, whereas salvage chemotherapy requires an inpatient stay. The ERG noted that the difference in costs between the 2 treatments was reflected in the administration costs included in the model. However, the ERG noted that the model did not assume any difference in quality of life between the 2 treatments to account for the different methods of administration. After technical engagement, the company updated its model to include a disutility value of -0.044 for high-intensity chemotherapy, which was sourced from a study by Wehler et al. (2018), because it was difficult to collect patient-reported outcomes from people in the salvage chemotherapy group in ADMIRAL. The company also updated some of the hospital costs to reflect this issue. The costs associated with the observed number of hospitalisations in the trial were spread across the event-free survival interval, including time on and off treatment. The clinical and patient experts explained that the benefit of taking an oral tablet at home compared with having chemotherapy in hospital would be important to patients. The committee was concerned that the potential quality-of-life benefits of oral gilteritinib, with less time in hospital, compared with inpatient chemotherapy with frequent debilitating complications, had not been adequately addressed. In its response to consultation the company updated its model to include additional disutilities for the first 3 cycles for all chemotherapy regimens using the Wehler study as a source. It also included increased costs for high-intensity chemotherapy to reflect that people on salvage chemotherapy would need inpatient treatment for the entire first 1-month cycle. The committee accepted the inclusion of additional disutilities. It also heard from the ERG that the new costs were applied in an unusual way in the model, which would overestimate the costs rather than reflect the total number of hospitalisation days observed in the trial. The committee noted that it is likely that the company's new approach overestimates the true costs of hospitalisation for the high-intensity chemotherapy regimens. The committee agreed it was not presented with good enough quality evidence to be able to accept the updated cost figures. It concluded that the increased costs for high-intensity chemotherapy should be excluded from the model, but the additional disutilities should be included.
3.12 The committee considered the advice about life-extending treatments for people with a short life expectancy in NICE's guide to the methods of technology appraisal. Median overall survival in the salvage chemotherapy group of ADMIRAL was 5.6 months. The clinical expert stated that median survival is around 2 to 3 months in this patient population, and the ERG's base case showed that modelled survival in the salvage chemotherapy and the best supportive care group was less than 2 years. Although the company's updated base case predicted that the mean overall survival in the blended comparator group was over 2 years, the committee agreed that this was likely to be because of the method the company used to model gilteritinib effectiveness after stem cell transplant (see section 3.6). Therefore, the committee concluded that the short life expectancy criterion was met. Both the company's and the ERG's base-case economic models showed that gilteritinib extended mean overall survival by over 3 months more than with salvage chemotherapy (in the ERG's model, 2.34 years more than best supportive care and 0.98 years more than salvage chemotherapy). ADMIRAL showed a median overall survival gain of 3.7 months for gilteritinib compared with salvage chemotherapy. The committee concluded that the extension to life criterion was also met, and that when its preferred assumptions were applied in the model, gilteritinib met the criteria to be considered as a life-extending treatment at the end of life.
3.13 The company submitted a revised base case after consultation. The ICER, incorporating corrections made by the ERG (it corrected some programming errors in the company's model, which resulted in a lower ICER), was £46,961 per quality-adjusted life year (QALY) gained, compared with combined salvage chemotherapy and best supportive care. All analyses include the patient access scheme for gilteritinib. However, the committee noted that the revised base case did not include all of its preferred assumptions. These were:
excluding best supportive care from the weighted comparator (see section 3.3)
including a cure point closer to 2 years than 3 years (see section 3.5)
excluding the gilteritinib maintenance therapy hazard ratio for overall survival and the cost of maintenance therapy from the model (see section 3.8)
including gilteritinib wastage of 7 days' supply (see section 3.9)
including drug costs in each cycle of the model (see section 3.10)
including additional disutilities during first 3 cycles for all chemotherapy regimens and excluding increased costs for high-intensity chemotherapy (see section 3.11).
3.14 Applying the committee's preferred assumptions (see section 3.13) and including all commercial arrangements in the model resulted in an ICER which was below £50,000 per QALY gained for gilteritinib compared with salvage chemotherapy (the ICER is confidential and cannot be reported here). The committee acknowledged that the modelling may not have included all benefits for gilteritinib (see section 3.11), and that doing so could possibly reduce the cost-effectiveness estimate, although this was not sufficiently quantified in the model. Based on the evidence presented to it, the committee concluded that, with the discount agreed in the commercial arrangement, the most plausible ICER was within the range that NICE normally considers an acceptable use of NHS resources for a life-extending treatment at the end of life. Therefore, it recommended gilteritinib as an option for treating relapsed or refractory FLT3‑mutation-positive AML in adults, although people whose disease responds to gilteritinib and who then go on to have a stem cell transplant should not restart gilteritinib after transplant.
3.15 No equality or social value judgement issues were identified.
3.16 The company, professional organisations and clinical experts considered that gilteritinib was innovative because it would be the first oral monotherapy targeted for relapsed or refractory FLT3‑positive AML. The committee agreed that these were important benefits of gilteritinib, but it concluded that it had not been presented with evidence of any demonstrable and distinct substantial additional benefits that could not be captured in the measurement of QALYs.