No conflict of interest is declared. J. Bryant developed the research protocol, assisted in development of the search strategy, assessed studies for inclusion in systematic review of clinical and cost-effectiveness, extracted data from and quality assessed the included studies, synthesized evidence, assisted in development of economic model, drafted and edited final report, and project managed the study. H. Brodin developed research protocol, assessed studies for inclusion in systematic review of cost-effectiveness, extracted data from and quality assessed the included studies, synthesized evidence, developed the economic model and collected data for the economic model, and drafted the report. E. Loveman developed the research protocol, assessed studies for inclusion in systematic review of clinical and cost-effectiveness, extracted data from and quality assessed the included studies, synthesized evidence, assisted in development of the economic model, and drafted the report. A. Clegg developed the research protocol, developed the search strategy, assessed studies for inclusion in systematic review of clinical and cost-effectiveness, extracted data from and quality assessed the included studies, synthesized evidence, assisted in development of the economic model, and drafted report.
Sudden cardiac death, usually due to ventricular arrhythmia, is a significant public health issue, occurring in approximately 75,000 to 100,000 people annually in the United Kingdom (41;33), 400,000 people in Europe (12), and up to 460,000 people in the United States (2). Most cases of sudden cardiac death occur as a consequence of a first episode of ventricular tachyarrhythmia, although increasing numbers of people are surviving a first arrhythmic event and are at high risk of further episodes. Prevention of sustained ventricular arrhythmias is either primary or secondary (preventing them from happening or recurring, respectively), and most patients are treated with antiarrhythmic drug therapy, which aims to suppress the development of arrhythmias. Some patients are treated with implantable cardioverter defibrillators (ICDs) which can actively sense and terminate life-threatening arrhythmias. Uncertainty exists about the clinical and cost-effectiveness of ICDs compared with conventional antiarrhythmic drug therapy.
Implantation rates of ICDs vary greatly between countries, with the total rate of implantation being approximately 35 per million per year (2002/2003) in England and Wales (19), 45 per million in Western Europe (43), and significantly more in the United States at approximately 184 per million per year (2000) (9). The use of ICDs in the management of arrhythmias is guided in the United Kingdom by recommendations from the National Institute for Health and Clinical Excellence (NICE) (32). We were commissioned by The National Health Service (NHS) Health Technology Assessment Program to undertake a systematic review and economic evaluation of ICDs in patients at risk of sudden cardiac death from arrhythmias to inform the process for NICE to update its guidance in 2004 in the light of new evidence and in particular to incorporate relevant newly emerging UK data. This study summarizes and updates the findings of the systematic review (7) and discusses its policy implications.
METHODS
Data Sources
Eleven electronic databases (including Medline, PubMed, Cochrane Database of Systematic Reviews, Cochrane Controlled Trials register, Embase) were searched from inception for periods up to January 2005 (search strategies available on request). Additional studies were identified through searching bibliographies of related publications and through contact with experts and manufacturers. Further details of the research methodology are available elsewhere (7).
Study Selection
Randomized controlled trials published in the English language were sought that compared ICD treatment with antiarrhythmic drug therapy (AAD); included people at risk of sudden cardiac death from arrhythmia in secondary or primary prevention categories (primary prevention including patients with a history of myocardial infarction, evidence of ventricular tachycardia, and left ventricular ejection fraction less than 35 percent; or patients with left ventricular ejection fraction less than equal to 30 percent and myocardial infarction or dilated cardiomyopathy); and used the primary patient-based outcome of mortality and the secondary outcome of quality of life. Systematic reviews of individual randomized trials were included. Published economic evaluations were also sought.
Quality Assessment and Data Extraction
The quality of systematic reviews and randomized controlled trials was assessed using criteria developed by the Center for Reviews and Dissemination (34) and by Jadad and colleagues (18), respectively. Published economic evaluations were assessed for internal (13) and external (37) validity. Inclusion criteria were applied independently by two reviewers. Decisions about quality criteria and data extraction were made by one reviewer and checked by a second, with any differences in opinion resolved through discussion.
Data Synthesis
Studies of clinical effectiveness were combined through narrative synthesis with full tabulation of included studies. Meta-analysis was not appropriate due to heterogeneity in patient characteristics, comparative interventions, and duration of trials.
Economic Evaluation
A decision analytic model with a 5-year time horizon was developed to examine the benefits and costs of ICDs compared with AAD. The model was constructed to show the cost-effectiveness of the two optional treatment arms (ICD or AAD therapy), based on probabilities of patients experiencing particular health states or treatment options during the period of the evaluation, and the consequences in terms of benefits to patients (survival and quality of life) and costs incurred. The economic evaluation adopted a UK NHS perspective for costs and benefits. It focused on hospital-based costs that were thought to represent the most significant component in the provision of the service. One generic base case model was specified for both secondary and primary prevention, as there are limited differences in treatment pathways.
The model was populated with data from a UK study (9) commissioned by the NHS HTA Programme to inform the ICD appraisal process supplemented with some trial data. Survival data, resource use, and costs for the ICD group were obtained from a representative sample of patients attending two UK NHS hospital trusts. Data for the AAD group were based on comparable Canadian patient-based data (10). The base case relative risk strategy for the model was derived from a synthesis of data from the two compatible sources (9;10) to give relative risk of death for UK patients. Quality of life utilities were obtained from a survey of the UK ICD patients using a generic health related quality of life measure, the EuroQol EQ5D, and used for both groups in the absence of conclusive evidence from the published literature to suggest any difference in quality of life between ICD and AAD patients. Costs were for 2001/2002, with costs and savings discounted at 6 percent and quality-adjusted life years (QALYs) at 1.5 percent.
A series of one-way sensitivity analyses were undertaken to determine the robustness of the base case model to several different scenarios, including the following: differences in survival by considering alternative relative risks of mortality for subgroups of patients, derived from clinical effectiveness data in the literature; variations in the assumptions about quality of life; and the impact of reduced device costs.
RESULTS
Quantity and Quality of Research
Clinical Effectiveness
Two systematic reviews (14;26), one meta-analysis (11), and eight randomized controlled trials (RCTs) met the inclusion criteria (Tables 1 and 2) (1;3;5;8;10;21;28;29). Three RCTs were secondary prevention trials (1;10;21), and five RCTs were primary prevention trials (3;5;8;28;29).
When judged using standard criteria for assessing methodological quality, the systematic reviews were found to be of good quality, clearly stating the research question, the search strategy, the inclusion/exclusion criteria, details of included studies, and appropriately summarizing the data, but neither provide details of the assessment of the validity of the included studies. The meta-analysis (11) pooled raw data from the same three secondary prevention trials (1;10;21) as did one of the systematic reviews (14).
All the RCTs were described as randomized, but none were double blind due to the nature of the intervention. Participants of the three secondary prevention trials were survivors of cardiac arrest, although details differed slightly between the studies, for example, in the degree of left ventricular dysfunction. Sample sizes were 1,016 (1) patients, 191 patients (21) and 659 patients (10), and length of follow-up ranged from 18 months (1) to 3 years (10). Participants in the primary prevention trials differed substantially from one trial to another. Two of the primary prevention trials included patients who had had a myocardial infarction with left ventricular ejection fraction equal to or less than either 35 percent (28) or 40 percent (2). Patients undergoing coronary artery bypass graft (CABG) surgery with left ventricular ejection fraction less than 35 percent were enrolled in one trial (5), whereas people with a history of previous myocardial infarction and left ventricular ejection fraction equal to or less than 30 percent were enrolled in another (29). Patients with dilated cardiomyopathy and left ventricular ejection fraction equal to or less than 30 percent were included in one (3). Follow-up ranged from 20 months (29) to 39 months (28).
Cost-Effectiveness
Eleven economic evaluations, published in twelve papers, were identified, none of which had both high internal and external validity (22–25;30;35;36;38–40;44;45). Few of the studies used patient data, and in most studies, numerous assumptions had to be made about outcomes and costs.
Assessment of Clinical Effectiveness
Mortality in Secondary Prevention
The evidence suggests that ICDs reduce mortality in patients with previous ventricular arrest or symptomatic sustained ventricular arrhythmias. Results suggest a 25 percent to 28 percent (11;14;26) reduction in relative risk of death with ICD that is due to a 50 percent reduction in arrhythmic death (Table 2). A significant reduction in total mortality with ICDs is reported as 39 percent at 1 year, 27 percent at 2 years, and 31 percent at 3 years, with an average additional life associated with ICDs of 2.7 months at 3 years (1). ICD is associated with a 23 percent nonsignificant risk reduction in all-cause mortality compared with AAD therapy in one trial (21) and a 19 percent in another (10).
Mortality in Primary Prevention
The evidence suggests that ICDs reduce mortality in patients who have not had a previous sudden cardiac death episode or previous ventricular arrhythmia but who have reduced left ventricular function due to coronary artery disease with asymptomatic nonsustained ventricular arrhythmia and sustained tachycardia that could be induced electrophysiologically; and in some patients with severe left ventricular dysfunction (ejection fraction ≤30 percent) after myocardial infarction (Table 2). Two systematic reviews report results favoring ICDs, with relative risks of .72 (95 percent confidence interval [CI], .63–.84) (14) and .66 (95 percent CI, .46–.96) (26) for all-cause mortality and .37 (95 percent CI, .27–.50) and .34 (95 percent CI, .23–.50), respectively, for arrhythmic death (Table 2). Three trials report survival benefits due to ICD use in cardiac arrest patients, with a reported 54 percent reduction in total mortality (28) and overall 5-year mortality rates of 24 percent in patients who received a defibrillator compared with 55 percent among those who did not (8). A 31 percent reduction in risk of death from any cause was reported in patients with a prior myocardial infarction and advanced left ventricular dysfunction who had prophylactic implantation of an ICD, with mortality rates of 14.2 percent in the ICD group and 19.8 percent in the AAD group (29). Mortality rates by 4 years of follow-up were 27 percent in people receiving an ICD and 24 percent in those who did not, in the trial in which patients received ICD at the time of CABG (5). The hazard ratio for death from any cause was 1.07 (95 percent CI, .81–1.42) (5). All cause mortality rates were not different between the ICD group or the usual care group in the trial considering the prophylactic use of ICDs in patients with recent onset dilated cardiomyopathy (3).
One primary prevention study (28) reported subgroup analyses which showed that the magnitude of survival benefit from ICD was directly related to severity of cardiac dysfunction, where defibrillator use was associated with a significant reduction in mortality rate in subsets of patients with three risk factors (ejection fraction less than 26 percent, QRS ≥.12 second, and a history of heart failure requiring treatment) (44).
Quality of Life
The two secondary prevention trials (1;10) reporting quality of life have inconsistent findings using a range of measures. One reports that ICD and AAD therapy are associated with similar self-perceived quality of life (27). The development of adverse symptoms is associated with significant impairment in quality of life in both groups, and the occurrence of sporadic shocks in ICD recipients (27). The other reports better quality of life with ICD therapy than AAD therapy, although this finding is not evident in patients who receive numerous shocks from their device (42). The one primary prevention trial that reported quality of life found that ICD patients had significantly lower levels of psychological well-being, perceptions of health and emotional role functioning compared with patients receiving conventional medical therapy (17;31).
Adverse Effects
Generally serious adverse events due to ICDs are infrequent. Complications that are reported include infection, hematomas and bleeding, lead dislodgement and migration, cardiac perforation, pleural effusion and pneumothorax, and device dysfunction/malfunction of generator. Adverse events associated with AAD therapy include pulmonary toxicity, thyroid dysfunction, insomnia, tremor, ataxia, and visual symptoms, and have lead to withdrawal from treatment for up to 25 percent of patients.
Assessment of Cost-Effectiveness
Cost-Effectiveness from the Literature
The published economic evaluations report a variation in the cost per life year gained (LYG) and cost per QALY associated with ICD use. In the nine studies of secondary prevention, the cost per LYG ranged from $17,100 (23) to C$213,500 (36) and the cost per QALY from $37,000 to $76,800 (38). Three studies assessing primary prevention report incremental cost-effectiveness ratios. Incremental cost per LYG ranged from $27,000 (30) to $36,700 (44) and incremental cost per QALY from $50,900 (44) to $558,000 (40). Although there is variation in the cost per QALY or LYG, generally costs appear to have increased between 1990 and 2004.
Cost-Effectiveness from the Economic Evaluation
As searches revealed that the cost-effectiveness evidence was lacking in internal validity and of limited relevance to the United Kingdom, a new economic evaluation was conducted. The discounted results of this current economic model, using the base case relative risk strategy, are shown in Table 3. Results of sensitivity analyses using different assumptions are shown in Table 4. Results show that an improved survival can be achieved with ICDs compared with drug therapy, but with high additional cost. The incremental cost-effectiveness is in several cases higher than £100,000 ($189,000) per QALY if the quality of life is assumed to be the same in both treatment groups, and more than £200,000 ($379,000) per QALY under the main base case, which assumes a relative mortality risk of .85. When assessing secondary prevention, the incremental cost per QALY ranged from £88,000 ($167,000) to £165,000 ($312,000), depending on mortality risk. For primary prevention, the incremental cost per QALY was £139,000 ($263,000), and for high risk patients with relative mortality risk of .46, the incremental cost per QALY was £52,000 ($98,000). The incremental cost per QALY for patients with a previous myocardial infarction and depressed left ventricular function was estimated at £123,900 ($233,000).
DISCUSSION
This review, guided by an expert advisory panel, considered systematically the evidence on the clinical and cost-effectiveness of ICDs for arrhythmias, and developed an economic evaluation. Eight RCTs, two systematic reviews, and a meta-analysis, plus twelve economic evaluations, were found that met the inclusion criteria of the review.
The RCTs are of variable quality when judged using Jadad quality score, and there are some methodological concerns that may impact on results. These concerns include factors such as the impossibility of double-blinding, due to the comparison of a device with drug therapy; randomization of patients being at the discretion of the surgeon at the time of surgery; crossovers of patients from one treatment group to the other, which may have compromised the intention to treat analysis; the fact that many ICD patients receive antiarrhythmic drug therapy to suppress tachyarrhythmias and treat underlying heart disease; patients serving as controls may not have received maximal medical therapy; and generalizability of trials to other settings. The systematic reviews and meta-analysis, although of good quality when judged using standard criteria, showed significant heterogeneity between studies, suggesting that results obtained by pooling data should be treated with caution.
Despite the limitations of the trials, the results of the studies meeting this review's criteria suggest that ICDs offer survival benefit for patients in both secondary and primary prevention categories. Prophylactic use of ICD is of benefit in some patients with severe left ventricular dysfunction after myocardial infarction, but does not appear to benefit patients with recent onset dilated cardiomyopathy and impaired left ventricular function.
Several other recent studies were identified that do not meet the inclusion criteria for the review. These report no statistically significant benefit from ICD (20) in the populations considered and unanticipated subgroup results (4). Another study suggests a benefit resulting in the use of cardiac resynchronization therapy combined with ICDs in some people with chronic heart failure (6). Such studies highlight the need to stratify patients and identify subgroups of patients who may derive benefit from prophylactic ICD.
The identified economic literature uses a range of methodologies, with different strengths and weaknesses, and reports a range of results. Despite the differences in the assumptions used in the different economic models, the results generated by the model in this review are in broad agreement with recent studies in that there are modest benefits due to ICDs at high cost. In general, results from the literature and the economic evaluation developed in this review show increasing incremental cost-effectiveness ratios over time as more data become available. The costs associated with ICDs are greater than previously thought and the benefits in terms of LYG are probably lower than previously estimated.
Despite previous research recommendations (32) and a UK NHS funded project (9) to collect data for an economic evaluation, data on quality of life and costs associated with ICDs and antiarrhythmic drug therapy remains limited. The economic model developed in this review therefore incorporated UK data for ICD patients and comparable data from a Canadian study for drug therapy patients and took a 5-year perspective. Although there are limitations with such an approach, it can be justified in that it uses the most up to date and relevant information available and takes a realistic time frame. Possible uncertainties have been explored through sensitivity analyses, which have focused on the key variables thought to be influential, namely quality of life utilities, the cost of the ICD, and the relative risk of mortality. These have shown that only when quality of life in the drug therapy group drops to .50 will the incremental cost-effectiveness be at a level at which technologies are usually regarded as being cost-effective. If device cost is reduced by 50 percent, the incremental cost-effectiveness is still in excess of £100,000 ($189,000). The results suggest that, for certain subgroups of patients, ICDs may be cost-effective. Only patients with any two or more risk factors including left ventricular function less than .26, QRS width ${\ge}$.12 seconds, and a history of heart failure requiring treatment, are likely to have a relative risk of mortality that will result in favorable cost-effectiveness.
CONCLUSIONS
The findings of this review suggest that ICDs offer some benefit to people at risk of arrhythmias and that the degree of survival benefit is related to the severity of cardiac dysfunction. Results of economic modeling suggest that the ICDs are unlikely to be cost-effective except for certain subgroups of patients at high risk of arrhythmias.
Such results pose a dilemma for policy makers who have to reconcile the needs of the individual patient with those of society when providing guidance on the use of ICDs, and highlight the importance of defining in which patient groups there may be survival benefit from ICD therapy. Information gaps relating to quality of life and cost-effectiveness associated with ICD use may cause uncertainty for those developing guidance and lead to delays in the decision-making process. After much deliberation, NICE issued its final appraisal determination in 2005 additionally recommending ICDs for patients in the primary prevention category who have a history of myocardial infarction and left ventricular ejection fraction of less than 30 percent and QRS duration ≥.12 seconds (16).
It is essential that further research is conducted to identify those patients for whom ICDs are most likely to be clinically and cost-effective. This must take into account developments in ICD and other cardiac technology such as resynchronization therapy, to inform future evidence-based policy and to ensure the appropriate use of ICDs and other devices and procedures (15).
CONTACT INFORMATION
Jackie Bryant, MSc (jsb1@soton.ac.uk), Senior Research Fellow, Hakan Brodin, PhD (hakan.brodin@fhi.se), Senior Research Fellow, Emma Loveman, PhD (Emma.Loveman@soton.ac.uk), Senior Research Fellow, Andrew Clegg, PhD (andy1@soton.ac.uk), Professor and Director of SHTAC, Southampton Health Technology Assessments Centre (SHTAC), Wessex Institute for Health Research and Development, University of Southampton, Mailpoint 728, Boldrewood, Bassett Crescent East, Southampton SO16 7PX, UK
This project was funded by the NHS R&D Health Technology Assessment Programme (project no. 03/32/01). The views and opinions expressed herein are those of the authors and do not necessarily reflect those of the Department of Health.
We are very grateful to the advisory panel, which provided expert advice and comments on the protocol and/or draft of the systematic review and economic evaluation. We also thank the staff at the Wessex Institute for Health Research and Development.
