Myocardial infarction (MI) is associated with acute risk of early malignant arrhythmias that can be easily treated during in-hospital intensive care by defibrillation, adjuvant antiarrhythmic therapy or even catheter ablation in resistant cases. Indeed, such management resulted in substantial improvement in MI survival rate. Despite the implementation of primary percutaneous coronary interventions, optimised medical treatment, delayed complete revascularisation if technically achievable (directed by the assessment of residual ischemia and regional myocardial viability) and cardiac resynchronisation therapy in selected cases, MI may frequently establish the substrate for fatal ventricular tachycardia/fibrillation (VT/VF) and sudden cardiac death (SCD) in the long term, significantly extending beyond the hospitalisation period.1
Secondary Prevention of Sudden Cardiac Death
When such serious clinical event occurs remotely (>48 hours) from acute MI, in the form of non-fatal cardiac arrest due to VT/VF, hemodynamically unstable VT, recurrent stable VT or unexplained syncope with clinical work-up suggesting a high probability that a VT/VF was the cause of the syncope, implantable cardioverter– defibrillator (ICD) is recommended for secondary prevention of SCD.2 Current recommendations are based on the results of three randomised clinical trials3–5 and their meta-analyses.6,7 In summary, ICD is recommended for secondary prevention of SCD in patients with structural heart disease who are receiving chronic optimal medical therapy and have significant left ventricular (LV) dysfunction (recommendation class I, level of evidence A). Such therapy is also reasonable for those with normal or near-normal LV ejection fraction (LVEF) (recommendation class IIa, level of evidence C).
The applicability of other stratifiers, including LVEF, in risk stratification for secondary prevention of SCD is significantly less supported by clinical data compared with primary prevention of SCD. Despite the fact that patients with relatively preserved LV function, even in the secondary preventative setting, may not have better survival when treated with the ICD compared with antiarrhythmic drugs,8 it is extremely unlikely that this matter will be investigated in a randomised fashion in future.
Advent of Primary Prevention Implantable Cardioverter–Defibrillator Trials
LV systolic dysfunction defined by depressed LVEF is a readily available (although not overriding) risk factor for SCD. This is why a number of randomised clinical trials were performed to establish its clinical utility in risk stratification of patients with structural heart disease. Specifically, the LVEF has been a major determinant for entry into these studies that were aimed at investigating the efficacy of ICD compared with antiarrhythmic or best medical therapy in primary prevention of SCD. Several invasive or non-invasive risk predictors were used as adjuvant risk markers in order to improve the selection of high-risk patients.
Historically, the Multicenter Automatic Defibrillator Implantation Trial (MADIT)9 was the first randomised clinical trial that compared ICD therapy with conventional care in a primary preventative setting, i.e. in patients with no prior history of life-threatening arrhythmias. Patients with ischaemic LV dysfunction (LVEF ≤35%), non-sustained VT and inducible, non-suppressible sustained ventricular arrhythmias during programmed electrical stimulation (PES) were enrolled. MADIT demonstrated 54% relative mortality reduction (p=0.009) in the ICD arm.
The Coronary Artery Bypass Graft Patch (CABG Patch)10 trial compared ICD therapy with usual care in patients undergoing coronary artery bypass grafting, having LV dysfunction (LVEF ≤35%) and the presence of late potentials assessed by signal-averaged electrocardiograph (ECG) recording. Although ICD significantly reduced the risk of SCD, no significant reduction in mortality was observed with ICD therapy, probably due to the impact of revascularisation.
The Multicentre Unsustained Tachycardia Trial (MUSTT)11 compared antiarrhythmic therapy with best medical therapy in patients with ischaemic LV dysfunction (LVEF ≤40%), non-sustained VT and inducible sustained ventricular arrhythmias during PES. ICD therapy could only be used in a non-randomised fashion after the failure of serial electrophysiological (EP) testing with antiarrhythmic drugs. The survival benefit in the EP-guided therapy group (significant 27% relative reduction [RR] in cardiac arrests/arrhythmic deaths compared with conventional care) was limited to patients who received an ICD. Drug therapy alone was not associated with a reduction in mortality compared with best medical therapy.
The MADIT II12 was the first trial that included patients primarily based on advanced LV dysfunction (LVEF ≤30%) late after MI without the need for invasive EP testing. Prophylactic ICD implantation was associated with significant reduction (31%; p=0.016) of all-cause mortality compared with the conventional therapy group. The survival benefit was similar in strata according to age, sex, LVEF, New York Heart Association (NYHA) class and the QRS width.
The Sudden Cardiac Death in Heart Failure Trial (SCD-HeFT)13 was designed to evaluate the hypothesis that amiodarone or an ICD would reduce the risk of death from any cause in a broad population of patients with mild to moderate ischaemic or non-ischaemic congestive heart failure (LVEF ≤35%) with NYHA class II–III. Patients receiving ICD had a significantly lower risk of death (RR of 23%; p=0.007) compared with those in the placebo arm. Compared with placebo, amiodarone was associated with a similar risk of death. The results did not vary according to either ischaemic or non-ischaemic causes of congestive heart failure, but they did vary according to NYHA class: ICD therapy had a significant benefit in patients in NYHA class II, but not in those in NYHA class III. However, this unanticipated subgroup effect, which was in conflict with the results of other ICD trials, was not assumed to be a sufficient basis for withholding ICD therapy from patients in NYHA class III.
Early Post-myocardial Infarction Implantable Cardioverter–Defibrillator Trials
While the benefit of ICD therapy in chronic post-MI patients was firmly established, patients early after MI attracted the attention. Due to the exponential shape of survival curves after MI, with the highest risk of SCD in the first month and later rapid decline, the hypothesis that early implementation of ICD therapy would enhance the mortality reduction seemed relevant.
The Defibrillator in Acute Myocardial Infarction Trial (DINAMIT)14 was designed to test whether prophylactic implantation of an ICD would reduce mortality in survivors of a recent MI (six to 40 days after MI) with LV dysfunction (LVEF ≤35%) and either elevated mean heart rate ≥80bpm or depressed heart rate variability (standard deviation of all normal to normal RR intervals [SDNN] ≤70ms) as assessed by 24-hour Holter monitoring performed at least three days after the MI. No difference was found in overall mortality between the ICD and conventional therapy arms. Although ICD therapy was associated with a reduction in the rate of death due to arrhythmia, it was offset by an increase in the rate of death from non-arrhythmic causes.
The Immediate Risk Stratification Improves Survival Trial (IRIS) randomised the population of patients with recent MI (six to 40 days) to ICD or conventional therapy.15 Patients were required to have LVEF ≤40% and a heart rate of ≥90bpm on ECG obtained within 48 hours after the MI. Alternatively, they had to meet the criterion of non-sustained VT (≥3 beats, ≥150bpm) during Holter ECG monitoring on days five to 31 after MI irrespective of LVEF. The results completely mirrored those observed in DINAMIT trial.
Why Did Early Post-myocardial Infarction Implantable Cardioverter–Defibrillator Interventions Fail?
Some plausible explanations may be offered for the failure of ICDs to alter the clinical outcome of patients early after MI, as summarised in part by Goldberger et al.16
First, malignant ventricular arrhythmias relatively early after MI are simply an epiphenomenon of cardiac instability due to multiple factors related to or associated with myocardial injury due to a recent index event. They may include the risk of in-stent restenosis/occlusion of infarct-related artery, unstable coronary plaques in remote coronary arteries, residual ischaemia, subclinical microinfarctions, process of adverse cardiac remodelling evidenced at multiple levels (gene expression, molecular, cellular, interstitial, macroscopic changes, haemodynamics and neurohumoral activation) leading to progression of heart failure. In such scenarios, the ICD therapy is not directed towards underlying pathophysiology and, therefore, is obviously symptomatic, ineffective and only able to delay death non-significantly and/or to change the mode of death. This concept is supported by a retrospective analysis of MADIT-II where no benefit from ICD implantation was observed when the ICD was implanted within 18 months (the lowest quartile) of the MI, with overall median implantation time of 60 months after the index event.17
Second, selection of patients for DINAMIT and IRIS trials was based not only on LVEF but also on other risk factors reflecting impaired cardiac autonomic regulations and/or predisposition to ventricular arrhythmias. It was believed that such inclusion criteria would help to select a high-risk population, increase the number of end-points and, consequently, find the appropriate target population for ICD therapy. Unfortunately, such selection may identify patients who are susceptible to primarily non-arrhythmic causes of death in the early post-MI period, which cannot be prevented by ICD implantation.
Third, there were significant differences in population characteristics between ‘historical’ ICD primary prevention trials and early post-MI ICD trials. For example, because of different enrolment periods (mid-enrolment of 1999 for MADIT-II, 2001 for DINAMIT and 2003 for IRIS), significant improvement is noticeable in medical therapy, namely beta-blockers and angiotensin-converting enzyme (ACE) inhibitors, in favour of trials performed later. The most striking difference can be found in reperfusion therapy, which was commenced in 87% patients in the IRIS trial (primary percutaneous coronary intervention [PCI] in the majority of cases), in 63% patients in the DINAMIT trial and probably at very low frequency in the MADIT-II trial (not reported), as many of the index events in this trial occurred in the early to mid- 1990s. That is why arrhythmogenic substrate and, consequently, the benefit from ICD therapy may differ remarkably between studies.
Fourth, it is also known that a significant variation in LVEF exists in the period of weeks/months after MI. LV systolic function detected in the early post-MI period generally tends to substantially improve. Thus, the DINAMIT and IRIS trials likely followed a significant proportion of patients with less impaired LV systolic function compared with that assessed at baseline.
Fifth, potential deleterious effects of ICD implantation, testing and subsequent appropriate or inappropriate shocks cannot be excluded; these effects may occur with greater frequency in the setting of healing versus healed MI.
Primary Prevention of Sudden Cardiac Death
All these trials, together with complementary meta-analyses,18,19 provided the basis for the evidence-based indication criteria for prophylactic ICD therapy that were summarised in the 2006 American College of Cardiology/American Heart Association Task Force on Practice Guidelines and the European Society of Cardiology (ACC/AHA) guidelines for the management of patients with ventricular arrhythmias and the prevention of SCD.2 Due to the inconsistency between trials in cut-off LVEF values for enrolment and averaged values of LVEF, and subgroup analyses that were discordant in their implications, the guideline-writing committee decided to construct recommendations to apply to patients with a LVEF less than or equal to a range of values. In general, it is important to realise that the recommendations are applicable to patients who have a reasonable expectation of survival with a good functional status for more than one year.
Specifically, ICD therapy is recommended for primary prevention to reduce total mortality by a reduction in SCD in patients with LV dysfunction due to prior MI (LVEF ≤30–40) and NYHA class II–III (recommendation class I, level of evidence A), and is reasonable in patients with LV dysfunction due to prior MI (LVEF ≤30–35%) and NYHA class I (recommendation class IIa, level of evidence B). Essentially, evaluation should be deferred until at least three months after revascularisation procedures and, more strictly, until at least 40 days after acute MI to allow adequate time for recovery of LV function. In cases of borderline EF (30–40%), EP testing (programmed ventricular stimulation) may be useful.
Non-invasive Electrocardiogram-based Risk-stratification Techniques
While LVEF is an excellent marker of risk for total mortality, no study has demonstrated that reduced LVEF is specifically related to SCD. Due to the relatively low cost-effectiveness of ICD therapy in the post-MI population selected primarily based on LVEF, and because this therapy carries considerable risk of complications and results in a decreased quality of life, extensive research continues to search for the holy grail marker or a combination of markers that would allow the identification of patients at risk of preventable arrhythmic death.
In the field of non-invasive electrophysiology, numerous ECG-based risk stratifiers have been investigated. They were proposed to characterise arrhythmogenic substrate during depolarisation (QRS duration, late ventricular potentials), repolarisation (QT duration, dispersion, variability and dynamicity, T-wave variability and alternans) and cardiac autonomic regulations (heart rate recovery, variability, turbulence and baroreflex sensitivity). The utility and caveats of all methods are comprehensively reviewed in the AHA/ACC/HRS scientific statement on non-invasive risk-stratification techniques for identifying patients at risk of SCD.20 Due to space limitations we will focus on heart rate turbulence (HRT) and microvolt T-wave alternans (MTWA): both appeared promising and a sufficient amount of contemporary clinical data exists confirming their predictive power.
HRT is a recently recognised electrocardiographic and autonomically mediated phenomenon (baroreflex-based short-term fluctuations in sinus cycle length) reflecting minute haemodynamic disturbances caused by isolated ventricular premature complexes (either spontaneous or induced).21
Standards of HRT measurement, its pathophysiological background and independent risk-stratification utility confirmed in several large-scale retrospective and prospective post-MI studies have been reported.22 HRT is conceptually related to deceleration capacity of heart rate, which later was proposed as a powerful post-MI risk stratifier that quantifies non-stationary, quasi-periodic heart rate decelerations in long-term ECG recordings by means of a so-called phase-rectified signal averaging algorithm.23
MTWA indicating beat-to-beat variations of the T-wave amplitude was shown to be an independent marker of risk for SCD. Alternative behaviour is a normal rate-dependent property of cardiomyocytes that develops at markedly slower heart rates in the setting of heart disease. The most widely applied method for non-invasive measurement of MTWA uses the spectral method during controlled HR elevation.24 Recently, a time-domain measurement technique referred to as the modified moving-average method has been proposed as an alternative to the spectral method for measuring MTWA.25 Several initial retrospective studies on MTWA suggested high negative predictive value of MTWA in patients with impaired LVEF,26 which could potentially be useful for risk stratification in MADIT-II-like patients, leading to improved cost-effectiveness of ICD therapy. Nevertheless, subsequent prospective trials in ICD-treated patients provided controversial results.27–29 However, the results were significantly biased (obscuring the predictive value of MTWA for SCD) when ICD-detected VTs and ICD therapies were used as surrogate end-points for SCD.30
An interesting observation was offered by the Risk Estimation Following Infarction, Non-invasive Evaluation (REFINE) study, which prospectively sought to determine whether combined assessment of autonomic modulation and cardiac electrical substrate identifies most patients at risk of serious events after MI with a less strict cut-off of LVEF <50%.31 MTWA and HRT combined with LVEF were able to identify a considerable proportion of high-risk patients (in terms of cardiac death/arrest, cardiac arrest and all-cause mortality) with reasonable sensitivity and positive predictive accuracy. Importantly, risk stratification was only effective when performed at 10–14 weeks after MI and completely useless in an earlier phase (two to four weeks post-MI).
Back to Basics
Even simple clinical characteristics might be very helpful in assessing who will and who will not benefit from primary preventative ICD implantation. Thus, the final decision of whether to implant an ICD for primary prevention must include an individualised assessment of competitive risk.
A simple stratification score was proposed in the MADIT-II population.32 The benefit was not present in a small proportion of patients with advanced renal failure (urea >17.9mmo/l and/or creatinine >221umol/l). The benefit was also absent in a larger subgroup of patients (29%) with none of following risk factors: >NYHA class >II, age >70 years, urea >9.3mmol/l, QRS >120ms and observable atrial fibrillation. The U-shaped curve of ICD efficacy constructed according to a five-factor risk score also indicated a pronounced ICD benefit in intermediate-risk patients and attenuated efficacy in lower- and higher-risk subsets.
Another example is modification of the previously validated Seattle Heart Failure Model,33 which may be useful in stratifying benefit from ICD implantation. This model, which is able to predict SCD and non- SCD rates based on routinely collected clinical variables, was applied to the SCD-HeFT population. When the population was grouped into quintiles according to four-year all-cause mortality, the ICD benefit was clearly missing in the highest-risk group.34
While ‘old-fashioned’ EP testing might not be predictive in patients with severely impaired LV function,35 it still has a role in patients with better preserved LV function.28,36 This stratification tool, which principally should be fairly specific for ventricular arrhythmias, was used in the Beta-Blocker Strategy Plus ICD (BEST-ICD) trial.37 In this trial, patients within five to 30 days of an acute MI with an LVEF <35%, pre-selected by the presence of non-invasive risk markers, were randomised to standard medical therapy or to an EP-study-guided ICD strategy. Although the trial was negative (terminated prematurely because of poor enrolment), the two-year mortality rate in the standard medical therapy group was 29.5 versus 20% (p<0.20) for the EPS-guided ICD strategy group.
Conundrum in Risk Stratification
Post-MI risk stratification aims at a moving target. Characteristics of post-MI populations are continuously changing because of advances in acute reperfusion techniques (from no therapy to thrombolysis and to primary percutaneous coronary interventions) and long-term medical treatment, including beta-blockers,38 ACE inhibitors, angiotensin-receptor blockers, lipid-lowering agents and aldosterone inhibitors. Such progress in therapy definitely affects the formation of arrhythmogenic substrate, changes the proportion of SCD, results in improved prognosis and, consequently, alters the predictive power of individual risk markers. This renders any risk-investigating study, which require a reasonably long follow-up period, somewhat outdated because of the significant changes in treatment strategies that occur in the meantime. The prolonged follow-up ensures a sufficient number of end-points and likely increases the positive predictive power of individual risk markers. On the other hand, the duration of follow-up itself competes with the predictive power of risk factors, because the result of the test that was used for risk stratification at baseline may no longer be valid if the medical condition changes during the follow-up. How often tests should be repeated is generally unknown. This matter is particularly pertinent in the early period after acute MI when significant LV remodelling takes place. Moreover, because the prognosis of post-MI patients has significantly improved in recent decades, screening processes to identify high-risk individuals are relatively demanding in terms of costs, especially when a multifactorial risk assessment is needed.
The majority of risk stratifiers are not able to predict individual modes of death, namely to differentiate between SCD and non-SCD. In addition, even in well-conducted clinical trials there is more or less uncertainty about the exact mode of death in particular patients, which adversely influences the correct association between risk predictors and the true end-point. Observational trials, which follow a certain proportion of patients with ICDs, suffer from a change of a patient’s natural course because of an implanted device.
Many risk stratifiers were retrospectively investigated in relatively small, single-centre populations with different cut-off values (frequently defined post hoc) and with conflicting, non-reproducible and non-validated conclusions. Some ECG-based risk-stratification tests that require stable sinus rhythm (heart rate variability), narrow QRS complex (late potentials) or a certain degree of exercise because of need for heart rate acceleration (MTWA) are not fully applicable to a non-selected post-MI population. The risk-stratification strategies may be also invalidated when technology that is not ordinarily available in clinical practice is required for use.
The final goal of risk stratification is not merely to identify individuals at high risk of death – cardiac death, not just arrhythmic death – but to identify those who are at high risk of preventable death, particularly those likely to benefit from a specific treatment. Despite the fact that numerous risk stratifiers and their combinations have proved to be effective in the selection of high-risk patients, the fact remains that a randomised clinical trial is the only way to prove survival benefit in high-risk patients using non-invasive test results. There are several ways in which risk-stratification strategies may be developed in the future.
First, only a fraction of the patients who qualify for a primary prevention ICD under current guidelines are implanted with these devices for several reasons, including the relatively low cost-effectiveness of the therapy and patient/physician reluctance. It seems reasonable to develop risk-stratification strategies with excellent negative predictive value that allow ICD implantation in low-risk patients to be deferred in order to significantly limit ‘overtreatment’ and prevent ‘undertreatment’ as much as possible. However, it might be argued that such an approach goes against class I indications according to established guidelines, and ethical/legal obstacles are likely to occur, especially when such strategies need to be prospectively validated in a randomised fashion. Furthermore, large-scale randomised trials cannot be performed without industry sponsorship that is conditioned by the fundamental expectation that the particular trial will expand (not restrict) currently accepted guidelines.
Second, most sudden deaths occur in patients with LVEF >35%. In other words, significantly impaired LVEF itself, although currently criticised for being too broad, in fact misses most patients who may benefit from an ICD. Therefore, the design of future ICD trials targeting post-MI patients with mild to moderate LV dysfunction is realistic. Undoubtedly, entry criteria based on a combination of already established risk stratifiers will have to be carefully specified in order to have enough power to detect the benefit from ICD intervention in an acceptably sized patient population that may potentially result in clinically applicable results in terms of cost-effectiveness.
Third, besides clinical and ECG-based risk factors, biomarkers,39 genetic determinants and the amount of myocardial scoring assessed by magnetic resonance imaging40 might provide important prognostic information in future.
Reduced LVEF has been a major determinant for entry into primary preventative ICD trials. Numerous other stratifiers are currently being investigated in order to determine the efficacy of ICD therapy. More specific selection of patients at risk of preventable cardiac death, based on more than simply LVEF, is crucial for the future development of cost-effective prophylactic treatment aimed at closing the gap between scientific evidence and the limited resources of healthcare systems.