METHODS

Electronic databases (Medline, PubMed, Embase, Cochrane systematic reviews database, Cochrane controlled trials register, Database of abstracts of reviews of effectiveness, and the National Research Register) were searched from January 1980 to December 1999 (search strategies are available on request). Additional studies were identified through searching bibliographies of related publications and through contact with experts.

We sought English language randomised controlled trials that compared ICD treatment with conventional treatment; included people at risk of sudden cardiac death from arrhythmia; and used patient based outcomes such as mortality, cardiac arrest/ventricular tachycardia, and quality of life measures.

We assessed the quality of systematic reviews and randomised controlled trials using criteria developed by the Centre for Reviews and Dissemination⁵ and Jadad and colleagues,⁶ respectively. Decisions about inclusion criteria, quality criteria, and data extraction were made by one reviewer and checked by a second, with disagreements resolved through discussion.

Studies were combined through narrative synthesis with full tabulation of included studies. Meta-analysis was not appropriate because of pronounced heterogeneity in patient characteristics and comparative interventions.

RESULTS

Seven randomised controlled trials, published between 1993 and 2000, met the inclusion criteria (table 1).^7–14 There were three primary prevention and four secondary prevention trials.

Most participants were men, with ages ranging from 57 to 67 years. Earlier trials used predominantly transthoracic devices whereas later trials used transvenous devices. The patients studied varied but in five trials they had a left ventricular ejection fraction of less than 40%. All the secondary prevention trials included cardiac arrest survivors and this was the sole entry criterion in two trials. The trials were large, ranging from 60 to over 1000 participants, and average follow up was 32 months. The trials were generally well conducted, although insertion of an ICD is practically impossible to double blind.

Two of the three primary prevention and all four secondary prevention studies found a survival advantage for patients treated with ICD (table 2). The secondary prevention trials showed absolute risk reductions (ARR) for ICD treatment ranging from 3.7% to 21.0%, and relative risk reductions (RRR) of 19.7% to 37.0%. In two of the primary prevention trials, the ARR with ICDs ranged from 22.8–24% (non-random evidence) and RRR from 54–56%. The reduction in total mortality was mainly because of fewer arrhythmic deaths.

There are some specific points to note in relation to the results in table 2.

• The ICD effectiveness results from MUSTT (randomised multicenter unsustained tachycardia trial) are based on non-randomised comparisons and so may be open to confounding.

• The CABG patch (coronary artery bypass graft patch) trial showed a non-significant increase (1.7%) in the risk of death in the ICD group. The risk of sudden cardiac death in this study was low and surgery in the control group may also have reduced the risk of death.

• Some of the ICDs used in the MADIT (multicenter automatic defibrillator implantation trial) and CASH (cardiac arrest study Hamburg) trials and in the trial by Wever and colleagues ¹³ were older transthoracic devices, which are associated with a greater morbidity and mortality. These earlier trials are also the smallest, the Wever trial having 6% of the study population included in the AVID (antiarrhythmics versus implantable defribrillators) trial.

• In CASH, the benefits of ICDs were more evident in the first five years after the index event and gradually declined, reaching an ARR of 10.6% at year 8.

• From non-randomised AVID evidence, there appears to be no advantage of one make of ICD over another.

Two other outcomes of ICD use need considering. Firstly, they can have unwanted effects, especially peri-implantation (for example, infection, bleeding, pneumothorax) (table 3). Secondly, patients' quality of life can be impaired as well as improved by ICDs. Three randomised controlled trials reported changes in quality of life. The CABG patch trial showed that patients with ICDs had lower levels of psychological wellbeing and reduced physical and emotional role functioning than controls at six months.¹⁵ Unpublished AVID data show that sporadic defibrillator shocks are associated with a significant reduction in mental wellbeing and an increase in patient concerns.¹⁶ Unpublished data from MADIT showed no difference in quality of life between ICD and controls, and found quality of life scores inversely correlating with number of shocks received. ¹⁷

DISCUSSION

In this review, guided by an advisory panel of experts, we have considered systematically the best evidence on the effectiveness of ICD treatment. Although we searched hard for all randomised controlled trials we may have missed some, leaving the review open to possible publication bias. Given this caveat, we found seven randomised controlled trials. Six of these showed that ICDs reduced total mortality in patients at high risk of sudden cardiac death from ventricular arrhythmias not occurring in association with reversible pathology.

The randomised controlled trials were generally well conducted but some methodological issues should be highlighted. Randomised controlled trials of ICDs pose special problems: comparing drugs and devices raises issues of blinding and compliance; the differential use of β blockers in ICD groups seen in two of the three trials for which data are available may have contributed to the apparent effectiveness of ICD (although there is evidence that they did not convey a survival advantage^18,19); and the evolution of devices over time makes the applicability of results from trials of older, transthoracic devices (which carry greater risks than transvenous devices) problematic. Two of the trials (CIDS and CASH) were underpowered to detect significant differences in outcomes, though this was partly addressed by the meta-analysis of three secondary prevention trials.¹⁸ Finally, the clinical characteristics of the patients included need to be considered carefully. For example, patients in CASH had a greater left ventricular ejection fraction and were relatively more healthy, so they would be expected to derive less benefit from ICDs than those in the AVID trial.²⁰

The patient groups that benefited in the trials included are listed in table 4. In addition there is widespread clinical consensus that patients with certain rare conditions also benefit from ICD treatment.^10,21–26 However, the optimal strategy for identifying those patients who could most benefit from ICDs is not clearly established. Techniques such as electrophysiological study, signal averaged ECGs, and heart rate variability have been used, although the evidence base for these is often weak.^27–29 Ongoing trials including those into treatment of cardiac failure with ICD, and elaboration of quality of life outcomes in patients with ICDs, will provide evidence that may have implications for those subgroups of patients in whom ICD are maximally effective.

The policy implications for ICD treatment are huge, with demand rising in most European countries. Recent NHS guidance, if implemented, will lead to an estimated increase in ICDs from 17 per million to 50 per million.³⁰ This is likely to be costly and to present policy makers with challenging decisions about value for money. For instance, speculative estimates of cost utility from a recent report⁴ show a cost per QALY (quality adjusted life years) of £21 300 to £108 800. But there is a tension between the utilitarian approach and the right to rescue for the individual. Eligible patients and their families may expect this treatment to be offered, perceiving it as a life saving benefit, but cost effectiveness remains a barrier.

Future research could help to inform evidence based decisions about the use of ICDs. In the first place, what is needed is information about the benefits and costs of ICDs over the longer term. As most costs occur early in treatment, cost effectiveness may become more favourable as patients survive longer, battery life of ICD extends beyond six to seven years, patient acceptability increases, cost of device is reduced, and improvements to efficacy occur. Secondly, we need to know more about current patterns of service use, equity of provision between different social groups, and the diffusion and effectiveness of different devices. Finally, we need to know more about the changes in patients' quality of life that ICDs bring.