Dual‑chamber pacemakers for symptomatic bradycardia due to sick sinus syndrome without atrioventricular block

Parallel group trials

4.2 The parallel group trials randomised participants to receive either a single‑chamber atrial pacemaker device or a dual‑chamber pacemaker device. The trials varied in size from 50 to 1415 randomised participants, the mean age was similar across the 3 parallel trials and between study arms (72–74 years), and either all or most of the people within each trial had the pacemakers programmed with the rate‑adaptive function activated. All parallel randomised controlled trials excluded patients if they had chronic atrial fibrillation, atrioventricular block, carotid sinus syndrome, vasovagal syncope, bundle branch block, surgery, short life expectancy, dementia or cancer. At baseline, most of the participants had their condition classed (according to the New York Heart Association [NYHA] class – used to classify the extent of heart failure according to the severity of symptoms) as I or II (96%) at the end of follow‑up with no or mild symptoms of heart failure. The Assessment Group conducted random‑effect model meta‑analyses for the parallel trials where appropriate, generating odds ratios.

4.3 Albertsen et al. (2008) compared the impact of dual‑chamber pacemaker devices (n=26) with single‑chamber atrial pacemaker devices (n=24) on left ventricular desynchronisation in people with sick sinus syndrome, including those with sinus arrest or sino‑atrial block, bradycardia‑tachycardia syndrome and sinus bradycardia. The study was based in Denmark, with a follow‑up of 12 months. The primary outcome was changes in left ventricular dyssynchrony (that is, the level of delay or difference in the timing of contraction in the different segments within the left ventricle) from baseline to 12 months of follow‑up. Secondary outcomes included the measurement of N‑terminal prohormone of brain natriuretic peptide (a hormone released in response to heart problems) and a 6‑minute walk test.

4.4 DANPACE (2011) compared rate‑responsive dual‑chamber pacemakers (n=708) with rate‑responsive single‑chamber atrial pacemakers (n=707) for treating sick sinus syndrome (including those with sino‑atrial block or sinus‑arrest, sinus bradycardia and bradycardia‑tachycardia). The study was based in Denmark, the UK and Canada, and there was a mean follow‑up of 5.4 years. The primary outcome was death by any cause. Secondary outcomes included paroxysmal or chronic atrial fibrillation, stroke, cardiovascular mortality, need for pacemaker re‑operation and quality of life.

4.5 Nielsen et al. (2003) compared rate‑responsive dual‑chamber pacemakers with rate‑responsive single‑chamber atrial pacemakers (n=54) for treating sick sinus syndrome (including those with sinus bradycardia, sino‑atrial block and bradycardia‑tachycardia syndrome). The trial had 2 dual‑chamber pacemaker arms, with different programmed atrioventricular block delay: short atrioventricular delay (less than 150 milliseconds, n=60) and long atrioventricular delay (a fixed delay of 300 milliseconds, n=63). The study was based in Denmark and mean follow‑up was 2.9 years. The primary outcome was changes in left atrial size and left ventricular size and function during follow‑up. Secondary outcomes were cardiographic (changes in left atrial volume, left ventricular volume and left ventricular ejection fraction) and clinical (atrial fibrillation, thromboembolism, all‑cause and cardiovascular mortality, and congestive heart failure). The Assessment Group noted that there were some imbalances between the trial arms for the subtypes of sick sinus syndrome.

Crossover trials

4.6 In the crossover trials, all participants had a dual‑chamber pacemaker device implanted, and were then randomised to either single‑ or dual‑ pacing modes and then later crossed over to the alternative pacing mode. Lau et al. (1994) and Schwaab et al. (2001) randomised participants before implant by pacing programme, and Gallick et al. (1994) randomised participants who recently had a pacemaker fitted. The Assessment Group reported that they were unable to undertake meta‑analyses on the crossover trials because of a lack of relevant data in all studies.

4.7 Gallick et al. (1994) compared dual‑chamber pacing mode with single‑chamber atrial pacing mode in people with sick sinus syndrome (including those with sinus node disease; n=12), using ventricular function to study the immediate effects of pacing mode during exercise. Outcomes were exercise and haemodynamic parameters. The trial measured haemodynamic effects during bicycle exercise, initially in 1 pacing mode and, after 0.5 to 1 hour rest, after which the exercise was repeated in the other pacing mode. Gallick et al. excluded people with evidence of atrioventricular node disease or who were unable to exercise. The study location was not reported, and the follow‑up was less than 1 day.

4.8 Lau et al. (1994) compared dual‑chamber, single‑chamber atrial and single‑chamber ventricular pacing modes for people with sick sinus syndrome (n=15), studying the effects of pacing modes and intrinsic conduction on physiological responses, arrhythmias, symptoms and quality of life. Participants spent 4 weeks in each pacing mode before crossing over to the other pacing mode. Lau et al. did not report specific exclusion criteria. The study location was not reported, and follow‑up was 3 months. Outcomes were Holter monitoring, ambulatory blood pressure monitoring, symptoms and quality‑of‑life assessments.

4.9 Schwaab et al. (2001) compared dual‑chamber with single‑chamber atrial pacing mode for people with bradycardia‑tachycardia syndrome (n=21). Participants spent 4 weeks in each pacing mode before crossing over to the other pacing mode. Participants had to have chronotrophic incompetence (that is, an inability of the heart to increase its rate appropriately with increased activity, leading to exercise intolerance), have experienced at least 2 documented episodes of atrial tachyarrhythmia and be on antiarrhythmic medication for prevention of atrial flutter or atrial fibrillation. The study was based in Germany, and the follow‑up period was 3 months. Outcomes included quality of life, left ventricular outflow, bicycle cardiopulmonary exercise testing (to assess outcomes including exercise duration), number of episodes and total duration of atrial tachyarrhythmia, incidence of atrioventricular block type I, II or III and maximum duration of the longest atrioventricular pause, and percentage of paced atrial and ventricular beats.

Parallel and crossover trial quality

4.10 The Assessment Group noted that the quality of the trials was generally high, with appropriate trial design and methodology, and that the trials appeared to be appropriately randomised with a low number of participants excluded or lost to follow‑up. The baseline characteristics were similar between the trial arms and across the parallel and crossover trials. In particular, DANPACE (2011) was considered to be a relatively large trial of good quality with a long follow‑up, which represents the best available evidence comparing dual‑chamber pacing with single‑chamber atrial pacing for people with sick sinus syndrome without evidence of atrioventricular block. However, the Assessment Group noted several limitations of the trials. General limitations included that data were not reported consistently across the trials. Trials were either open label, or blinding to pacing modes was unclear (possibly increasing the risk of bias to subjective outcomes such as quality of life and exercise capacity) and the programmed atrioventricular block delay in the dual‑pacing mode differed greatly between the trials and study arms, adding heterogeneity to the trials. In addition, new technologies in this area develop rapidly, therefore the implants used in the trials may now be superseded, limiting applicability of the results to current devices. Weaknesses of the parallel trials included that Albertsen et al. (2008) and Nielsen et al. (2003) had small sample sizes and short follow‑up in comparison with DANPACE (2011; giving them little weight in meta‑analyses), and both DANPACE and Nielsen et al. were under‑powered to show a statistically significant difference in the primary outcome (all‑cause mortality in DANPACE, and changes in left atrial size and left ventricular size and function in Nielsen et al.) because they were terminated early. Recruitment for Nielsen et al. was stopped after randomisation of 177 patients (from a target of 450) because recruitment for DANPACE had started. Recruitment for DANPACE was stopped after randomisation of 1415 patients (of a target of 1900) because of the increasing use of dual‑chamber pacemakers with additional features that were not permitted in the trial. Limitations of the crossover trials included that the trials had a small number of participants (n=12–21) and short durations (up to 3 months), which limited the possible outcomes and reduced the power to detect differences between pacing modes.

Change in pacing mode

4.12 People in the parallel group trials were randomised to receive a single‑ or dual‑chamber device. During the trial, some participants changed pacing mode from the one to which they were randomised. Some people randomised to the dual‑chamber pacemaker device arm changed to single‑chamber atrial or single‑chamber ventricular mode. Some people randomised to the single‑chamber atrial pacemaker device arm changed to a dual‑chamber pacemaker device (if they developed atrioventricular block) or a single‑chamber ventricular device (if they developed atrial fibrillation). A meta‑analysis of all 3 parallel group trials comparing 857 people randomised to dual‑chamber pacemaker devices with 785 people randomised to single‑chamber atrial pacemaker devices, reported that statistically significantly fewer people with a dual‑chamber pacemaker changed pacing mode than those in the single‑chamber pacemaker group (odds ratio [OR] 0.50, 95% confidence interval [CI] 0.37 to 0.67). Most people who changed from a single‑chamber atrial pacemaker changed to a dual‑chamber pacemaker, primarily because of the development of high degree atrioventricular block, or Wenckebach block during implantation. However, there were also a small number of people who switched from single‑chamber atrial pacing to single‑chamber ventricular pacing, primarily because of persistent atrial fibrillation. The Assessment Group noted that in DANPACE, the results for change in pacing mode and re‑operation were probably conservative because the incidence of atrioventricular block leading to these outcomes would have continued to increase over time beyond the follow‑up period of the trial.

Atrial and ventricular pacing

4.13 The proportion of atrial and ventricular pacing varied greatly between the studies, study arms and pacing mode, which may have been associated with differences in the level of atrial fibrillation across the trials. In the parallel trials, dual‑chamber devices were associated with between 57% and 67% atrial pacing, and ventricular pacing between 17% and 90%. For single‑chamber devices, the rate of atrial pacing ranged from 53% to 69%, and ventricular pacing ranged from 3% to 99% (Albertsen et al. [2008], for those who had upgraded from single to dual). In the crossover trials, the rate of dual‑chamber atrial pacing was only reported by Schwaab et al. (2001; 95%), and the rate of ventricular pacing was reported by Lau et al. (1994; 64%) and Schwaab et al. (99%). For single‑chamber devices, the rate of atrial pacing was not reported, and the rate of ventricular pacing was only reported by Schwaab et al. (95%). The Assessment Group noted that although the dual‑chamber pacemakers in DANPACE (2011) were programmed in a way intended to reduce unnecessary ventricular pacing, ventricular pacing was still 65% (with a range of ±33%), which may have offset some of the benefit of implanting a dual‑chamber pacemaker.

All‑cause mortality

4.14 All‑cause mortality was reported in DANPACE (2011) and Nielsen et al. (2003), and for both studies there were no statistically significant results. DANPACE (for which the primary outcome was all‑cause mortality) had an unadjusted hazard ratio for single‑ compared with dual‑chamber pacemakers of 1.06 (95% CI 0.88 to 1.29), an adjusted (for patient characteristics) hazard ratio of 0.94 (95% CI 0.77 to 1.14), and there were no statistically significant results in any of the pre‑defined subgroups (age, gender, hypertension, left ventricular ejection fraction, history of atrial fibrillation, previous myocardial infarction, PQ interval, diabetes, NYHA classification; p>0.45). For Nielsen et al., the Assessment Group derived an odds ratio from the trial for dual‑ compared with single‑chamber pacemakers of 1.47 (0.64 to 3.38). In a meta‑analysis of both trials (DANPACE and Nielsen et al.), there was no statistically significant difference in all‑cause mortality between dual‑ (n=831) and single‑chamber atrial (n=761) pacemakers (OR 0.97, 95% CI 0.67 to 1.41).

Heart failure

4.15 The outcome measures used as a proxy for heart failure varied between the 3 parallel studies, including:

There was no statistically significant difference between dual and single atrial pacing for the outcome of heart failure. DANPACE (2011) conducted predefined subgroup analyses for single‑ compared with dual‑chamber pacemakers for a younger (75 years or under) and older (over 75 years) population, which showed that in younger people, those with a single‑chamber pacemaker were at a statistically significantly lower risk of developing heart failure than those with dual‑chamber pacemakers (hazard ratio [HR] 0.72, 95% CI 0.53 to 1.00, p=0.05), whereas in the older subgroup, those with single‑chamber pacemakers were at a statistically significantly higher risk than those with dual‑chamber pacemakers (HR 1.34, 95% CI 1.01 to 1.80, p=0.05). All other subgroup analyses were non‑significant (p>0.31).

Atrial fibrillation

4.16 DANPACE (2011) and Nielsen et al. (2003) reported results on the incidence of atrial fibrillation, diagnosed by standard 12‑lead electrocardiogram (ECG) at planned follow‑up visits. In DANPACE, atrial fibrillation was defined as either paroxysmal (the first diagnosis of atrial fibrillation detected in the ECG and verified by the pacemaker telemetry at a planned follow‑up visit) or chronic (atrial fibrillation at 2 consecutive follow‑up visits and at all subsequent follow‑up visits). DANPACE showed that the risk of paroxysmal atrial fibrillation was statistically significantly lower for dual‑ compared with single‑chamber pacemakers (OR 0.75, 95% CI 0.59 to 0.96), although there was no statistically significant difference for chronic atrial fibrillation (OR 0.96, 95% CI 0.68 to 1.33). In addition, subgroup analyses of paroxysmal atrial fibrillation in DANPACE showed that dual‑chamber pacing was associated with statistically significantly lower paroxysmal atrial fibrillation in subgroups of people without a prior history of atrial fibrillation, higher BMI, and a dilated left atrium at baseline (p<0.05). The Assessment Group noted that Nielsen et al. reported conflicting results: that the risk of developing atrial fibrillation (paroxysmal and chronic combined) was statistically significantly higher for dual‑ compared with single‑chamber pacemakers (OR 3.19, 95% CI 1.05 to 9.67).The Assessment Group noted that both DANPACE and Nielsen were good‑quality trials but stated that, because DANPACE was the larger trial (almost 10 times the size of Nielsen et al.) and had a longer mean follow‑up, it was reasonable to have more confidence in the results of DANPACE.

Stroke

4.17 DANPACE (2011) and Nielsen et al. (2003) reported the effectiveness of single‑ compared with dual‑chamber pacemakers for stroke. DANPACE reported a non‑statistically significant unadjusted hazard ratio for stroke of 1.13 (95% CI 0.72 to 1.80, p=0.59) for people with single atrial pacing compared with dual pacing, and an adjusted (for patient characteristics) hazard ratio of 1.11 (95% CI 0.70 to 1.77, p=0.65). The results of the Nielsen et al. study were also not statistically significant (p=0.32). A meta‑analysis of both studies reported no statistically significant difference between dual‑ and single‑chamber pacemaker devices for stroke (OR 0.93, 95% CI 0.60 to 1.45).

Exercise capacity

4.18 Exercise capacity was reported in the parallel trial Albertsen et al. (2008) and in the crossover trials Gallick et al. (1994) and Schwaab et al. (2001). Albertsen et al. reported that at 12 months' follow‑up: people with a single atrial pacemaker walked statistically significantly further than patients with a dual‑chamber pacemaker (p<0.05) based on the 6‑minute walking test to assess either exercise tolerance, capacity or both, measuring the distance a person is able to walk over a total of 6 minutes on a hard, flat surface. However, the Assessment Group noted significant uncertainty in this result, which almost did not reach the clinically important difference of 54–80 metres, and that no statistical significance had been demonstrated at baseline. Schwaab et al. reported a statistically significantly better exercise capacity with single atrial pacing mode compared with dual pacing mode, for bicycle exercise duration and workload (p<0.05). Gallick et al. reported no statistically significant difference between the pacing modes when using the upright bicycle exercise to test exercise capacity (p=0.74) The Assessment Group noted that Gallick et al. was a very short‑term study, with both pacing modes tested in the same day with 0.5 to 1 hour rest in between, which may partly explain the difference in the result from Schwaab et al.

Further surgery

4.19 The need for pacemaker re‑operation during follow‑up was an outcome in DANPACE (2011), in which there were statistically significantly more participants in the single‑chamber pacemaker arm needing a re‑operation compared with the dual‑chamber pacemaker arm (unadjusted HR 1.99, 95% CI 1.53 to 2.59; adjusted [for patient characteristics] HR 2.00, 95% CI 1.54 to 2.61). The only statistically significant difference in reason for re‑operation was need for surgical change of mode of pacing (dual n=4, single n=66, p<0.001), primarilyfrom single‑ to dual‑chamber pacemaker because of the development of high‑grade atrioventricular block.

Adverse effects of pacemaker implantation

4.20 Only Albertsen et al. (2008) and DANPACE (2011) reported data on adverse effects linked to pacemaker implantation. Albertsen et al. considered complications around device implantation and did not report any adverse events in either dual‑ or single‑chamber atrial pacemakers. DANPACE did not report adverse effects at implantation, but did consider the indications for re‑operation during follow‑up. Of 1415 patients in both arms, 240 had 1 or more re‑operations during the follow‑up period. The more frequent indications for re‑operation were battery depletion (dual n=42, single n=59), lead complications (dual n=30, single n=37) and need for change of pacing mode (dual n=4, single n=66). Less common indications for re‑operation were surgical or mechanical complications (dual n=7, single n=10), infection (dual n=3, single n=3), skin erosion (dual n=3, single n=1) or device failure (dual n=2, single n=2). The only indication that was statistically significantly different between the dual and single atrial pacemaker arm was surgical change in pacing mode.

Health‑related quality of life

4.21 Quality of life was studied in the crossover trials Lau et al. (1994) and Schwaab et al. (2001). Lau et al. used the Visual Analogue Scale (VAS) for general wellbeing, the Specific Activity Scale (SAS) functional questionnaire, 12‑item General Health Questionnaire (GHQ; a measure of current mental health), symptom questionnaire and Somatic Symptoms Inventory (SSI) adapted for local use from the Bradford Somatic Inventory (which assessed adequacy of daily life activities, emotional adjustment, social interactions [frequency, range and quality], work adjustment, sleep, fatigue and appetite). Schwaab et al. used 3 different self‑administered questionnaires relevant to this appraisal: the VAS for general wellbeing, VAS Karolinska questionnaire (contains 16 questions on cardiovascular symptoms relevant to pacemaker patients) and SAS functional questionnaire. The Assessment Group noted that results for both general wellbeing and functional status were similar across Lau et al. and Schwaab et al., with no statistically significant difference between dual‑ or single‑chamber pacing modes in either trial. For multi‑dimensional measures, there was no statistically significant difference between the pacing modes for tests of mental wellbeing (12‑GHQ, SSI), or for most symptoms in either Lau et al. or Schwaab et al. Schwaab et al. reported people experiencing less dizziness on single atrial pacing than dual‑chamber pacing (p<0.05); however, Lau et al. did not find a difference for the same symptom. Schwaab et al. was the only included trial that used the multi‑dimensional quality‑of‑life questionnaires (self‑perceived health status) with a section on cognitive function and reported no statistically significant difference between single atrial and dual pacing modes. The Assessment Group noted that there was a substantial amount of uncertainty around all quality‑of‑life results, because both trials were relatively small and had limited follow‑up.

Identified economic evaluations

4.23 The Assessment Group identified 12 papers of relevance in its systematic literature review, which included 2 cost–utility analyses that were relevant to the population in this appraisal: 1 NHS‑based (Castelnuovo et al. [2005], developed for TA88) and 1 non‑NHS‑based (Oddershede et al. [2014], partly based on DANPACE 2011). Castelnuovo et al. compared dual‑ with single‑chamber (atrial or ventricular) pacemakers over a 10‑year time horizon for several subpopulations with bradycardia. The Castelnuovo paper found single‑chamber atrial pacemakers to dominate (that is, were more effective and less costly than) dual‑chamber pacemakers for sick sinus syndrome. Oddershede et al. considered (from a Danish healthcare perspective) the cost‑utility of dual‑ compared with single‑chamber atrial pacemakers in people with sick sinus syndrome and no atrioventricular block. Depending on the risk stratification, adjustment, and pooling of data used, the cost effectiveness of dual‑ compared with single‑chamber atrial pacemakers (assessed by calculating net monetary benefit) ranged from £460 to £7847 when assuming a maximum acceptable incremental cost‑effectiveness ratio (ICER) of £20,000 per quality‑adjusted life year (QALY) gained, and −£1238 to £10,615 when assuming a maximum acceptable ICER of £30,000 per QALY gained.

Model overview

4.24 The Assessment Group constructed a Markov cohort model to estimate the cost effectiveness of rate‑responsive dual‑chamber pacemakers compared with rate‑responsive single‑chamber atrial pacemakers in people with symptomatic bradycardia as a result of sick sinus syndrome without atrioventricular block. The Assessment Group conducted the economic analysis from the perspective of the NHS and Personal Social Services, and the model had a cycle length of 1 month. Costs and health effects were discounted at an annual rate of 3.5%.

Model structure

4.25 The model developed by the Assessment Group contained multiple health states. Patients entered the model in the 'requiring a pacemaker' health state, and then transitioned into either the 'with dual‑chamber pacemaker' or 'with single‑chamber atrial pacemaker' health state. In each arm, patients could then experience adverse events and move on to the 'atrial fibrillation', 'stroke' or 'heart failure' health states. In the single‑chamber pacemaker arm only, patients could move on to the 'atrioventricular block' health state, and from there transition to the 'with dual‑chamber pacemaker (after re‑operation)' health state. The Assessment Group noted that all re‑operations were assumed to occur in the single‑chamber atrial arm only (because the need to change pacing mode to dual‑chamber pacing was the only reason for re‑operation with a statistically significant difference between the 2 arms [see sections 4.12 and 4.19]), and all re‑operations after the atrioventricular block health state were assumed to be an upgrade to a dual‑chamber pacemaker. Only 1 instance of re‑operation was permitted in the model. Patients who moved to the 'heart failure' and 'stroke' health states from the single‑chamber arm could have a re‑operation; however, they only incurred the costs of re‑operation, and remained in the 'heart failure' or 'stroke' health state. In the atrial fibrillation health state, modelled patients could need reprogramming of the device to act as a ventricular pacemaker in the dual‑chamber arm, or re‑operation in the single atrial arm to a single‑chamber ventricular device. However, in both arms, patients could transition from 'atrial fibrillation' to 'heart failure' or 'stroke'. Patients were at risk of death in each health state.

4.26 The Assessment Group populated the model with a cohort that had the same baseline characteristics of the DANPACE (2011) trial: baseline age was 73 years, and 65% of participants were female. However, rather than using the prevalence of comorbidities from the DANPACE trial of atrial fibrillation (44%), stroke (8%) and heart failure (12%), these were assumed to be zero for all patients on entry to the model. The Assessment Group stated that this assumption was made to simplify the model.

Modelled treatment effectiveness

4.27 The treatment effect for dual‑ compared with single‑chamber pacemakers was predominantly informed by results from the DANPACE (2011) trial only. The Assessment Group stated that it did not combine the available trial data because of heterogeneity between the trials as a result of different patient populations (for example, prior history of atrial fibrillation) and different device programming used (for example, different proportions of ventricular pacing). For the risk of heart failure and stroke for people with atrial fibrillation, the Assessment Group carried out targeted literature searches.

4.28 Re‑operation: Re‑operation was assumed to only occur in the single‑chamber arm of the model (see section 4.25). All re‑operations following the development of atrioventricular block were assumed to be an upgrade to a dual‑chamber pacemaker. For people with single‑chamber atrial devices, the need to change pacing mode was predominantly a result of the development of atrioventricular block needing upgrade to a dual‑chamber device (see sections 4.12 and 4.19). Therefore, the Assessment Group used the difference in event rates for re‑operations between the single‑ and the dual‑chamber arms in DANPACE (2011) that were because of a need for surgical change of pacing mode to estimate the risk of people in the single‑chamber atrial arm developing atrioventricular block per patient per month, and applied this as a constant risk for the lifetime of the model. Because the Kaplan–Meier plot suggested a non‑linear relationship, re‑operation as a time‑dependent parameter (rather than assuming constant risk) was used in sensitivity analyses. To derive the transition probabilities for the model, the Assessment Group converted the difference in monthly event rates into a monthly probability of atrioventricular block for those with single‑chamber pacemakers, generating a value of 0.142. People with single‑chamber pacemakers in the stroke or heart failure health states could develop atrioventricular block at the same monthly rate as those not in the heart failure or stroke health states (0.142), and have a re‑operation to receive a dual‑chamber pacemaker. However, they remained in the stroke or heart failure health states, incurring the cost of re‑operation but no change in utility. Re‑operation was also possible for some people in the single‑chamber atrial arm who developed atrial fibrillation and still needed pacing, where one‑third of modelled patients had a re‑operation to surgically change pacing mode and receive a single‑chamber ventricular pacemaker. However, these patients remained in the atrial fibrillation health state and only incurred the costs of the health state with no change in utility.The Assessment Group assumed that at 96 months' post‑implantation, based on an 8‑year battery life, all patients who had not yet experienced re‑operation or developed atrial fibrillation received a replacement of their existing pacemaker.

4.29 Atrial fibrillation, heart failure and stroke (all patients): For the dual‑chamber pacemaker arm, the Assessment Group derived event rates for atrial fibrillation, heart failure and stroke from DANPACE (2011) and Riahi et al. (2012) (summary statistics from DANPACE). For the single‑chamber arm, the Assessment Group derived event rates for atrial fibrillation, heart failure and stroke by taking the relevant hazard ratios from DANPACE and Riahi et al. and applying these to the associated event rates derived for the dual‑chamber arm. The Assessment Group then transformed the event rates into monthly probabilities. The following monthly probabilities were used in the model: paroxysmal atrial fibrillation (dual 0.4%, single 0.5%), chronic atrial fibrillation (dual and single 0.18%), heart failure (dual 0.42%; single 0.46%) and stroke (dual and single 0.08%).

4.30 Heart failure and stroke (those with atrial fibrillation): The Assessment Group carried out targeted literature searches to estimate the increased risk of heart failure or stroke for those who also have atrial fibrillation. For heart failure, the Assessment Group did not identify any studies; therefore, it assumed in the model that the risk of heart failure was the same in people with and without atrial fibrillation. For stroke, the Assessment Group identified a paper by Gallagher et al. (2014), which reported the results of a population‑based cohort study of people with atrial fibrillation. Gallagher et al. presented incidence rates of stroke, adjusted for covariates such as risk of stroke, age and smoking status for various scenarios of warfarin therapy (currently exposed, recently exposed, history of exposure or no history of exposure). The Assessment Group considered the incidence rate of stroke in people currently exposed to warfarin therapy (0.9 per 100 person years) the most suitable to inform the risk of stroke in people with atrial fibrillation, because current guidelines recommend effective anticoagulation therapy for stroke prevention in people with paroxysmal and persistent atrial fibrillation. The Assessment Group noted this was similar to the monthly rate for those without atrial fibrillation.

4.31 Mortality: The Assessment Group assumed (based on pooled estimates of all‑cause and cardiovascular mortality identified in the clinical literature review) that the risk of death was consistent across all treatment arms, but varied by age and health state. People in the 'with pacemaker' health states had the same risk of death as the age‑ and gender‑matched UK general population (because the Assessment Group found no evidence of higher mortality risks for people needing or already implanted with a pacemaker [this assumption was tested in a scenario analysis]). The Assessment Group considered 2 overarching forms of mortality in the model: case fatality and all‑cause mortality:

4.32 Adverse events: The Assessment Group did not incorporate the effect of adverse events into the model because it did not identify any statistically significant differences in adverse events in the literature other than adverse events leading to re‑operation.

4.33 Health‑related quality of life: The Assessment Group carried out a systematic review to identify health‑related quality‑of‑life evidence, identifying 6 relevant studies reporting generic, preference‑based measures of quality of life. The Assessment Group noted that all of the health‑related quality‑of‑life studies identified for inclusion reported time trade‑off utility data collected directly from patients and, of these, 5 (Fleischmann et al. 2006; Fleischmann et al. 2009; Shulka et al. 2005; Link et al. 2004 and Lamas et al. 2002) reported the results of quality‑of‑life analyses carried out with the MOST clinical trial, and 1 (Lopez‑Jimenez et al. 2002) from the PASE trial. The MOST trial was a UK‑based randomised controlled trial comparing 2010 people with sick sinus syndrome who were randomised to either dual‑ or single‑chamber ventricular pacing. PASE was a randomised controlled trial of 407 people comparing dual‑ with single‑chamber ventricular pacing for people with bradycardia, with a primary end point of health‑related quality of life. In both trials, patients received dual‑chamber pacemakers before randomisation to either dual‑ or single‑chamber ventricular mode. In addition to the identified utility studies, the Assessment Group carried out a targeted literature search for utility associated with stroke (with or without atrial fibrillation), identifying a study by Luengo‑Fernandez et al. (2013), which evaluated quality of life after transient ischaemic attack and stroke. People in the dual and single 'requiring pacemaker' health states were assumed to have a utility of 0.725 and those in the dual and single 'with pacemaker' health states had a utility of 0.825 (both utilities taken from Fleischmann et al. [2006], which considered the impact of pacemaker device and mode on quality of life). People in the 'atrial fibrillation' health state had a utility of 0.805 (Fleischmann et al. [2009], which considered the impact of atrial fibrillation on quality of life and functional status after implantation). 'Stroke' (with or without atrial fibrillation) was associated with a utility of 0.64 in month 1 and 0.70 thereafter (Luengo‑Fernandez et al. [2013],). 'Heart failure' (with or without atrial fibrillation) was associated with a utility of 0.64 (taken from Lopez‑Jimenez et al. [2002], which described the results of the PASE study). Death was assumed to be associated with a utility of 0.

Costs

4.34 The Assessment Group identified studies for UK‑specific resource use and costing studies of atrial fibrillation, heart failure and stroke from the following sources:

In addition, the resource use and costs were based on standard UK sources (NHS reference costs 2012/13, Electronic market information tool [eMit] or British national formulary [BNF]) for the unit costs applied within the Assessment Group economic model.

4.35 Device and implantation costs: The Assessment Group obtained procedure costs (including hardware costs) associated with implantation of a single‑ (£1875) or dual‑chamber (£2438) device from a weighted average of episode costs associated with relevant Healthcare Resource Group (HRG) codes (NHS reference costs 2012/13). The Assessment Group assumed that upgrade procedures cost the same as an initial implantation of a dual‑chamber device. In sensitivity analyses, the Assessment Group used spell level (rather than episode level) data for each HRG code. The Assessment Group assumed that at 96 months' post‑implantation, all patients who had not yet experienced re‑operation or developed atrial fibrillation received a replacement dual‑chamber device.

4.36 Monitoring costs: Modelled patients received follow‑up checks from a cardiologist (£86) after pacemaker implantation, for which the Assessment Group used HRG codes. Based on expert clinical opinion, the Assessment Group assumed initial follow‑up to be 1 week after implantation, and a second follow‑up at 2 months post‑implantation and subsequent annual visits. Therefore, the Assessment Group applied the cost of a follow‑up visit on entry into the dual and single 'requiring pacemaker' and 'with pacemaker' health states. The cost of a follow‑up visit was also applied annually to all patients in the dual and single 'with pacemaker' health states.

4.37 Episode costs: Modelled patients were exposed to the risk of heart failure and stroke, with or without the presence of atrial fibrillation:

The Assessment Group applied the episode cost of heart failure to people entering the 'heart failure' and 'atrial fibrillation and heart failure' health states. The Assessment Group applied the episode cost of stroke to people entering the 'stroke' health state, and the episode cost of stroke after atrial fibrillation to people entering the 'atrial fibrillation and stroke' health state. If people developed atrial fibrillation and needed their device reprogramming to ventricular mode, they accrued the following costs:

The Assessment Group applied (based on expert clinical opinion) the cost of reprogramming and device replacement to one‑third of people developing atrial fibrillation. The impact of this assumption was tested in a sensitivity analysis.

4.38 Long‑term costs: After the onset of heart failure, stroke or atrial fibrillation, the Assessment Group assumed modelled patients accrued costs over the long term, for example, medication, hospitalisation and primary care costs.

Results of the economic analyses

4.39 After consultation on the Assessment Group report, an error was noted in the calculation of total UK direct costs of cardiovascular disease, which had been underestimated in both arms of the model. This reduced the base‑case ICER, and all ICERs in scenario and sensitivity analyses. The original base‑case ICERs are described in detail in the overview and an Erratum has been produced to describe the corrections, both of which are available in the committee papers. This document contains only the corrected ICERs.

4.40 The deterministic base‑case ICER for dual‑ compared with single‑chamber atrial pacemakers was £6056 per QALY gained (incremental costs £269, incremental QALYs 0.04), and the mean probabilistic ICER across 1000 simulations was £6068 per QALY gained (incremental costs £277, incremental QALYs 0.05). The probability of dual‑chamber pacemakers being cost effective at a maximum acceptable ICER of £20,000 per QALY gained was 72.9%, and at £30,000 per QALY gained was 78.7%. The Assessment Group noted that in 66% of simulations, dual‑chamber devices had both greater costs and greater QALYs than single atrial pacemakers, and in 24.1% of simulations dual‑chamber pacemakers dominated single‑chamber atrial pacemakers (that is, produced more QALYs at a lower cost). Single‑chamber devices dominated dual‑chamber devices in 9% of cases.

4.41 The Assessment Group carried out one‑way sensitivity analyses on the following parameters by using the lower and upper limits of the 95% confidence intervals: age, efficacy values, utility values, costs, all‑cause mortality and heart failure hospitalisation. The Assessment Group presented a tornado diagram of the 10 most influential results in the deterministic sensitivity analysis, which showed the variance from the base case, and noted that many of the parameters tested had minimal impact on the deterministic cost‑effectiveness results. The Assessment Group noted that the parameters most likely to increase the deterministic ICER over £20,000 per QALY gained were:

4.42 The Assessment Group conducted structural sensitivity analyses. When using a 5‑year time horizon, the deterministic ICER was £14,261 per QALY gained, and the probabilistic ICER was £13,837 per QALY gained. When using Kaplan–Meier data as the basis for re‑operation, the ICER was £3425 per QALY gained.

4.43 The Assessment Group conducted scenario analyses on the base‑case results, including varying efficacy sources, cost estimates (using spell costs of pacemaker device and implantation [dual cost: £4142.11; single cost: £3362.18]), discount rates and differing risk of developing heart failure by age. Most of the scenarios explored had a minor impact on the resulting ICER with the exception of:

4.44 To address the uncertainty in the appraisal arising from the lack of a specific list price for pacemaker devices, the Assessment Group conducted a threshold analysis, in which the costs of dual‑chamber pacemakers were increased until the ICER reached £20,000 per QALY gained. When the price of dual‑chamber pacemakers was increased by £495 to £2933 (and the price of single‑chamber atrial pacemakers remained at £1875), the ICER was £19,992 per QALY gained (incremental costs £888.28, incremental QALYs 0.04).