Dabigatran etexilate for the prevention of stroke and systemic embolism in atrial fibrillation

3.1 The manufacturer's submission included three trials that directly compared dabigatran with dose-adjusted warfarin: RE-LY, PETRO and 1160.49. The PETRO and 1160.49 trials were both dose-finding studies with safety data collection as the primary objective. The main evidence for clinical effectiveness presented in the manufacturer's submission was based on the RE-LY randomised controlled trial.

3.2 RE-LY was a non-inferiority trial in which two blinded doses of dabigatran (110 mg twice daily and 150 mg twice daily) were compared with open-label warfarin (with a target international normalised ratio [INR] of 2.0 to 3.0) for the prevention of stroke and systemic embolism in people with non-valvular atrial fibrillation and at least one additional risk factor for stroke. The RE-LY trial included people with documented atrial fibrillation and at least one of the following additional risk factors: history of stroke, transient ischaemic attack, or systemic embolism; left ventricular ejection fraction of less than 40%; symptomatic heart failure; age 75 years or older; age 65 years or older with diabetes mellitus, documented coronary artery disease or hypertension. People were excluded from the RE-LY trial if they had a severe, disabling stroke in the previous 6 months or any stroke within the previous 14 days, conditions associated with increased risk of bleeding, or a contraindication to warfarin treatment.

3.3 The RE-LY study took place in 44 countries including the UK and a total of 18,113 people were randomised across the three treatment arms in a 1:1:1 ratio. People recruited into the study were randomised within 14 days of the screening visit and were randomly allocated to dabigatran 110 mg twice daily (n=6015), dabigatran 150 mg twice daily (n=6076) or warfarin (n=6022). Minimum follow-up was 1 year, and median follow-up was 23.7 months. The mean age of people in the study was 71.5 years and 63.6% were male. Risk of stroke at baseline was classified according to CHADS₂ score, which is used to estimate the risk of stroke in people with atrial fibrillation to determine whether they need anticoagulation treatment. The score was calculated by giving one point each for the presence of congestive heart failure, hypertension or diabetes mellitus, and age 75 years or older. Two points were given if people had already had an ischaemic stroke or transient ischaemic attack.

3.4 The primary outcome in the study was incidence of all types of stroke (including haemorrhagic stroke) or systemic embolism. To show non-inferiority in the RE-LY trial, the upper limits of the confidence interval (CI) of the hazard ratio (HR) for dabigatran versus warfarin had to be less than the margin specified. Two margins were used in the manufacturer's submission, 1.46 and 1.38, of which 1.38 was specified as the preferred margin of non-inferiority by the US Food and Drug Administration (FDA).

3.5 The reduction in relative risk of stroke or systemic embolism compared with warfarin was 10% for dabigatran 110 mg and 35% for dabigatran 150 mg. Dabigatran 150 mg twice daily was associated with a lower incidence of stroke or systemic embolism compared with warfarin and this was statistically significant (HR=0.65, 95% CI 0.52 to 0.81). A statistically significant beneficial effect of dabigatran 150 mg twice daily was also demonstrated in terms of a reduced incidence of ischaemic stroke (HR=0.75, 95% CI 0.58 to 0.97) and vascular mortality (HR=0.85, 95% CI 0.72 to 0.99). A reduction in all-cause mortality was also observed and, although it did not reach statistical significance, it showed dabigatran 150 mg twice daily to be non-inferior to warfarin (HR=0.88, 95% CI 0.77 to 1.00). There were no statistically significant differences between dabigatran 110 mg twice daily and warfarin in the incidence of stroke or systemic embolism, ischaemic stroke or vascular mortality. Both doses of dabigatran were associated with an increased risk of acute myocardial infarction compared with warfarin but this was not statistically significant (HR=1.29, 95% CI 0.96 to 1.75 [110 mg twice daily]; HR=1.27, 95% CI 0.94 to 1.71 [150 mg twice daily]).

3.6 The manufacturer's submission included post hoc subgroup analyses of people older and younger than 80 years of age. In both age groups, there were no statistically significant differences between either dose of dabigatran and warfarin in the incidence of ischaemic stroke, systemic embolism and myocardial infarction. However, the manufacturer did report a statistically significant reduction in the incidence of transient ischaemic attack (HR=0.45, 95% CI 0.23 to 0.89) in people older than 80 years receiving dabigatran 110 mg twice daily, compared with warfarin.

3.7 The manufacturer's submission reported results from pre-planned subgroup analyses of people naive to vitamin K antagonists such as warfarin (defined as treatment for 2 months or less in a person's lifetime) and people who have previously used vitamin K antagonists (defined as treatment for more than 2 months during a person's lifetime). In both groups, dabigatran 150 mg twice daily was associated with a statistically significant reduction in the incidence of stroke or systemic embolism compared with warfarin (HR=0.63, 95% CI 0.46 to 0.87 [vitamin K antagonist-naive group] and HR=0.63, 95% CI 0.49 to 0.89 [vitamin K antagonist-experienced group]). No statistically significant differences were reported for the lower, 110 mg twice daily, dose of dabigatran compared with warfarin.

3.8 For adverse events, the manufacturer reported a statistically significant reduction in the incidence of haemorrhagic stroke for both doses of dabigatran compared with warfarin (HR=0.31, 95% CI 0.17 to 0.56 [dabigatran 110 mg twice daily] and HR=0.26, 95% CI 0.14 to 0.49 [dabigatran 150 mg twice daily]). Both doses of dabigatran were also associated with statistically significantly fewer life-threatening bleeds compared with warfarin (HR=0.67, 95% CI 0.54 to 0.82 [dabigatran 110 mg twice daily] and HR=0.80, 95% CI 0.66 to 0.98 [dabigatran 150 mg twice daily]). Both doses of dabigatran were associated with fewer cases of intracranial haemorrhage (including haemorrhagic stroke) than warfarin (HR=0.30, 95% CI 0.19 to 0.45 [dabigatran 110 mg twice daily]; HR=0.41, 95% CI 0.28 to 0.61 [dabigatran 150 mg twice daily]). Treatment with dabigatran 110 mg was also associated with a statistically significant reduction in major bleeding compared with warfarin. In contrast, both doses of dabigatran were associated with a significantly higher rate of gastrointestinal bleeding compared with warfarin (HR=1.35, 95% CI 1.19 to 1.53 [dabigatran 110 mg twice daily] and HR=1.52, 95% CI 1.35 to 1.72 [dabigatran 150 mg twice daily]). Dabigatran 150 mg twice daily was associated with a significantly higher incidence of major gastrointestinal bleeding (HR=1.47, 95% CI 1.17 to 1.85) and life-threatening gastrointestinal bleeding (HR=1.62, 95% CI 1.17 to 2.26). The manufacturer reported that more people in the dabigatran groups discontinued the study drug (22.0% in the dabigatran 110 mg twice daily group and 22.8% in the dabigatran 150 mg twice daily group), compared with those on warfarin (17.9%). More people in the dabigatran groups also discontinued study medication because of outcome events; however discontinuations caused by major bleeds were similar for all treatments.

3.9 The manufacturer reported a statistically significant reduction in the incidence of haemorrhagic stroke in the post hoc subgroup analyses of people younger than 80 years compared with warfarin for both doses of dabigatran (HR=0.33, 95% CI 0.16 to 0.65 [dabigatran 110 mg twice daily]; HR=0.21, 95% CI 0.09 to 0.47 [dabigatran 150 mg twice daily]) and in people older than 80 years receiving dabigatran 110 mg twice daily (HR=0.26, 95% CI 0.07 to 0.91). However, the reduction in incidence of haemorrhagic stroke in people older than 80 years for dabigatran 150 mg twice daily compared with warfarin was not statistically significant (HR=0.93, 95% CI 0.81 to 1.07).

3.10 Health-related quality of life data were collected in a sub-study of the RE-LY trial (1440 of the 18,113 people enrolled in the RE-LY study completed the EQ-5D questionnaire as part of the quality of life sub-study). The manufacturer reported that the sub-study was reasonably representative of the overall RE-LY population with patients having similar demographic and disease characteristics. The manufacturer stated that analysing the EQ-5D data for specific events of interest was not possible and the quality of life sub-study was unable to provide utility values for event-driven health states to use in the economic model. However, background utility values could be derived from the quality of life sub-study for people being treated with warfarin and dabigatran, the details of which are academic-in-confidence and are not reported here.

3.11 The manufacturer performed a mixed-treatment comparison of dabigatran, aspirin monotherapy and aspirin plus clopidogrel. The treatments considered by the manufacturer to be relevant in this analysis were dabigatran 150 mg twice daily, dabigatran 110 mg twice daily, dose-adjusted warfarin, aspirin, aspirin plus clopidogrel, and placebo. An additional sequential regimen of dabigatran was used in the mixed-treatment comparison. This was intended to reflect the use of dabigatran according to the licensed regimen which is 150 mg twice daily in people up to the age of 80 years, and then 110 mg twice daily in those aged 80 years and older. Results from the RE-LY trial and the mixed-treatment comparison were very similar for both dabigatran doses compared with dose-adjusted warfarin.

3.12 The manufacturer's economic evaluation was based on a cost–utility analysis designed to compare the costs and outcomes of dabigatran with treatments used in the UK (warfarin, aspirin and aspirin plus clopidogrel). The manufacturer developed a Markov model that used three levels of disability (independent, moderate and severe) and death to define health states. A hypothetical cohort entered the model at risk of specified clinical events and was on one of the treatments under comparison. They moved between health states when a clinical event occurred and their disability status changed. The clinical events considered were ischaemic stroke, intracranial haemorrhage, haemorrhagic stroke, extracranial haemorrhage, systemic embolism, transient ischaemic attack and acute myocardial infarction. All clinical outcomes were associated with acute costs and disutility. Further longer-term costs and disutility beyond the acute stage were associated with ischaemic stroke, haemorrhagic stroke and intracranial haemorrhage. The model permitted one clinical event per 3-month cycle over a lifetime horizon. The model also allowed for a switch to second-line treatment or a discontinuation of treatment.

3.13 The manufacturer presented two economic models: a single-dose model and a sequential regimen model. In the single-dose model, the cohort with atrial fibrillation received either 110 mg twice daily or 150 mg twice daily throughout their treatment. In the sequential regimen model, the cohort was divided by age and modelled separately. The model for people younger than 80 years assumed that treatment began with dabigatran 150 mg twice daily, and switched to dabigatran 110 mg twice daily when the age of 80 years was reached. The model for people aged 80 years or older at baseline assumed a dose of dabigatran 110 mg twice daily throughout. Therefore, the sequential regimen model resulted in two sets of outputs: a sequential regimen model for people starting treatment younger than 80 years (incorporating a life-time horizon including the switch to 110 mg twice daily at 80 years) and a sequential regimen model for those starting treatment at 80 years or older.

3.14 The event risk for all treatment strategies was applied to the baseline risk of events in people treated with warfarin in the RE-LY trial. Therefore, treatment effects were converted into relative risks and applied to the warfarin arm of the RE-LY trial. The relative risks for the various clinical events while on treatment with dabigatran 110 mg twice daily and 150 mg twice daily were obtained from the RE-LY trial. In the sequential regimen model, the relative risks were derived from the post hoc subgroup analyses of people older and younger than 80 years of age. The relative risks for aspirin, aspirin plus clopidogrel and placebo were obtained from the mixed-treatment comparison.

3.15 The manufacturer's economic evaluation focused on health-related quality of life associated with disability and disutility caused by the various clinical events. The baseline utility value for people with atrial fibrillation in the base-case analyses was taken from the RE‑LY quality of life sub-study. Utility values associated with clinical events and disability status were derived from published sources.

3.16 The manufacturer's model considered resource costs associated with anti-thrombotic treatment, acute event costs, and long-term follow‑up costs resulting from disability. These costs were derived from the national payment by results tariff, systematic reviews and a manufacturer-sponsored study based on the Oxford Vascular study (OXVASC) cohort. The cost of dabigatran was £2.52 (excluding VAT) per day for either the 110 mg twice daily or 150 mg twice daily doses. Treatment with warfarin, aspirin and aspirin plus clopidogrel was assumed to cost £0.04, £0.09, and £0.26 per day, respectively. Treatment with dabigatran was not considered to need any monitoring, but the cost of INR monitoring for warfarin was estimated to be £414.90 per annum. The model assumed an NHS perspective and costs and benefits were discounted at 3.5% per annum.

3.17 The manufacturer reported pairwise cost-effectiveness results for dabigatran compared with warfarin. The incremental cost-effectiveness ratios (ICERs) for the dabigatran sequential regimen in which people started treatment when younger than 80 years and continued for the rest of their lives, and the sequential regimen in which people started treatment when older than 80 years were £7314 and £7873 per QALY gained respectively, compared with warfarin. The ICERs for dabigatran 150 mg and 110 mg twice daily compared with warfarin were £6264 and £18,691 per QALY gained respectively.

3.18 The manufacturer performed structural, univariate and probabilistic sensitivity analyses to reflect uncertainty in the model inputs and assumptions. The structural sensitivity analysis explored the cost effectiveness of dabigatran by varying INR cost (±25%), time horizon (2, 10 and 15 years), and discount rate (0 to 6%). The cost effectiveness of dabigatran was highly sensitive to the time horizon specified. A 2-year time horizon resulted in ICERs of £75,891 and £23,403 per QALY gained, respectively, for the dabigatran sequential regimen in people starting treatment when younger than 80 years and the dabigatran sequential regimen in people starting treatment when older than 80 years, compared with warfarin. For dabigatran 150 mg and 110 mg twice daily, the ICERs were £75,601 and £108,736 per QALY gained, respectively.

3.19 In the univariate sensitivity analysis, the cost effectiveness of the dabigatran sequential regimen in people starting treatment when younger than 80 years was most sensitive to risk of ischaemic stroke. Setting the relative risk for ischaemic stroke to the 95% upper confidence limits increased the base-case ICER compared with warfarin from £7314 to £17,100 per QALY gained. The cost effectiveness of the dabigatran sequential regimen in people starting treatment when older than 80 years was most sensitive to risk of ischaemic stroke and high baseline CHADS₂ scores. Setting the relative risks for ischaemic stroke to the 95% upper confidence limits increased the base-case ICER for the dabigatran sequential regimen in people older than 80 years compared with warfarin from £7873 to £46,509 per QALY gained. The ICER for the dabigatran sequential regimen in people starting treatment when older than 80 years compared with warfarin increased from the base-case estimate of £7873 to £21,129 per QALY gained for a group with a CHADS2 score of 5. The ICER for dabigatran 150 mg twice daily compared with warfarin was robust to the parameters and ranges tested by the manufacturer, and the highest ICER was £10,234 per QALY gained. The cost effectiveness of dabigatran 110 mg twice daily in relation to warfarin was highly sensitive to high baseline CHADS2 scores, risk of ischaemic stroke and risk of intracranial haemorrhage.

3.20 In the probabilistic sensitivity analysis, the ICERs for the dabigatran sequential regimens in people starting treatment when younger than 80 years and in people starting treatment when older than 80 years compared with warfarin were £7811 and £11,912 per QALY gained respectively. The probabilistic ICERs for dabigatran 150 mg and 110 mg twice daily compared with warfarin were £7940 and £15,867 per QALY gained respectively.

3.21 The ERG noted that the manufacturer's submission included two generally well-conducted systematic reviews: the first was of dabigatran trials in the relevant indication, and the second was of all potentially relevant pharmacological interventions for the prevention of stroke in people with atrial fibrillation. The ERG commented that the RE‑LY trial was of good quality and that the manufacturer appropriately concentrated on the results from this trial. The ERG highlighted the limitations of non-inferiority trials, such as establishing the non-inferiority margin and the population on which to base analyses. Overall, the ERG felt that adequate measures were taken by the manufacturer to reduce the impact of potential bias associated with non-inferiority trials.

3.22 The ERG commented that the results of the RE-LY trial showed both doses of dabigatran to be non-inferior to dose-adjusted warfarin in the prevention of stroke or systemic embolism. The ERG noted that a submission from the manufacturer to the FDA indicated that dabigatran 150 mg twice daily reduced the risk of stroke or systemic embolism compared with warfarin in people with good INR control (HR=0.68, 95% CI 0.50 to 0.92 for time in therapeutic INR range 65% or above; HR=0.70, 95% CI 0.51 to 0.96 for time in therapeutic INR range 68% or above). The ERG also highlighted that an analysis in the submission produced for the FDA showed a greater benefit of dabigatran in people with poor INR control than in those whose INR was well controlled (the threshold being the centre-level median of 67%). The FDA report concluded that, although the results showed efficacy of dabigatran in people who had INR control above the centre-level median, the results did not show superiority over warfarin. The submission further subdivided people by INR control (less than 58.5%, 58.5% or above, less than 66.8%, 66.8% or above, and less than 74.2%). This demonstrated that the greatest benefit of dabigatran was in the lowest quartile of INR control and that, in people with good INR control with warfarin, little or no additional benefit in terms of effectiveness would be gained with dabigatran.

3.23 A key uncertainty highlighted by the ERG was the generalisability of the results to people with atrial fibrillation in the NHS. The ERG commented that the definition of moderate or high risk of stroke or systemic embolism in the manufacturer's submission differed slightly to the definition in 'The management of atrial fibrillation' (NICE clinical guideline 36). The ERG commented that the population in the manufacturer's submission seemed to be at higher risk of stroke because the definition of moderate risk included those aged 75 years and over with no additional risk factors, whereas NICE clinical guideline 36 defines moderate risk as people aged 65 years and over with no additional risk factors. The ERG commented that including the potentially large subgroup of people over 65 years with atrial fibrillation but with no other risk factors for stroke would have been useful, and would reflect NICE clinical guideline 36 more closely and reduce the overall risk level of the population. The clinical specialists advising the ERG noted that the threshold for treatment with warfarin seems to be decreasing, therefore decreasing the risk of stroke in the eligible atrial fibrillation population, making the population in the RE-LY trial less representative of clinical practice over time.

3.24 The ERG commented that the general approach taken by the manufacturer to estimate lifetime cost effectiveness was appropriate and met the requirements of the NICE reference case. The ERG noted that the model included most of the relevant clinical events in atrial fibrillation; however, pulmonary embolism was not included in the model. The ERG commented that excluding pulmonary embolism is potentially an optimistic approach in favour of dabigatran because dabigatran is associated with higher rates of pulmonary embolism than warfarin.

3.25 The ERG noted that, although the manufacturer's submission considered the atrial fibrillation population to be heterogeneous, reflected by the distribution of CHADS₂ scores, the manufacturer assumed that all people would be treated the same. The ERG commented that this may be an over-simplification of the decision problem and does not allow the potential impact of clinical heterogeneity on cost effectiveness to be considered.

3.26 The ERG highlighted that acute myocardial infarction and systemic embolism were assumed by the manufacturer to be associated with acute costs and disutility, but not with any ongoing or long-term consequences. The ERG considered this assumption to be over-simplistic and that the effect of including long-term consequences of acute myocardial infarction and systemic embolism on the cost effectiveness of dabigatran is uncertain. The ERG commented that dabigatran was associated with higher discontinuation rates than warfarin in the first 2 years of the trial, which could suggest that people tend to tolerate warfarin better than dabigatran.

3.27 The two main weaknesses of the manufacturer's model were considered by the ERG to be related to the sequence of treatments and the cost of anticoagulation monitoring. The ERG commented that the full set of relevant sequences of treatment was not fully investigated by the manufacturer. For example, the ERG considered that starting treatment with dabigatran and subsequently switching to warfarin would be a reasonable treatment sequence, but the manufacturer's model assumed that a person could not switch to warfarin if dabigatran was the first treatment. In addition, the ERG stated that the cost of anticoagulation monitoring was a key driver of the model in terms of resources and costs, and that it was likely that the average cost of monitoring had been overestimated in the model, biasing the results in favour of dabigatran. The ERG also highlighted that its clinical advisers were concerned about the high variability of monitoring costs in practice. This heterogeneity was not considered in the manufacturer's submission. The ERG commented that uncertainty around the monitoring costs was also inadequately modelled in the manufacturer's submission.

3.28 The ERG carried out exploratory cost-effectiveness analyses by subgroups according to INR control with warfarin. The ICER for dabigatran 150 mg twice daily compared with warfarin in people with perfect INR control (that is, in target INR range 100% of the time for the entire duration of treatment) was £60,895 per QALY gained. Dabigatran 110 mg twice daily was dominated by warfarin because it was associated with greater costs but lower health benefits. The group of people with poor INR control was also evaluated by the ERG. The ICER for dabigatran 150 mg twice daily compared with warfarin for people with an INR below 2 was £740 per QALY gained. For people with an INR above 3, warfarin was dominated by dabigatran 150 mg twice daily. The ERG did not include pairwise cost-effectiveness results for dabigatran in the sequential regimen compared with warfarin. The ERG concluded that INR control is a key parameter in the economic evaluation.

3.29 The ERG used three approaches to calculate the variable costs of INR monitoring, which it considered had been overestimated in the manufacturer's model. The alternative costs used by the ERG were £279.36, £241.54 and £115.14, instead of £414.90 as assumed by the manufacturer. Adjusting the model to test each individual cost assumption increased the ICER for dabigatran 150 mg twice daily compared with warfarin to £10,528, £11,720 and £15,701 per QALY gained respectively.

3.30 The ERG considered that the disutility of dabigatran captured by the RE-LY quality of life sub-study had not been fully reflected in the manufacturer's cost-effectiveness analysis. The disutility associated with dabigatran treatment was tested by the ERG but it did not change the overall conclusions about the cost effectiveness of this intervention.

3.31 The ERG commented that treatment with dabigatran was associated with an increased incidence of dyspepsia compared with warfarin treatment, but that the model assumed that the cost of dyspepsia was only accrued in the first cycle. The ERG considered that a more conservative approach would be to assume that costs of dyspepsia continue throughout treatment. This caused the ICER for dabigatran 150 mg twice daily compared with warfarin to increase slightly from £6262 per QALY to £6659 per QALY gained.

3.32 The ERG highlighted that disability and mortality risk after stroke is considered to be treatment dependent in the manufacturer's model. Therefore the ERG explored the model assuming that disability caused by stroke is independent of treatment. The ICER for dabigatran 150 mg twice daily compared with warfarin increased from £6262 to £8393 per QALY gained.

3.33 The ERG presented analyses using an alternative set of assumptions to those provided by the manufacturer. The ERG's alternative base case assumed:

3.34 By introducing these assumptions, the ICER for dabigatran 150 mg twice daily compared with warfarin increased from £6264 to £24,173 per QALY gained in the ERG's alternative base-case analysis.