Implantable cardioverter defibrillators for the treatment of arrhythmias and cardiac resynchronisation therapy for the treatment of heart failure: systematic review and economic evaluation

Diagnosis of sudden cardiac death

As SCD can happen without warning, it is important for general practitioners and secondary care providers to be aware of risk factors so that patients at high risk of SCD can be identified and referred for cardiac evaluation. A range of diagnostic tests may be used to identify risk of SCD. An electrocardiogram (ECG) can detect abnormalities in the heart’s electrical activity and may reveal evidence of heart damage from CHD, or signs of a previous or current heart attack. Electrophysiological testing is sometimes used to identify the origins of an arrhythmia and programmed electrical stimulation (PES) of the heart may be used to stimulate the heart to induce the arrhythmia. An electrophysiological or PES study may be used before implantation of an ICD to confirm the need for an ICD or for diagnostic work-up. Other tests that may be used to identify SCD risk include ultrasound echocardiography and cardiac magnetic resonance imaging (to image or film different parts or the whole of the heart), blood tests (to check concentrations of chemicals involved in heart function, e.g. potassium and magnesium) and cardiac catheterisation (e.g. if blood samples from within the heart are required, or to inject dye for angiographic studies).

Implantable devices for sudden cardiac death

Ventricular arrhythmias, particularly sustained VT and VF, are life-threatening events. For patients who meet specified treatment criteria, the NICE guidance issued in 2006 [technology appraisal (TA)9542] recommends that ICD (or CRT-D) therapy is recommended for primary prevention (prevention of a first life-threatening arrhythmic event) and secondary prevention (prevention of an additional life-threatening event in survivors of sudden cardiac events or patients with recurrent unstable rhythms) of SCD. Patients with sustained ventricular arrhythmias associated with haemodynamic compromise in the presence of LVSD should be considered for ICD therapy after reversible factors are addressed. Patients with LVSD and who have recently had a MI or patients who have a cardiac condition that is associated with a high risk of sudden death should also be considered for ICD therapy in addition to OPT. OPT (as described below) is used as an adjunct or provided for those patients for whom an ICD would not be appropriate (e.g. those with a severely limited prognosis).

Specific recommendations of the NICE guidance42 (which does not cover non-ischaemic dilated cardiomyopathy) are that ICDs may be used as primary prevention for patients who have a history of previous (≤ 4 weeks) MI and either left ventricular dysfunction with a left ventricular ejection fraction (LVEF) < 35% (no worse than NYHA class III) and non-sustained VT on Holter (24-hour ECG) monitoring and inducible VT on electrophysiological testing or left ventricular dysfunction with a LVEF of < 30% (no worse than NYHA class III) and a QRS duration of ≥ 120 milliseconds; or who have a familial cardiac condition with a high risk of sudden death, including long QT syndrome, hypertrophic cardiomyopathy, Brugada syndrome or arrhythmogenic right ventricular dysplasia, or have undergone surgical repair of congenital heart disease.

Implantable cardioverter defibrillators as secondary prevention for arrhythmias are recommended for individuals who present, in the absence of a treatable cause, with one of the following: survived a cardiac arrest due to either VT or VF; spontaneous sustained VT causing syncope or significant haemodynamic compromise; sustained VT without syncope or cardiac arrest and who have an associated reduction in ejection fraction (LVEF < 35%) (no worse than NYHA class III). 42

Optimal pharmacological therapy for sudden cardiac death

Chronic prophylactic antiarrhythmic drug (AAD) therapy is aimed at suppressing the development of arrhythmias in patients at high risk of SCD. The class III drugs such as amiodarone are used for specific indications. These drugs may enhance the maintenance of sinus rhythm but cannot terminate an arrhythmia once it is initiated. A meta-analysis based on 8522 patients from 15 trials found that amiodarone reduced the risk of SCD by 29% and cardiovascular death (CVD) by 18% in patients at risk of SCD. 50 However, amiodarone therapy was neutral with respect to all-cause mortality and was associated with a high discontinuation rate and significant end-organ adverse reactions including hepatic, pulmonary and thyroid toxicity, with a two- and fivefold increased risk of pulmonary and thyroid toxicity respectively50 Other drugs that may be included in the OPT of SCD are angiotensin-converting enzyme (ACE) inhibitors (recommended for all patients with LVSD to improve ventricular geometry and function), aldosterone receptor antagonists (for people resistant to other drug therapy) and beta-blockers (to reverse ventricular remodelling) among others. 51

Diagnosis of heart failure

The NICE clinical guideline CG108, Chronic Heart Failure: Management of Chronic Heart Failure in Adults in Primary and Secondary Care,44 provides a diagnostic pathway for HF, the key elements of which are shown in Figure 2. Serum natriuretic peptides (SNPs; protein substances secreted by the wall of the heart when it is stretched or under increased pressure) should be measured in people with suspected HF without MI, although the guideline cautions that levels of SNPs can be reduced by certain conditions (e.g. obesity) or treatments (e.g. diuretics, ACE inhibitors, beta-blockers). Conversely, other conditions [e.g. left ventricular hypertropy, renal dysfunction, chronic obstructive pulmonary disease (COPD)] can cause high levels of SNPs. Therefore, an ECG and other tests (e.g. chest radiography, blood tests, urinalysis, spirometry) may be required to evaluate other possible diagnoses. Transthoracic Doppler two-dimensional echocardiography is used to assess the function (systolic and diastolic) of the left ventricle, to detect intracardiac shunts and to exclude important valve disease. If a poor image is obtained, other imaging methods (e.g. radionuclide angiography, cardiac magnetic resonance imaging or transoesophageal Doppler two-dimensional echocardiography) can be considered.

FIGURE 2.

Key elements in the NICE HF guideline diagnostic pathway. 52 BNP, B-type natriuretic peptide; NTproBNP, N-terminal pro-B-type natriuretic peptide.

Management of heart failure

A patient presenting with the typical signs and symptoms of HF should receive specialist assessment including echocardiography. 44 If HF is diagnosed the goals of treatment are to reduce mortality and improve the health outcome of the patient. In clinical practice, pharmacological agents are routinely used as the first-line therapy in managing HF44 (details of OPT for HF are given in Optimal pharmacological therapy for heart failure).

In addition to drug therapy, according to the NICE clinical guideline,44 individuals should be encouraged to participate in exercise-based cardiac rehabilitation (including a psychological and educational component), to give up smoking if applicable or be referred to a smoking cessation service, and to abstain from alcohol consumption if they have alcohol-related HF. Similarly, the European Society of Cardiology recommends that individuals with HF should be enrolled in a multidisciplinary care management programme. 53

Implantable devices for heart failure

As the severity of HF symptoms increases, a patient’s symptoms may no longer be controlled by OPT or lifestyle changes. There are multiple syndromes associated with HF that could predispose patients to the need for further intervention. In patients with HF, the existence of a modifiable risk factor such as arrhythmias may constitute a rationale for the use of multiple interventions. The NICE pathway for chronic HF52 indicates that, when symptoms are not controlled by OPT, treatment with CRT-P or CRT-D can be considered for patients meeting specific criteria.

Current NICE guidance issued in 2007 (TA12043) recommends CRT-P as a treatment option for individuals with HF who fulfil all of the following criteria: are currently experiencing or have recently experienced NYHA class III–IV symptoms; are in sinus rhythm – either with a QRS duration of ≥ 150 milliseconds estimated by standard ECG or with a QRS duration of 120–149 milliseconds estimated by ECG and mechanical dyssynchrony that is confirmed by echocardiography; have a LVEF of ≤ 35%; are receiving OPT. CRT-D may be considered for individuals who fulfil the criteria for implantation of a CRT-P device and who also separately fulfil the criteria for the use of an ICD device (see Implantable devices for sudden cardiac death).

Comments received from a clinical expert indicate that CRT is increasingly being considered for people without symptoms with the aim of improving prognosis by modifying the natural history of HF. Another interventional procedure that may be considered for patients with severe refractory symptoms is cardiac transplant. For those awaiting a donor heart, short-term circulatory support with a left ventricular assist device may be indicated. 54

Optimal pharmacological therapy for heart failure

Optimal medical drug therapy for HF can include ACE inhibitors, diuretics (for the relief of congestive symptoms and fluid retention), beta-blockers, aldosterone antagonists, digoxin (if symptoms continue despite the use of ACE inhibitors), amiodarone, anticoagulants (to reduce the risk of stroke), aspirin (to reduce the risk of vascular events), statins (to reduce the risk of MI and stroke), inotropic agents (to stimulate the heart muscle) and calcium channel blockers (for comorbid hypertension and angina).

The NICE 2010 clinical guideline44 suggests that medical drug therapy for HF has two aims – first, to improve morbidity (by reducing symptoms, improving exercise tolerance, reducing hospital admissions and improving QoL) and, second, to improve prognosis (by reducing all-cause mortality or HF-related mortality). According to the guideline, first-line treatment should include both ACE inhibitors and beta-blockers licensed for HF for all individuals with HF due to LVSD.

If an individual remains symptomatic despite optimal therapy with an ACE inhibitor and a beta-blocker, second-line treatment recommendations are to add one of the following: an aldosterone antagonist licensed for HF [especially if the patient has moderate to severe HF (NYHA class III–IV) or has had an MI within the past month] or an angiotensin II receptor antagonist (also known as an angiotensin receptor blocker or ARB) licensed for HF [especially if the patient has mild to moderate HF (NYHA class II–III)] or hydralazine in combination with nitrate [especially if the patient is of African or Caribbean origin and has moderate to severe HF (NYHA class III–IV)]. 44

Pharmacological recommendations for all types of HF include diuretics, calcium channel blockers, amiodarone, anticoagulants, aspirin and inotropic agents (such as dobutamine, milrinone or enoximone). ACE inhibitor therapy should not be initiated in individuals with a clinical suspicion of haemodynamically significant valve disease. 44

Characteristics of the included studies

Study characteristics are summarised in Tables 8–10 and participant characteristics are summarised in Tables 11–13. Additional details can be found in Appendix 7.

Risk of bias

The risk of bias in the included trials is summarised in Table 16 and further details for each trial can be found in the data extraction tables in Appendix 7. All 13 trials were unclear on risk of bias associated with randomisation. In fact eight trials did not report details of either randomisation or allocation concealment; therefore, the risk of selection bias (differences between known and unknown baseline characteristics of the groups) is unclear. Five trials (CIDS,84 MADIT I,99 IRIS,97 DINAMIT,95 CABG Patch75) did not report the randomisation method, although sufficient details were reported to establish that the allocation sequence was adequately concealed and they were judged to have a low risk of selection bias.

It was not possible to blind participants and personnel (health-care providers) in these trials as one group received surgery. This could bias the results because of differences in behaviours across intervention groups or differences in the care provided, such as administration of co-interventions. The trials were therefore judged to have a high risk of performance bias. Cause of death was determined or reviewed by a committee blinded to treatment group in the AVID,71 DEFINITE,90 DINAMIT, 95 AMIOVIRT,69 IRIS97 and SCD-HeFT105 trials. Outcome assessors were not blinded in the other trials but mortality was judged unlikely to be influenced by lack of blinding and so the trials were considered to have a low risk of detection bias for this outcome. Unblinded trials reporting QoL69,71,75,84,90,101,105 were judged to have a high risk of detection bias for this outcome.

Risk of attrition bias (differences between groups in withdrawals from the study) was low in seven of the trials69,81,90,95,97,99,101 and unclear in three trials. 82,84,89 In the AVID,71 CABG Patch75 and SCD-HeFT105 trials, risk of attrition bias was judged to be low for mortality but high or unclear for QoL outcomes.

Risk of selective reporting bias (differences between reported and unreported findings) was considered to be low in six studies. 69,71,81,89,99,105 Five studies listed outcomes in a protocol or methods section that were not then reported. 82,84,90,95,97 Risk of selective reporting bias was unclear in two studies (MADIT II,101 CABG Patch75).

Risk of other sources of bias was judged to be high in DINAMIT95 as block randomisation in an unblinded trial can lead to prediction of allocation. The authors of the CASH study81 note that centres were reluctant to enrol patients for potential ICD therapy in the early phase of the study and to deny ICD therapy in the late phase of the study. The effect of this is unclear. Seven of the trials were stopped early;69,71,75,82,89,99,101 however, simulation evidence suggests that inclusion of stopped-early trials in meta-analyses does not lead to substantial bias. 65

Methodological comments

All-cause mortality

All 13 trials comparing the use of ICDs with the use of AADs in people at increased risk of SCD from ventricular arrhythmias reported measures of all-cause mortality as their primary outcome measure. 69,71,75,81,82,84,89,90,95,97,99,101,105 Four trials71,81,84,89 assessed the use of ICDs compared with the use of AADs in people at increased risk of SCD from previous ventricular arrhythmias. All four trials showed beneficial effects on crude mortality rates for those receiving an ICD, although only the AVID trial71 (ICD arm 15.8%, AAD arm 24.0%, p < 0.012, follow-up 18.2 months) and the main DEBUT trial89 (ICD arm 0%, AAD arm 13.8%, p < 0.02, follow-up 3 years) found statistically significant differences. A separate pilot study for the DEBUT trial89 had previously shown no significant difference between the ICD arm and the AAD arm (ICD arm 0%, AAD arm 30.0%, p = 0.07, follow-up maximum 3 years). In the other two studies differences were either not statistically significant or not assessed. The CASH trial81 reported an all-cause mortality rate of 36.4% for the ICD group compared with 44.4% for the AAD group (p-value not stated, follow-up 57 months). The CIDS trial84 reported a crude mortality rate of 25.3% for the ICD group and 29.6% for the AAD group over the 3-year follow-up, equating to an annual crude mortality rate of 8.3% for the ICD group and 10.2% for the AAD group, a risk ratio reduction (RRR) of 19.7% (95% CI –7.7% to 40.0%, p = 0.142) (Table 17). A meta-analysis of the four studies (including the DEBUT pilot study89) using a random-effects model showed a statistically significant benefit for ICDs compared with AADs with a RR of 0.75 (95% CI 0.61 to 0.93, p = 0.010), with limited heterogeneity (χ² = 5.89, df = 4, I² = 32%) (Figure 4).

FIGURE 4.

All-cause mortality.

Of the nine trials69,75,82,90,95,97,99,101,105,146 including people who had not suffered a life-threatening arrhythmia but who were at increased risk, three showed statistically significant benefits for all-cause mortality in the ICD + OPT group compared with the different comparators (see Table 17). In the MADIT I trial,99 15.8% of participants receiving an ICD + OPT died compared with 38.6% of participants receiving OPT (mean follow-up 27 months), equating to a hazard ratio (HR) of 0.46 (95% CI 0.26 to 0.82, p = 0.009) (see Table 17). The MADIT II trial101 also found significant benefits, with 14.2% of those with an ICD + OPT dying compared with 19.8% of those who received OPT only (mean follow-up 20 months), a HR of 0.69 (95% CI 0.51 to 0.93, p = 0.016). Post-trial follow-up in the MADIT II study101 found continued benefit of ICDs at 8 years (HR 0.66, 95% CI 0.56 to 0.78, p = 0.001); analysis was undertaken on an efficacy basis by including data on crossovers and validated in an intention-to-treat (ITT) analysis. 149 The SCD-HeFT trial,105 which had a longer follow-up period (mean 45.5 months), reported that 22.0% of people who received an ICD + OPT died compared with 28.4% of those receiving amiodarone + OPT and 28.8% of those receiving placebo + OPT. HRs showed that the difference between the ICD + OPT group and the placebo + OPT group was statistically significant (HR 0.77, 97.5% CI 0.62 to 0.96, p = 0.007), whereas that between the amiodarone + OPT group and the placebo + OPT group was not statistically significant (HR 1.06, 97.5% CI 0.86 to 1.30, p = 0.53). 105 A meta-analysis of the two MADIT trials99,101 using a random-effects model showed a statistically significant benefit for those receiving ICDs + OPT compared with OPT alone, with a RR of 0.57 (95% CI 0.33 to 0.97, p = 0.04), although there was some apparent heterogeneity (χ² = 3.54, df = 1, I² = 72%), which may reflect differences in disease severity (see Figure 4).

The other six trials, which included people with cardiomyopathy,69,82,90 in the early period post MI95,97 or scheduled for a CABG,78 found no statistically significant differences between groups for all-cause mortality. The AMIOVIRT trial69 reported all-cause mortality after a mean follow-up of 2 years, finding that 11.8% of those with an ICD + OPT had died compared with 13.5% of those receiving amiodarone + OPT (p = 0.8). The CAT trial82 reported all-cause mortality at 1 year, showing no significant difference between groups (ICD + OPT 8% vs. OPT 3.7%, p = 0.3672). Longer mean follow-up to 5.5 years showed a limited difference between groups, with 26% of the ICD + OPT group and 31.5% of the OPT group dying (p-value not stated). The DEFINITE trial90 found that 12.2% of participants receiving an ICD + OPT and 17.5% of those receiving OPT had died at a mean follow-up of 29 months, a HR of 0.65 (95% CI 0.40 to 1.06, p = 0.08) (see Table 17). Combining these three cardiomyopathy trials using random-effects meta-analysis confirmed that there was no significant difference between the treatments, with a RR of 0.77 (95% CI 0.52 to 1.15, p = 0.20), with no heterogeneity (χ² = 1.73, df = 2, I² = 0%) (see Figure 4). The effect of combining the three cardiomyopathy trials with the non-ischaemic CHF subgroup of the SCD-HeFT trial105 is assessed in Subgroup analyses reported by included randomised controlled trials. The DINAMIT95 and IRIS97 trials assessed the effects of ICDs + OPT compared with OPT in those who were in the early period post MI. The DINAMIT trial95 reported that 18.7% of participants receiving an ICD + OPT and 17.0% of those receiving OPT had died by 30 months’ follow-up, resulting in a HR of 1.08 (95% CI 0.76 to 1.55, p = 0.66). Similarly, the IRIS trial97 found no significant difference in all-cause mortality between the ICD + OPT group (26.1%) and the OPT group (25.8%), reflected in a HR of 1.04 (95% CI 0.81 to 1.35, p = 0.15). Meta-analysis of the DINAMIT95 and IRIS97 trials confirmed that there was no significant difference between the treatments, with a RR of 1.04 (95% CI 0.86 to 1.25, p = 0.69), with no heterogeneity (χ² = 0.19, df = 1, I² = 0%) (see Figure 4). The CABG Patch trial,78 which included people who were scheduled for a CABG, reported a mortality rate of 22.9% for those receiving an ICD + OPT compared with 21.2% for those receiving OPT (p-value not stated), a RR of 1.08 (95% CI 0.85 to 1.38, p = 0.53) (see Figure 4).

Total cardiac deaths

Only two trials in people at increased risk of SCD due to previous ventricular arrhythmias, specifically the AVID72 and CIDS84 trials, assessed the effects of ICDs compared with AADs on total cardiac deaths (Table 18). Although both studies found lower crude rates for those receiving an ICD, neither reported whether the effect was statistically significant (AVID:72 ICD 12.4%, AAD 18.5%, p-value not stated; CIDS:84 ICD 20.4%, AAD 25.1%; p-value not stated). In addition, the CIDS trial84 found no statistically significant difference between the interventions with regard to annual crude mortality rates (ICD 6.7%, AAD 8.6%, RRR 23.4%, 95% CI –5.7 to 44.5, p = 0.104). However, a meta-analysis of the two studies using a random-effects model showed that ICDs had a statistically significant effect compared with AADs, with a RR of 0.74 (95% CI 0.61 to 0.91, p = 0.004) and no apparent heterogeneity (χ² = 0.84, df = 1, I² = 0%) (Figure 5).

FIGURE 5.

Total cardiac deaths.

Eight trials69,78,82,95,97,99,101,108 in people who had not suffered a life-threatening arrhythmia but who were at increased risk assessed the effects of ICDs + OPT compared with either OPT, amiodarone + OPT or placebo + OPT on total cardiac deaths (see Table 18). Of these, only the MADIT II trial,103 which included people remote from MI (ICD + OPT 10.6%, OPT 16.3%, p < 0.01), and the SCD-HeFT trial,108 which included people with mild to moderate HF (ICD + OPT 14.7%, placebo + OPT 19.7%, amiodarone + OPT 19.2%; HR 0.76, 95% CI 0.60 to 0.95, p = 0.018), found statistically significant benefit for those receiving an ICD + OPT. A similar difference was identified in the MADIT I trial,99 which included people remote from MI (ICD + OPT 11.6%, OPT 26.7%); however, statistical significance was not stated. A meta-analysis of the MADIT I99 and II103 trials using a random-effects model showed a statistically significant benefit for ICDs + OPT, with a RR of 0.59 (95% CI 0.42 to 0.83, p = 0.003) and limited heterogeneity (χ² = 1.3, df = 1, I² = 23%) (see Figure 5).

The DINAMIT95 (ICD + OPT 13.9%, OPT 14.3%, p-value not stated) and IRIS97 (ICD + OPT 21.3%, OPT 21.9%, p-value not stated) trials, which included those with a recent MI, the AMIOVIRT trial,69 which included those with cardiomyopathy (ICD + OPT 7.8%, amiodarone + OPT 9.6%, p-value not stated) and the CABG Patch trial,78 which included people scheduled for a CABG (ICD + OPT 17.0%, OPT 17.4%; HR 0.97, 95% CI 0.71 to 1.33, p = 0.84) found limited differences in total cardiac deaths between those receiving ICD + OPT and those receiving either OPT or amiodarone + OPT (see Table 18). In contrast, the CAT trial,82 which included people with cardiomyopathy, reported higher total cardiac mortality among those receiving an ICD + OPT than among those receiving OPT (ICD + OPT 8%, OPT 0%), although the statistical significance was not stated. When these trials were meta-analysed by patient group using random-effects models, the lack of any statistically significant benefit was evident. Combining the DINAMIT95 and IRIS97 trials produced a RR of 0.97 (95% CI 0.79 to 1.20, p = 0.8) with no apparent heterogeneity (χ² = 0, df = 1, I² = 0%) (see Figure 5). The meta-analysis of the AMIOVIRT69 and CAT82 trials resulted in a RR of 2.03 (95% CI 0.17 to 23.62, p = 0.57) with some moderate heterogeneity (χ² = 2.59, df = 1, I² = 61%) (see Figure 5).

Sudden cardiac death/arrhythmic deaths

Sudden cardiac and arrhythmic death rates were lower among people receiving an ICD than among those receiving AADs in the four trials72,81,84,89 that included people at increased risk of SCD as a result of previous ventricular arrhythmias (Table 19). Both the CASH81 [ICD 13.0%, 95% CI 7.9 to 19.6; AAD (either amiodarone or metoprolol) 32.8%, 95% CI 27.2 to 41.8] and DEBUT89 (ICD 0%; AAD 13.8%) trials reported lower rates of SCD for those receiving an ICD than for those receiving AADs, although only the CASH trial81 showed a statistically significant difference. Similarly, the AVID72 and CIDS84 studies showed benefit for people receiving an ICD compared with AADs with regard to crude rate of arrhythmic deaths (AVID:72 ICD 4.7%, AAD 10.8%; CIDS84: ICD 9.2%, AAD 13.0%), although neither demonstrated a statistically significant difference. The CIDS trial84 also showed no statistically significant difference when comparing the interventions for annual crude mortality rate [ICD 3.0%, AAD 4.5%, RRR 32.8%; 95% CI –7.2 to 57.8, p = 0.094]. Combining the four studies through a random-effects meta-analysis showed a statistically significant benefit for the ICD group compared with the AAD group, with a RR of 0.49 (95% CI 0.34 to 0.69, p < 0.0001) and limited heterogeneity (χ² = 5.47, df = 4, I² = 27%) (Figure 6).

FIGURE 6.

Sudden cardiac deaths/arrhythmic deaths.

All nine trials69,78,82,90,95,97,99,103,108 that included people who had not suffered a life-threatening arrhythmia but who were at increased risk reported sudden cardiac or arrhythmic death as an outcome (see Table 19). Although eight of the trials69,78,90,95,97,99,103,108 showed benefit for those receiving an ICD + OPT compared with OPT, amiodarone + OPT or placebo + OPT, only four90,95,97,103 identified showed a statistically significant effect. The DINAMIT95 and IRIS97 trials highlighted the benefits of ICDs + OPT compared with OPT for people who had had a recent MI, reporting HRs of 0.42 (95% CI 0.22 to 0.83, p = 0.009) and 0.55 (95% CI 0.31 to 1.00, p = 0.049) respectively (see Table 19). When meta-analysed, a combined RR of 0.45 (95% CI 0.31 to 0.64, p < 0.0001) resulted, with no heterogeneity reported (χ² = 0.03, df = 1, I² = 0%) (see Figure 6).

The MADIT I99 (ICD + OPT 3.2%, OPT 12.9%, p-value not stated) and MADIT II103 (ICD + OPT 3.8%, OPT 10.0%, p < 0.01) trials, which included people remote from MI, showed lower rates of sudden cardiac or arrhythmic death among those receiving an ICD + OPT than among those receiving OPT. Meta-analysis using a random-effects model showed a significant benefit for ICDs + OPT with a RR of 0.36 (95% CI 0.23 to 0.55, p < 0.00001) and no heterogeneity (χ² = 0.42, df = 1, I² = 0%) (see Figure 6).

The AMIOVIRT,69 CAT82 and DEFINITE90 trials, which included people with cardiomyopathy, reported differing outcomes. The DEFINITE trial90 found that significantly fewer people who received an ICD + OPT (1.3%) died from sudden cardiac or arrhythmic death than those receiving OPT (6.1%), reflected in a HR of 0.20 (95% CI 0.06 to 0.71, p = 0.006) (see Table 19). Although the AMIOVIRT trial69 also found benefit for those receiving an ICD + OPT (2.0%) compared with those receiving amiodarone + OPT (3.8%), the benefit was not statistically significant (p = 0.7). The CAT trial82 reported no sudden cardiac or arrhythmic deaths in either the ICD + OPT group or the OPT group. A random-effects meta-analysis of the three trials showed an overall statistically significant benefit for participants who received an ICD + OPT compared with the comparator treatment, with a RR of 0.26 (95% CI 0.09 to 0.77, p = 0.02) and no heterogeneity (χ² = 0.41, df = 1, I² = 0%) (see Figure 6).

The CABG Patch trial,78 which included people who were scheduled for CABG surgery, reported lower rates of sudden cardiac and arrhythmic death in the ICD + OPT group (3.4%) than in the OPT (6.2%), although the difference was marginally insignificant (HR 0.55, 95% CI 0.29 to 1.03, p = 0.06) (see Table 19). In contrast, the SCD-HeFT trial108 found significantly lower sudden cardiac or arrhythmic mortality in the group receiving an ICD + OPT (4.6%) than in the group receiving amiodarone + OPT (9.5%) or the group receiving placebo + OPT (11.6%), with a RR of 0.44 (95% CI 0.31 to 0.61, p < 0.00001) (see Figure 6).

Non-arrhythmic cardiac deaths

Two trials72,84 that included people at increased risk of SCD because of previous ventricular arrhythmias reported rates of non-arrhythmic deaths. The AVID72 and CIDS84 trials assessed the effects of ICDs compared with the effects of AADs on crude non-arrhythmic cardiac deaths, with neither stating whether there was any statistically significant benefit (AVID72: ICDs 7.7%, AAD 7.7%; CIDS84: ICDs 11.3%, AAD 12.1%) (Table 20). The CIDS trial84 also reported annual crude mortality rates (ICDs 3.7%, AAD 4.2%), which resulted in a non-significant RRR of 13.5% (95% CI –35.4 to 44.7, p = 0.526). A random-effects meta-analysis confirmed the lack of a statistically significant difference between the groups, reporting a RR of 0.97 (95% CI 0.72 to 1.31, p = 0.83) and no heterogeneity (χ² = 0.06, df = 1, I² = 0%) (Figure 7).

FIGURE 7.

Non-arrhythmic cardiac deaths.

Implantable cardiac defibrillator + OPT appeared to have a limited effect on the occurrence of non-arrhythmic cardiac deaths compared with OPT, amiodarone + OPT or placebo + OPT in people who had not suffered a life-threatening arrhythmia but who were at increased risk (see Table 20). In people who had had a recent MI, the DINAMIT95 and IRIS trials97 found a statistically significant benefit for those receiving OPT only compared with those receiving an ICD + OPT, reporting HRs of 1.72 (95% CI 0.99 to 2.99, p = 0.05) and 1.92 (95% CI 1.29 to 2.84, p = 0.001) respectively. Combining the studies using a random-effects meta-analysis confirmed the statistically significant benefit for people receiving OPT, with a RR of 1.77 (95% CI 1.30 to 2.40, p = 0.0002) and no apparent heterogeneity (χ² = 0, df = 1, I² = 0%) (see Figure 7).

The effect of the different interventions on non-arrhythmic cardiac deaths in other patient subgroups was more equivocal. The MADIT I99 and MADIT II103 trials, which included people remote from MI, reported contrasting mortality rates (MADIT I:99 ICD + OPT 7.4%, OPT 12.9%; MADIT II:103 ICD + OPT 5.8%, OPT 4.3%). Meta-analysing these data using a random-effects model showed no statistically significant difference between the ICD + OPT group and the OPT group (RR 0.95, 95% CI 0.41 to 2.18, p = 0.9; χ² = 2.77, df = 1, I² = 64%) (see Figure 7). Similar variation was reported by the three trials assessing non-arrhythmic cardiac deaths among people with cardiomyopathy. The AMIOVIRT69 (ICD + OPT 5.9%, amiodarone + OPT 5.8%), CAT82 (ICD + OPT 8%, OPT 0%) and DEFINITE90 (ICD + OPT 3.9%, OPT 4.8%) trials reported differing mortality rates; when these data were meta-analysed there were no statistically significant differences between the groups (RR 1.13, 95% CI 0.42 to 3.03, p = 0.81; χ² = 2.71, df = 2, I² = 26%) (see Figure 7). Similarly, the CABG Patch trial,78 which included those who were scheduled for CABG surgery (RR 1.26, 95% CI 0.87 to 1.82, p = 0.21), and the SCD-HeFT trial,108 which included those with mild to moderate HF (RR 1.14, 95% CI 0.88 to 1.48, p = 0.32) found no statistically significant differences between the groups (see Figure 7).

Other causes of death: non-cardiac deaths

Two trials72,84 in people at increased risk of SCD because of previous ventricular arrhythmias assessed non-cardiac causes of death as an outcome (Table 21). The AVID72 and CIDS84 trials found no statistically significant difference between ICDs and AADs for the outcome of other non-cardiac causes of death (AVID:72 ICD 3.4%, AAD 5.5%, RR 1.78, 95% CI 0.98 to 3.26, p = 0.053; CIDS:84 non-cardiac vascular: ICD 0.9%, AAD 0.6%, RRR –36.6%, 95% CI –719.8 to 77.2, p = 0.732; non-vascular: ICD 4.0%, AAD 3.9%, RRR 4.5%, 95% CI –106.1 to 55.7, p = 0.908) (see Table 21), reflected in a random-effects meta-analysis (RR 0.79, 95% CI 0.45 to 1.37, p = 0.40; χ² = 1.51, df = 1, I² = 34%) (Figure 8). The CIDS trial84 presented annual crude death rates for the ICD and AAD groups for non-cardiac vascular (ICD 0.3%, AAD 0.2%) and non-vascular (ICD 1.3%, AAD 1.4%) causes,84 finding limited differences.

FIGURE 8.

Other causes of death: non-cardiac deaths.

Eight trials69,78,82,95,97,99,103,108 in people who had not suffered a life-threatening arrhythmia but who were at increased risk assessed the effects of ICDs + OPT on other non-cardiac causes of death compared with the different comparator treatments, finding no statistically significant benefit (see Table 21). Meta-analyses using random-effects models of the DINAMIT95 and IRIS97 trials in people with a recent MI (RR 1.39, 95% CI 0.86 to 2.27, p = 0.18; χ² = 0.70, df = 1, I² = 0%), the MADIT I99 and MADIT II103 trials in people remote from MI (RR 1.06, 95% CI 0.58 to 1.95, p = 0.84; χ² = 0.55, df = 1, I² = 0%) and the AMIOVIRT69 and CAT82 trials in people with cardiomyopathy (RR 0.65, 95% CI 0.13 to 3.29, p = 0.60; χ² = 0.75, df = 1, I² = 0%) all found no statistically significant effects (see Figure 8). Similarly, the CABG Patch trial78 in people who were scheduled for CABG surgery (RR 1.50, 95% CI 0.82 to 2.73, p = 0.19) and the SCD-HeFT108 trial, which included people with mild-to moderate HF (RR 0.92, 95% CI 0.66 to 1.27, p = 0.60), reported no statistically significant differences between groups in deaths from other non-cardiac causes (see Figure 8).

Cumulative mortality

The cumulative mortality risk for both total and arrhythmic mortality was assessed annually for up to 3 years’ follow-up in the CIDS trial in people at increased risk of sudden cardiac death as a result of previous ventricular arrhythmias. 84 Rates were consistently lower for those receiving an ICD compared with those receiving AADs, with a RRR for total mortality of 15.4% in year 1, 29.7% in year 2 and 13.7% in year 3 and a RRR for arrhythmic mortality of 29.9% in year 1, 31.4% in year 2 and 17.8% in year 3 (Table 22).

Four trials in people who had not suffered a life-threatening arrhythmia but who were at increased risk reported other mortality outcomes. 75,90,97,105 The IRIS trial97 in people with a recent MI presented cumulative death rates annually up to 3 years (see Table 22). Although this trial found lower mortality rates for those receiving an ICD + OPT (year 1 10.6%, year 2 15.4%, year 3 22.4%) than for those receiving OPT (year 1 12.5%, year 2 18.2%, year 3 22.9%), the differences were not found to be statistically significant (p = 0.76). Similarly, the DEFINITE trial,90 which included people with cardiomyopathy (year 1: ICD + OPT 2.6%, OPT 6.2%; year 2: ICD + OPT 7.9%, OPT 14.1%), and the SCD-HeFT trial,105 which included people with mild to moderate HF (Kaplan–Meier estimate, 5 years: ICD + OPT 0.289; amiodarone + OPT 0.340; placebo + OPT 0.361), also reported lower all-cause mortality following implantation of an ICD (p-values not stated). In contrast, the CABG Patch trial,75 which included people scheduled for CABG surgery, reported higher actuarial mortality at 4 years’ follow-up in those receiving an ICD + OPT (27%) than in those receiving OPT (24%), although the difference was not statistically significant (HR 1.07, 95% CI 0.81 to 1.42, p = 0.64) (see Table 22).

Survival

Differences in mortality were reflected in the survival outcomes reported by the AVID71,72 and CASH81 trials in people at increased risk of SCD as a result of previous ventricular arrhythmias. The AVID trial reported statistically significant differences in overall survival during the 3 years of follow-up (p < 0.02),71 survival free of cardiac death at 2 years (p = 0.0042)72 and survival to arrhythmic death at 2 years (p = 0.0002),72 favouring ICDs compared with AAD (Table 23). Survival free of non-arrhythmic cardiac death did not differ significantly between those receiving ICDs and those receiving AADs (p = 0.8039). 72 Despite the CASH trial81 finding benefits for ICDs compared with AADs for overall survival (HR 0.766, p = 0.081) and survival free of cardiac arrest (HR 0.481, p = 0.072), the differences were not statistically significant. In contrast, the CASH trial81 did report a significant benefit for survival free of sudden death for people who received an ICD compared with those who received AADs (HR 0.423, p = 0.005). The DEBUT trial89 reported a mean survival time for the AAD group of 26.2 [standard error of the mean (SEM) 1.4] months (no deaths in the ICD group).

Only the AMIOVIRT69 and CAT82 trials, which included people with cardiomyopathy, reported survival (see Table 23). The AMIOVIRT trial69 presented overall and arrhythmia-free survival rates for the ICD + OPT group and the amiodarone + OPT group at 1 and 3 years’ follow-up, showing no statistically significant difference (overall survival p = 0.1, arrhythmia-free survival p = 0.8). The CAT trial82 presented cumulative survival data for the ICD + OPT group and the OPT group for up to 6 years’ follow-up, finding no statistically significant difference between the groups (p = 0.554) (see Table 23).

Heart failure hospitalisations

Only the AVID study,71 which included people at increased risk of SCD because of previous ventricular arrhythmias, reported the proportion of patients rehospitalised annually for up to 3 years’ follow-up. Significantly higher rates were reported for the ICD group than for the AAD group (p = 0.04) (Table 24). For both groups, the rehospitalisation rate was > 55% at year 1, > 65% at year 2 and > 75% at year 3.

The MADIT II trial,101 which included people remote from MI, reported the proportion of hospitalisations because of HF (ICD + OPT 19.9%, OPT 14.9%, p-value not stated) and the number of patients hospitalised per 1000 months of follow-up (ICD + OPT 11.3, OPT 9.4, p = 0.09), with higher rates among those receiving an ICD + OPT (see Table 24).

Symptoms/complications related to arrhythmias

The CAT82 and AMIOVIRT69 trials, which included people with cardiomyopathy, reported the occurrence of syncope. Some 12% of people with an ICD + OPT had syncope during VTs in the CAT trial82 and 3.9% of ICD + OPT patients and 5.8% of amiodarone + OPT patients had syncope in the AMIOVIRT study69 (Table 25). The MADIT II trial,103 which included people remote from MI, reported the number of adverse cardiac events in the week before SCD (ICD + OPT 28, OPT 49), with comparable rates of syncope and angina pectoris (4% for both), lower rates of MI for the ICD + OPT group (ICD + OPT 4%, OPT 10%) and higher rates of ventricular arrhythmia (ICD + OPT 25%, OPT 10%) and CHF (ICD + OPT 43%, OPT 16%) for the ICD + OPT group.

Quality of life

Two trials in people at increased risk of SCD because of previous ventricular arrhythmias, the AVID74 and CIDS87 trials, reported results from substudies using a range of generic and condition-specific measures of QoL (Table 26). The AVID trial74 assessed QoL using the Short Form questionnaire-36 items (SF-36) physical component summary (PCS) and mental component summary (MCS), the 46-item patient concerns checklist and the cardiac version of the QL index. Follow-up was for 12 months and assessments were made of the impact of adverse symptoms and ICD shocks. Comparison of PCS scores at baseline and 12 months’ follow-up showed no statistically significant differences between the ICD group and the AAD group (baseline: ICD 37.4, AAD 36.5, p = 0.3; 12 months: ICD 40.0, AAD 38.0, p = 0.3). In contrast, the ICD group had a lower (worse) mean score on the MCS at baseline than the AAD group, which was statistically significant (p = 0.006), although any difference had disappeared by 12 months’ follow-up. Scores on the patient concerns checklist did not differ significantly between the ICD group and the AAD group at baseline (ICD 15.9, AAD 16.2, p = 0.06) or at 12 months’ follow-up (p = 0.1). On the QL index the scores for the ICD and AAD groups were similar at baseline (ICD 22.1, AAD 21.9, p-value not stated) and at 12 months’ follow-up (scores and p-values not stated).

The effects of adverse symptoms and ICD shocks were assessed in the AVID trial74 using PCS scores, MCS scores and patient concerns using multivariate analysis including age, sex, race, index arrhythmia, ejection fraction, history of HF and use of beta-blockers at hospital discharge (see Table 26). Adverse symptoms led to a statistically significant worsening of PCS scores (p < 0.001), MCS scores (p = 0.002) and patient concern scores (p < 0.001) for the ICD group and PCS scores (p = 0.009) and patient concern scores (p = 0.03) for the AAD group. The occurrence of ICD shocks had a similar adverse effect on QoL, with statistically significant worsening of PCS scores (p = 0.03), MCS scores (p = 0.04) and patient concern scores (p < 0.001).

A substudy of the CIDS trial87 reported the effects of ICDs and AADs on three domains of the Mental Health Inventory (MHI) and seven domains of the Nottingham Health Profile (NHP), with an additional assessment of the consequences of ICD shocks on these measures (see Table 26). At 12 months’ follow-up the ICD group had shown a significantly greater improvement than the AAD group on the MHI domains of ‘total index’ (p = 0.001), ‘psychological distress’ (p = 0.001) and ‘psychological well-being’ (p = 0.03) and the NHP domains of ‘energy level’ (p = 0.0001), ‘physical mobility’ (p = 0.002), ‘emotional reactions’ (p = 0.002), ‘sleep disturbance’ (p = 0.02) and ‘lifestyle impairment’ (p = 0.005). It was notable that none of the domains on the MHI and the NHP improved for the AAD group between baseline and 12 months’ follow-up, with the domains of ‘energy level’ and ‘physical mobility’ deteriorating.

The effects of ICD shocks on QoL were assessed in the CIDS trial87 on the different domains of the MHI and the NHP through univariate comparisons between groups in terms of the numbers of shocks (i.e. ICD no shocks, ICD one to four shocks, ICD five or more shocks and AAD group without an ICD) (see Table 26). It was evident that the ICD five or more shocks group, like the AAD group without an ICD, did not experience the significant improvements in QoL that were reported by the ICD groups with less than five shocks. At 12 months’ follow-up the ICD five or more shocks group scored significantly worse (p < 0.05) than both the ICD no shocks group and the ICD one to four shocks group on the MHI ‘total index’ and ‘psychological distress’ domains, than the ICD one to four shocks group on the MHI ‘psychological well-being‘ domain and than the ICD no shocks group on the NHP ‘emotional reactions’ domain. Although the ICD five or more shocks group did not differ significantly from the AAD group without an ICD on any of the MHI and NHP domains, the ICD no shocks and ICD one to four shocks groups had significantly better (p < 0.05) QoL than the AAD group without an ICD on the MHI ‘total index’ and ‘psychological distress’ domains and the NHP ‘energy level’, ‘physical mobility’ (ICD no shocks only), ‘emotional reactions’ and ‘lifestyle impairment’ domains.

Five trials69,80,94,104,107 in people who had not suffered a life-threatening arrhythmia but who were at increased risk assessed QoL. The MADIT II trial104 assessed QoL in those remote from their MI through the Health Utilities Index 3 (HUI3), reporting the mean score, mean annual change and overall mean score (including death) for those alive at assessment annually to 3 years’ follow-up (see Table 26). The mean annual change in HUI3 scores showed a worsening in HRQoL for the ICD + OPT group compared with the OPT group annually, with a statistically significantly change in years 2 (p = 0.05) and 3 (p = 0.10). 104 Despite these changes, comparison of the HUI3 scores for the different interventions showed that they were not significantly different during follow-up, even when mortality was taken into account (valuing death as 0). 104

The AMIOVIRT study69 in people with cardiomyopathy assessed changes in QoL using the Quality of Well-Being Scale (QWBS) and the State–Trait Anxiety Inventory (STAI) (see Table 26). Comparison of the ICD + OPT group with the amiodarone + OPT group at 1 year of follow-up showed no statistically significant difference between the groups for well-being on the QWBS (p = 0.5) or for anxiety on the STAI (p = 0.4). Although the DEFINITE trial94 in people with cardiomyopathy assessed QoL using the Short Form questionnaire-12 items (SF-12) MCS and PCS and the Minnesota Living with Heart Failure Questionnaire (MLWHFQ), stating that no statistically significant differences were found between the ICD + OPT group and the OPT group, no data were reported.

The CABG Patch trial80 in people scheduled for a CABG assessed HRQoL using measures of perception of health, ability to function and psychological well-being at 6 months’ follow-up (see Table 26). On all measures of HRQoL the group receiving OPT reported a higher QoL than the ICD + OPT group, with statistically significant differences for the measures of perception of health transition (p = 0.030), emotional role function (p = 0.003), mental health (p = 0.004), satisfaction with appearance (p = 0.008) and satisfaction with scar (p = 0.040). 80 With 38.5% of people with an ICD + OPT having received a shock in the 6 months before completing the QoL instrument, the CABG Patch trial80 assessed the effects on QoL scores. On 10 of the 12 measures the OPT group had a higher QoL than the ICD + OPT group when the device either fired or did not fire. The scores for the ICD + OPT group when the device did not fire were similar to those of the OPT group, with no statistically significant differences (p-values not stated). In contrast, the ICD + OPT group when the device did fire had a lower QoL, with statistically significant differences (p = 0.05) for perception of health transition, physical limitations, physical role functioning, emotional role functioning, mental health and satisfaction with appearance.

The SCD-HeFT trial107 in people with HF reported QoL through a comparison of the Duke Activity Status Index (DASI), the Mental Health Inventory 5 (MHI-5), the MLWHFQ and the global health status of the ICD + OPT, amiodarone + OPT and placebo + OPT groups at baseline and 3, 12 and 30 months’ follow-up (see Table 26). The effects on QoL were compared between those experiencing shocks and those not receiving a shock in the ICD + OPT group using the SF-36. Using the DASI there were no clinical (4-point difference) or statistically significant differences in median or mean scores between the groups at baseline and 3, 12 and 30 months. On the MHI-5, outcomes were more equivocal. Although the differences in the median and mean scores comparing the ICD + OPT group and amiodarone + OPT group separately with the placebo + OPT group were below clinically meaningful levels (i.e. 5-point difference), some were statistically significant. Comparison of the median scores showed that the ICD + OPT group had significantly better scores than the placebo + OPT group at 3 months (p = 0.01) and 12 months (p = 0.003). By 30 months the scores for the ICD + OPT group had declined to baseline levels. Similarly, the mean scores for the ICD + OPT group differed significantly from those for the placebo + OPT group at 3 and 12 months (p = 0.05). Although the amiodarone + OPT group had a significantly higher MHI-5 score at baseline than the placebo + OPT group (p = 0.05), these differences disappeared during subsequent follow-up.

Similar improvements for the ICD + OPT group were reported on the MLWHFQ in the SCD-HeFT trial,107 resulting in significantly better scores for the ICD + OPT group than for the placebo + OPT group at 3 (p = 0.006) and 30 (p = 0.05) months (see Table 26). However, these differences were thought to be clinically insignificant (5-point change). In contrast, a comparison using a time trade-off utility measure showed that the health status of the ICD + OPT group and the placebo + OPT group declined from baseline with no statistically significant difference at 30 months’ follow-up (p = 0.18).

The effects of ICD shocks on QoL were assessed in the SCD-HeFT trial using the SF-36 (see Table 26). 107 A comparison of the changes in scores for those who had received a shock within 1 month of a scheduled QoL assessment and those who had not received a shock showed a significant decrease in the QoL of those who received a shock with regard to their relative perceptions of general health (p = 0.002), physical function (p < 0.001), emotional function (p = 0.02), social function (p = 0.009) and self-related health (p = 0.009). 107

Adverse events

All four trials71,81,84,89 comparing the use of ICDs with AADs in people at increased risk of SCD because of previous ventricular arrhythmias reported adverse events (Table 27). Reported adverse events differed between the trials, limiting comparisons. Only the total number of adverse events and mortality rates were compared between the interventions in the DEBUT trial89 and the AVID71 and CASH81 trials respectively. The DEBUT trial89 reported that 29.7% of the ICD group and 13.8% of the AAD group suffered adverse events (p-value not stated). The AVID trial71 compared deaths within 30 days of initiation of therapy or by hospital discharge if 30 days after therapy began, finding no statistically significant difference between the ICD group (2.4%) and the AAD group (3.5%) (p = 0.27). In contrast, the CASH trial81 found significantly (p = 0.029) higher mortality rates during the perioperative period for the ICD group (5.1%) than for the AAD group (1.1%). The only other comparison between interventions was in the AVID trial,71 finding that the use of thyroid replacement medication was higher for the AAD group at year 1 (10.0%) and year 2 (16.0%) than in the ICD group (1.0% years 1 and 2) (p-value not stated).

Analysis of the adverse events reported for the ICD groups in the four trials71,81,84,89 showed that these tended to be limited in occurrence (see Table 27). The most frequent were those related to the placement and operation of the device itself, including defibrillation discharges caused by superventricular tachycardia or sinus tachycardia (19.0%);89 T-wave oversensing (8.1%);89 ICD product discomfort (7.6%);84 ICD permanently or temporarily explanted because of infection, heart transplantation or patient preference (5.2%);84 device dysfunction (5.1%);81 pocket erosion requiring removal of the ICD (2.7%);89 dislodgement or migration of system leads (3.0%);81 ICD dislodgement/fracture (2.4%);84 bleeding requiring reoperation or transfusion (1.2%);71 and unsuccessful first attempt at ICD implantation without thoracotomy (1.0%). 71 Other adverse events included haematoma or seroma (6.1%);81 serious haematoma (2.6%);71 pleural effusion (3.0%);81 infection (2.0–4.6%);71,84 and pneumothorax (1.6%). 71

Adverse events reported for the AAD groups differed between the four trials (see Table 27). 71,81,84,89 The CIDs trial84 found that > 10% of people receiving amiodarone reported insomnia (19.3%), ataxia (17.2%), tremor (15.4%), visual symptoms (14.5%) or photosensitivity (10.3%). Other adverse events reported in the CIDs trial84 included skin discolouration (6.3%) and pulmonary infiltrate (5.7%). In the CASH trial81 10% of people receiving amiodarone (9.8%) or metoprolol (10.3%) had to discontinue drug treatment. The AVID trial71 reported that 5% of the AAD group had suspected pulmonary toxicity at 2 years. Other adverse events reported by the AVID,71 CASH81 and DEBUT89 trials affected < 5% of participants (see Table 27).

All nine trials69,75,82,90,95,97,99,101,105 comparing ICDs + OPT with the differing comparator treatments in people who had not suffered a life-threatening arrhythmia but who were at increased risk reported adverse events, with six trials69,82,90,95,97,101 focused predominantly on those related to the placement of ICDs (see Table 27). The type of adverse events reported differed between the trials, making comparisons difficult. Adverse events were thought to affect between 5%105 and 61%75 of people receiving an ICD, depending on the definition of an adverse event or complication and the period of follow-up. Only three trials75,99,105 reported adverse events for the different comparator treatments, with rates varying from 11.9% to 55%.

Mortality rates associated with implantation of an ICD appeared low, with no deaths reported by four trials82,95,99,101 and crude death rates ranging from 1.6% to 5.4% in the IRIS97 and CABG Patch75 trials respectively. Deaths among those receiving the comparator treatments were reported only in the CABG Patch trial,75 with a crude death rate for the OPT group of 4.4%.

Lead-, electrode- or defibrillator generator-related problems were reported in five trials,82,90,97,99,101 affecting between 1.8% and 14.0% of people. In the IRIS trial,97 these led to a surgical revision rate of 2.4%. Surgical or device-related infections were reported in four trials,75,82,90,99 affecting between 0.4% and 12.3% of people in the ICD group. A further three trials81,82,101 reported infection leading to surgical intervention or device removal/replacement, which occurred in 0.7–4% of people.

Other non-device-specific adverse events were reported by four trials. 75,82,90,99 In the MADIT I99 and SCD-HeFT75 trials only syncope (5%) and hypothyroidism (6%) affected ≥ 5% of people in the comparator groups. The CABG Patch trial75 reported adverse events in the postoperative period and during long-term follow-up for both the ICD + OPT group and the OPT group, focusing predominantly on changes in underlying cardiac conditions. In the postoperative period the CABG Patch trial75 reported event rates of ≥ 5% for the ICD + OPT group and ≥ 4% for the OPT group for atrial fibrillation (ICD + OPT 22.9%, OPT 20.7%), new or severe HF (ICD + OPT 15.7%, OPT 12.6%), conduction defect (ICD + OPT 14.1%, OPT 14.5%), sustained VT (ICD + OPT 5.8%, OPT 6.8%), shock (ICD + OPT 9.2%, OPT 7.5%), pneumonia (ICD + OPT 8.5%, OPT 4.0%) and renal failure (ICD + OPT 6.7%, OPT 4.8%). 75 Events during long-term follow-up that affected ≥ 5% of the ICD + OPT group and the OPT group included new or worsening HF (ICD + OPT 42.5%, OPT 42.5%), angina pectoris (ICD + OPT 27.0%, OPT 27.5%), ventricular arrhythmias (ICD + OPT 19.4%, OPT 14.3%) and atrial fibrillation (ICD + OPT 14.7%, OPT 10.1%).

Subgroup analyses reported by included randomised controlled trials

Six trials71,75,90,97,103,105 reported prespecified subgroup analyses, although it should be noted that the trials were not powered to detect differences in subgroups.

The report of the AVID trial,71 which included people at increased risk of SCD because of previous ventricular arrhythmias, presented in a figure four prespecified subgroup analyses for all-cause mortality (age, LVEF, cause of arrhythmia and qualifying arrhythmia). No subgroup differed significantly from the others or the overall population. For most of the subgroups the 95% CIs crossed 1.0, apart from those for LVEF ≤ 35%, cause of arrhythmia coronary artery disease and VF rhythm, which favoured ICD. Subgroup analyses for the index arrhythmia were also reported (baseline: VF n = 455; VT n = 561). 72 ICDs improved survival free of arrhythmic death for people whose presenting arrhythmia was VT (p = 0.025) or VF (p = 0.0019). For non-arrhythmic cardiac death there were no statistically significant differences in survival between the ICD group and the AAD group for people presenting with either VT (p = 0.72) or VF (p = 0.98).

The IRIS trial,97 which included people in the early period post MI, prespecified 13 subgroup analyses for all-cause mortality, nine of which were presented in a figure [age, sex, CHF on admission, criterion of inclusion (for definitions see Appendix 7), ST-elevation MI, early reperfusion for ST-elevation MI, number of vessels, smoking and NYHA class at discharge] and four of which were not presented but described as similar in the two study groups (diabetes, hypertension, lipid abnormalities and number of risk factors). For most of the subgroups the 95% CIs crossed 1.0, apart from those for thrombolytic therapy for early reperfusion for ST-elevation MI (favoured control, data in figure only) and left main artery (favoured ICD, data in figure only).

In people remote from their MI, the MADIT II trial103 reported prespecified subgroup analyses for all-cause mortality using baseline characteristics, five of which were presented in a figure only (age, sex, ejection fraction, NYHA class or QRS interval) and seven of which were not presented (hypertension, diabetes, LBBB, atrial fibrillation, the interval since the most recent MI, type of ICD, and blood urea nitrogen level). The HRs in all of the subgroups were similar, with no statistically significant interactions.

The DEFINITE trial,90 which included people with cardiomyopathy, presented six prespecified subgroup analyses for all-cause mortality in a figure only (age, sex, LVEF, QRS interval, NHYA class and history of atrial fibrillation). None of the differences between subgroups were statistically significant. For most of the subgroups the 95% CIs crossed 1.0, apart from those for men (RR 0.49, 95% CI 0.27 to 0.90, p = 0.018), NYHA class III (RR 0.37, 95% CI 0.15 to 0.90, p = 0.02) and LVEF ≥ 20% (favoured ICD, data in figure only).

The CABG Patch trial,75 which included people who were scheduled for CABG surgery, evaluated 10 prespecified subgroups (age, sex, HF, NYHA class, LVEF, diabetes mellitus, QRS complex duration, use of ACE inhibitors, use of class I or class III AADs and use of beta-adrenergic-blocking drugs). HRs for the ICD group compared with the control group were found to be similar among the subgroups for all-cause mortality (data not reported).

The SCD-HeFT trial, which included people with mild to moderate HF, reported prespecified subgroup analyses for all-cause mortality105 and cause of death108 according to cause of CHF (ischaemic or non-ischaemic) and NYHA class (II or III) and for all-cause mortality according to race. 106 Table 28 presents the results for ICDs compared with placebo; subgroup results for the comparisons between amiodarone and placebo are reported in Appendix 7.

There was no significant interaction between ICD therapy and the cause of CHF for all-cause mortality (p = 0.68). 105 The HRs for those with ischaemic and non-ischaemic CHF were 0.79 (97.5% CI 0.60 to 1.04, p = 0.05) and 0.73 (97.5% CI 0.50 to 1.07, p = 0.06) respectively. Similarly, there was no significant interaction between ICD therapy and the cause of CHF for each of the specified modes of death108 (see Table 28). A significant reduction in sudden death presumed to be ventricular tachyarrhythmic was found for both ischaemic (HR 0.43, 95% CI 0.27 to 0.67) and non-ischaemic (HR 0.34, 95% CI 0.17 to 0.70) causes of CHF, whereas no significant reduction in other modes of death was found for either subgroup (see Table 28).

There was a statistically significant interaction between ICD therapy and NYHA class (p < 0.001). 105 Compared with placebo, ICDs reduced the risk of death in people in NYHA class II (HR 0.54, 97.5% CI 0.40 to 0.74, p < 0.001), but not in those in NYHA class III (HR 1.16, 97.5% CI 0.84 to 1.61, p = 0.30). The interaction between ICD therapy and NYHA class was statistically significant for cardiac mortality (p = 0.0004) and sudden death presumed to be ventricular tachyarrhythmic (p = 0.0091), but not for HF (p = 0.29) or non-cardiac (p = 0.11) deaths. 108 ICD therapy reduced the risk of cardiac mortality (HR 0.50, 95% CI 0.36 to 0.70) and sudden tachyarrhythmic death (HR 0.26, 95% CI 0.15 to 0.44) in people in NYHA class II, but not in those in NYHA class III (HR 1.17, 95% CI 0.84 to 1.64, and HR 0.73, 95% CI 0.41 to 1.29 respectively).

There was no significant interaction between ICD therapy and race (p = 0.53); ICD therapy reduced the risk of death in both racial groups (African American: HR 0.65, 95% CI 0.43 to 0.99; white: HR 0.73 95% CI 0.58 to 0.90). 106

Combining data from the SCD-HeFT105 non-ischaemic CHF subgroup with data from the three cardiomyopathy trials (AMIOVIRT,69 CAT,82 DEFINITE90) was considered appropriate by clinical experts. SCD-HeFT105 did not report the number of events for all-cause mortality occurring in each of the ischaemic and non-ischaemic subgroups; therefore, these were estimated by reviewers and data from the non-ischaemic subgroup were combined in a meta-analysis (Figure 9). The SCD-HeFT non-ischaemic subgroup strongly influenced the analysis and a statistically significant effect in favour of ICD therapy with no statistical heterogeneity was found (RR 0.74, 95% CI 0.58 to 0.93, p = 0.01). This is in contrast to the non-significant result of the meta-analysis of the three cardiomyopathy trials alone (see Figure 4).

FIGURE 9.

All-cause mortality, cardiomyopathy RCTs and SCD-HeFT non-ischaemic CHF subgroup.

Ventricular arrhythmia/cardiac arrest (secondary prevention)

• Four RCTs compared the effectiveness of ICDs and AADs. Average length of follow-up ranged from 18 months to 57 months and sample size ranged from 66 to 1016. The proportion of participants with CHF differed in the trials. In two trials 100% of participants had CHF, with > 80% in NYHA classes I and II. In the other two trials between 60% and 90% had CHF with approximately 50% in both trials in NYHA classes I and II. LVEF also varied, ranging from 30% to 70% across all four studies.

• All four RCTs assessed all-cause mortality as the primary outcome measure, which when combined through meta-analysis showed a statistically significant benefit for ICDs compared with AADs (RR 0.75, 95% CI 0.61 to 0.93, p = 0.01). Differences were found in the four RCTs for the outcome of sudden cardiac/arrhythmic deaths, with a statistically significant benefit for ICDs compared with AADs when combined through meta-analysis (RR 0.49, 95% CI 0.34 to 0.69, p < 0.0001).

• Meta-analysis of two trials showed a statistically significant benefit for ICDs compared with AAD for the outcome of total cardiac deaths (RR 0.74, 95% CI 0.61 to 0.91, p = 0.004); however, no differences were found for the outcomes of non-arrhythmic cardiac deaths (RR 0.97, 95% CI 0.72 to 1.31, p = 0.83) or other non-cardiac causes of death (RR 0.79, 95% CI 0.45 to 1.37, p = 0.40). Two RCTs reported different measures of survival, finding a statistically significant benefit for ICDs compared with AADs for overall survival at 3 years (difference 11%, p < 0.02), survival free of cardiac death at 2 years (difference 4%, p = 0.004), survival to arrhythmic death at 2 years in one trial (difference 5%, p = 0.0002) and survival free of sudden death at 57 months in the other trial (HR 0.423, p = 0.005). One RCT found lower cumulative mortality annually over 3 years’ follow-up with ICDs (difference: year 1 14.5%, year 2 1.7%, year 3 4.1%).

• Two RCTs assessed QoL through separate substudies using a range of measures. In one RCT there were no significant between-group differences at follow-up. A second RCT found that QoL improved significantly in the ICD group on three domains of the MHI and five domains of the NHP, whereas there were no changes in the OPT group. In this trial the QoL of those experiencing five or more ICD shocks did not differ significantly from that of the OPT group when analysed using the MHI and the NHP. The no shocks and one to four shocks ICD groups showed significant improvements on the MHI and NHP compared with the OPT group.

• One trial reported prespecified subgroup analyses for all-cause mortality. The subgroups for age, LVEF, cause of arrhythmia and qualifying arrhythmia did not differ significantly from each other or the overall population for all-cause mortality.

People with a recent myocardial infarction (within 6–41 days or ≤ 31 days)

• Two RCTs compared ICDs + OPT with OPT. Length of follow-up was 30 or 37 months and sample size ranged from 674 to 898. About 60% of participants in both trials were in NYHA class II, but the majority of the remaining participants had NYHA class III symptoms in one trial and NYHA class I symptoms in the other trial. Similarly, mean LVEF differed between the studies (28% and 35%), reflecting different eligibility criteria.

• Meta-analysis of the two trials found no difference between the groups in all-cause mortality (RR 1.04, 95% CI 0.86 to 1.25, p = 0.69), total cardiac deaths (RR 0.97, 95% CI 0.79 to 1.20, p = 0.8) and non-cardiac deaths (RR 1.39, 95% CI 0.86 to 2.27, p = 0.18). Those who received an ICD + OPT had a lower risk of SCD (RR 0.45, 95% CI 0.31 to 0.64, p < 0.0001) but a higher risk of non-arrhythmic cardiac death (RR 1.77, 95% CI 1.30 to 2.40, p = 0.0002). One trial reporting cumulative mortality found no statistically significant difference between the groups. QoL was not reported.

• One trial reported prespecified subgroup analyses for all cause-mortality. No significant differences were found for the 13 prespecified subgroups.

People with remote myocardial infarction (> 3 weeks or > 1 month previously)

• Two RCTs compared ICDs + OPT with OPT, although the pharmacological therapy in one of these may not be considered optimal by current standards. Average length of follow-up was 27 and 20 months and sample size was 196 and 1232 respectively. About two-thirds of participants had NYHA class II or III symptoms and one-third had NYHA class I symptoms. Mean LVEF differed between the studies (about 26% and 23%), reflecting different eligibility criteria.

• Meta-analysis of the two trials found a reduction in all-cause mortality (RR 0.57, 95% CI 0.33 to 0.97, p = 0.04), total cardiac deaths (RR 0.59, 95% CI 0.42 to 0.83, p = 0.003) and SCD (RR 0.36, 95% CI 0.23 to 0.55, p < 0.00001) in the ICD + OPT group compared with the OPT group. There was no difference in non-arrhythmic cardiac death (RR 0.95, 95% CI 0.41 to 2.18, p = 0.9) or non-cardiac death (RR 1.06, 95% CI 0.58 to 1.95, p = 0.84) between the groups. One trial reporting hospitalisations found a higher rate per 1000 months’ follow-up among those who received an ICD (11.3 vs. 9.4, p = 0.09), with higher HF hospitalisations (19.9% vs. 14.9%, p-value not reported).

• In one trial that assessed QoL using the HUI3, scores were lower in the ICD + OPT group than in the OPT group at baseline. Differences between the groups were not statistically significant at 3 years’ follow-up.

• One trial reported prespecified subgroup analyses for all-cause mortality. The HRs in all 12 of the subgroups were similar, with no statistically significant interactions.

People with non-ischaemic or idiopathic dilated cardiomyopathy

• Three RCTs compared ICD + OPT with OPT or ICD + OPT with amiodarone + OPT. Mean follow-up was between 24 months (two RCTs) and 29 months and sample size ranged from 103 to 458. One trial enrolled people with recent onset of disease. Over half to two-thirds of participants were in NYHA class II; in one trial the remaining participants were in NYHA class III, but in two trials around 15–21% were in NYHA class I. Mean LVEF ranged from 21% to 25%.

• Meta-analysis found no significant difference between ICDs and OPT or amiodarone in all-cause mortality (RR 0.77, 95% CI 0.52 to 1.15, p = 0.20), total cardiac deaths (RR 2.03, 95% CI 0.17 to 23.62, p = 0.57), non-arrhythmic cardiac death (RR 1.13, 95% CI 0.42 to 3.03, p = 0.81) or non-cardiac death (RR 0.65, 95% CI 0.13 to 3.29, p = 0.60). However a reduction was found in rate of SCDs (RR 0.26, 95% CI 0.09 to 0.77, p = 0.02) with ICDs.

• Two trials reported no significant difference in survival between groups.

• Two trials reported no significant differences in QoL, assessed using the QWBS and STAI or the SF-12 MCS and PCS and MLWHFQ.

• One trial reported six prespecified subgroup analyses for all-cause mortality. None of the differences between subgroups was statistically significant.

• Meta-analysis of the three cardiomyopathy trials and the non-ischaemic CHF subgroup of the SCD-HeFT trial found a statistically significant reduction in all-cause mortality (RR 0.74, 95% CI 0.58 to 0.93, p = 0.01) with ICDs compared with OPT or amiodarone.

People scheduled for coronary artery bypass graft surgery

• One trial compared ICD + OPT with OPT, although the pharmacological therapy would not be considered optimal by current standards. Mean follow-up was 32 months and 900 participants were randomised. The majority of participants were in NYHA class II or III and mean LVEF was 27%.

• No significant difference was found between groups in all-cause mortality (RR 1.08, 95% CI 0.85 to 1.38, p = 0.53), total cardiac deaths (HR 0.97, 95% CI 0.71 to 1.33, p = 0.84), non-arrhythmic cardiac death (HR 1.24, 95% CI 0.84 to 1.84, p = 0.28), non-cardiac death (RR 1.50, 95% CI 0.82 to 2.73, p = 0.19) or actuarial mortality at 4 years’ follow-up (HR 1.07, 95% CI 0.81 to 1.42, p = 0.64). The rate of SCD was lower in the ICD group but this did not reach statistical significance (HR 0.55, 95% CI 0.29 to 1.03, p = 0.06).

• HRQoL was higher among those receiving OPT than among those receiving an ICD + OPT for all measures and this was statistically significant for some perception of health transition, emotional role function, mental health, satisfaction with appearance and satisfaction with scar.

• HRs for the ICD group compared with the control group for all-cause mortality were found to be similar among 10 prespecified subgroups.

A broad population of people with mild to moderate heart failure

• One three-arm trial compared ICDs, amiodarone and placebo; all participants received OPT. Mean follow-up was 46 months and 2521 participants were randomised. Over two-thirds of participants were in NYHA class II, with the remaining participants in NYHA class III. Mean LVEF was 25%.

• All-cause mortality was significantly lower in the ICD + OPT group than in the placebo + OPT group (HR 0.77, 97.5% CI 0.62 to 0.96, p = 0.007). A significant reduction in total cardiac deaths (HR 0.76, 95% CI 0.60 to 0.95, p = 0.018) and SCD (compared with the placebo and amiodarone groups combined, RR 0.44, 95% CI 0.31 to 0.61, p < 0.00001) in favour of ICD was also found. There was no statistically significant difference between the ICD group and the placebo and amiodarone groups combined in the number of non-arrhythmic cardiac deaths (RR 1.14, 95% CI 0.88 to 1.48, p = 0.32) or deaths from non-cardiac causes (RR 0.92, 95% CI 0.66 to 1.27, p = 0.60).

• Little difference was found in QoL assessed using the DASI. Statistically significant differences in MHI scores and global health status at 3 and 12 months were not maintained at 30 months, and the difference in MHI score was not clinically meaningful. A significant decrease in perceptions of QoL was found using the SF-36 among people who had received an ICD shock within the previous month compared with those who had not received a shock.

• There was no interaction between ICD therapy (p = 0.68) and the cause of CHF (ischaemic or non-ischaemic) for all-cause mortality or other specified modes of death. There was a statistically significant interaction between ICD therapy and NYHA class: compared with placebo, ICDs reduced the risk of all-cause mortality, cardiac mortality and sudden death presumed to be ventricular tachyarrhythmic in people with NYHA class II, but not in those with NYHA class III. The interaction between ICD therapy and NYHA class was not statistically significant for HF (p = 0.29) or non-cardiac (p = 0.11) deaths.

Adverse events

• Adverse events were reported by all four RCTs that included those with previous ventricular arrhythmias. Up to 30% of people in the ICD groups reported adverse events, with most related to the placement and operation of the device. Rates in the OPT group appeared lower.

• The nine RCTs that included people who had not suffered a life-threating arrhythmia reported adverse event rates in the ICD group of between 5% and 61%, depending on the definition of adverse event and length of follow-up. Adverse event rates for the comparator treatment group were between 11.9% and 55% in the three RCTs reporting this. Lead-, electrode- or defibrillator generator-related problems affected 1.8–14% of people in the five trials that reported this.