Azacitidine for treating acute myeloid leukaemia with more than 30% bone marrow blasts

Clinical effectiveness

4.3 The committee noted that the key clinical effectiveness evidence was from the AZA‑AML‑001 trial, in which azacitidine was compared with a combined conventional care regimen. The committee discussed the use of a combined conventional care regimen in the clinical trial rather than a single comparator. The committee understood that there was some uncertainty about which patients would be most likely to receive each of the comparators and that clear criteria did not exist. Nevertheless the committee considered that treatment allocation was associated with performance status. The committee considered that although there was no clear distinction between groups in the individual conventional care regimens these could be regarded as separate patient groups seen in clinical practice. The committee concluded that using the combined conventional care regimen in the clinical trial was not optimal for decision-making. However, in this instance its use was partly justified by the uncertainties in treatment allocation and the relatively small numbers in 2 of the 3 treatment groups in the clinical trial (see sections 4.4 and 4.6). The committee concluded that it was appropriate to not only consider the outcomes of the combined conventional care arm, but also the outcomes for the individual treatment subgroups.

4.4 The committee asked the company to clarify the proportions of patients randomised to the individual components of the conventional care regimen. The committee heard that the anticipated patient distribution between intensive chemotherapy, low-dose chemotherapy and best supportive care was 50:30:20 respectively (see section 3.17). The committee was told that the actual patient distribution for the individual components of the conventional care regimen was 18:64:18. The committee heard from the company that it considered the 18% of people randomised to intensive chemotherapy treatment to reflect current clinical practice in the NHS. However, the proportions of patients randomised to low-dose chemotherapy and best supportive care in the trial appeared different to current practice in the NHS. The company noted that it is difficult to identify the exact proportions of patients receiving each type of treatment because there is no widely accepted risk algorithm that clinicians use in England to decide which patients will receive treatment. The committee concluded that there were some differences in treatment allocation in the clinical trial and in NHS clinical practice which could affect the generalisability of the results to NHS clinical practice.

4.5 The committee discussed the results from AZA‑AML‑001. The committee noted that while this trial demonstrated gains favouring azacitidine in its primary efficacy end point of overall survival, it failed to reach statistical significance when comparing azacitidine with the combined conventional care regimen. Also there were no statistically significant differences for the secondary outcomes. The company noted that in this difficult-to-treat and heterogeneous patient group, clinical trials often fail to reach statistical significance for overall survival. Furthermore, the company considered that the use of subsequent treatments in the trial had confounded the results. The committee expressed concern about the absence of statistically significant results for the end points when using intention-to-treat analysis, but concluded that the use of subsequent treatments would have affected the results and agreed it was appropriate to consider analyses that controlled for treatment switching.

4.6 The committee then considered the results of AZA‑AML‑001 that compared azacitidine with the individual components of the conventional care regimen. The committee understood that the trial was not powered for these comparisons, which were based on small patient numbers. The committee noted that for overall survival azacitidine was more effective than intensive chemotherapy, low-dose chemotherapy and best supportive care. However, only the comparison with best supportive care reached statistical significance. The committee noted that these results were not always intuitive, for example for the comparison with low-dose chemotherapy the hazard ratio was closest to 1.00, but the difference in median overall survival was the biggest. The committee heard from the evidence review group (ERG) that this was because the trial arms converged at about 2 years, which indicated that proportional hazards models were not appropriate. The committee also noted that for event-free survival and relapse-free survival the individual chemotherapy conventional care regimens appeared to do better than azacitidine. The committee concluded that the degree to which azacitidine was more effective than any of the individual conventional care regimens was very uncertain.

4.7 The committee discussed the company analyses that adjusted for treatment switching and baseline covariates. The committee understood that during the trial no crossover between any treatment groups was allowed and once randomised to a conventional care regimen patients could not change to a different conventional care regimen. However, patients who stopped the study treatment could receive subsequent therapy during study follow-up. The committee was told that 67 patients in the azacitidine arm and 75 patients in the conventional care arm received further treatments after stopping treatment. The company therefore adjusted overall survival outcomes for treatment switching using a range of different methodological approaches. The committee considered that the approaches used by the company were all susceptible to bias because they relied on the proportional hazards assumption, which the committee did not consider was appropriate. The committee noted that the ERG had concluded that the more complex methods of adjusting for treatment switching had not added anything to the simpler censoring methods. However, the committee noted that the assumptions behind the simpler methods preferred by the ERG were also not supported by the data or by NICE technical support documents. The committee considered that a more robust method such as the Branson and Whitehead adjustment method would have been more appropriate. The company and ERG both noted that they had not considered this method. The committee concluded that there were limitations in the approaches used by both the company and the ERG and that an approach that adjusted for treatment switching in both arms, but which did not rely on an assumption of proportional hazards, would have been more appropriate.

Cost effectiveness

4.8 The committee considered the cost-effectiveness evidence presented by the company and the exploratory analyses presented by the ERG. It heard that the ERG considered the model to be simple and transparent. The committee considered the company's model in detail, and discussed 4 key areas of concern about the company's approach: the extrapolation of overall survival and adjustment for treatment switching, implementation errors in the model, health resource use estimates and costs of the relapsed and progressive disease state.

4.9 The committee discussed the implementation errors in the company model identified by the ERG. These mainly related to the formula used to calculate healthcare resource use, but also to the extrapolation of outcomes. The committee understood that 3 of the errors identified related only to sensitivity analyses, and that of the others, 8 increased the estimate of the incremental cost-effectiveness ratio (ICER) and 1 reduced it. The committee noted that the cumulative effect of the corrections was to increase the company base-case ICER from £21,000 to £63,000 per quality-adjusted life year (QALY) gained. The company told the committee that they accepted these model errors and the implementation changes needed to correct them. The committee accepted these changes made by the ERG and concluded that the base-case ICER in the company model was £63,000 per QALY gained.

4.10 The committee discussed the modelling of the number of treatment cycles. It heard from the ERG that, instead of using a mean of 8.8 cycles of treatment with azacitidine, the company had used a maximum number of 8 treatment cycles, which meant that the mean number of treatment cycles in the model was incorrectly calculated to be 5.6 cycles. The committee agreed with the ERG that the number of cycles in the model should reflect AZA‑AML‑001, in which a mean 8.8 cycles were received. The committee heard from the ERG that similar errors were made in the calculation of treatment cycles for the comparator treatments. The committee concluded that it was appropriate for the number of cycles of treatment in the model to reflect the number of cycles of treatment in the clinical trial. The committee noted that this change increased the ICER from £63,000 to £132,000 per QALY gained. The company accepted the ERG's changes and accepted the resulting increase in the ICER.

4.11 The committee discussed the ERG's further analysis relating to costs in the relapsed and progressive disease state. The committee heard that costs for the disease states were calculated using a healthcare resource-use questionnaire. The questionnaire was sent to 2 separate groups of 7 clinicians for azacitidine and the conventional care regimen. The methodology used by the company meant that in the relapsed and progressive disease health state in the model, different resource estimates were applied to the azacitidine group and the conventional care group despite both groups being treated with best supportive care. The committee noted that the main difference in estimates was for inpatient days; the company estimated that people in the azacitidine arm would have 1.73 inpatient days and 2.61 days in the conventional care regimen arm. The committee noted that this was one of the most expensive categories of healthcare resource use in the company model, so using an assumption of differing resource use led to an accumulation of differential costs in the groups. The committee considered that because all patients in this disease state would be receiving best supportive care, the resource use in the 2 groups would be the same. The committee heard from the ERG that the ICER was very sensitive to both the absolute resource use assumed and the difference in resource use between the azacitidine and conventional care groups. To test the sensitivity to the changes in costs, the ERG had set healthcare resource use in the progressive disease state to £0 and the ICER increased from £63,000 to approximately £74,000 per QALY gained. The committee concluded that equalising costs in both the azacitidine arm and the conventional care regimen arm for the progressive disease state was justified and noted that this change increased the ICER for £63,000 to £160,000 per QALY gained. However, the committee noted that the ERG had equalised the cost to the higher estimate of resource use, and that it might have been more reasonable to take an average of the resource use estimates for the 2 groups. The committee noted that the combined effect of correcting the model implementation errors, the number of treatment cycles and equalising the costs in the relapsed and progressive disease health state increased the ICER from £63,000 to £240,000 per QALY gained. However, it agreed that because of the use of the higher estimate of resource use, the ICER for azacitidine was likely to be somewhat overestimated.

4.12 The committee discussed the changes made by the ERG for treatment switching, adjusting for baseline covariates and the assumption of proportional hazards. The committee recalled the approaches used by the company and the ERG to adjust for treatment switching and the limitations associated with these (see section 3.6). The committee did not consider that an assumption of proportional hazards was appropriate and accepted the changes made by the ERG to use Kaplan–Meier curves from each trial arm rather than fitting curves that assumed proportional hazards. The committee noted that individually these changes did not affect the ICER, but that when considered with the changes to cost estimates they had a larger effect. The committee was not convinced that the ERG's changes to adjust for treatment switching using censor at switch was any better than the methods used by the company. The committee understood that the company and therefore the ERG had presented an analysis that adjusted for baseline covariates. However, it understood that there were some limitations associated with adjusting because of small sample sizes in some groups. The committee concluded that an assumption of proportional hazards was not appropriate; it would have liked to have seen an estimate of the ICER that adjusted for treatment switching in both groups and did not use an assumption of proportional hazards. Nevertheless, considering the data presented, the committee concluded that issues associated with the extrapolation and survival modelling were of less importance than the issues associated with cost estimates, and these limitations did not prevent it from making a decision about the cost effectiveness of azacitidine.

4.13 The committee discussed the cost-effectiveness results presented by the company and the ERG's exploratory analyses. The committee considered the most plausible ICER for azacitidine compared with the conventional care regimen. It noted that the company's base-case ICER was £21,000 per QALY gained. With corrections to the implementation errors in the model, this increased to £63,000 per QALY gained. The committee accepted the changes to the number of treatment cycles in the company model and the equalising of costs in the relapse and progressive disease state (see sections 3.22 and 3.23). It noted that with these changes, the ICER increased further to £240,000 per QALY gained, but recognised that this ICER is likely to be somewhat overestimated because of the higher estimate of resource use used by the ERG. The committee also accepted the use of the Kaplan–Meier curves, which would lead to further increases in the ICER and reflected that there was considerable uncertainty due to the confounded overall survival data (section 4.7). However, the committee concluded that a reasonable estimate of the most plausible ICER was approximately £240,000 per QALY gained.

4.14 The committee considered the innovative nature of azacitidine. The committee heard from the company that it considers azacitidine to be an innovative treatment option both in its clinical effectiveness and because it provides another treatment option in an area of high unmet need for a difficult-to-treat population. The committee discussed the innovative nature of azacitidine. It concluded that azacitidine should not be considered a step change in the treatment of acute myeloid leukaemia and that there was a high degree of uncertainty about its clinical effectiveness relative to current conventional chemotherapy treatments.

4.15 The committee considered supplementary advice from NICE that should be taken into account when appraising treatments that may extend the life of patients with a short life expectancy. For this advice to be applied, all the following criteria must be met.

4.16 The committee noted the evidence presented by the company that showed that people with acute myeloid leukaemia have a life expectancy of less than 24 months. The company presented evidence to show that azacitidine offers an extension to life, of at least an additional 3 months, compared with current NHS treatment. The company presented a median overall survival benefit of 3.8 months. The committee heard from the ERG that the estimates of extension to life were neither plausible nor robust because the azacitidine and conventional care regimen curves converged over time. The ERG suggested that using a restricted mean estimate was more appropriate than using a median estimate. The committee accepted the ERG's restricted mean overall survival estimate of 2.5 months. The committee concluded that azacitidine did not meet the criteria to be considered a life-extending, end-of-life treatment.

4.17 Taking into account that the most plausible ICER for azacitidine (approximately £240,000 per QALY gained, see section 4.11), that azacitidine should not be considered a step change (see section 4.14), and that azacitidine did not meet the criteria to be considered a life-extending, end-of-life treatment (see section 4.16), the committee concluded that it could not recommend azacitidine for treating acute myeloid leukaemia with more than 30% bone marrow blasts in people of 65 years or older who are not eligible for haematopoietic stem cell transplant as a cost-effective use of NHS resources.

4.18 The committee was aware of the new arrangements for the Cancer Drugs Fund recently agreed by NICE and NHS England. Under the new arrangements, drugs may be given a conditional recommendation by NICE and made available to NHS patients through the Cancer Drugs Fund. Such a drug will remain available within the Cancer Drugs Fund, normally for up to 2 years, while the company gathers more evidence. The committee considered that the most plausible ICER for azacitidine was substantially higher than the range normally considered a cost-effective use of NHS resources, and so azacitidine did not have the potential to satisfy the criteria for routine use (see section 4.13). The committee also considered that, given the uncertainties in the evidence it had seen, collecting outcomes data from people in the NHS would not be enough to inform an update of the guidance or cause the committee to reconsider its recommendations. The committee also heard from the company that it did not intend to submit a case for azacitidine's inclusion in the Cancer Drugs Fund. For these reasons, the committee therefore concluded that it could not recommend azacitidine for inclusion in the Cancer Drugs Fund.

4.19 The committee was aware of NICE's position statement on the Pharmaceutical Price Regulation Scheme (PPRS) 2014, and in particular the PPRS payment mechanism. It accepted the conclusion 'that the 2014 PPRS payment mechanism should not, as a matter of course, be regarded as a relevant consideration in its assessment of the cost effectiveness of branded medicines'. The committee heard nothing to suggest that there is any basis for taking a different view about the relevance of the PPRS to this appraisal. It therefore concluded that the PPRS payment mechanism was not relevant in considering the cost effectiveness of the technology in this appraisal.

4.20 The committee considered whether its recommendations were associated with any potential issues related to the equality legislation and the requirement for fairness. It considered a consultee comment received during consultation which highlighted that around 75% of people with acute myeloid leukaemia are over the age of 60 years, and that these people may be less able to tolerate the toxicity and side effects associated with current treatments. The committee was aware that age is a protected characteristic as defined by the Equalities Act. The committee was also aware that azacitidine has a marketing authorisation for people of 65 years or older, and that it was this subgroup who were enrolled in the AZA‑AML‑001 trial that both underpinned the marketing authorisation for azacitidine (see section 4.1) and was the main focus of the company's submission. However, the committee was aware that it could only make recommendations in accordance with the marketing authorisation for azacitidine. However, the committee agreed that its recommendations were not made because of the age of the patients, but rather because azacitidine within its marketing authorisation was not cost effective, and that it had not identified any special factors which would require or justify making a positive recommendation despite the very high ICERs.

4.21 NICE technology appraisal guidance on azacitidine for the treatment of myelodysplastic syndromes, chronic myelomonocytic leukaemia and acute myeloid leukaemia recommends azacitidine as an option for acute myeloid leukaemia with 20% to 30% bone marrow blasts. The committee considered a consultee comment received during consultation, which highlighted that an inequitable situation would be created by not recommending azacitidine for treating acute myeloid leukaemia with more than 30% bone marrow blasts. The committee was aware that severity of disease is not a protected characteristic as defined by the Equalities Act, and that it did not fall within NICE's obligations to avoid discrimination while carrying out its functions. The committee noted that azacitidine had been considered by NICE separately for acute myeloid leukaemia with 20% to 30% and more than 30% bone marrow blasts because of the timing of the regulatory marketing authorisation approval process. However, the committee agreed that it would have been preferable to have developed a single piece of guidance for azacitidine in this indication.

Summary of appraisal committee's key conclusions