Apixaban for preventing stroke and systemic embolism in people with non-valvular atrial fibrillation

4.1 The Appraisal Committee reviewed the data available on the clinical and cost effectiveness of apixaban, having considered evidence on the nature of non-valvular atrial fibrillation and the value placed on the benefits of apixaban by people with the condition, those who represent them, and clinical specialists. It also took into account the effective use of NHS resources.

4.2 The Committee heard from clinical specialists and patient experts that the current standard treatment for people with non-valvular atrial fibrillation who need anticoagulation is warfarin or the newer oral anticoagulants rivaroxaban or dabigatran. The clinical specialists explained that the majority of people receiving an anticoagulant currently receive warfarin. In addition, some people who meet criteria for anticoagulation are currently receiving the antiplatelet agent aspirin inappropriately because of clinical reluctance to prescribe warfarin. The Committee heard that warfarin is an effective treatment but that it is associated with a number of problems. The patient experts explained that repeated INR monitoring tests with warfarin can cause pain and scarring and limit a person's choice of leisure or other activities and that warfarin can have a greater impact on a person's quality of life than atrial fibrillation itself. They also highlighted that warfarin has multiple interactions with food, alcohol and drugs that can cause further inconvenience that make adherence to treatment difficult. Overall, the patient experts considered that making the day-to-day choices about lifestyle needed in order to take and monitor warfarin appropriately has a substantial impact on a person's quality of life. The Committee accepted the limitations of warfarin therapy and the considerable effect it may have on the people who take it, and recognised the potential benefits of apixaban for people with atrial fibrillation.

4.3 The Committee considered the clinical-effectiveness data from the ARISTOTLE trial comparing apixaban with warfarin. It considered that the ARISTOTLE trial was of good quality and it discussed whether the results were generalisable to people diagnosed with atrial fibrillation in the NHS. The Committee noted that the mean time in therapeutic range for those in the warfarin arm was 62.2% (the median was 66%) and asked the clinical specialists whether this was representative of what would be achieved in clinical practice in the UK. The clinical specialists explained that there is variation in time in therapeutic range achieved between centres. One clinical specialist quoted a benchmarking study using a computerised dose adjustment system in which a mean time in therapeutic range of over 70% was achieved. Another clinical specialist stated that the time in therapeutic range observed in ARISTOTLE reflected what is generally seen in the UK, not what is observed in centres achieving the best time in therapeutic range, and that centres should aim for a time in therapeutic range for each individual of 70% and above. One clinical specialist also highlighted that in their experience people treated for atrial fibrillation tended to be older and more likely to be on non-steroidal anti-inflammatory drugs (NSAIDS), which can impact on bleeding complications, than the ARISTOTLE population. The Committee noted the potential differences between the trial population and people treated for atrial fibrillation in the UK but concluded that the characteristics of the people who participated in ARISTOTLE were broadly generalisable to the UK population.

4.4 The Committee considered the results of the ARISTOTLE trial. It noted that apixaban was more effective than warfarin in reducing the primary efficacy outcome of all stroke (ischaemic and haemorrhagic), and systemic embolism. The Committee noted that the primary efficacy outcome was a composite of the effectiveness outcomes (ischaemic stroke and systemic embolism) and a bleeding outcome (haemorrhagic stroke). The Committee heard from the clinical specialists that there is debate about the primary outcomes to use in trials of anticoagulants for atrial fibrillation, but that it is common to use composite outcomes, such as the primary efficacy outcome in ARISTOTLE. The Committee considered the individual components of the composite outcome. It heard from the clinical specialists that once an embolus leaves the heart it is a matter of chance whether it flows to the brain, resulting in ischaemic stroke, or to the rest of the body, causing systemic embolism. The proportion of each was therefore not a treatment effect. The Committee also heard from the clinical specialists that a particular benefit conferred by the new anticoagulants compared with warfarin was the reduction in haemorrhagic strokes. This was also shown in the ARISTOTLE trial, in which there was a statistically significant reduction in haemorrhagic stroke with apixaban compared with warfarin, whereas for the other individual components of the composite end points (ischaemic stroke and systemic embolism) there was no statistically significant difference. The Committee concluded that apixaban was more clinically effective than warfarin for the primary efficacy outcome of reducing stroke and systemic embolism.

4.5 The Committee considered the results of the manufacturer's subgroup analyses from ARISTOTLE. It noted that the subgroup analysis by CHADS₂ score comprised 3 groups: people with a CHADS₂ score of 1 or less, people with a CHADS₂ score of 2 and people with a CHADS₂ score of 3 or over. The Committee was aware that the ERG had concerns that because people with CHADS₂ scores of 3 to 6 had been grouped together, it was not possible to comment on potential variation in treatment effect for these subgroups. The Committee concluded that there was no biologically plausible reason to indicate that the relative treatment effect would be dependent on baseline risk and that the mean CHADS₂ score of 2.1 in the trial was a reasonable reflection of the UK population currently on anticoagulant therapy.

4.6 The Committee noted that the manufacturer presented data for subgroups based on INR control using quartiles of centre time in therapeutic range. It further noted that there was a numerically lower rate of stroke or systemic embolism with apixaban compared with warfarin in all analyses broken down by centre time in therapeutic range, but that ARISTOTLE was not statistically powered to demonstrate superiority across subgroups. The Committee concluded that the evidence from subgroups based on centre time in therapeutic range was not sufficiently robust to use to formulate guidance based on an individual's time in therapeutic range.

4.7 The Committee considered the adverse events reported in ARISTOTLE. The Committee noted that for the primary safety outcome of major bleeding (using the International Society on Thrombosis and Haemostasis definition), treatment with apixaban resulted in fewer bleeding events than warfarin, including a reduced rate of intracranial bleeding. The Committee recognised that this has a high mortality rate and a large impact on a person's quality of life, and is the most feared bleeding outcome for people taking any type of anticoagulant. The Committee noted however that there were no statistically significant differences in the rates of gastrointestinal bleeding between apixaban and warfarin. The Committee concluded that apixaban resulted in fewer bleeds than warfarin and it recognised the particular importance of the effects of apixaban in reducing the risk of intracranial bleeding for people with atrial fibrillation when compared with warfarin.

4.8 The Committee noted that all anticoagulants are associated with a risk of bleeding and discussed the management of atrial fibrillation in people who experience a bleed while taking warfarin or apixaban. The Committee heard from the clinical specialists that people taking warfarin who experience a bleed may be given vitamin K to reverse the effects of warfarin. However, there are no standard treatments to reverse the effects of apixaban (or the other newer oral anticoagulants rivaroxaban and dabigatran) and that this is an area of ongoing research. Current clinical opinion is that the newer oral anticoagulants have moderate half-lives and that people who have bleeds while taking these drugs should stop treatment. The Committee also heard from the patient experts and the clinical specialists that reversing the effect of warfarin with vitamin K may take several hours, but that there are other approaches, such as using a prothrombin concentrate, that are fast-acting. The Committee concluded that there is a standard approach to reverse significant bleeding for a person taking warfarin, but that there is uncertainty about the most effective way to stop active bleeding when a person is taking apixaban.

4.9 The Committee noted that the manufacturer had included evidence on the efficacy of apixaban compared with aspirin for people for whom vitamin K antagonist treatment was unsuitable, but that this was not part of the scope issued by NICE. The Committee understood that the manufacturer's rationale for including aspirin as an additional comparator reflected the recommendation in NICE clinical guideline 36 that people who need anticoagulation but for whom warfarin is unsuitable should be offered aspirin. The Committee was also aware that aspirin had not been included as a comparator in the apixaban scope because, since the publication of NICE clinical guideline 36, dabigatran and rivaroxaban had been recommended for use by NICE and these were now alternative treatments to warfarin. The Committee heard from the clinical specialists that aspirin is a less effective treatment than the anticoagulants but is still being prescribed for some people with atrial fibrillation despite publication of NICE technology appraisal guidance 249 and 256. The Committee further heard that although AVERROES was a useful trial, the population for whom warfarin was unsuitable was very mixed, including people for whom warfarin was clinically unsuitable and those who were unwilling to take it. The Committee noted that the ERG did not consider that AVERROES met the inclusion criteria for this appraisal and had focused its critique on ARISTOTLE. The Committee agreed that the comparators defined in the final scope were appropriate and that the key trial for this appraisal was ARISTOTLE. However, it noted with interest that the evidence presented in AVERROES showed that apixaban was associated with a reduced rate of stroke or systemic embolism compared with aspirin and an increased rate of bleeding events overall, but not an increased rate of major bleeds.

4.10 The Committee discussed the indirect clinical-effectiveness evidence for apixaban compared with dabigatran (both the 110 mg twice daily dose and 150 mg twice daily dose) and rivaroxaban. The Committee noted that the manufacturer presented 2 network meta-analyses for vitamin K antagonist-suitable and -unsuitable populations (which included aspirin) respectively. The Committee noted that the ERG considered the second meta-analysis to be flawed as there were no specific data available for rivaroxaban or dabigatran for a warfarin-unsuitable population and the network meta-analysis included a mixed population including people for whom warfarin was suitable and unsuitable. The Committee considered that only the first network meta-analysis, relating to the warfarin-suitable population, was appropriate to the decision problem. The Committee noted that the population in the study comparing rivaroxaban with warfarin (ROCKET-AF) had a higher mean baseline CHADS₂ score than the population in the study comparing dabigatran with warfarin (RE-LY) or ARISTOTLE. The Committee additionally noted that the mean time in therapeutic range in the warfarin arm was lower in ROCKET-AF than in RE-LY or ARISTOTLE. The Committee considered that the differences in baseline characteristics between the study populations meant that there was uncertainty surrounding the results of the network meta-analysis. The Committee noted that the network meta-analysis did not detect any difference between apixaban, rivaroxaban and dabigatran in the rate of stroke or systemic embolism; showed a lower rate of all bleeding outcomes with apixaban compared with rivaroxaban and of all bleeding outcomes except intracranial haemorrhage and clinically relevant non-major bleeding (which was not measured in RE-LY) compared with dabigatran 150 mg; a lower rate of 'any bleeding' compared with dabigatran 110 mg; and a lower rate of myocardial infarction with apixaban compared with dabigatran (both doses). The Committee noted that the network meta-analysis had shown broadly similar outcomes and some differences between apixaban, rivaroxaban and dabigatran, but because of differences in the trial populations the results of the network meta-analysis should be viewed with caution. It also noted that some of the criteria in the network meta-analysis were in fact not a direct treatment effect, such as the proportion of ischaemic stroke compared with systemic embolism, and evidence was lacking that the severity of an ischaemic or haemorrhagic stroke was treatment specific for the new agents. The Committee concluded that the network meta-analysis results should be interpreted in the light of these uncertainties and were not sufficiently robust to reliably differentiate between apixaban, rivaroxaban and dabigatran.

4.11 The Committee considered the manufacturer's economic model and the exploratory analyses performed by the ERG. It agreed with the ERG that the general modelling approach was reasonable and consistent with other analyses of atrial fibrillation treatments. The Committee noted the discussion on the proportion of ischaemic stroke compared with systemic embolism being unrelated to the treatment (see 4.4). It also questioned whether the severity of an ischaemic or haemorrhagic stroke was treatment specific (see 3.13). In a previous appraisal the Committee had heard from experts that it was plausible and that there is evidence that strokes on warfarin were likely to be more severe than on dabigatran, but the clinical specialists for the appraisal of apixaban did not put forward any evidence that this had been substantiated and were of the opinion that at least for the newer agents, there was no biologically plausible reason or evidence that the severity of strokes would differ between apixaban, rivaroxaban and dabigatran. The Committee concluded that although the general modelling approach was appropriate, weaknesses included the assumption that whether a person experienced an ischaemic stroke or systemic embolism was treatment related, and that there is currently insufficient evidence to support the assumption in the model that the severity of an ischaemic or haemorrhagic stroke was dependent on the specific anticoagulant agent they had received.

4.12 The Committee considered the utility values used in the model. The Committee noted that ARISTOTLE had not assessed health-related quality of life and that the utility values used in the manufacturer's model were identified through a systematic review. The Committee questioned whether the manufacturer's assumption of a permanent utility decrement following a myocardial infarction was appropriate. However, it accepted the views of the clinical specialists that disutility following a myocardial infarction would not be expected to change substantially after 6 months. The Committee concluded that the utilities used in the model were appropriate.

4.13 The Committee considered the costs used in the model. It noted that the estimates for stroke and systemic embolism were based on a cohort study of a population living in the Oxford area of the UK and that the costs for ischaemic stroke were lower than those for haemorrhagic stroke. The Committee questioned whether the study had been able to estimate haemorrhagic stroke costs accurately given the lower incidence of this event than ischaemic stroke in the population, and whether the higher haemorrhagic stroke costs assumed in the model could have driven the cost-effectiveness results. The Committee heard from the ERG that the haemorrhagic stroke costs were consistent with those used in NICE technology appraisal guidance 249 and 256 and that, as a small proportion of people had a haemorrhagic stroke in the model, other factors such as discontinuation rates drove the cost-effectiveness results to a greater extent than the cost of haemorrhagic stroke. The Committee also noted that an INR monitoring cost of £248 was assumed by the manufacturer, and that this was consistent with the monitoring costs used in NICE technology appraisal guidance 249 (and was updated for inflation). The Committee concluded that the costs used in the model were appropriate.

4.14 The Committee considered the results of the economic model. It noted that the manufacturer's base-case deterministic and probabilistic ICERs for apixaban compared with warfarin were £11,000 and £16,900 per QALY gained respectively, and that the ERG's revised deterministic base case, (see 3.30) resulted in an ICER of £12,800 per QALY gained. The Committee noted that only one of the sensitivity analyses performed by the ERG (in which alternative second-line treatments rather than aspirin were considered, see 3.31) influenced the results substantially. The Committee accepted the ERG's comment that this analysis should be interpreted with caution because the main driver of the ICER was discontinuation rates on first-line treatment. The Committee noted that the ERG's sensitivity analysis assuming stroke severity was independent of treatment had a modest effect on the ICER compared with warfarin (the ICER increased to £12,300 per QALY gained when stroke severity was assumed to be the same for all of the anticoagulants). The Committee concluded that apixaban had been shown to be cost effective compared with warfarin, the most plausible ICER being less than £20,000 per QALY gained, and could be recommended as an option for preventing stroke and systemic embolism for people with non-valvular atrial fibrillation who have 1 or more risk factors for stroke.

4.15 The Committee noted that in the manufacturer's model dabigatran and rivaroxaban had higher ICERs compared with warfarin than the ICER for apixaban compared with warfarin. In addition, in the incremental analysis dabigatran 110 mg twice daily was dominated by the dabigatran blend and the dabigatran blend and rivaroxaban were extendedly dominated by apixaban. However, the Committee was concerned that there was considerable uncertainty about the relative treatment effects and cost effectiveness of apixaban, rivaroxaban and dabigatran arising from differences in the baseline characteristics of the people included in the trials and the relative treatment effects attributed to the individual anticoagulants that informed the network meta-analysis. The Committee concluded that there was insufficient evidence to distinguish between the cost effectiveness of apixaban, dabigatran and rivaroxaban at this time.

4.16 Finally, the Committee concluded that the decision about whether to start treatment with apixaban should be made after an informed discussion between the clinician and the person about the risks and benefits of apixaban compared with warfarin, dabigatran and rivaroxaban. For people who are taking warfarin, the potential risks and benefits of switching to apixaban should be considered in light of their level of INR control.