Dronedarone for the treatment of non-permanent atrial fibrillation

The Committee discussed the evidence on the effect of dronedarone on atrial fibrillation recurrence. It considered that the randomised controlled trials (ATHENA, EURIDIS, ADONIS and DIONYSOS) demonstrated that dronedarone was more effective than placebo at reducing atrial fibrillation recurrence, but less effective than amiodarone. It discussed the results of the meta-analyses and mixed treatment comparison conducted by the manufacturer that indicated that all antiarrhythmic drugs, including dronedarone, were effective at reducing atrial fibrillation recurrence, but that dronedarone was the least effective. The Committee concluded that dronedarone reduced atrial fibrillation recurrence compared with placebo, but that it appeared to be less effective for atrial fibrillation recurrence than other antiarrhythmic drugs.

The Committee discussed the evidence on the effect of dronedarone on ventricular rate. It noted that the licensed indication for dronedarone was to prevent recurrence of atrial fibrillation or to lower ventricular rate, but that the only studies that assessed ventricular rate in people with non-permanent atrial fibrillation were the EURIDIS and ADONIS trials. (Note that this refers to the licensed indication at the time of the appraisal and not the indication amended in 2011.) It noted that these trials reported a lower ventricular rate in the dronedarone group than the placebo group. The Committee was also aware that ventricular rate was not included in the manufacturer's economic model. It heard from clinical specialists that the ERATO trial had shown that dronedarone reduces ventricular rate compared with placebo. However, the Committee noted that this study was in people with permanent atrial fibrillation, which is not a licensed indication for dronedarone. The Committee considered that there was insufficient evidence to reach a definitive conclusion about the benefits of dronedarone for reducing ventricular rate in people with non-permanent atrial fibrillation.

The Committee discussed the trial evidence on the effect of dronedarone on mortality. It considered the results of the ATHENA trial, which reported lower cardiovascular mortality in the dronedarone group compared with the placebo group but no statistically significant difference in all-cause mortality (for the whole trial population). The Committee concluded that a reduction in all-cause mortality with dronedarone compared with placebo had not been demonstrated by the ATHENA trial for the whole trial population. The Committee discussed the post-hoc subgroup analysis of the ATHENA trial that reported a lower risk of all-cause mortality in the dronedarone group compared with the placebo group for people with a CHADS₂ score of 4 or more. It discussed the use of the CHADS₂ score to predict mortality. It heard from clinical specialists that the CHADS₂ score was a useful method of assessing stroke risk in people with atrial fibrillation to determine the need for anticoagulation treatment, but that it was not used to predict mortality. It noted published evidence that reported an association between CHADS₂ score and risk of all-cause mortality for people who have had a stroke. However, the Committee considered that no evidence had been presented to validate the use of CHADS₂ score for more generally predicting all-cause mortality in people with atrial fibrillation.

The Committee discussed the issue of all-cause mortality in relation to the DIONYSOS trial, which provided the only direct head-to-head comparison of dronedarone with another antiarrhythmic drug, and reported fewer deaths in the dronedarone group than in the amiodarone group. However, the Committee noted the short follow-up of the trial and the small number of deaths in the study, and it considered whether all the deaths in the amiodarone group were because of cardiovascular causes (noting that the causes of death were marked commercial in confidence). Therefore the Committee considered that no conclusion about the relative effect of dronedarone and amiodarone on mortality could be made on the basis of this trial.

The Committee discussed the results of the manufacturer's mixed treatment comparison, which reported a lower risk of all-cause mortality for dronedarone compared with placebo. It noted the ERG's criticism about the methodology used for the mixed treatment comparison. It also noted that the CIs for this analysis crossed the null effect value, indicating that the difference was not statistically significant. The Committee considered that the mixed treatment comparison was largely based on the difference in all-cause mortality between the treatment arms in the ATHENA trial, which itself was not statistically significant. The Committee concluded that there was considerable uncertainty about the effect of dronedarone on all-cause mortality. It was not persuaded that a reduction in the risk of all-cause mortality for dronedarone compared with placebo had been demonstrated by the mixed treatment comparison.

The Committee then discussed the evidence on the risk of all-cause mortality for the other antiarrhythmic drugs. It noted results from the mixed treatment comparison showing that sotalol and amiodarone were associated with a higher risk of mortality than placebo. It discussed the results of the ERG's reanalysis of the mixed treatment comparison and also considered evidence from two published meta-analyses of antiarrhythmic drugs. The Committee noted that all of these analyses reported a trend towards increased all-cause mortality with sotalol and amiodarone compared with placebo, albeit less than that reported in the manufacturer's mixed treatment comparison. It also noted that only the hazard ratio for the comparison of sotalol with placebo for all-cause mortality had a CI that did not cross 1.0, indicating a statistically significant difference between these drugs. The Committee was aware that there were limited data on class 1c drugs. It considered comments received at consultation and from the clinical specialists that amiodarone, sotalol and class 1c drugs were associated with an increased risk of mortality. Overall, the Committee accepted that the risk of mortality with the other antiarrhythmic drugs was likely to be higher than with dronedarone.

The Committee considered the evidence on the effect of dronedarone on the risk of stroke. It heard from clinical specialists that stroke was a known complication of arrhythmias such as atrial fibrillation. Therefore, drugs that are more effective in reducing atrial fibrillation might be expected to have a greater long-term benefit in relation to stroke prevention. The Committee discussed the post-hoc analysis of the ATHENA trial and noted that the ATHENA investigators concluded that 'the observation of a reduced rate of stroke in patients receiving dronedarone cannot be considered a definitive conclusion'. It discussed the mixed treatment comparison that resulted in reduced risk of stroke with dronedarone compared with placebo. It noted that this analysis was based on a small number of studies with very few events and no studies had any prospective collection of data on stroke incidence. The Committee concluded there was considerable uncertainty about the effect of dronedarone on the risk of stroke. It was not persuaded that a reduction in the risk of stroke with dronedarone compared with other antiarrhythmic drugs had been demonstrated.

The Committee discussed the adverse events associated with dronedarone. It was aware of the ANDROMEDA trial in which dronedarone was associated with an increased risk of mortality in people with severe congestive heart failure and noted that having atrial fibrillation was not an inclusion criterion for this trial. It was also aware that the SPC states that dronedarone is contraindicated in people with unstable NYHA class 3 and 4 heart failure, and that it is not recommended in people with stable, recent NYHA class 3 heart failure and people with left ventricular ejection fraction less than 35%. Note that this refers to the licensed indication at the time of the appraisal and not the indication amended in 2011.) The Committee noted that the most common adverse events reported across the trials of dronedarone were gastrointestinal. It discussed the possibility of serious adverse events such as pulmonary fibrosis, thyroid disease and torsades de pointes (an arrhythmia). It noted that there were very few cases reported in the randomised controlled trials, but that these trials were all relatively short term. The Committee considered evidence from the DIONYSOS trial that showed that people in the dronedarone group had fewer adverse events than those in the amiodarone group. However, it noted that these results were also based on short-term data. It noted there was no direct evidence comparing the adverse effects of dronedarone with sotalol or class 1c drugs. The Committee heard from patient experts that they did not consider that the adverse events associated with dronedarone would impact significantly on quality of life for people with atrial fibrillation. It noted comments from patients and clinical specialists received during consultation on the first ACD that all current antiarrhythmic drugs had side effects that had a significant impact on quality of life, but particularly amiodarone, with long-term use. Overall, the Committee concluded that the adverse effect profile of dronedarone was likely to be more favourable than amiodarone.

The Committee specifically considered the balance between the better short-term side-effect profile and the lower effectiveness of dronedarone compared with amiodarone. It heard from patient experts that some people with atrial fibrillation might prefer to take an antiarrhythmic drug that has better tolerability, despite it having less effect on atrial fibrillation recurrence. It also heard that effectiveness of an antiarrhythmic drug for reducing atrial fibrillation recurrence and its tolerability could be more important to some people with atrial fibrillation than longer-term benefits such as a reduction in the risk of stroke or death. The Committee was aware of the potential value placed on dronedarone by patients when it examined the cost-effectiveness analyses.

The Committee discussed the economic analysis provided by the manufacturer. It noted the ERG's conclusion that the approach used was, in general, appropriate and in accordance with the NICE reference case. However, the Committee was concerned about some of the key assumptions in the model, in particular that there was a beneficial effect of dronedarone on mortality and that the ATHENA trial was an appropriate source of data for the baseline risk of events. It also had concerns about the modelled costs of dronedarone and other antiarrhythmic drugs and the utilities used in the model. In addition, the Committee noted that the economic analysis was based on pair-wise comparisons of two treatments and did not use incremental analyses to compare all treatments simultaneously. It also noted that the economic analysis did not evaluate all possible uses of dronedarone.

The Committee discussed the use of the placebo arm of the ATHENA trial to inform the baseline event rates in the model. It considered that this was based on an assumption that the ATHENA population was representative of people with atrial fibrillation in the UK. The Committee noted that the ATHENA trial included people who were older and had higher cardiovascular risk than people in the other dronedarone trials. It considered that the ATHENA population represented a higher-risk group than the more general population of people with atrial fibrillation in the UK in whom dronedarone would be used. Therefore, the Committee concluded that the cost-effectiveness estimates it had been presented with were only relevant to the population represented by the ATHENA trial (that is, people with additional cardiovascular risk).

The Committee noted that in the manufacturer's base-case analysis, the ICERs ranged from £1,900 to £20,800 per QALY gained depending on the type of atrial fibrillation, the presence of structural heart disease, left ventricular dysfunction and coronary heart disease, and the comparator. It considered the ERG's revisions to the model to be appropriate, involving correction of coding errors and use of a lifetime time horizon for mortality benefits, which resulted in decreased ICERs ranging from £1,700 to £19,000 per QALY gained. However, it noted that these figures did not incorporate changes in other key assumptions such as the mortality benefit associated with dronedarone.

The Committee discussed the ERG's analyses exploring the main factors that influenced the cost effectiveness of dronedarone. It considered that the relative effect of dronedarone and antiarrhythmic drugs on mortality was a key factor in the economic analysis. It again considered comments received during consultation on the first ACD and from the clinical specialists about the likely excess mortality associated with other antiarrhythmic drugs. In light of the uncertainty about the effect of dronedarone on mortality, the Committee discussed several exploratory scenario analyses conducted by the ERG in which the relative all-cause mortality effects of each antiarrhythmic drug were varied. It considered the scenario in which dronedarone had no effect on mortality compared with placebo, and other antiarrhythmic drugs were associated with some increase in mortality (as calculated in the ERG's meta-analysis), to be the most appropriate, given the uncertainty about the mortality effect of dronedarone. It noted the ICERs from this analysis were below £15,000 per QALY gained for the use of dronedarone as a second-line treatment alternative to sotalol, class 1c drugs and amiodarone. It also noted the ICERs from this analysis were above £50,000 per QALY gained for the use of dronedarone as part of first-line treatment in addition to standard baseline therapy (in people with a CHADS₂ score of 4 or more). The Committee concluded that these cost-effectiveness estimates were the most plausible of all those presented.

The Committee discussed the costs and utilities included in the economic analysis. It noted that lower initiation and monitoring costs were attributed to dronedarone than other antiarrhythmic drugs. The Committee also noted the ERG's criticism that the utilities used in the model appeared to exceed general population values for healthy states. It considered the ERG's analyses that used revised drug administration costs and utilities, and concluded that these changes had a marginal impact on the cost-effectiveness estimates of dronedarone.

The Committee noted comments received at consultation on the first ACD that the economic analysis had not taken into account the potential cost savings of the reduced cardiovascular hospitalisations associated with dronedarone (shown in the ATHENA trial). It considered that the costs and effects of hospitalisation were included in the analysis through the modelling of events such as atrial fibrillation recurrence and stroke for which hospitalisation costs may be incurred. The Committee also considered comments received at consultation on the first ACD that treatment with dronedarone would be stopped if it was not effective and this was not considered in the economic evaluation. It noted that the submission provided by the manufacturer did not specifically evaluate treatment stopping rules and also did not consider the full range of potential treatment sequences for dronedarone. However, the Committee considered that some element of treatment discontinuation had been accounted for in the modelling of withdrawal because of adverse events or lack of efficacy.

In light of the above considerations, the Committee discussed the cost effectiveness of dronedarone as a first-line treatment for atrial fibrillation (in addition to standard baseline therapy usually including beta-blockers). It noted that cost-effectiveness evidence for dronedarone as a first-line therapy had only been provided for the subgroup of people with a CHADS₂ score of 4 or more. In addition to concerns about the validity of the CHADS₂ score in this context, the Committee considered that the beneficial effect of dronedarone on all-cause mortality assumed in the manufacturer's submission was not proven. It noted that when this effect was removed from the economic analysis the cost per QALY gained was above £50,000. The Committee concluded that the use of dronedarone in people with a CHADS₂ score of 4 or more (in addition to standard baseline therapy) for the first-line treatment of atrial fibrillation could not be considered a cost-effective use of NHS resources. The Committee also noted that no evidence had been provided for first-line treatment with dronedarone other than in people with a CHADS₂ score of 4 or more. Therefore it could not make any conclusions about the first-line use of dronedarone in other people with non-permanent atrial fibrillation. The Committee concluded that dronedarone could not be recommended as a first-line treatment for atrial fibrillation (in addition to standard baseline therapy usually including beta-blockers).

The Committee discussed the cost effectiveness of dronedarone as a second-line treatment for people whose atrial fibrillation is not controlled by standard baseline therapy (that is, as an alternative to the antiarrhythmic drugs: amiodarone, sotalol and class 1c agents). It considered that a beneficial effect of dronedarone on all-cause mortality was not proven; however, it accepted that the risk of mortality with the other antiarrhythmic drugs was likely to be higher than with dronedarone. It considered that when this scenario was modelled, the costs per QALY gained were within an acceptable range. The Committee noted that these cost-effectiveness estimates were largely based on data from the ATHENA trial, which included people who had a higher risk of a major cardiovascular event, and it was uncertain whether these data were applicable to people in England and Wales who would receive second-line treatment for atrial fibrillation. Therefore the Committee concluded that using dronedarone as a second-line alternative to amiodarone, class 1c drugs, or sotalol for the treatment of non-permanent atrial fibrillation could be considered a cost-effective use of NHS resources in people who have the same characteristics as the population in the ATHENA trial, that is, they have at least one of the following additional cardiovascular risk factors: hypertension requiring drugs of at least two different classes, diabetes, previous transient ischaemic attack, stroke or systemic embolism, left atrial diameter at least 50 mm, left ventricular ejection fraction less than 40% or age 70 years or older. The Committee was mindful that there might be some overlap between people with cardiovascular risk factors and those in whom dronedarone is contraindicated (with unstable NYHA class 3 or 4 heart failure) or not recommended (with left ventricular ejection fraction less than 35%). Therefore the Committee considered it important to emphasise in its recommendations that dronedarone should not be used in people with unstable NYHA class 3 or 4 heart failure and to refer to the recommendation in the SPC about the use of dronedarone in people with left ventricular ejection fraction less than 35%.