When is an ICD Indicated in the
Primary prevention of Sudden Death:
A Review of the Evidence

David Newman

Associate Professor Faculty of Medicine
St. Michael's Hospital, University of Toronto,
Toronto, Canada

INTRODUCTION
Sudden cardiac death accounts for approximately 50% of all deaths from cardiovascular causes. In patients dying within 60 minutes of the onset of symptoms, 90% of the deaths follow an episode of malignant ventricular tachyarrhythmia. Currently, less than 5% of cardiac arrest victims survive to hospital discharge. This small number explains the expanding role of primary prevention, namely efforts to identify patients at risk for fatal ventricular arrhythmias and treat them prophylactically. Patients with known underlying structural heart disease, most commonly coronary disease or heart failure, have the highest risk for sudden cardiac death. Epidemiological data indicates that structural coronary artery diseases and their consequences are the cause of 80% of fatal arrhythmias with idiopathic dilated cardiomyopathy accounting for the second largest number of sudden deaths from cardiac causes. This review, will mainly deal with the data that exists regarding prophylactic defibrillator implantation in these two disease entities. The generic issues with respect to trials design, intervention choices, analytic issues and endpoint assessment will be discussed before the specific clinical trial data followed by some summary conclusions. The rare entities in which clinical 'artistry' is ahead of evidence-based approaches will be discussed briefly.

Although relevant, generic health policy issues will not be discussed in detail. For example, prophylactic ICD therapy could have implications for driving licensure. Regulatory authority disclosure is not generally mandated for all patients at risk who could receive a device. Since an ICD is not designed to prevent an arrhythmia, but only to treat its consequences, routine reporting to driving authorities for prophylactic implantation should not be required.

1. Epidemiology and/general comments
Clinicians faced with the task of trying to prevent premature death due to malignant arrhythmias are faced with a well known paradox. This paradox states that there is an inverse relationship between the proportion of patients at risk for malignant cardiac arrhythmias and their absolute number in the general population. Patients who have recovered from an out-of-hospital cardiac arrest in the setting of poor left ventricular function with inducible ventricular arrhythmias have a one-year risk for recurrence in the order of 80%. However, the number of such patients in the overall population is extraordinarily small. Sudden cardiac death in the whole population is a societal epidemiological problem; the Framingham data demonstrated that in 50% of cases, the first manifestation of coronary artery disease is sudden death. Primary prevention using ICD therapy is challenged to reach the appropriate cut point between what is practical and reasonable as filters for risk stratification in the population at risk. From the viewpoint of cardiac disease generally it is a secondary goal established after the identification of some cardiac abnormality. The goal is to make the intervention efficient at the same time as it can be statistically proven to be effective.

The single most important factor in post-MI patients remains the presence and degree of left ventricular systolic dysfunction expressed by the ejection fraction. The degree of functional impairment from heart failure, expressed by NYHA classification, is also a powerful predictor of cardiac death. The caveat is that, as the severity of heart failure increases, the proportion of deaths attrituble to pump failure also increases with a corresponding decrease in arrhythmia related sudden death. Over the years, electropharmacologically-guided invasive EP studies have fallen progressively out of favour, replaced by a degree of therapeutic nihilism. This is largely related to the remarkable technical and clinical advances achieved with empiric ICD deployment.

A complete review of the alternative, medical therapy with amiodarone is beyond the scope of this paper. It should be mentioned that there is an amiodarone-beta blocker synergy which was not appreciated at the time of the secondary prevention ICD trials, which established ICD superiority over amiodarone. The adjunctive role, prognostically, of mechanical revascularization interventions will be discussed below. Suffice to say that anti-ischemic, antiplatelet, ACE/ARB, LDL lowering therapy as well as aldosterone receptor blockade all interact towards the prevention of sudden death. Beta blockers, in particular, are required for all patients at risk for malignant arrhythmias.

2. Design Issues in Primary Prevention Trials
A. Selection Bias
The standard tool in support of prophylactic ICD therapy is the randomized controlled trial. This requires randomization and unambiguous endpoints ideally arbitrated by a committee blinded to the intervention. With two exceptions, MIRACLE (resynchronization pacing in heart failure) and NAVPAC II (pacing for vasovagal syncope), device trials do not have double or even single blinding of investigators to the intervention which can be obviated (with difficulty) by a blinded outcome adjudication process. Ideally, a registry is also available to assess the selection bias that went into patient recruitment and thereby address generalizabilty of the results. Among the prophylactic ICD trials that have been performed only the MUSTT trial failed to have randomization to an ICD intervention. The MUSTT trial (described in detail below) was an ICD case series embedded in a randomized trial. Reasonable clinical inferences have been made from the data even if, from a strictly scientific viewpoint, however, these were hypothesis generating observations.

B. Entry Filters
I) EF 'plus'
Left ventricular ejection fraction remains the single most important prognosticator for possible ICD use. At issue is whether the efficiency of this filter can be increased still further by employing other measures to enhance prognostic risk. The optimal method by which ejection fraction should be measured is unknown. In the MADIT-2 trial, both semi-qualitative (catheterization or echo using Simpson's or other rules) and quantitative measures were used. Blood-pool nuclear quantitative measures of ejection fraction are not operator dependent; more clearly look at the complicated geometry of a damaged ventricle with no anatomic presumptions and cover the entire three-dimensional structure. The COMPANION trial, borrowing from the heart failure research trials' methodologies, also used a heart failure hospitalization within the last year. This enriches the population in both morbidity and poor prognosis. Although the COMPANION entry window was an EF of 35%, perhaps related to the use of this selection filter, the mean EF in the trial was 22%.

II) ECG and Holter based
Generally ECG and Holter based methods remain more intriguing than used. The DINAMIT trial in 300, <30d post -MI patients used impaired scalar hear rate variability measures. No trials have used other measures such as T wave alternans, heart rate turbulence or alternate autonomic related measures such as barroreceptor slope. The MUSTT trial used non-sustained ventricular tachycardia plus EF < 35% as its added filter.

QRS width has been shown to independently predict mortality in a 'dose-dependent' fashion in the vesnarinone CHF trial. Its value as a mortality filter has been established retrospectively form the MADIT-2 trial and prospectively from the COMPANION trial. It is possible that the greater absolute 1 year mortality benefit observed in COMPANION (40% vs. 30%) was related to this QRS duration filter.

III) EP inducibility
The negative predictive value of invasive EP demonstrated inducibility in post-MI patients has been reassessed by the MUSTT registry data. In that data post MI patients with EF<35% who were not inducible still had 2 and 5 year sudden death rates of 12 and 24% respectively. The same data set and others however, showed that untreated patients with inducible VT had a 50% higher sudden death rate than those not inducible. This has also recently been observed in an analysis from the secondary prevention AVID registry. The final role of inducibilty testing therefore is still unclear; it may help to buttress the decision to implant an ICD (positive predictive value) with less reassurance than was previously thought if the patient is found to be non-inducible, especially if other risks are present ( lower negative predictive value).

IV) Surgical (complete) Revascularization
The only negative prophylactic ICD trial yet published is the CABG-PATCH trial. The CABG-PATCH trial randomized 900 post-MI patients to receive an ICD post bypass surgery. It did not have a benefit on mortality despite reasonable selection filters. CABG-PATCH patients had to have an ejection fraction of 35% and a positive signal-averaged ECG as a surrogate of inducible VT prediction. This suggests that there is an adjunctive prognostic benefit with revascularization. An alternative and contrary explanation notes that CABG-PATCH patients follow-up only started after the significant intervention of successful bypass surgery which itself presumably is a selection filter in favour of healthier patients. Indeed, half of MADIT-2 patients, who only had EF <30% as their selection filter, had undergone prior revascularization required to be at least three months prior to study entry. It is reasonable to suggest that any patient receiving prophylactic ICD in the setting of coronary disease should at least have the magnitude of inducible ischemia assessed and ideally be found to be is either modest or not amenable to mechanical relief.

C. Interventions
Implantable defibrillators now come in a variety of modes: single chamber, dual chamber, and with cardiac resynchronization therapy (CRT), all of which can be combined with a rate-responsive sensor. All of this will have an effect on secondary outcomes depending on design. Pending the DANPACE data, (N=5000, bradycardia platform AAI vs. DDD) it may be that the proportion of time ventricular paced may have an effect on mortality and on morbidity (atrial fibrillation induction or worsened heart failure). This has been inferred from both the MOST (percentage of time paced in RV apex), and the CTOPP data (pace dependency analysis), but was not established in the recently presented UK-PACE trial; all trials with bradycardia platforms. The DAVID trial randomized 506 patients to have 'forced' ventricular pacing via a short AV delay, showed that ventricular pacing in a dual-chamber ICD platform can exacerbate heart failure and increase hospitalization.

It should be recalled that the MADIT-2 prophylactic ICD trial also had an increased incidence of heart failure hospitalizations in the ICD arm. This could be due to the half the patients in this trial given VVI-ICDs where an unknown amount of ventricular pacing occurred, or alternatively that ICD related prolongation of life increased the opportunity for morbid events to occur. For the moment, the issue of the optimal ICD functionality to have in prophylactic ICD platforms is not totally established.

The recent COMPANION trial suggests that in patients with significant heart failure, and heart failure hospitalization within the last year, the addition of resynchronization therapy to the ICD platform will have a significant effect on morbidity measures. From a purely scientific viewpoint, upcoming trials such as the RAFT trial (which aims to randomize ICD recipients between investigator chosen single or dual chamber ICD and CRT+ ICD) will answer the question more formally. It is possible that the clinical as opposed to scientific interpretation of the COMPANION trial data may impact participation in RAFT. Unlike COMPANION, RAFT will allow atrial fibrillation to be present in either arm.

D. Outcome Measures
I) Mortality
In the early 1990s there was significant debate as to whether mortality is the sole or even appropriate arbitrator of efficacy for ICD therapies. This debate initially arose within the secondary prevention trials. There was a concern that competing causes of largely pump failure mortality will undermine ICD efficacy for presumed arrhythmic death. Cause-specific mortality especially if pre-established and determined using a blinded endpoint assessment committee remains useful. The most common method uses the Hinkle-Thaler criteria. One preliminary analysis using ICD stored telemetry data from the CIDS trial validated Hinkle-Thaler criteria as an outcome measure in these patients. Despite these efforts, and as articulated in a NASPE consensus conference statement on this matter, presumed arrhythmic death and other surrogates of cause-specific mortality assignment are unlikely ever to be sufficiently persuasive as a primary outcome measure.

II) Quality of life/morbidity and economic endpoints
The costs of ICDs have led to sophisticated pharmaco-economic evaluations and modeling. In these evaluations, the collection of careful cost data is an arduous task for which a variety of health economic models exist. Generally, the dialysis benchmark is used with a cost-per-life-year saved significantly over $100,000 USD per year generally felt to be exorbitant. In the last 10 years, there has been a significant interest in health-related quality of life as an alternative primary outcome measure, either alone or combined with economic cost (the QUALY or quality of life adjusted cost per year gained). There are a variety of well validated tools that have been applied. There are both generic as well as some limited disease-specific ICD quality-of-life instruments. An intriguing model is the health trade-off quality-of-life model in which patients essentially calculate how much more life they would like to have in their current situation.

The use of a combined morbidity and mortality outcome measure may have opposite directions and is a challenge of interpretation in clinical data. This is especially so if, as in the recent COMPANION trial, the study is over-sampled for a morbidity outcome and barely sized for a mortality outcome. In such a situation, the combination of morbidity (all-cause hospitalization) combined with a mortality outcome, may obscure the interpretation of benefit.

III) Inappropriate Shock
The secondary prevention CIDS ICD quality of life study, found that having more than six ICD shocks (be they appropriate or inappropriate) undermined the quality of life advantage seen in those randomized to ICD over amiodarone therapy. The actuarial first year estimate of inappropriate shock frequency in secondary prevention data is 20-40%. As prophylactic ICD implantation volumes progressively increase, the impact of inappropriate shock on morbidity may increase.

Some retrospective studies have suggested that a history of atrial fibrillation, relative youth, or concomitant antiarrhythmic medication (perhaps related to atrial fibrillation) are all predictors of inappropriate therapy. There are few prospective studies designed to assess strategies to decrease inappropriate therapy. One small 80-patient study suggested that the addition of an atrial sensor to the detection algorithm did not have an effect of decreasing inappropriate therapy. In contrast, the 280-patient ASTRID trial found that the one-year actuarial incidence of inappropriate therapy decreased from 48% to 24%. The ASTRID investigators concluded that an atrial-based algorithm for tachycardia discrimination enhances the specificity of discrimination. Whether template-based discrimination algorithms (all of which are variants on morphology such as the Medtronic wavelet function) will enhance discrimination without the need for atrial based sensing is unknown.

Specific Trials (see Table 1)

Table 1

MADIT-1
The first and most important proof of concept trial was MADIT I. This had a sample of only 200 patients and never provided any data with respect to the number of patients screened to arrive at the patient population studied, who had significant filters to entry ( EF <35%, inducible VT not suppressed by procainamide infusion). Although it led to a change in FDA approved indications coincident with its first presentation, it was not compelling and provided the justification for MADIT-2.

MADIT-2
MADIT-2 was in many ways, at the opposite extreme. It had a remarkably simple selection filter; an ejection fraction <30% in 1,232 post -MI patients who did not need, or already had, revascularization. MADIT-2 showed a 30% mortality reduction with ICD therapy. It is not surprising that this is a less powerful effect than what was seen with MADIT I with therefore a corresponding increase in cost per life you saved. MADIT-2 has fuelled the single largest increase in device utilization rates per million in all jurisdictions. MADIT-2 did show a slight increase in heart failure related hospitalization in patients randomized to ICD therapy perhaps related to the non-specific effect of prolonged life or an iatrogenic issue related to the single chamber ICD used in half the ICD population. Post-hoc analysis showed that all of the benefit was observed in those who had QRS duration over 120 ms. Although tempting, the role of QRS duration alone combined with low EF, as a selection filter has never been prospectively assessed.

MUSTT
The MUSTT trial recruited 704 post-MI patients with an EF <35% and non-sustained VT (a monitor observed ventricular triplet in 20% of the patients) over a 10-year period. Patients were randomized to EP guided therapy or usual care. The EP guided therapy arm was left up to the discretion of individual clinicians with respect to the timing and deployment of ICD therapy. This is an example of a trial that led to a conclusion without randomization of the intervention. It was found that those randomized to EP guided therapy who did not receive an ICD did significantly worse. As in many trials that used amiodarone in a portion of the control arm, the synergy of amiodarone plus beta-blocker combined use, now known from the EMIAT and CAMIAT analyses, was not appreciated . The most compelling support for ICD therapy from the MUSTT trial is the observation that the magnitude of benefit of those randomized to the EP-guided arm had a 'dose-response' increase according to centre preference for ICD therapy as their method for 'EP-guided' intervention.

CABG-PATCH
The CABG-PATCH trial randomized 900 patients to peri-operative ICD therapy (in the epicardial patch era). All patients had to have a pre-operative left ventricular EF < 35% and positive late potentials on a signal averaged ECG, a reasonable if imperfect surrogate of VT inducibilty. It stands alone as the only negative prophylactic ICD trial and as a result receives significant attention even if the medical therapies used are dated by current standards. The explanation for the discordant data is unclear. The differences between CABG-PATCH and all other trials may relate to design since all follow-up began only after selection for, and survival from surgery. Alternatively, and as supported by secondary prevention data, CABG-PATCH suggests that there is significant protective and non-specific antiarrhythmic effect that accrues for complete revascularization.

COMPANION
The recently presented (ACC 2003) COMPANION trial has solidified further the utility of prophylactic ICD therapy. The COMPANION trial attempted to establish the relative importance of resynchronization pacing therapy to an ICD platform. As a result it is the first prophylactic ICD trial to also have a significant emphasis on indices related to morbidity. The COMPANION investigators hypothesized that an ICD+CRT should produce a double effect, on mortality from the ICD, and morbidity from the CRT components of the combined system. In COMPANION, 1,604 patients were asymmetrically randomized between usual care, CRT only, and ICD+CRT. It was found that all of the morbidity benefits were due to the CRT arm, with a 35% decrease in heart failure hospitalization and a 20% decrease in all hospitalization. The mortality benefit however, was largely ( but not solely) related to the ICD platform with a 20% decrease in mortality in CRT only, compared to a 40% decrease in mortality in those randomized to the CRT + ICD arm. These reductions in mortality were only significant for ICD+CRT. Twelve month absolute actuarial mortality fell from 20% to 15% to 10% in usual care, CRT only, and CRT + ICD respectively. Being the most recent entry, COMAPNION is also relevant to conventional clinical practice with 70% of patients on beta blockers, over 50% on spironolactone and 85% on ACE/ARB. Post COMPANION, it remains to be seen whether further trials will be performed answering more scientifically the question with respect to the role of CRT therapy in ICD systems. The proposed RAFT trial intends to randomize CHF patients, including those with atrial fibrillation between clinician chosen ICD platform and ICD+CRT.

Non-Ischemic (idiopathic) dilated cardiomyopathy (IDC)
Almost all of the compelling information on prophylactic ICD therapy relates to patients with ischemic cardiomyopathy. There are three specific data sets with respect to dilated cardiomyopathy available in the literature. The recently completed COMPANION trial had 45% of its sample (N of 720) having non-ischemic dilated cardiomyopathy. Although not all the data has yet been presented, the multivariate analysis by sub-groups showed that the hazard ratio of 0.4 for risk reduction of mortality by ICD therapy occurred with a similar magnitude regardless of heart failure aetiology. The confidence interval surrounding this hazard ratio crossed the line of identity in the dilated myopathy group, presumably related to smaller number of outcomes in this population. This is now the most potent data in support of ICD platforms in patients receiving bi-ventricular pacing for heart failure and it undermines the utility of bradycardia platform only CRT therapy in this population.

Two small trials specifically addressing the utility of ICD therapy in only non-ischemic dilated cardiomyopathy patients have both been negative. The AMIOVERT trial had complex design issues with a modest sample size of only 80 patients. Similarly, the CAT trial also had negative results with respect to the utility of ICD therapy in this population but again with a sample size of only 100 patients. The SCDHeFT trial of 2,500 patients has, like COMPANION, close to half of its sample size represented by non-ischemic dilated cardiomyopathy patients. This trial is due to be presented at the 2004 American Heart Association annual meetings and will provide important information therapy on the utility of ICD therapy in this population. SCDHeFT randomized patients between amiodarone or placebo and ICD. Its entry filters required an EF of 35% and a less than two-year window since the onset of >NYHA class II heart failure. All ICD therapy in SCDHeFT is VVI only, with the trial recruiting for a longer time period. The data and safety monitoring board of SCDHeFT have prolonged the trial's follow-up phase so as to enrich the sample in outcome events. Until SCDHeFT data is presented, digested and understood, the COMPANION data set remains the largest single data set supporting, albeit with some reservation, the utility of ICD therapy in the non-ischemic dilated cardiomyopathy population who would also benefit from CRT therapy.

Special Disease Entities
Channelopothies
There is relatively little evidence-based randomized controlled data on the utility of prophylactic ICD therapy in more rare sub-groups. The channelopothies (long-QT syndrome and Brugada syndrome) have received the greatest attention and effect younger patients. The international long-QT registry data would suggest that almost all patients with long-QT syndrome (the exception being the less common LQT3 genotype) will have had their clinical event by the age of 40 and that there is a female predisposition to malignant events in this population. This registry has also established the utility of beta blocker therapy. It is not known if ICD therapy is superior to beta blocker therapy in this population. It must be recalled that in this relatively young population current ICD battery longevity in 5-6 years, coupled with the psychological and health-related quality of life consequences of inappropriate shocks mandate individualized decision making in first degree effected relatives who are beta blocker naive. Whether Brugada syndrome patients can be risk-stratified by invasive EP studies is controversial. The data by Priori et al contrasts with the clinical recommendations of Brugada et al. At issue is the specificity of the induction of non-sustained or sustained polymorphic ventricular arrhythmias in patients who have a native or provoked Brugada phenotype but no events. It is possible that an ILR implant prior to ICD decisions could be useful to obtain more information on the magnitude of ambient ectopy and risk stratification in this population.

Chagas disease
Chagas disease is a unique form of non-ischemic dilated cardiomyopathy for which there is little controlled data with respect to ICD utilization. Certainly the Argentinean GESSICA trial of amiodarone vs. medical therapy, which had a significant proportion of patients with Chagas disease suggested that this group has risks and benefits similar to the dilated cardiomyopathy population; however, for the moment therapy must remain individualized with respect to ICD decisions.

Hypertrophic cardiomyopathy
Hypertrophic cardiomyopathies have an enhanced risk for sudden death. It is unknown if pharmacologic or surgical relief of the outflow tract gradient has an impact on malignant ventricular arrhythmias in this disease. There was initial promise that genotypic information could predict predicting phenotypic (sudden death) expression. This long awaited (or feared) promise remains elusive. As in congenital long QT syndromes, there is sufficient overlap as well as variation in the genotypic abnormalities to make this approach unlikely to be clinically feasible. As a result, most ICD decisions, are in those with familial cardiomyopathy, based on the non-specific tendency in a given kindred for the phenotypic expression to be that of a malignant arrhythmia. In patients who do not have a familial cardiomyopathy, prophylactic ICD decisions are generally challenging and not often performed. Efforts in such patients to risk stratify by other phenotypic markers such as LV septal thickness over 25mm, non-sustained ventricular tachycardia on Holter or syncope are used but have had relatively limited prospective validation.

Post-Tetralogy of Fallot repair
A burgeoning population is patients with repaired congenital heart disease, in particular the post-Tetralogy of Fallot repair group. For now, the risk stratification of these patients is based on ambient ectopy as well as evidence of right ventricular pressure and volume overload. The latter is common especially in the early era of surgical repair, consequent to post-operative free pulmonary insufficiency. Further risk stratification among this particular sub-group is difficult and challenging given the very low event rate in young patients at risk. Theoretically non-endovascular subcutaneous staged 'shock boxes', could have an important role in this population.

SUMMARY AND CONCLUSIONS
Prophylactic ICD therapy clearly has a role, especially in the post-coronary-disease population. Two significant issues that remain outstanding are whether the efficiency of this proven intervention could be increased by employing further selection filters beyond ejection fraction only. The other remaining compelling question is the device platform to be deployed in patients with heart failure symptomatology. For those who are at the extreme of heart failure symptomatology (meeting COMPANION criteria with heart failure hospitalization within the last 12 months, QRS duration >130 ms, LVEDD >60 with EF <35), prophylactic ICD therapy coupled to CRT therapy is reasonable.

In patients without COMPANION entry criteria, whether the MADIT-2 data alone is sufficient to warrant prophylactic ICD implantation remains controversial only because of health economic issues. Quantitative ejection fraction, absence of inducible ischemia or absence of mechanical opportunities for resolving significant amounts of inducible ischemia are reasonable filters to consider. Although it is preferred to see patients with QRS durations over 120ms, it is difficult to use this as a sole criterion to exclude. Invasive studies may buttress a decision, or perhaps help in borderline cases, but do not as a routine get performed. The use of enhanced selection filters speaks to the inefficiency of an intervention not to its efficacy. One strategy, employed by the on- going AIRBAG registry (Biotroinc device) would use a simple ICD 'shock box' to deliver a limited number of life-saving shocks upon which upgrading to a more sophisticated device at greater cost can be considered.

As always and, at the end of an evidence-based review, it is important to emphasize that an individual discussion to weigh patient specific issues is always needed. Clinically, evidence based medicine must always inform clinical thinking and is not be its sole arbitrator.

FORM DEACTIVATED SINCE
December 1st., 2003

Your questions, contributions and commentaries will be answered
by the lecturer or experts on the subject in the Arrhythmias list.
Please fill in the form and press the "Send" button.